1 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3 * config/aarch64/aarch64-simd.md
4 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
6 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
10 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
14 2018-07-24 Jakub Jelinek <jakub@redhat.com>
17 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
18 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
19 and abs_d == d, do the power of two handling if profitable.
21 2018-07-24 Richard Biener <rguenther@suse.de>
23 * match.pd: Add BIT_FIELD_REF canonicalizations.
25 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
28 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
30 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
32 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
33 terminated STRING_CST object.
35 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
37 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
39 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
41 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
42 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
43 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
44 the elements into a register.
45 (rs6000_split_v4si_init_di_reg): Delete.
46 (rs6000_split_v4si_init): Delete.
47 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
48 (vsx_init_v4si): Rewrite as a define_expand.
50 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
52 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
53 zero_extend argument from memory): New.
55 2018-07-22 Martin Sebor <msebor@redhat.com>
58 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
59 diagnosing calls with unknown arguments unless -Walloca-larger-than
60 is restricted to less than PTRDIFF_MAX bytes.
62 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
64 * doc/gcov.texi (Invoking Gcov): Editorial changes.
66 2018-07-20 David Malcolm <dmalcolm@redhat.com>
68 * pretty-print.c (text_info::set_location): Remove redundant
69 "line_table" parameter from call to rich_location::set_range.
71 2018-07-20 Martin Sebor <msebor@redhat.com>
74 * builtins.c (expand_builtin_alloca): Adjust.
75 * calls.c (alloc_max_size): Simplify.
76 * cgraphunit.c (cgraph_node::expand): Adjust.
77 * common.opt (larger_than_size, warn_frame_larger_than): Remove
79 (frame_larger_than_size): Same.
80 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
81 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
83 * doc/invoke.texi (GCC Command Options): Document option arguments.
84 Explain byte-size arguments and suffixes.
85 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
86 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
87 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
88 * doc/options.texi (UInteger): Expand.
89 (Host_Wide_Int, ByteSize): Document new properties.
90 * final.c (final_start_function_1): Include sizes in an error message.
91 * function.c (frame_offset_overflow): Same.
92 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
93 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
94 Diagnose unbounded alloca calls only for limits of less than
96 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
97 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
99 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
100 only for limits of less than PTRDIFF_MAX.
101 * langhooks-def.h (lhd_handle_option): Change function argument
103 * langhooks.c (lhd_handle_option): Same.
104 * langhooks.h (handle_option): Same.
105 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
107 (var_type, var_type_struct): Same.
108 (var_set): Handle ByteSize flag.
109 * optc-gen.awk: Add comments to output to ease debugging. Make
110 use of HOST_WIDE_INT where appropriate.
111 * opts-gen-save.awk: Use %lx to format unsigned long.
112 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
113 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
114 arguments. Parse bytes-size suffixes.
115 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
116 (enum_value_to_arg): Same.
117 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
118 (handle_option): Adjust.
119 (generate_option): Change function argument to HOST_WIDE_INT.
120 (cmdline_handle_error): Adjust.
121 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
122 (set_option): Change function argument to HOST_WIDE_INT.
123 (option_enabled): Handle cl_host_wide_int.
124 (get_option_state): Handle CLVC_SIZE.
125 (control_warning_option): Same.
126 * opts.c (common_handle_option): Change function argument to
127 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
128 OPT_Wvla_larger_than_.
129 * opts.h (enum cl_var_type): Add an enumerator.
130 * stor-layout.c (layout_decl): Print a more meaningful warning.
131 * toplev.c (output_stack_usage): Adjust.
133 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
135 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
136 call to inline_expand_builtin_string_cmp.
137 (expand_builtin_strcmp): Likewise.
138 (expand_builtin_strncmp): Likewise.
139 (inline_string_cmp): Delete the last parameter, change char_type_node
140 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
142 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
143 the inlining expansion on target where the type of the call has same or
144 narrower precision than unsigned char.
146 2018-07-20 David Malcolm <dmalcolm@redhat.com>
148 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
149 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
150 * common.opt (fsave-optimization-record): New option.
151 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
152 * doc/invoke.texi (-fsave-optimization-record): New option.
153 * dumpfile.c: Include "optinfo-emit-json.h".
154 (struct kv_pair): Move to coretypes.h.
155 (optgroup_options): Make non-static.
156 (dump_context::end_scope): Call
157 optimization_records_maybe_pop_dump_scope.
158 * dumpfile.h (optgroup_options): New decl.
161 * optinfo-emit-json.cc: New file.
162 * optinfo-emit-json.h: New file.
163 * optinfo.cc: Include "optinfo-emit-json.h".
164 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
165 (optinfo_enabled_p): Check optimization_records_enabled_p.
166 (optinfo_wants_inlining_info_p): Likewise.
167 * optinfo.h: Update comment.
168 * profile-count.c (profile_quality_as_string): New function.
169 * profile-count.h (profile_quality_as_string): New decl.
170 (profile_count::quality): New accessor.
171 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
172 and optinfo_emit_json_cc_tests.
173 * selftest.h (selftest::json_cc_tests): New decl.
174 (selftest::optinfo_emit_json_cc_tests): New decl.
175 * toplev.c: Include "optinfo-emit-json.h".
176 (compile_file): Call optimization_records_finish.
177 (do_compile): Call optimization_records_start.
178 * tree-ssa-live.c: Include optinfo.h.
179 (remove_unused_scope_block_p): Retain inlining information if
180 optinfo_wants_inlining_info_p returns true.
182 2018-07-20 Richard Biener <rguenther@suse.de>
185 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
186 to cover -flto-partition=none.
188 2018-07-20 Martin Liska <mliska@suse.cz>
190 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
191 (get_decl_source_range): Remove unused function.
193 2018-07-20 Richard Biener <rguenther@suse.de>
195 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
196 (struct vn_phi_s): Likewise.
197 (struct vn_reference_s): Likewise.
198 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
199 for searching the slot of an entry known to be in the hash itself.
200 (vn_phi_hasher::equal): Likewise.
201 (vn_reference_hasher::equal): Likewise.
202 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
204 (optimistic_info, current_info): Remove, keeping only valid_info.
205 (vn_reference_lookup_1): Remove fallback lookup.
206 (vn_reference_lookup_2): Likewise.
207 (vn_nary_op_lookup_1): Likewise.
208 (vn_phi_lookup): Likewise.
209 (vn_nary_build_or_lookup_1): Make sure to not chain the built
211 (vn_reference_insert): Adjust, chain the inserted hash element
212 at last_inserted_ref.
213 (vn_reference_insert_pieces): Likewise.
214 (visit_reference_op_call): Likewise.
215 (vn_nary_op_insert_into): Chain the inserted hash element at
217 (vn_nary_op_insert_pieces): Adjust.
218 (vn_nary_op_insert): Likewise.
219 (vn_nary_op_insert_stmt): Likewise.
220 (vn_phi_insert): Adjust, chain the inserted hash element at
222 (process_scc): Remove clearing and copying the optimistic
223 table. Instead remove elements inserted during an optimistic
224 iteration from the single table we maintain.
225 (init_scc_vn): Adjust.
226 (free_scc_vn): Likewise.
227 (sccvn_dom_walker::record_cond): Likewise.
228 (sccvn_dom_walker::after_dom_children): Likewise.
230 2018-07-19 Martin Sebor <msebor@redhat.com>
232 PR tree-optimization/84047
233 PR tree-optimization/83776
234 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
235 (check_array_bounds): Call it.
237 2018-07-19 Martin Sebor <msebor@redhat.com>
239 * align.h (align_flags): Use member initialization.
241 2018-07-19 David Malcolm <dmalcolm@redhat.com>
243 * Makefile.in (OBJS): Add optinfo.o.
244 * coretypes.h (class symtab_node): New forward decl.
245 (struct cgraph_node): New forward decl.
246 (class varpool_node): New forward decl.
247 * dump-context.h: New file.
248 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
250 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
251 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
252 (set_alt_dump_file): Likewise.
253 (dump_context::~dump_context): New dtor.
254 (dump_gimple_stmt): Move implementation to...
255 (dump_context::dump_gimple_stmt): ...this new member function.
256 Add the stmt to any pending optinfo, creating one if need be.
257 (dump_gimple_stmt_loc): Move implementation to...
258 (dump_context::dump_gimple_stmt_loc): ...this new member function.
259 Start a new optinfo and add the stmt to it.
260 (dump_gimple_expr): Move implementation to...
261 (dump_context::dump_gimple_expr): ...this new member function.
262 Add the stmt to any pending optinfo, creating one if need be.
263 (dump_gimple_expr_loc): Move implementation to...
264 (dump_context::dump_gimple_expr_loc): ...this new member function.
265 Start a new optinfo and add the stmt to it.
266 (dump_generic_expr): Move implementation to...
267 (dump_context::dump_generic_expr): ...this new member function.
268 Add the tree to any pending optinfo, creating one if need be.
269 (dump_generic_expr_loc): Move implementation to...
270 (dump_context::dump_generic_expr_loc): ...this new member
271 function. Add the tree to any pending optinfo, creating one if
273 (dump_printf): Move implementation to...
274 (dump_context::dump_printf_va): ...this new member function. Add
275 the text to any pending optinfo, creating one if need be.
276 (dump_printf_loc): Move implementation to...
277 (dump_context::dump_printf_loc_va): ...this new member function.
278 Start a new optinfo and add the stmt to it.
279 (dump_dec): Move implementation to...
280 (dump_context::dump_dec): ...this new member function. Add the
281 value to any pending optinfo, creating one if need be.
282 (dump_context::dump_symtab_node): New member function.
283 (dump_context::get_scope_depth): New member function.
284 (dump_context::begin_scope): New member function.
285 (dump_context::end_scope): New member function.
286 (dump_context::ensure_pending_optinfo): New member function.
287 (dump_context::begin_next_optinfo): New member function.
288 (dump_context::end_any_optinfo): New member function.
289 (dump_context::s_current): New global.
290 (dump_context::s_default): New global.
291 (dump_scope_depth): Delete global.
292 (dumpfile_ensure_any_optinfo_are_flushed): New function.
293 (dump_symtab_node): New function.
294 (get_dump_scope_depth): Reimplement in terms of dump_context.
295 (dump_begin_scope): Likewise.
296 (dump_end_scope): Likewise.
297 (selftest::temp_dump_context::temp_dump_context): New ctor.
298 (selftest::temp_dump_context::~temp_dump_context): New dtor.
299 (selftest::verify_item): New function.
300 (ASSERT_IS_TEXT): New macro.
301 (ASSERT_IS_TREE): New macro.
302 (ASSERT_IS_GIMPLE): New macro.
303 (selftest::test_capture_of_dump_calls): New test.
304 (selftest::dumpfile_c_tests): Call it.
305 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
306 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
307 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
309 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
310 (dump_node, dump_bb): Move these unrelated decls.
311 (class dump_manager): Add leading comment.
312 * optinfo.cc: New file.
313 * optinfo.h: New file.
315 2018-07-19 Michael Collison <michael.collison@arm.com>
316 Richard Henderson <rth@redhat.com>
318 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
319 (subti): Handle op1 zero.
320 (subvti4, usub4ti4): New.
321 (*sub<GPI>3_compare1_imm): New.
322 (sub<GPI>3_carryinCV): New.
323 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
324 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
326 2018-07-19 Michael Collison <michael.collison@arm.com>
327 Richard Henderson <rth@redhat.com>
329 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
330 (addti3): Create simpler code if low part is already known to be 0.
331 (addvti4, uaddvti4): New.
332 (*add<GPI>3_compareC_cconly_imm): New.
333 (*add<GPI>3_compareC_cconly): New.
334 (*add<GPI>3_compareC_imm): New.
335 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
336 handle constants within this pattern..
337 (*add<GPI>3_compareV_cconly_imm): New.
338 (*add<GPI>3_compareV_cconly): New.
339 (*add<GPI>3_compareV_imm): New.
340 (add<GPI>3_compareV): New.
341 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
342 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
343 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
344 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
345 with 'comparison' operator.
346 (*add<GPI>3_compareV_cconly_imm): Ditto.
347 (*add<GPI>3_compareV_cconly): Ditto.
348 (*add<GPI>3_compareV_imm): Ditto.
349 (add<GPI>3_compareV): Ditto.
350 (add<mode>3_carryinC): Ditto.
351 (*add<mode>3_carryinC_zero): Ditto.
352 (*add<mode>3_carryinC): Ditto.
353 (add<mode>3_carryinV): Ditto.
354 (*add<mode>3_carryinV_zero): Ditto.
355 (*add<mode>3_carryinV): Ditto.
357 2018-07-19 Michael Collison <michael.collison@arm.com>
358 Richard Henderson <rth@redhat.com>
360 * config/aarch64/aarch64-modes.def (CC_V): New.
361 * config/aarch64/aarch64-protos.h
362 (aarch64_addti_scratch_regs): Declare
363 (aarch64_subvti_scratch_regs): Declare.
364 (aarch64_expand_subvti): Declare.
365 (aarch64_gen_unlikely_cbranch): Declare
366 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
367 for signed overflow using CC_Vmode.
368 (aarch64_get_condition_code_1): Handle CC_Vmode.
369 (aarch64_gen_unlikely_cbranch): New function.
370 (aarch64_addti_scratch_regs): New function.
371 (aarch64_subvti_scratch_regs): New function.
372 (aarch64_expand_subvti): New function.
374 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
376 * config/aarch64/aarch64-option-extensions.def: New entry for profile
378 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
379 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
382 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
385 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
386 address check not strict.
388 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
390 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
392 (store_pair_lanes<mode>): Likewise.
393 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
394 enum value 'ADDR_QUERY_LDP_STP_N'.
395 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
396 (aarch64_print_address_internal): Add declaration.
397 (aarch64_print_ldpstp_address): Remove.
398 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
399 (aarch64_print_operand): Change printing of 'y'.
400 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
401 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
402 'true' rather than '1'.
403 * config/aarch64/constraints.md (Uml): Likewise.
404 (Uml): Rename to Umn.
407 2018-07-19 Richard Biener <rguenther@suse.de>
409 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
411 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
412 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
413 (vn_reference_hasher): Likewise.
414 (struct vn_tables_s): Remove obstack and alloc-pool members.
415 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
416 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
417 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
418 (vn_reference_insert_pieces): Likewise.
419 (alloc_vn_nary_op_noinit): Adjust.
420 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
422 (shared_lookup_phiargs): Remove.
423 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
424 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
425 (visit_reference_op_call): Likewise.
426 (copy_nary, copy_phi, copy_reference): Remove.
427 (process_scc): Rewind the obstack when iterating. Do not
428 copy the elements to valid_info but just move them from one
429 hashtable to the other.
430 (allocate_vn_table): Adjust.
431 (free_vn_table): Likewise.
432 (init_scc_vn): Likewise.
433 (free_scc_vn): Likewise.
435 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
438 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
439 indirect_return as function type attribute.
440 (ix86_attribute_table): Change indirect_return to function
442 * doc/extend.texi: Update indirect_return attribute.
444 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
446 * wide-int.h (widest2_int): New.
447 * gimple-fold.c (arith_overflowed_p): Use it.
448 * tree.h (widest2_int_cst): New.
449 * tree-vrp.c (wide_int_binop_overflow): Rename from
451 Rewrite to work on trees.
452 (extract_range_from_multiplicative_op_1): Abstract code to...
453 (wide_int_range_min_max): ...here.
454 (wide_int_range_cross_product): ...and here.
455 (extract_range_from_binary_expr_1): Abstract overflow code to...
456 (wide_int_range_mult_wrapping): ...here.
457 * tree-vrp.h (wide_int_range_cross_product): New.
458 (wide_int_range_mult_wrapping): New.
460 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
461 Julia Koval <julia.koval@intel.com>
463 * config/i386/x86-tune-costs.h (skylake_memcpy,
464 skylake_memset): Replace rep_prefix with unrolling for size 512.
466 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
469 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
472 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
474 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
475 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
476 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
477 previously known as "PowerPC AltiVec Built-in Functions". Move
478 some material to new subsubsections "PowerPC AltiVec Built-in
479 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
481 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
482 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
483 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
484 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
486 2018-07-18 Richard Biener <rguenther@suse.de>
488 PR tree-optimization/86557
489 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
492 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
494 * config/s390/s390.c (s390_function_profiler): Generate CFI.
496 2018-07-17 Jeff Law <law@redhat.com>
498 * config/arm/arm.c (get_label_padding): Update for recent
499 changes to label_to_alignment.
501 PR tree-optimization/86010
502 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
504 * config/mips/mips.c (vr4130_align_insns): Update for recent
505 changes to label_to_alignment.
507 * config/frv/frv.c (frv_label_align): Update for recent changes
508 to label_to_alignment.
510 * config/nios2/nios2.c (nios2_label_align): Update for recent
511 changes which dropped ALIGN_LABELS_LOG.
513 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
515 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
516 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
518 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
520 * config/arc/arc.c (arc_label_align): Use align_labels instead of
521 deprecated align_labels_log.
523 2018-07-17 Richard Biener <rguenther@suse.de>
526 * dwarf2out.c (init_sections_and_labels): Always generate
527 a debug_line_str_section for early LTO debug.
528 (dwarf2out_finish): Reset debug_line_str_hash output early.
529 Bump counter for extra dwarf5 .debug_loc labels to not conflict
531 (dwarf2out_early_finish): Output debug_line_str.
533 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
535 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
536 index register on z196 or later.
538 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
540 * config/s390/s390.c (s390_default_align): Set default function
542 (s390_override_options_after_change): Call s390_default align.
543 (s390_option_override_internal): Call s390_default_align.
544 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
546 2018-07-17 Jakub Jelinek <jakub@redhat.com>
549 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
550 to _looptemp_ clauses, other than the first two.
552 2018-07-17 Martin Liska <mliska@suse.cz>
554 * opts.c: Do not enable OPT_falign_* for -Os.
556 2018-07-17 Martin Liska <mliska@suse.cz>
558 * align.h (MAX_CODE_ALIGN): New.
559 (MAX_CODE_ALIGN_VALUE): New.
560 * common/config/i386/i386-common.c (ix86_handle_option):
561 (MAX_CODE_ALIGN): Moved to align.h.
562 * final.c (MAX_CODE_ALIGN): Likewise.
563 * opts.c (parse_and_check_align_values):
564 (MAX_CODE_ALIGN): Likewise.
565 (MAX_CODE_ALIGN_VALUE): Likewise.
567 2018-07-17 Martin Liska <mliska@suse.cz>
569 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
570 in order to fulfil coding style.
571 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
572 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
573 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
574 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
575 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
576 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
577 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
578 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
579 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
581 2018-07-17 Martin Liska <mliska@suse.cz>
584 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
585 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
586 align_flags of label_to_alignment.
587 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
589 * config/m68k/m68k.c: Do not use removed align_labels_value and
591 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
592 (LOOP_ALIGN): Likewise.
593 (LABEL_ALIGN): Likewise.
594 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
595 Remove not used macro.
596 (rs6000_loop_align): Change return type to align_flags.
597 (rs6000_loop_align_max_skip): Remove.
598 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
599 Change return type to align_flags.
600 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
601 Remove not used macro.
602 (rs6000_loop_align): Change return type to align_flags.
603 (rs6000_loop_align_max_skip): Remove.
604 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
605 * config/rx/rx-protos.h (rx_align_for_label): Make it
607 * config/rx/rx.c (rx_align_for_label): Change return type
609 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
611 into align_flags class.
612 (LABEL_ALIGN): Likewise.
613 (LOOP_ALIGN): Likewise.
614 * config/s390/s390.c (s390_label_align): Use align_flags
616 (s390_asm_output_function_label): Likewise.
617 * config/sh/sh.c (sh_override_options_after_change):
618 Use align_flags class directly without macros.
619 (find_barrier): Likewise.
620 (barrier_align): Likewise.
621 (sh_loop_align): Likewise.
622 * config/spu/spu.c (spu_option_override):
623 Use align_flags_tuple::get_value instead of removed macros.
624 (spu_sched_init): Likewise.
625 * config/spu/spu.h (GTY): Likewise.
626 * config/visium/visium.c (visium_option_override):
627 Set "8" as default secondary alignment.
628 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
629 in order to guarantee secondary alignment of 8.
630 * coretypes.h: Include align.h header file.
631 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
632 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
633 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
634 * doc/tm.texi.in: Likewise.
635 * final.c (struct label_alignment): Remove not used structure.
636 (LABEL_ALIGN): Change type to align_flags.
637 (LOOP_ALIGN): Likewise.
638 (JUMP_ALIGN): Likewise.
639 (default_loop_align_max_skip): Remove.
640 (default_label_align_max_skip): Likewise.
641 (default_jump_align_max_skip): Likewise.
642 (default_label_align_after_barrier_max_skip):
643 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
644 (LABEL_TO_MAX_SKIP): Remove.
645 (label_to_alignment): Return align_flags type instead of integer.
646 (label_to_max_skip): Remove.
647 (align_fuzz): Use align_flags type.
648 (compute_alignments): Use align_flags type and use align_flags::max
649 to combine multiple alignments.
650 (grow_label_align): Grow vec instead of C array.
651 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
652 (shorten_branches): Use align_flags type and use align_flags::max
653 to combine multiple alignments.
654 (final_scan_insn_1): Remove usage of secondary alignment that comes
655 from label alignment, but instead use proper secondary alignment
656 which is computed in grow_label_align.
657 * flags.h (struct align_flags_tuple): Move to align.h.
658 (struct align_flags): Likewise.
659 (state_align_loops): Rename to align_loops.
660 (state_align_jumps): Rename to align_jumps.
661 (state_align_labels): Rename to align_labels.
662 (state_align_functions): Rename to align_functions.
663 (align_loops_log): Remove.
664 (align_jumps_log): Remove.
665 (align_labels_log): Remove.
666 (align_functions_log): Remove.
667 (align_loops_max_skip): Remove.
668 (align_jumps_max_skip): Remove.
669 (align_labels_max_skip): Remove.
670 (align_functions_max_skip): Remove.
671 (align_loops_value): Remove.
672 (align_jumps_value): Remove.
673 (align_labels_value): Remove.
674 (align_functions_value): Remove.
675 * output.h (label_to_alignment): Change return type to align_flags.
676 (label_to_max_skip): Remove.
677 * target.def: Remove loop_align_max_skip, label_align_max_skip,
678 jump_align_max_skip macros.
679 * targhooks.h (default_loop_align_max_skip): Remove.
680 (default_label_align_max_skip): Likewise.
681 (default_jump_align_max_skip): Likewise.
682 (default_label_align_after_barrier_max_skip): Remove.
683 * toplev.c (read_log_maxskip): Use ::normalize function.
684 (parse_N_M): Remove not used argument and also call ::normalize.
685 (parse_alignment_opts): Do not pass unused arguments.
686 * varasm.c (assemble_start_function): Use directly align_functions
687 instead of removed macros.
688 * system.h: Do not poison removed macros.
690 2018-07-17 Jakub Jelinek <jakub@redhat.com>
693 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
694 and cond temporaries don't have reference type if iterator has
695 pointer type. For init use &for_pre_body instead of pre_p if
696 for_pre_body is non-empty.
698 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
700 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
701 double-double modes to SFmode directly directly.
702 (trunc<mode>sf2_fprs): Delete.
704 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
706 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
707 for conversions between IFmode and the decimal floating point modes.
708 (init_float128_ieee): Use the correct names for conversions between
709 KFmode and the decimal floating point modes.
711 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
713 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
714 for the conversions between TDmode and IFmode.
715 (init_float128_ieee): Use more correct names for the conversions
716 between TDmode and KFmode.
718 2018-07-16 Jakub Jelinek <jakub@redhat.com>
720 PR tree-optimization/86526
721 * builtins.c (expand_builtin_memcmp): Formatting fixes.
722 (inline_expand_builtin_string_cmp): Likewise.
723 (inline_string_cmp): Likewise. Use c_readstr instead of
724 builtin_memcpy_read_str. Add unit_mode temporary.
726 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
729 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
730 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
732 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
734 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
735 Alphabetize prototypes of built-in functions, separating out
736 built-in functions that are listed in this section but should be
739 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
742 * expmed.c (emit_store_flag): Do not emit setcc followed by a
743 conditional move when trapping comparison was split to a
744 non-trapping one (and vice versa).
746 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
748 * config/s390/s390.c (s390_function_profiler): Generate nops
749 instead of profiler call sequences.
750 * config/s390/s390.opt: Add the new option.
752 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
754 * config/s390/s390.c (s390_function_profiler): Generate
755 __mcount_loc section.
756 * config/s390/s390.opt: Add the new option.
758 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
760 * common.opt: Add the new warning.
761 * config/s390/s390.c (s390_function_profiler): Emit "brasl
762 %r0,__fentry__" when -mfentry is specified.
763 (s390_option_override_internal): Disallow -mfentry for 31-bit
765 * config/s390/s390.opt: Add the new option.
767 2018-07-16 Richard Biener <rguenther@suse.de>
770 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
771 for function-local FUNCTION_DECL and RESULT_DECL immediately.
773 2018-07-16 Martin Liska <mliska@suse.cz>
776 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
779 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
781 * config/arc/arcHS.md: Update ARCHS scheduling rules.
783 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
785 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
787 * config/arc/arc-cpus.def (hs4x): New cpu.
789 * config/arc/arc-tables.opt: Regenerate.
790 * config/arc/arc.c (arc_sched_issue_rate): New function.
791 (TARGET_SCHED_ISSUE_RATE): Define.
792 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
793 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
795 (attr tune): Add ARCHS4x tune values.
796 (attr tune_dspmpy): Define.
797 (*tst): Correct instruction type.
798 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
799 * config/arc/arcHS4x.md: New file.
800 * config/arc/fpu.md: Update instruction type attributes.
801 * config/arc/t-multilib: Regenerate.
803 2018-07-16 Tom de Vries <tdevries@suse.de>
806 * var-tracking.c (vt_initialize): Fix pre_dec handling.
808 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
810 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
813 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
815 PR tree-optimization/86514
816 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
817 conversion to a boolean type from a type with greater precision.
819 2018-07-16 Tom de Vries <tdevries@suse.de>
821 * var-tracking.c (vt_initialize): Print adjusted insn slim if
822 dump_flags request TDF_SLIM.
824 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
826 * fold-const.c (int_const_binop_1): Abstract...
827 (wide_int_binop): ...wide int code here.
828 (poly_int_binop): ...poly int code here.
829 Abstract the rest of int_const_binop_1 into int_const_binop.
830 * fold-const.h (wide_int_binop): New.
831 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
832 Remove useless PLUS/MINUS_EXPR case.
833 (zero_nonzero_bits_from_vr): Move wide int code...
834 (zero_nonzero_bits_from_bounds): ...here.
835 (extract_range_from_binary_expr_1): Move mask optimization code...
836 (range_easy_mask_min_max): ...here.
837 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
838 (range_easy_mask_min_max): New.
840 2018-07-15 Jeff Law <law@redhat.com>
843 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
846 2018-07-14 Jim Wilson <jimw@sifive.com>
848 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
850 2018-07-14 Paul Koning <ni1d@arrl.net>
852 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
854 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
856 * lto-streamer-out.c (copy_function_or_variable): Dump info about
859 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
860 Steve Munroe <munroesj52@gmail.com>
862 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
863 (_mm_andnot_si128): Likewise.
864 (_mm_or_si128): Likewise.
865 (_mm_xor_si128): Likewise.
867 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
870 * builtins.c (expand_builtin_memcmp): Inline the calls first
871 when result_eq is false.
872 (expand_builtin_strcmp): Inline the calls first.
873 (expand_builtin_strncmp): Likewise.
874 (inline_string_cmp): New routine. Expand a string compare
875 call by using a sequence of char comparison.
876 (inline_expand_builtin_string_cmp): New routine. Inline expansion
877 a call to str(n)cmp/memcmp.
878 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
880 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
882 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
884 * config/arm/driver-arm.c: Include arm-native.h.
885 (host_detect_local_cpu): Use auto-generated data tables.
886 (vendors, arm_cpu_table): Delete. Move part information to ...
887 * config/arm/arm-cpus.in: ... here.
888 * config/arm/parsecpu.awk (gen_native): New function.
889 (vendor, part): New CPU fields.
890 (END): Add support for building the native CPU detection tables.
891 * config/arm/t-arm (arm-native.h): Add build rule.
892 (driver-arm.o): Add dependency on arm-native.h.
894 2018-07-13 Richard Biener <rguenther@suse.de>
897 * match.pd (addr1 - addr2): Allow either of the operand to
900 2018-07-13 Tom de Vries <tdevries@suse.de>
902 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
905 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
907 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
908 arrays instead of numbered variables.
910 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
912 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
913 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
915 2018-07-13 Richard Biener <rguenther@suse.de>
918 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
919 instead of get_context_die.
921 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
922 Richard Biener <rguenther@suse.de>
925 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
926 that the loop latch destination where phi is defined.
928 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
930 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
931 (riscv_expand_epilogue): Add assertion to check interrupt mode.
932 (riscv_set_current_function): Extract getting interrupt type to new
934 (riscv_get_interrupt_type): New function.
935 (riscv_merge_decl_attributes): New function, checking interrupt type is
937 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
939 2018-07-12 Paul Koning <ni1d@arrl.net>
941 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
944 2018-07-12 Paul Koning <ni1d@arrl.net>
946 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
947 zero reference, add doloop_end instead.
948 * doc/md.texi (decrement_and_branch_until_zero): Remove.
949 (Looping patterns): Remove decrement_and_branch_until_zero. Add
950 detail for doloop_end.
952 2018-07-12 Martin Sebor <msebor@redhat.com>
955 * attribs.c (decl_attributes): Reject conflicting attributes before
956 calling attribute handlers.
958 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
960 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
962 (gcc::dump_manager::get_dump_file_name): likewise.
963 (dump_begin): Likewise.
964 * dumpfile.h (dump_begin): Update prototype.
965 (gcc::dump_manager::get_dump_file_name,
966 gcc::dump_manager::get_dump_file_name): Update prototype.
968 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
970 * internal-fn.h (vectorizable_internal_fn_p): New function.
971 * tree-vect-slp.c (compatible_calls_p): Likewise.
972 (vect_build_slp_tree_1): Remove nops argument. Handle calls
973 to internal functions.
974 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
976 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
978 * fold-const.h (inverse_conditions_p): Declare.
979 * fold-const.c (inverse_conditions_p): New function.
980 * match.pd: Use inverse_conditions_p. Add folds of view_converts
981 that test the inverse condition of a conditional internal function.
982 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
983 * internal-fn.c (internal_fn_mask_index): Handle conditional
985 (vectorized_internal_fn_supported_p): New function.
986 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
987 (any_pred_load_store): Replace with...
988 (need_to_predicate): ...this new variable.
989 (redundant_ssa_names): New variable.
990 (ifcvt_can_use_mask_load_store): Move initial checks to...
991 (ifcvt_can_predicate): ...this new function. Handle tree codes
992 for which a conditional internal function exists.
993 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
994 instead of ifcvt_can_use_mask_load_store. Update after variable
996 (predicate_load_or_store): New function, split out from
997 predicate_mem_writes.
998 (check_redundant_cond_expr): New function.
999 (value_available_p): Likewise.
1000 (predicate_rhs_code): Likewise.
1001 (predicate_mem_writes): Rename to...
1002 (predicate_statements): ...this. Use predicate_load_or_store
1003 and predicate_rhs_code.
1004 (combine_blocks, tree_if_conversion): Update after above name changes.
1005 (ifcvt_local_dce): Handle redundant_ssa_names.
1006 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1007 general conditional functions.
1008 * tree-vect-stmts.c (vectorizable_call): Likewise.
1010 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1011 Alan Hayward <alan.hayward@arm.com>
1012 David Sherwood <david.sherwood@arm.com>
1014 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1015 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1016 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1017 plus and minus and convert them into IFN_COND_FMA-based sequences.
1018 (convert_mult_to_fma): Handle conditional plus and minus.
1020 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1022 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1023 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1024 (cond_fnms_optab): New optabs.
1025 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1027 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1028 * internal-fn.h (get_conditional_internal_fn): Declare.
1029 (get_unconditional_internal_fn): Likewise.
1030 * internal-fn.c (cond_ternary_direct): New macro.
1031 (expand_cond_ternary_optab_fn): Likewise.
1032 (direct_cond_ternary_optab_supported_p): Likewise.
1033 (FOR_EACH_COND_FN_PAIR): Likewise.
1034 (get_conditional_internal_fn): New function.
1035 (get_unconditional_internal_fn): Likewise.
1036 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1037 (gimple_match_op::gimple_match_op): Add a new overload for 5
1039 (gimple_match_op::set_op): Likewise.
1040 (gimple_resimplify5): Declare.
1041 * genmatch.c (decision_tree::gen): Generate simplifications for
1043 * gimple-match-head.c (gimple_simplify): Define an overload for
1044 5 operands. Handle calls with 5 arguments in the top-level overload.
1045 (convert_conditional_op): Handle conversions from unconditional
1046 internal functions to conditional ones.
1047 (gimple_resimplify5): New function.
1048 (build_call_internal): Pass a fifth operand.
1049 (maybe_push_res_to_seq): Likewise.
1050 (try_conditional_simplification): Try converting conditional
1051 internal functions to unconditional internal functions.
1052 Handle 3-operand unconditional forms.
1053 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1054 Define ternary equivalents of the current rules for binary conditional
1056 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1058 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1059 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1060 (optab): Handle them.
1061 (SVE_COND_FP_TERNARY): New int iterator.
1062 (sve_fmla_op, sve_fmad_op): New int attributes.
1063 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1064 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1065 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1067 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1069 * target.def (preferred_else_value): New target hook.
1070 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1071 * doc/tm.texi: Regenerate.
1072 * targhooks.h (default_preferred_else_value): Declare.
1073 * targhooks.c (default_preferred_else_value): New function.
1074 * internal-fn.h (conditional_internal_fn_code): Declare.
1075 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1076 (get_conditional_internal_fn): Use it.
1077 (conditional_internal_fn_code): New function.
1078 * gimple-match.h (gimple_match_cond): New struct.
1079 (gimple_match_op): Add a cond member function.
1080 (gimple_match_op::gimple_match_op): Update all forms to take a
1082 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1083 for the suboperation, but don't specify a particular else_value.
1084 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1085 (visit_nary_op, visit_reference_op_load): Pass
1086 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1087 * gimple-match-head.c: Include tree-eh.h
1088 (convert_conditional_op): New function.
1089 (maybe_resimplify_conditional_op): Likewise.
1090 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1091 (gimple_resimplify2): Likewise.
1092 (gimple_resimplify3): Likewise.
1093 (gimple_resimplify4): Likewise.
1094 (maybe_push_res_to_seq): Return null for conditional operations.
1095 (try_conditional_simplification): New function.
1096 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1098 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1100 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1102 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1104 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1106 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1108 (hash_tree): Do not hash DECL_FCONTEXT
1109 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1110 Do not stream DECL_FCONTEXT.
1111 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1112 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1114 2018-07-12 Richard Biener <rguenther@suse.de>
1117 * dwarf2out.c (gen_block_die): Only output blocks when they have
1118 at least one !DECL_IGNORED_P variable.
1120 2018-07-12 Richard Biener <rguenther@suse.de>
1123 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1124 Remove -mieee-fp handling.
1126 2018-07-12 Richard Biener <rguenther@suse.de>
1128 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1129 left-over from last patch.
1131 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1133 PR tree-optimization/86492
1134 * gimple-ssa-store-merging.c
1135 (imm_store_chain_info::coalesce_immediate_stores): Call
1136 check_no_overlap even for the merge_overlapping case. Formatting fix.
1138 2018-07-12 Richard Biener <rguenther@suse.de>
1141 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1142 move possibly trapping operations into the conditional.
1144 2018-07-12 Richard Biener <rguenther@suse.de>
1146 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1147 (vn_lookup_simplify_result): Remove recursion limit applied
1149 (vn_nary_build_or_lookup_1): Adjust.
1150 (try_to_simplify): Likewise.
1151 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1153 (gimple_resimplify2): Likewise.
1154 (gimple_resimplify3): Likewise.
1155 (gimple_resimplify4): Likewise.
1157 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1159 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1160 Use __mmask64 type instead of __mmask8 for __M argument.
1161 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1162 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1164 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1165 __mmask16 for __M argument.
1166 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1167 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1168 to __mmask16 instead of __mmask8.
1169 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1170 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1171 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1172 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1173 instead of __mmask16 for __U argument.
1174 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1175 __mmask16 instead of __mmask8 for __U argument.
1176 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1178 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1180 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1182 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1184 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1186 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1188 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1190 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1192 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1194 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1196 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1197 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1198 return type as well as __M argument type and all casts from __mmask8
1200 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1201 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1202 return type as well as __M argument type and all casts from __mmask8
1204 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1205 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1206 return type as well as __M argument type and all casts from __mmask8
1208 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1209 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1210 return type as well as __M argument type and all casts from __mmask8
1212 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1213 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1216 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1218 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1219 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1222 2018-07-11 Paul Koning <ni1d@arrl.net>
1224 * doc/md.texi (define_subst): Document how multiple occurrences of
1225 the same argument in the replacement pattern are handled.
1227 2018-07-11 Paul Koning <ni1d@arrl.net>
1229 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1231 (Common Type Attributes): Add "mode" attribute.
1233 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1235 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1236 stream DECL_ORIGINAL_TYPE.
1237 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1238 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1239 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1240 Do not walk original type.
1241 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1243 (write_ts_decl_non_common_tree_pointers): Do not stream
1245 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1246 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1248 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1250 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1251 thread twice from the same starting edge.
1253 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1255 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1256 * gimple.c (gimple_call_nonnull_result_p): ...here...
1257 (gimple_call_nonnull_arg): ...and here.
1258 * gimple.h (gimple_call_nonnull_result_p): New.
1259 (gimple_call_nonnull_arg): New.
1261 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1263 * config/arm/arm-cpus.in: Move information from fpu field of each
1264 cpu definition to the isa field.
1265 * config/arm/parsecpu.awk (fpu): Delete match rule.
1266 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1268 2018-07-11 Richard Biener <rguenther@suse.de>
1271 * dwarf2out.c (init_sections_and_labels): Use
1272 output_asm_line_debug_info consistently.
1273 (dwarf2out_early_finish): Likewise.
1274 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1277 2018-07-11 Richard Biener <rguenther@suse.de>
1279 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1280 Rework father_bb setting in a way to avoid propagating constants
1281 multiple times on a loop body.
1283 2018-07-10 Mark Wielaard <mark@klomp.org>
1286 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1288 2018-07-10 Richard Biener <rguenther@suse.de>
1290 * hash-map.h (hash_map::iterator::operator*): Return
1291 references to key and value.
1293 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1296 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1297 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1298 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1299 (gimplify_omp_for): For composite loops, move outer
1300 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1301 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1302 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1303 TREE_LIST for both the original class iterator and the "last" helper
1304 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1305 loop, remember has_decl_expr from outer composite loops for the
1306 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1308 2018-07-09 Martin Sebor <msebor@redhat.com>
1312 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1313 accessing implicitly initialized array elements.
1314 * expr.c (string_constant): Handle string initializers of
1315 character arrays within aggregates.
1316 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1317 Store element offset. As a special case, handle zero size.
1318 (fold_nonarray_ctor_reference): Same.
1319 (fold_ctor_reference): Add argument. Store subobject offset.
1320 * gimple-fold.h (fold_ctor_reference): Add argument.
1322 2018-07-09 Paul Koning <ni1d@arrl.net>
1324 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1325 (pdp11_insn_cost): New function.
1326 (pdp11_md_asm_adjust): New function.
1327 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1328 (pdp11_rtx_costs): Update to match machine better.
1329 (output_addr_const_pdp11): Correct format mismatch warnings.
1330 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1331 * config/pdp11/pdp11.md: General change to add base_cost and/or
1332 length attributes for use by new pdp11_insn_cost function.
1333 (MIN_BRANCH): Correct definition.
1335 (doloop_end): Use standard pattern name for looping pattern.
1336 (doloop_end_nocc): New.
1337 (movsf): Add another constraint alternative.
1338 (zero_extendqihi2): Add constraint alternatives for not in place
1340 (zero_extendhisi2): Remove.
1341 (shift patterns): Add CC handling variants.
1345 (define_peephole2): New peephole to recognize mov that sets CC for
1348 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1350 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1351 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1354 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1357 * real.c (real_nextafter): Return true if result is denormal.
1359 2018-07-09 Martin Liska <mliska@suse.cz>
1361 * common.opt: Add back wrongly removed attribute.
1363 2018-07-09 Richard Biener <rguenther@suse.de>
1366 * dwarf2out.c (gen_block_die): For an early generated DIE
1367 always output high/low PC attributes.
1369 2018-07-09 Tom de Vries <tdevries@suse.de>
1371 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1372 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1373 onto VAR_DECL with abstract origin.
1375 2018-07-07 Jim Wilson <jimw@sifive.com>
1377 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1379 2018-07-07 Tom de Vries <tdevries@suse.de>
1381 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1383 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1385 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1387 (combine_bound): Use wide-int overflow calculation instead of
1389 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1391 * fold-const.c (int_const_binop_2): Same.
1392 (extract_muldiv_1): Same.
1393 (fold_div_compare): Same.
1394 (fold_abs_const): Same.
1396 * poly-int.h (add): Same.
1400 * predict.c (predict_iv_comparison): Same.
1401 * profile-count.c (slow_safe_scale_64bit): Same.
1402 * simplify-rtx.c (simplify_const_binary_operation): Same.
1403 * tree-chrec.c (tree_fold_binomial): Same.
1404 * tree-data-ref.c (split_constant_offset_1): Same.
1405 * tree-if-conv.c (idx_within_array_bound): Same.
1406 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1407 * tree-ssa-phiopt.c (minmax_replacement): Same.
1408 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1409 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1410 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1411 * wide-int.cc (wi::add_large): Same.
1412 (wi::mul_internal): Same.
1413 (wi::sub_large): Same.
1414 (wi::divmod_internal): Same.
1415 * wide-int.h: Change overflow type to overflow_type for neg, add,
1416 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1417 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1418 mul_internal, divmod_internal.
1419 (overflow_type): New enum.
1420 (accumulate_overflow): New.
1422 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1424 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1425 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1427 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1429 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1430 argument is checked for zero before entering loop, avoid checking again.
1432 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1434 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1435 * gimplify.h (generic_expr_could_trap_p): Likwise.
1436 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1438 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1440 PR tree-optimization/86401
1441 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1442 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1443 (fold_bit_and_mask): ... here. New helper function for match.pd.
1444 * fold-const.h (fold_bit_and_mask): Declare.
1445 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1447 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1450 * target.def (translate_mode_attribute): New hook.
1451 * targhooks.h (default_translate_mode_attribute): Declare.
1452 * targhooks.c (default_translate_mode_attribute): New function.
1453 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1454 * doc/tm.texi: Regenerate.
1455 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1456 (rs6000_translate_mode_attribute): New function.
1458 2018-07-06 Paul Koning <ni1d@arrl.net>
1460 * doc/md.texi (define_split): Document DONE and FAIL.
1461 (define_peephole2): Ditto.
1463 2018-07-05 Jeff Law <law@redhat.com>
1465 PR tree-optimization/86010
1466 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1467 both the head and tail of mem* and str* calls.
1469 2018-07-05 Jim Wilson <jimw@sifive.com>
1471 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1473 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1475 * config/aarch64/aarch64-simd.md: correct flags text for
1476 MIN_EXPR replacement.
1478 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1480 * configure: Regenerated.
1482 2018-07-05 Carl Love <cel@us.ibm.com>
1484 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1485 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1486 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1487 VSX_BUILTIN_DOUBLEL_V4SF.
1489 2018-07-05 Martin Sebor <msebor@redhat.com>
1492 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1493 than its domain to compute its the upper bound of a char array.
1495 2018-07-05 Nathan Sidwell <nathan@acm.org>
1497 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1498 * doc/cpp.texi: Update comment.
1499 * doc/tm.texi: Rebuilt.
1500 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1501 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1502 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1503 headers affected by extern "C".
1504 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1505 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1506 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1507 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1508 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1509 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1510 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1511 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1512 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1513 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1514 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1515 config/sparc/sp64-elf.h, config/spu/spu.h,
1516 config/stormy16/stormy16.h, config/v850/v850.h,
1517 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1518 define NO_IMPLICIT_EXTERN_C.
1519 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1521 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1524 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1525 instead of GET_MODE_SIZE when comparing Units.
1527 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1530 * rtlanal.c (set_noop_p): Constrain on mode change,
1531 include hard-reg-set.h
1533 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1535 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1537 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1540 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1543 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1545 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1547 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1550 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1552 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1554 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1557 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1559 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1561 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1562 neg_*_op* variables.
1564 2018-07-04 Martin Liska <mliska@suse.cz>
1566 * tree-switch-conversion.c: Define
1567 max_ratio_for_speed and max_ratio_for_size constants.
1569 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1570 Martin Liska <mliska@suse.cz>
1575 * common.opt: Rename align options with 'str_' prefix.
1576 * common/config/i386/i386-common.c (set_malign_value): New
1578 (ix86_handle_option): Use it to set -falign-* options/
1579 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1580 type from int to string.
1581 * config/aarch64/aarch64.c: Update default values from int
1583 * config/alpha/alpha.c (alpha_override_options_after_change):
1585 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1586 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1587 max skip conditionally.
1588 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1589 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1590 max skip conditionally.
1591 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1592 max skip conditionally.
1593 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1594 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1595 max skip conditionally.
1596 * config/i386/i386.c (struct ptt): Change type from int to
1598 (ix86_default_align): Set default values.
1599 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1600 max skip conditionally.
1601 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1602 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1603 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1604 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1605 max skip conditionally.
1606 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1607 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1608 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1609 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1610 max skip conditionally.
1611 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1612 * config/ia64/ia64.c (ia64_option_override): Set default values
1613 for alignment options.
1614 * config/m68k/m68k.c: Handle new str_align_* options.
1615 * config/mips/mips.c (mips_set_compression_mode): Change
1617 (mips_option_override): Set default values for options.
1618 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1620 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1622 * config/rx/rx.c (rx_option_override): Likewise.
1623 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1624 (LABEL_ALIGN): Use align_labels_log.
1625 (LOOP_ALIGN): Use align_loops_align.
1626 * config/s390/s390.c (s390_asm_output_function_label): Use new
1628 * config/sh/sh.c (sh_override_options_after_change):
1629 Change type of constants.
1630 * config/spu/spu.c (spu_sched_init): Likewise.
1631 * config/sparc/sparc.c (sparc_option_override): Set default
1633 * config/visium/visium.c (visium_option_override): Likewise.
1634 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1635 emit p2align format with last argument if it's not needed.
1636 * doc/invoke.texi: Document extended format of -falign-*.
1637 * final.c: Use align_labels alignment.
1638 * flags.h (struct target_flag_state): Change type to use
1640 (struct align_flags_tuple): New.
1641 (struct align_flags): Likewise.
1642 (align_loops_log): Redefine macro to use new types.
1643 (align_loops_max_skip): Redefine macro to use new types.
1644 (align_jumps_log): Redefine macro to use new types.
1645 (align_jumps_max_skip): Redefine macro to use new types.
1646 (align_labels_log): Redefine macro to use new types.
1647 (align_labels_max_skip): Redefine macro to use new types.
1648 (align_functions_log): Redefine macro to use new types.
1649 (align_loops): Redefine macro to use new types.
1650 (align_jumps): Redefine macro to use new types.
1651 (align_labels): Redefine macro to use new types.
1652 (align_functions): Redefine macro to use new types.
1653 (align_functions_max_skip): Redefine macro to use new types.
1654 (align_loops_value): New macro.
1655 (align_jumps_value): New macro.
1656 (align_labels_value): New macro.
1657 (align_functions_value): New macro.
1658 * function.c (invoke_set_current_function_hook): Propagate
1659 alignment values from flags to global variables default in
1661 * ipa-icf.c (sem_function::equals_wpa): Use
1662 cl_optimization_option_eq instead of memcmp.
1663 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1665 (cl_optimization_stream_in): Likewise.
1666 * optc-save-gen.awk: Support strings in cl_optimization.
1667 * opth-gen.awk: Likewise.
1668 * opts.c (finish_options): Remove error checking of invalid
1670 (MAX_CODE_ALIGN): Remove.
1671 (MAX_CODE_ALIGN_VALUE): Likewise.
1672 (parse_and_check_align_values): New function.
1673 (check_alignment_argument): Likewise.
1674 (common_handle_option): Use check_alignment_argument.
1675 * opts.h (parse_and_check_align_values): Declare.
1676 * toplev.c (init_alignments): Remove.
1677 (read_log_maxskip): New.
1678 (parse_N_M): Likewise.
1679 (parse_alignment_opts): Likewise.
1680 (backend_init_target): Remove usage of init_alignments.
1681 * toplev.h (parse_alignment_opts): Declare.
1682 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1684 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1685 * tree.c (cl_option_hasher::equal): New.
1686 * varasm.c: Use new global macros.
1688 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1690 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1691 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1692 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1693 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1694 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1695 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1696 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1697 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1698 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1699 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1700 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1701 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1702 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1704 2018-07-04 Martin Liska <mliska@suse.cz>
1705 Jonathan Wakely <jwakely@redhat.com>
1707 * coverage.c: Use correct type.
1708 * doc/invoke.texi: Language correction.
1710 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1713 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1714 ENDBRANCH for non-tail call which may return via indirect branch.
1715 * doc/extend.texi: Document indirect_return attribute.
1717 2018-07-03 Martin Sebor <msebor@redhat.com>
1719 PR tree-optimization/86274
1720 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1722 (format_floating): Correct handling of infinities and NaNs.
1724 2018-07-03 Martin Sebor <msebor@redhat.com>
1726 * print-tree.c (print_real_cst): New function.
1727 (print_node_brief): Call it.
1728 (print_node): Ditto.
1730 2018-07-03 Jeff Law <law@redhat.com>
1732 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1733 into a single pattern.
1735 * config/h8300/h8300.md (ors code_iterator): New.
1736 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1737 a single pattern and single splitter.
1738 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1739 (iorqi3_1, xorqi3_1): Likewise.
1740 (iorqi3, xorqi3 expanders): Similarly.
1742 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1743 (movmd_internal) into a single pattern using the P mode iterator.
1744 (movmd splitters): Similarly.
1745 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1746 (movsd splitters): Similarly.
1748 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1749 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1750 right table based on the size of the operand.
1751 * config/h8300/h8300.md (length_table): Corresponding changes. All
1752 references to "addb", "addw" and "addl" changed to "add".
1753 (btst patterns): Merge two variants into a single pattern.
1754 (tstqi, tsthi): Likewise.
1755 (addhi3_incdec, addsi3_incdec): Likewise.
1756 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1757 (mulhi3, mulsi3): Likewise.
1758 (udivhi3, udivsi3): Likewise.
1759 (divhi3, divsi3): Likewise.
1760 (andorqi3, andorhi3, andorsi3): Likewise.
1762 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1765 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1766 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1768 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1770 PR tree-optimization/85694
1771 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1772 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1774 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1775 (<u>avg<mode>3_ceil): New patterns.
1777 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1779 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1780 scan-tree-dump directive.
1781 * gcc.dg/vect/slp-perm-2.c: Likewise.
1782 * gcc.dg/vect/slp-perm-3.c: Likewise.
1783 * gcc.dg/vect/slp-perm-5.c: Likewise.
1784 * gcc.dg/vect/slp-perm-6.c: Likewise.
1785 * gcc.dg/vect/slp-perm-7.c: Likewise.
1786 * gcc.dg/vect/slp-perm-8.c: Likewise.
1788 2018-07-03 Marek Polacek <polacek@redhat.com>
1791 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1793 2018-07-03 Richard Biener <rguenther@suse.de>
1796 * tree-ssa-structalias.c (find_func_clobbers): Properly
1797 handle indirect calls.
1799 2018-07-03 Jeff Law <law@redhat.com>
1801 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1802 (shifts): New code iterator.
1803 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1804 expander. Fix HImode handling on H8/SX.
1805 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1806 (subqi3, subhi3, subsi3 expanders): Likewise.
1807 (andqi3, andhi3, andsi3 expanders): Likewise.
1808 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1809 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1810 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1811 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1812 (zero_extendqihi2, zero_extendqisi2): Likewise.
1813 (extendqihi2, extendqisi2): Likewise.
1814 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1815 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1816 (rotlqi3_1, rotlhi3_1): Likewise.
1817 (logicalhi3_sn, logicalsi3_sn): Likewise.
1818 (logicalhi3, logicalsi3): Likewise.
1820 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1822 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1823 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1824 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1825 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1826 type to append_pattern_def_seq instead of creating a stmt_vec_info
1828 (build_mask_conversion): Likewise. Remove vinfo argument.
1829 (vect_add_conversion_to_patterm): Likewise, renaming to...
1830 (vect_add_conversion_to_pattern): ...this.
1831 (vect_recog_mask_conversion_pattern): Update call to
1832 build_mask_conversion. Pass the vector type to
1833 append_pattern_def_seq here too.
1834 (vect_recog_gather_scatter_pattern): Update call to
1835 vect_add_conversion_to_pattern.
1837 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1839 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1840 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1841 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1842 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1843 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1844 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1845 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1846 append_pattern_def_seq instead of new_pattern_def_seq.
1847 (vect_recog_divmod_pattern): Do both of the above.
1848 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1851 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1853 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1854 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1855 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1856 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1857 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1858 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1859 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1860 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1861 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1862 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1863 parameter with a single stmt_vec_info.
1864 (vect_recog_func_ptr): Likewise.
1865 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1866 (vect_try_gather_scatter_pattern): ...this.
1867 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1868 the stmt_vec_info of the statement to be matched. Don't clear
1869 STMT_VINFO_RELATED_STMT.
1870 (vect_pattern_recog): Update call accordingly.
1872 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1874 PR tree-optimization/85694
1875 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1876 (uavgM3_ceil): Document new optabs.
1877 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1878 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1880 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1881 (savg_ceil_optab): New optabs.
1882 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1883 (vect_vect_recog_func_ptrs): Add it.
1884 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1885 constant directly from the associated lhs.
1887 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1889 * tree-vect-patterns.c (vect_split_statement): New function.
1890 (vect_convert_input): Use it to try to split an existing cast.
1892 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1894 * poly-int.h (print_hex): New function.
1895 * dumpfile.h (dump_dec, dump_hex): Declare.
1896 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1897 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1898 min_input_precision, operation_precision and operation_sign.
1899 * tree-vect-patterns.c (vect_get_range_info): New function.
1900 (vect_same_loop_or_bb_p, vect_single_imm_use)
1901 (vect_operation_fits_smaller_type): Delete.
1902 (vect_look_through_possible_promotion): Add an optional
1903 single_use_p parameter.
1904 (vect_recog_over_widening_pattern): Rewrite to use new
1905 stmt_vec_info infomration. Handle one operation at a time.
1906 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1907 (vect_truncatable_operation_p, vect_set_operation_type)
1908 (vect_set_min_input_precision): New functions.
1909 (vect_determine_min_output_precision_1): Likewise.
1910 (vect_determine_min_output_precision): Likewise.
1911 (vect_determine_precisions_from_range): Likewise.
1912 (vect_determine_precisions_from_users): Likewise.
1913 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1914 (vect_vect_recog_func_ptrs): Put over_widening first.
1916 (vect_pattern_recog): Call vect_determine_precisions.
1918 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1920 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1921 statements that have been replaced by further pattern statements.
1922 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1924 2018-07-03 Richard Biener <rguenther@suse.de>
1926 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1927 always set *dt. Dump vectype in vectype overload.
1928 * dumpfile.h (dump_gimple_expr): New function.
1929 (dump_gimple_expr_loc): Likewise.
1930 * dumpfile.c (dump_gimple_expr): New function.
1931 (dump_gimple_expr_loc): Likewise.
1933 2018-07-02 Jeff Law <law@redhat.com>
1935 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1936 the H8/300, H8/300H and H8/S variants into a single pattern.
1937 (movhi_h8300, movqi_h8300hs): Similarly.
1938 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1939 (QHI mode iterator): New.
1941 * config/h8300/h8300.md: Remove trailing whitespace.
1943 2018-07-02 Jim Wilson <jimw@sifive.com>
1945 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1946 instead of emit_insn for interrupt returns.
1947 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1948 (riscv_sret, riscv_uret): Likewise.
1950 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1952 * pretty-print.c (selftest::test_pp_format): Move save and restore
1953 of quotes to class auto_fix_quotes, and add an instance.
1954 * selftest.c: Include "intl.h".
1955 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1956 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1957 * selftest.h (selftest::auto_fix_quotes): New class.
1959 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1961 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1962 (aarch64_sve_prepare_conditional_op): Remove.
1963 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1964 Allow aarch64_simd_reg_or_zero as select operand; remove
1965 the aarch64_sve_prepare_conditional_op call.
1966 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1967 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1968 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1969 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1970 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1971 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1972 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1973 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1974 and a splitters to match all of the *_any patterns.
1975 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1977 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1978 (SVE_COND_FP_BINARY_REV): Remove.
1979 (sve_int_op_rev, sve_fp_op_rev): New.
1980 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1981 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1982 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1983 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1984 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1985 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1986 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1987 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1988 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1990 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1991 Remove match_dup 1 from the inner unspec.
1992 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1994 * config/aarch64/aarch64.md (movprfx): New attr.
1995 (length): Default movprfx to 8.
1996 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1997 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1998 (*<su>mul<SVE_I>3_highpart): Likewise.
1999 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2000 (*v<ASHIFT><SVE_I>3): Likewise.
2001 (*<su><MAXMIN><SVE_I>3): Likewise.
2002 (*<su><MAXMIN><SVE_F>3): Likewise.
2003 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2004 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2005 (*div<SVE_F>4): Likewise.
2007 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2009 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2012 2018-07-02 Richard Biener <rguenther@suse.de>
2014 PR tree-optimization/86363
2015 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2016 memset argument refers to a non-variable address.
2018 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2020 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2021 {PLUS,MINUS}_EXPR code to...
2022 (adjust_symbolic_bound): ...here,
2023 (combine_bound): ...here,
2024 (set_value_range_with_overflow): ...and here.
2026 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2028 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2030 (extract_range_from_abs_expr): ...here.
2032 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2034 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2035 -fno-omit-frame-pointer when not optimizing.
2037 2018-07-02 Martin Liska <mliska@suse.cz>
2040 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2041 (propagate_nothrow): Likewise.
2043 2018-07-02 Martin Liska <mliska@suse.cz>
2046 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2048 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2050 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2051 function in r262149, changing "loc" param from source_location to
2052 const dump_location_t &.
2053 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2054 declaration, as above.
2056 2018-07-01 Paul Koning <ni1d@arrl.net>
2058 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2059 -munit-asm, -mgnu-asm, -mdec-asm.
2060 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2061 (pdp11_output_labelref): New.
2062 (pdp11_output_def): New.
2063 (pdp11_output_addr_vec_elt): New.
2064 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2065 %# and %@ format codes.
2066 (pdp11_option_override): New.
2067 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2068 (pdp11_output_ident): New.
2069 (pdp11_asm_named_section): New.
2070 (pdp11_asm_init_sections): New.
2071 (pdp11_file_start): New.
2072 (pdp11_file_end): New.
2073 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2074 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2075 %o, like %c but octal.
2076 (pdp11_option_override): New.
2077 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2079 (DATA_SECTION_ASM_OP): Ditto.
2080 (READONLY_DATA_SECTION_ASM_OP): New.
2081 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2082 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2083 (ASM_OUTPUT_LABELREF): Ditto.
2084 (ASM_OUTPUT_DEF): Ditto.
2085 (ASM_OUTPUT_EXTERNAL): New.
2086 (ASM_OUTPUT_SOURCE_FILENAME): New.
2087 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2088 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2089 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2090 %# and %@ format codes.
2091 * config/pdp11/pdp11.opt (mgnu-asm): New.
2092 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2093 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2094 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2096 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2098 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2099 dereferencing path[] beyond its length.
2101 (debug_all_paths): New.
2102 (rewire_first_differing_edge): New.
2103 (adjust_paths_after_duplication): New.
2104 (duplicate_thread_path): Call adjust_paths_after_duplication.
2106 (thread_through_all_blocks): Add new argument to
2107 duplicate_thread_path.
2109 2018-06-30 Jim Wilson <jimw@sifive.com>
2111 * config/riscv/predicates.md (p2m1_shift_operand): New.
2112 (high_mask_shift_operand): New.
2113 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2114 pattern using p2m1_shift_operand.
2115 (lshsi3_zero_extend_3+2): New combiner pattern using
2116 high_mask_shift_operand.
2118 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2120 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2122 (vect_recog_rotate_pattern): ...here.
2123 (vect_convert_input): Try to insert casts of invariants in the
2125 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2126 preheader to be empty.
2128 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2130 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2131 vector type. If given, install it in the new statement's
2133 (vect_element_precision): New function.
2134 (vect_unpromoted_value): New struct.
2135 (vect_unpromoted_value::vect_unpromoted_value): New function.
2136 (vect_unpromoted_value::set_op): Likewise.
2137 (vect_look_through_possible_promotion): Likewise.
2138 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2139 (vect_widened_op_tree, vect_convert_input): Likewise.
2140 (vect_convert_inputs, vect_convert_output): Likewise.
2141 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2142 to handle the optional cast of the multiplication result and
2143 vect_widened_op_tree to detect the widened multiplication itself.
2144 Do not require the input and output of promotion casts to have
2145 the same sign, but base the signedness of the operation on the
2146 input rather than the result. If the pattern includes two
2147 promotions, check that those promotions have the same sign.
2148 Do not restrict the MULT_EXPR handling to a double-width result;
2149 handle quadruple-width results and wider. Use vect_convert_inputs
2150 to convert the inputs to the common type.
2151 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2152 to handle the optional cast of the ABS result. Also allow a sign
2153 change or a sign extension between the ABS and MINUS.
2154 Use vect_widened_op_tree to detect the widened subtraction and use
2155 vect_convert_inputs to convert the inputs to the common type.
2156 (vect_handle_widen_op_by_const): Delete.
2157 (vect_recog_widen_op_pattern): New function.
2158 (vect_recog_widen_mult_pattern): Use it.
2159 (vect_recog_widen_shift_pattern): Likewise.
2160 (vect_recog_widen_sum_pattern): Use
2161 vect_look_through_possible_promotion to handle the promoted
2164 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2166 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2167 the containing gimple_seq *.
2168 * gimple-iterator.h (gsi_for_stmt): Declare it.
2169 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2170 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2171 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2172 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2173 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2175 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2177 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2178 statement being replaced is part of an existing pattern
2179 definition sequence, inserting the new pattern statements before
2181 (vect_pattern_recog_1): Don't return a bool. If the statement
2182 is already part of a pattern, instead apply pattern matching
2183 to the pattern definition statements. Don't clear the
2184 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2185 (vect_pattern_recog): Don't break after the first match;
2186 continue processing the pattern definition statements instead.
2187 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2189 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2191 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2192 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2193 (vect_recog_widen_sum_pattern): Use it.
2195 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2197 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2198 phi is not a pattern statement and has not been replaced by
2199 a pattern statement.
2200 * tree-vect-patterns.c (type_conversion_p): Don't check
2201 STMT_VINFO_IN_PATTERN_P.
2202 (vect_recog_vector_vector_shift_pattern): Likewise.
2203 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2204 the pattern statement rather than the original statement; check
2205 directly for a WIDEN_MULT_EXPR here.
2206 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2207 vect_is_simple_use to return the pattern statement rather
2208 than the original statement; use is_pattern_stmt_p to check
2209 for such a pattern statement.
2210 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2211 to return the pattern statement rather than the original statement;
2212 don't do the same transformation here.
2213 (vect_is_simple_use): If the defining statement has been replaced
2214 by a pattern statement, return the pattern statement instead.
2215 Remove the corresponding (local) transformation from the vectype
2218 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2220 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2221 end and default to null.
2222 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2223 (vectorizable_reduction): Update calls accordingly, dropping the
2224 gimple ** argument if the passed-back statement isn't needed.
2225 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2226 (vect_recog_rotate_pattern): Likewise.
2227 (vect_recog_mask_conversion_pattern): Likewise.
2228 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2229 (vect_mask_constant_operand_p): Likewise.
2230 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2231 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2232 (get_group_load_store_type, get_load_store_type): Likewise.
2233 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2234 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2235 (vectorizable_conversion, vectorizable_assignment): Likewise.
2236 (vectorizable_shift, vectorizable_operation): Likewise.
2237 (vectorizable_store, vect_is_simple_cond): Likewise.
2238 (vectorizable_condition, vectorizable_comparison): Likewise.
2239 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2240 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2241 and move it to the end. Cope with null def_stmt_outs.
2243 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2245 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2247 2018-06-29 Jeff Law <law@redhat.com>
2249 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2250 displacements for TARGET_V850E2V3 and newer.
2251 (TARGET_LRA_P): Remove. Defaults to LRA now.
2252 * config/v850/v850.md (sign23byte_load): Remove.
2253 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2254 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2256 2018-06-29 Martin Liska <mliska@suse.cz>
2259 * coverage.c (coverage_init): Mangle full path name.
2260 * doc/invoke.texi: Document the change.
2261 * gcov-io.c (mangle_path): New.
2262 * gcov-io.h (mangle_path): Likewise.
2263 * gcov.c (mangle_name): Use mangle_path for path mangling.
2265 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2267 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2268 if starting source register is not even.
2270 2018-06-29 Martin Liska <mliska@suse.cz>
2272 PR tree-optimization/86263
2273 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2274 Make edge redirection.
2276 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2278 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2279 (dump_scope_depth): New variable.
2280 (get_dump_scope_depth): New function.
2281 (dump_begin_scope): New function.
2282 (dump_end_scope): New function.
2283 * dumpfile.h (get_dump_scope_depth): New declaration.
2284 (dump_begin_scope): New declaration.
2285 (dump_end_scope): New declaration.
2286 (class auto_dump_scope): New class.
2287 (AUTO_DUMP_SCOPE): New macro.
2288 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2291 2018-06-29 Richard Biener <rguenther@suse.de>
2293 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2294 compute_all_dependences succeeds.
2295 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2296 exceed --param loop-max-datarefs-for-datadeps.
2298 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2300 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2302 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2305 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2306 alternative 0 in preferred_for_speed attribute.
2308 2018-06-28 Paul Koning <ni1d@arrl.net>
2310 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2311 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2312 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2313 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2314 for shift insn patterns.
2316 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2318 * cgraph.c (cgraph_node::get_body): Replace assignments to
2319 "dump_file" with calls to set_dump_file.
2320 * dumpfile.c (alt_dump_file): Make static, and group with...
2321 (alt_flags): ...this definition.
2322 (dumps_are_enabled): New variable.
2323 (refresh_dumps_are_enabled): New function.
2324 (set_dump_file): New function.
2325 (set_alt_dump_file): New function.
2326 (gcc::dump_manager::dump_start): Replace assignments to
2327 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2329 (gcc::dump_manager::dump_finish): Likewise.
2330 * dumpfile.h (alt_dump_file): Delete decl.
2331 (dumps_are_enabled): New variable decl.
2332 (set_dump_file): New function decl.
2333 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2335 * tree-nested.c (lower_nested_functions): Replace assignments to
2336 "dump_file" with calls to set_dump_file.
2338 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2340 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2341 goto_locus of each outgoing edge of each basic block.
2343 2018-06-28 Richard Biener <rguenther@suse.de>
2345 * dwarf2out.c (decl_scope_table): Remove.
2346 (push_decl_scope): Likewise.
2347 (pop_decl_scope): Likewise.
2348 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2349 (gen_struct_or_union_type_die): Likewise.
2350 (gen_tagged_type_die): Likewise.
2351 (dwarf2out_init): Do not initialize decl_scope_table.
2352 (dwarf2out_c_finalize): Do not free it.
2354 2018-06-28 Richard Biener <rguenther@suse.de>
2356 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2357 deciding whether to not re-use a DIE.
2359 2018-06-28 Richard Biener <rguenther@suse.de>
2361 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2362 DW_AT_abstract_origin attribute.
2364 2018-06-28 Martin Liska <mliska@suse.cz>
2366 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2367 Use newly introduced constants.
2368 * tree-switch-conversion.h (struct jump_table_cluster):
2369 Define max_ratio_for_size and max_ratio_for_speed.
2371 2018-06-28 Martin Liska <mliska@suse.cz>
2373 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2374 Add new checking assert to catch invalid state.
2375 (jump_table_cluster::can_be_handled): Handle single case
2377 (jump_table_cluster::is_beneficial): Bail out for such case.
2378 (bit_test_cluster::find_bit_tests):
2379 Add new checking assert to catch invalid state.
2380 (bit_test_cluster::can_be_handled): Handle single case
2382 (bit_test_cluster::is_beneficial): Bail out for such case.
2383 (switch_decision_tree::analyze_switch_statement):
2386 2018-06-28 Martin Liska <mliska@suse.cz>
2388 * common.opt: Introduce -completion option.
2389 * gcc.c (driver_handle_option): Handle it.
2390 (driver::main): Print completions if completion
2392 * opt-suggestions.c (option_proposer::get_completions):
2394 (option_proposer::suggest_completion): Likewise.
2395 (option_proposer::find_param_completions): Likewise.
2396 (verify_autocompletions): Likewise.
2397 (test_completion_valid_options): Likewise.
2398 (test_completion_valid_params): Likewise.
2399 (in_completion_p): Likewise.
2400 (empty_completion_p): Likewise.
2401 (test_completion_partial_match): Likewise.
2402 (test_completion_garbage): Likewise.
2403 (opt_proposer_c_tests): Likewise.
2404 * opt-suggestions.h: Declare new functions.
2405 * opts.c (common_handle_option): Handle OPT__completion_.
2406 * selftest-run-tests.c (selftest::run_tests): Add
2407 opt_proposer_c_tests.
2408 * selftest.c (assert_str_startswith): New.
2409 * selftest.h (assert_str_startswith): Likewise.
2410 (opt_proposer_c_tests): New.
2411 (ASSERT_STR_STARTSWITH): Likewise.
2413 2018-06-28 Martin Liska <mliska@suse.cz>
2415 * Makefile.in: Add opt-suggestions.o.
2416 * gcc-main.c: Include opt-suggestions.h.
2417 * gcc.c (driver::driver): Likewise.
2418 (driver::~driver): Remove m_option_suggestions.
2419 (driver::build_option_suggestions): Moved to option_proposer.
2420 (driver::suggest_option): Likewise.
2421 (driver::handle_unrecognized_options): Use option_proposer.
2422 * gcc.h (class driver): Add new memver m_option_proposer.
2423 * opt-suggestions.c: New file.
2424 * opt-suggestions.h: New file.
2426 2018-06-28 Martin Liska <mliska@suse.cz>
2428 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2429 (auto_string_vec::~auto_string_vec): Likewise.
2431 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2433 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2434 prevent_decl_creation_for_types fields up and add reset_location field.
2435 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2436 statement if id->reset_location is true.
2437 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2438 id->reset_location is true.
2439 (copy_phis_for_bb): Force input_location on the arguments if
2440 id->reset_location is true.
2441 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2442 is set on the function to be inlined.
2444 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2446 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2448 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2450 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2451 registers for Pmode.
2452 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2453 hard registers for the clobbered pseudo.
2455 2018-06-27 Paul Koning <ni1d@arrl.net>
2457 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2458 mutually exclusive options.
2459 * config/pdp11/constraints.md (h): New constraint.
2460 (O): Update definition to match shift code generation.
2461 (D): New constraint.
2462 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2464 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2466 (output_jump): Change arguments.
2467 (pdp11_fixed_cc_regs): New function.
2468 (pdp11_cc_mode): Ditto.
2469 (pdp11_expand_shift): Ditto.
2470 (pdp11_assemble_shift): Ditto.
2471 (pdp11_small_shift): Ditto.
2472 (pdp11_branch_cost): Remove.
2473 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2475 (pdp11_register_move_cost): Update for CC registers.
2476 (pdp11_rtx_costs): Add case for LSHIFTRT.
2477 (pdp11_output_jump): Add CCNZ mode conditional branches.
2478 (notice_update_cc_on_set): Remove.
2479 (pdp11_cc_mode): New function.
2480 (simple_memory_operand): Correct pre/post decrement case.
2481 (no_side_effect_operand): New function.
2482 (pdp11_regno_reg_class): Add CC_REGS class.
2483 (pdp11_fixed_cc_regs): New function.
2484 (pdp11_small_shift): New function.
2485 (pdp11_expand_shift): New function to expand shift insns.
2486 (pdp11_assemble_shift): New function to output shifts.
2487 (pdp11_branch_cost): Remove.
2488 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2489 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2490 (WCHAR_TYPE): Ditto.
2491 (PTRDIFF_TYPE): Ditto.
2492 (ADJUST_INSN_LENGTH): New macro.
2493 (FIXED_REGISTERS): Add CC registers.
2494 (CALL_USED_REGISTERS): Ditto.
2496 (REG_CLASS_NAMES): Ditto.
2497 (REG_CLASS_CONTENTS): Ditto.
2498 (SELECT_CC_MODE): Use new function.
2499 (TARGET_FLAGS_REGNUM): New macro.
2500 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2501 (cc0_reg_rtx): Remove.
2502 (CC_STATUS_MDEP): Remove.
2503 (CC_STATUS_MDEFP_INIT): Remove.
2504 (CC_IN_FPU): Remove.
2505 (NOTICE_UPDATE_CC): Remove.
2506 (REGISTER_NAMES): Add CC registers.
2507 (BRANCH_COST): Change to constant 1.
2508 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2510 * config/pdp11/pdp11.opt (mbcopy): Remove.
2511 (mbcopy-builtin): Remove.
2512 (mbranch-cheap): Remove.
2513 (mbranch-expensive): Remove.
2514 * config/pdp11/predicates.md (expand_shift_operand): Update to
2515 match shift code generation.
2516 (ccnz_operator): New predicate.
2517 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2518 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2519 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2521 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2522 description of O constraint.
2524 2018-06-27 Jeff Law <law@redhat.com>
2525 Austin Law <austinklaw@gmail.com>
2527 * config/v850/v850.md (addsi3_set_flags): New pattern.
2528 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2529 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2530 (zero_extendhisi2_v850_set_flags): Likewise.
2531 (zero_extendqisi2_v850_set_flags): Likewise.
2532 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2533 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2534 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2536 * config/v850/v850-protos.h (notice_update_cc): Remove.
2537 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2538 (v850_print_operand): Handle 'D' and "d".
2539 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2540 Add handling of arithmetic/logical operations compared against zero.
2541 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2542 Do not look at v850_compare_op, instead get mode from last argument.
2543 (v850_gen_compare): Remove
2544 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2545 after reload for prologue insns.
2546 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2548 (construct_save_jarl): Likewise.
2549 (TARGET_FLAGS_REGNUM): Define.
2550 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2551 (NOTICE_UPDATE_CC): Remove.
2552 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2553 than cc0. Conditionalize on reload_completed.
2554 (cmpsi_insn, setfcc_insn): Likewise.
2555 (tst1 splitter): Turn into define_and_split which sets the flags
2557 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2558 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2559 (cstoresf4, cstoredf4): Clobber the flags.
2560 (cmpsi, cmpsf, cmpdf): Remove expanders.
2561 (setf_insn): Remove pattern.
2562 (addsi3): Turn into define_and_split which clobbers the flags after
2563 reload and a suitable pattern (addsi3_clobber_flags) for use after
2565 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2566 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2567 (ashrsi3, ashrsi3_v850e2): Likewise.
2568 (bins): Clobber the flags.
2569 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2570 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2571 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2572 (call_value_internal_short, call_value_internal_long): Likewise.
2573 (callt_save_interrupt, callt_return_interrupt): Likewise.
2574 (save_interrupt, return_interrupt): Likewise.
2575 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2576 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2577 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2578 (All FP comparisons): Only allow after reload has completed.
2580 (divh, divhu): Tweak output template.
2581 (branch_z_normal, branch_z_invert): Remove
2582 (branch_nz_normal, branch_nz_invert): Likewise.
2583 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2585 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2586 * config/v850/v850.c (notice_update_cc): Remove.
2587 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2588 (CC_NO_CARRY): Likewise.
2589 (NOTICE_UPDATE_CC): Define to nothing.
2590 * config/v850/v850.md: Remove block comment on cc0 handling
2591 Remove "cc" attribute from all patterns. Remove cc_status handling
2592 from all patterns. Minor formatting fixes.
2594 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2596 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2597 (cortex-a76.cortex-a55): Likewise.
2598 * config/aarch64/aarch64-tune.md: Regenerate.
2599 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2600 cortex-a76.cortex-a55.
2602 2018-06-27 Jeff Law <law@redhat.com>
2604 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2605 (MULTILIB_DIRNAMES): Similarly.
2607 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2609 * gimple.h (gimple_return_retbnd): Delete.
2610 (gimple_return_set_retbnd): Likewise.
2611 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2612 gimple_return_set_retbnd.
2613 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2614 gimple_return_retbnd and adjust.
2615 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2616 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2617 Explicitly return NULL in a couple more cases. Move assertion
2618 on debug statements and remove unreachable code.
2619 (reset_debug_binding): Do not test id->retbnd.
2620 (expand_call_inline): Do not set it.
2622 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2624 * configure.ac: Add --disable-gcov option.
2625 * configure: Regenerate.
2626 * Makefile.in: Honour @enable_gcov@.
2627 * doc/install.texi: Document --disable-gcov.
2629 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2631 * config/arm/arm-cpus.in (cortex-a76): New entry.
2632 (cortex-a76.cortex-a55): Likewise.
2633 * config/arm/arm-tables.opt: Regenerate.
2634 * config/arm/arm-tune.md: Likewise.
2635 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2636 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2637 cortex-a76.cortex-a55.
2639 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2642 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2644 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2646 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2648 (EPILOGUE_USES): Likewise.
2650 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2652 * tree-inline.c (remap_location): New function extracted from...
2653 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2654 (copy_phis_for_bb): ...here. Call remap_location.
2655 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2657 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2659 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2660 unaligned vsx for 16B memset.
2662 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2665 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2666 ieee128_float_type_node to long_double_type_node unless
2667 TARGET_LONG_DOUBLE_128 is set.
2669 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2671 * cfgloop.c (get_loop_location): Convert return type from
2672 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2673 by implicit construction from rtx_insn *, and using
2674 dump_user_location_t::from_function_decl for the fallback case.
2675 * cfgloop.h (get_loop_location): Convert return type from
2676 location_t to dump_user_location_t.
2677 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2678 dump_printf_loc to pass in a dump_location_t rather than a
2679 location_t, via the gimple stmt.
2680 * coverage.c (get_coverage_counts): Update calls to
2681 dump_printf_loc to pass in dump_location_t rather than a
2683 * doc/optinfo.texi (Dump types): Convert example of
2684 dump_printf_loc from taking "locus" to taking "insn". Update
2685 description of the "_loc" calls to cover dump_location_t.
2686 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2688 (dump_user_location_t::dump_user_location_t): New constructors,
2689 from gimple *stmt and rtx_insn *.
2690 (dump_user_location_t::from_function_decl): New function.
2691 (dump_loc): Make static.
2692 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2693 const dump_location_t &.
2694 (dump_generic_expr_loc): Delete.
2695 (dump_printf_loc): Convert param "loc" from location_t to
2696 const dump_location_t &.
2697 (selftest::test_impl_location): New function.
2698 (selftest::dumpfile_c_tests): New function.
2699 * dumpfile.h: Include "profile-count.h".
2700 (class dump_user_location_t): New class.
2701 (struct dump_impl_location_t): New struct.
2702 (class dump_location_t): New class.
2703 (dump_printf_loc): Convert 2nd param from source_location to
2704 const dump_location_t &.
2705 (dump_generic_expr_loc): Delete.
2706 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2707 const dump_location_t &.
2708 * gimple-fold.c (fold_gimple_assign): Update call to
2709 dump_printf_loc to pass in a dump_location_t rather than a
2710 location_t, via the gimple stmt.
2711 (gimple_fold_call): Likewise.
2712 * gimple-loop-interchange.cc
2713 (loop_cand::analyze_iloop_reduction_var): Update for change to
2714 check_reduction_path.
2715 (tree_loop_interchange::interchange): Update for change to
2717 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2718 change in return-type of find_loop_location.
2719 (graphite_regenerate_ast_isl): Likewise.
2720 * graphite-optimize-isl.c (optimize_isl): Likewise.
2721 * graphite.c (graphite_transform_loops): Likewise.
2722 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2723 pass in a dump_location_t rather than a location_t, via the
2725 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2726 * ipa.c (walk_polymorphic_call_targets): Likewise.
2727 * loop-unroll.c (report_unroll): Convert "locus" param from
2728 location_t to dump_location_t.
2729 (decide_unrolling): Update for change to get_loop_location's
2731 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2732 location_t to dump_user_location_t.
2733 (grid_find_single_omp_among_assignments_1): Updates calls to
2734 dump_printf_loc to pass in a dump_location_t rather than a
2735 location_t, via the gimple stmt.
2736 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2737 location_t to dump_location_t. Updates calls to dump_printf_loc
2738 to pass in a dump_location_t rather than a location_t, via the
2740 (grid_inner_loop_gridifiable_p): Likewise.
2741 (grid_dist_follows_simple_pattern): Likewise.
2742 (grid_gfor_follows_tiling_pattern): Likewise.
2743 (grid_target_follows_gridifiable_pattern): Likewise.
2744 (grid_attempt_target_gridification): Convert initialization
2745 of local "grid" from memset to zero-initialization; FIXME: does
2746 this require C++11? Update call to dump_printf_loc to pass in a
2747 optinfo_location rather than a location_t, via the gimple stmt.
2748 * profile.c (read_profile_edge_counts): Updates call to
2749 dump_printf_loc to pass in a dump_location_t rather than a
2751 (compute_branch_probabilities): Likewise.
2752 * selftest-run-tests.c (selftest::run_tests): Call
2754 * selftest.h (dumpfile_c_tests): New decl.
2755 * tree-loop-distribution.c (pass_loop_distribution::execute):
2756 Update for change in return type of find_loop_location.
2757 * tree-parloops.c (parallelize_loops): Likewise.
2758 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2759 "locus" from location_t to dump_user_location_t.
2760 (canonicalize_loop_induction_variables): Likewise.
2761 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2762 for change in return type of find_loop_location.
2763 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2764 to dump_printf_loc to pass in a dump_location_t rather than a
2765 location_t, via the stmt.
2766 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2768 * tree-vect-loop-manip.c (find_loop_location): Convert return
2769 type from source_location to dump_user_location_t.
2770 (vect_do_peeling): Update for above change.
2771 (vect_loop_versioning): Update for change in type of
2773 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2774 from location_t to dump_user_location_t.
2775 (vect_estimate_min_profitable_iters): Update for change in type
2777 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2778 location_t to dump_location_t.
2779 (vect_slp_bb): Update for change in type of vect_location.
2780 * tree-vectorizer.c (vect_location): Convert from source_location
2781 to dump_user_location_t.
2782 (try_vectorize_loop_1): Update for change in vect_location's type.
2783 (vectorize_loops): Likewise.
2784 (increase_alignment): Likewise.
2785 * tree-vectorizer.h (vect_location): Convert from source_location
2786 to dump_user_location_t.
2787 (find_loop_location): Convert return type from source_location to
2788 dump_user_location_t.
2789 (check_reduction_path): Convert 1st param from location_t to
2790 dump_user_location_t.
2791 * value-prof.c (check_counter): Update call to dump_printf_loc to
2792 pass in a dump_user_location_t rather than a location_t; update
2793 call to error_at for change in type of "locus".
2794 (check_ic_target): Update call to dump_printf_loc to
2795 pass in a dump_user_location_t rather than a location_t, via the
2798 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2800 * config/s390/s390.h (enum processor_flags): Do not use
2801 default tune parameter when -march was specified.
2803 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2806 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2807 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2810 2018-06-26 Richard Biener <rguenther@suse.de>
2812 PR tree-optimization/86287
2814 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2815 (vect_analyze_loop): Initialize n_stmts.
2817 2018-06-26 Richard Biener <rguenther@suse.de>
2820 * fold-const.c (fold_convertible_p): Pointer extension
2823 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2826 * dwarf2out.c (loc_list_has_views): Adjust comments.
2827 (dw_loc_list): Split single cross-partition range with
2830 2018-06-25 Jeff Law <law@redhat.com>
2832 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2833 on -mbig-switch by default.
2835 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2837 (const_float_0_operand): Remove unused predicate.
2838 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2839 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2840 (recipsf2): New expander. Original pattern now called
2842 (recipdf2, recipdf2_insn): Similarly.
2843 (rsqrtsf2, rsqrtsf2_insn): Similarly
2844 (rsqrtdf2, rsqrtdf2_insn): Similarly
2846 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2848 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2849 Simplify logic for FreeBSD (twice).
2851 2018-06-25 Martin Sebor <msebor@redhat.com>
2853 PR tree-optimization/86204
2854 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2855 a strnlen result if it's less than the length of the string.
2857 2018-06-25 Martin Sebor <msebor@redhat.com>
2859 PR tree-optimization/85700
2860 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2861 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2862 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2864 2018-06-25 Martin Sebor <msebor@redhat.com>
2866 * doc/extend.texi (Zero-length arrays): Update and clarify.
2868 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2870 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2871 added IEEE/IBM long double multilib support on PowerPC little
2872 endian Linux systems.
2873 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2874 (MULTILIB_DEFAULTS): Likewise.
2875 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2877 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2878 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2879 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2881 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2884 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2885 (REORDER_45): Likewise.
2887 2018-06-25 Jeff Law <law@redhat.com>
2889 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2890 dividend to 32 bits. Adjust length.
2891 (udivmodhi4): Cleanup output template. Fix length.
2893 2018-06-25 Carl Love <cel@us.ibm.com>
2895 * config/rs6000/vsx.md: Change word selector to prefered location.
2897 2018-06-25 Richard Biener <rguenther@suse.de>
2899 PR tree-optimization/86304
2900 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2901 epilogue-if-converted loops as well.
2903 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2905 * lto-section-out.c (lto_begin_section): Do not print section
2906 name for noaddr and unnumbered dumps.
2908 2018-06-25 Richard Biener <rguenther@suse.de>
2910 * tree-vectorizer.h (struct vec_info_shared): New structure
2911 with parts split out from struct vec_info and loop_nest from
2912 struct _loop_vec_info.
2913 (struct vec_info): Adjust accordingly.
2914 (struct _loop_vec_info): Likewise.
2915 (LOOP_VINFO_LOOP_NEST): Adjust.
2916 (LOOP_VINFO_DATAREFS): Likewise.
2917 (LOOP_VINFO_DDRS): Likewise.
2918 (struct _bb_vec_info): Likewise.
2919 (BB_VINFO_DATAREFS): Likewise.
2920 (BB_VINFO_DDRS): Likewise.
2921 (struct _stmt_vec_info): Add dr_aux member.
2922 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2923 (DR_MISALIGNMENT_UNINITIALIZED): New.
2924 (set_dr_misalignment): Adjust.
2925 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2926 (vect_analyze_loop): Adjust prototype.
2927 (vect_analyze_loop_form): Likewise.
2928 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2929 Compute dependences lazily.
2930 (vect_record_base_alignments): Use shared datarefs/ddrs.
2931 (vect_verify_datarefs_alignment): Likewise.
2932 (vect_analyze_data_refs_alignment): Likewise.
2933 (vect_analyze_data_ref_accesses): Likewise.
2934 (vect_analyze_data_refs): Likewise.
2935 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2936 constructor parameter for shared part.
2937 (vect_analyze_loop_form): Pass in shared part and adjust.
2938 (vect_analyze_loop_2): Pass in storage for the number of
2939 stmts. Move loop nest finding to the caller. Compute
2941 (vect_analyze_loop): Pass in shared part.
2942 (vect_transform_loop): Verify shared datarefs are unchanged.
2943 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2944 constructor parameter for shared part.
2945 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2946 (vect_slp_bb): Verify shared datarefs are unchanged before
2948 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2950 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2951 DR_MISALIGNMENT_UNINITIALIZED.
2952 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2953 parameter for shared part.
2954 (vec_info::~vec_info): Adjust.
2955 (vec_info_shared::vec_info_shared): New.
2956 (vec_info_shared::~vec_info_shared): Likewise.
2957 (vec_info_shared::save_datarefs): Likewise.
2958 (vec_info_shared::check_datarefs): Likewise.
2959 (try_vectorize_loop_1): Construct shared part live for analyses
2960 of a single loop for multiple vector sizes.
2961 * tree-parloops.c (gather_scalar_reductions): Adjust.
2963 2018-06-25 Richard Biener <rguenther@suse.de>
2965 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2966 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2967 (vect_analyze_data_refs): Remove similar code from here and
2968 simplify accordingly.
2970 2018-06-25 Richard Biener <rguenther@suse.de>
2972 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2973 for reverse storage order accesses rather than asserting
2974 they cannot happen here.
2976 2018-06-25 Tom de Vries <tdevries@suse.de>
2979 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2980 Use data16 instead of .byte for insn prefix.
2982 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2985 * parser.c (make_char_string_pack): Pass this literal chars
2986 through cpp_interpret_string.
2987 (cp_parser_userdef_numeric_literal): Check the result of
2988 make_char_string_pack.
2990 2018-06-24 Maya Rashish <coypu@sdf.org>
2992 * ginclude/stddef.h: Simplify conditions around avoiding
2993 re-definition of __size_t.
2995 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2997 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2998 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3000 2018-06-22 Maya Rashish <coypu@sdf.org>
3002 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3003 made to OpenBSD and NetBSD through the years.
3005 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3007 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3008 behavior of vec_pack (vector double, vector double) to match
3009 behavior of vec_float2 (vector double, vector double).
3011 2018-06-22 Olivier Hainque <hainque@adacore.com>
3013 * gimplify.c (gimplify_function_tree): Prevent creation
3014 of a trampoline for the address of the current function
3015 passed to entry/exit instrumentation hooks.
3017 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3020 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3023 2018-06-22 Martin Liska <mliska@suse.cz>
3025 PR tree-optimization/86263
3026 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3027 Bail out if is_enabled is false.
3028 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3030 (jump_table_cluster::is_enabled): New function.
3032 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3034 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3035 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3036 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3037 (lto_input_ts_binfo_tree_pointers): Likewise.
3038 * tree-streamer-out.c (streamer_write_tree_bitfields,
3039 write_ts_binfo_tree_pointers): Likewise.
3040 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3042 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3044 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3046 2018-06-22 Martin Liska <mliska@suse.cz>
3048 * symbol-summary.h (get): Make it pure and inline move
3049 functionality from ::get function.
3050 (get): Remove and inline into ::get and ::get_create.
3051 (get_create): Move code from ::get function.
3053 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3056 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3057 -x assembler-with-cpp.
3059 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3061 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3062 _FILE_OFFSET_BITS=64 for C++.
3064 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3066 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3067 conversion insn that shows up when pr85657-3.c is compiled using
3068 IEEE 128-bit long double.
3069 (neg<mode>2_internal): Use the correct mode to check whether the
3070 mode is IBM extended.
3071 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3072 multiply and divide external functions from being created more
3075 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3077 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3079 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3080 the edge can be forwarded.
3081 (cfg_layout_merge_blocks): Likewise.
3083 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3085 * except.c (finish_eh_generation): Commit edge insertions only after
3086 the EH edges have been redirected from post-landing to landing pads.
3088 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3090 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3091 create_tmp_var_for to create the FRAME decl.
3092 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3094 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3096 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3097 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3098 debug statement when resetting its value.
3099 (expand_call_inline): Copy the locus of the call onto the assignment
3100 of the return value, if any. Use local variable in more cases.
3102 2018-06-21 Martin Liska <mliska@suse.cz>
3104 * ipa-pure-const.c (propagate_nothrow): Use
3105 funct_state_summaries->get.
3106 (dump_malloc_lattice): Likewise.
3107 (propagate_malloc): Likewise.
3109 2018-06-21 Richard Biener <rguenther@suse.de>
3111 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3112 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3113 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3115 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3116 BLOCK_ABSTRACT_ORIGIN unconditionally.
3118 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3120 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3122 * ipa-reference.c (ipa_reference_c_finalize): Delete
3123 ipa_ref_opt_sum_summaries and set it to NULL.
3125 2018-06-21 Tom de Vries <tdevries@suse.de>
3127 PR tree-optimization/85859
3128 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3129 test with comment from bb_no_side_effects_p.
3131 2018-06-21 Richard Biener <rguenther@suse.de>
3133 PR tree-optimization/86232
3134 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3135 max for constant niter.
3137 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3139 * config/aarch64/aarch64-simd.md
3140 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3142 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3144 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3145 Make opernads of the unspec commutative.
3147 2018-06-21 Richard Biener <rguenther@suse.de>
3149 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3150 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3151 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3152 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3154 (vect_record_base_alignments): Do not record base alignment
3155 for gathers or scatters.
3156 (vect_compute_data_ref_alignment): Drop return value that is always
3157 true. Bail out early for gathers or scatters.
3158 (vect_enhance_data_refs_alignment): Bail out early for gathers
3160 (vect_find_same_alignment_drs): Likewise.
3161 (vect_analyze_data_refs_alignment): Remove dead code.
3162 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3163 (vect_analyze_data_refs): For possible gathers or scatters do
3164 not create an alternate DR, just check their possible validity
3165 and mark them. Adjust DECL_NONALIASED handling to not rely
3167 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3168 update inits of gathers or scatters.
3169 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3170 Also copy gather/scatter flag to pattern vinfo.
3172 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3174 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3175 behavior of vec_packsu (vector unsigned long long, vector unsigned
3176 long long) to match behavior of vec_packs with same signature.
3178 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3179 Thomas Schwinge <thomas@codesourcery.com>
3180 Cesar Philippidis <cesar@codesourcery.com>
3182 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3183 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3184 (gimplify_adjust_omp_clauses): Likewise.
3185 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3186 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3187 (gimplify_omp_target_update): Update handling of acc update and
3189 * omp-low.c (install_var_field): Remove unused parameter
3190 base_pointers_restrict.
3191 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3192 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3194 (omp_target_base_pointers_restrict_p): Delete.
3195 (scan_omp_target): Update call to scan_sharing_clauses.
3196 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3198 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3199 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3200 (convert_local_omp_clauses): Likewise.
3201 * tree-pretty-print.c (dump_omp_clause): Likewise.
3202 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3204 (omp_clause_code_name): Likewise.
3206 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3209 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3212 PR tree-optimization/86231
3213 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3214 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3216 2018-06-20 Tom de Vries <tdevries@suse.de>
3218 PR tree-optimization/86097
3219 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3220 iv type if signedness of iv type is not the same as that of *nit.
3222 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3224 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3225 EDGE_EH edges, verify they are all EDGE_EH.
3227 2018-06-20 Maya Rashish <coypu@sdf.org>
3229 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3231 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3233 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3234 * config/aarch64/aarch64.c (xgene1_tunings): Add
3235 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3236 (aarch64_mode_valid_for_sched_fusion_p):
3237 Allow 16-byte modes.
3238 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3239 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3241 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3243 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3244 * config/aarch64/iterators.md (VQ2): New mode iterator.
3246 2018-06-20 Martin Liska <mliska@suse.cz>
3248 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3249 Change default ratio from 10 to 8.
3251 2018-06-20 Martin Liska <mliska@suse.cz>
3253 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3255 (bit_test_cluster::find_bit_tests): Likewise.
3256 (switch_decision_tree::analyze_switch_statement): Find clusters.
3257 * tree-switch-conversion.h (struct jump_table_cluster): Document
3260 2018-06-20 Martin Liska <mliska@suse.cz>
3262 * tree-switch-conversion.c (switch_conversion::collect):
3263 Record m_uniq property.
3264 (switch_conversion::expand): Bail out for special conditions.
3265 (group_cluster::~group_cluster): New.
3266 (group_cluster::group_cluster): Likewise.
3267 (group_cluster::dump): Likewise.
3268 (jump_table_cluster::emit): New.
3269 (switch_decision_tree::fix_phi_operands_for_edges): New.
3270 (struct case_node): Remove struct.
3271 (jump_table_cluster::can_be_handled): New.
3272 (case_values_threshold): Moved to header.
3273 (reset_out_edges_aux): Likewise.
3274 (jump_table_cluster::is_beneficial): New.
3275 (bit_test_cluster::can_be_handled): Likewise.
3276 (add_case_node): Remove.
3277 (bit_test_cluster::is_beneficial): New.
3278 (case_bit_test::cmp): New.
3279 (bit_test_cluster::emit): New.
3280 (expand_switch_as_decision_tree_p): Remove.
3281 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3282 (fix_phi_operands_for_edge): Likewise.
3283 (switch_decision_tree::analyze_switch_statement): New.
3284 (compute_cases_per_edge): Move ...
3285 (switch_decision_tree::compute_cases_per_edge): ... here.
3286 (try_switch_expansion): Likewise.
3287 (switch_decision_tree::try_switch_expansion): Likewise.
3288 (record_phi_operand_mapping): Likewise.
3289 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3290 (emit_case_decision_tree): Likewise.
3291 (switch_decision_tree::emit): Likewise.
3292 (balance_case_nodes): Likewise.
3293 (switch_decision_tree::balance_case_nodes): Likewise.
3294 (dump_case_nodes): Likewise.
3295 (switch_decision_tree::dump_case_nodes): Likewise.
3296 (emit_jump): Likewise.
3297 (switch_decision_tree::emit_jump): Likewise.
3298 (emit_cmp_and_jump_insns): Likewise.
3299 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3300 (emit_case_nodes): Likewise.
3301 (switch_decision_tree::emit_case_nodes): Likewise.
3302 (conditional_probability): Remove.
3303 * tree-switch-conversion.h (enum cluster_type): New.
3305 (struct cluster): Likewise.
3306 (cluster::cluster): Likewise.
3307 (struct simple_cluster): Likewise.
3308 (simple_cluster::simple_cluster): Likewise.
3309 (struct group_cluster): Likewise.
3310 (struct jump_table_cluster): Likewise.
3311 (struct bit_test_cluster): Likewise.
3312 (struct min_cluster_item): Likewise.
3313 (struct case_tree_node): Likewise.
3314 (case_tree_node::case_tree_node): Likewise.
3315 (jump_table_cluster::case_values_threshold): Likewise.
3316 (struct case_bit_test): Likewise.
3317 (struct switch_decision_tree): Likewise.
3318 (struct switch_conversion): Likewise.
3319 (switch_decision_tree::reset_out_edges_aux): Likewise.
3321 2018-06-20 Martin Liska <mliska@suse.cz>
3323 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3324 (hoist_edge_and_branch_if_true): Likewise.
3325 (expand_switch_using_bit_tests_p): Likewise.
3326 (struct case_bit_test): Likewise.
3327 (case_bit_test_cmp): Likewise.
3328 (emit_case_bit_tests): Likewise.
3329 (switch_conversion::switch_conversion): New class.
3330 (struct switch_conv_info): Remove old struct.
3331 (collect_switch_conv_info): More to ...
3332 (switch_conversion::collect): ... this.
3333 (check_range): Likewise.
3334 (switch_conversion::check_range): Likewise.
3335 (check_all_empty_except_final): Likewise.
3336 (switch_conversion::check_all_empty_except_final): Likewise.
3337 (check_final_bb): Likewise.
3338 (switch_conversion::check_final_bb): Likewise.
3339 (create_temp_arrays): Likewise.
3340 (switch_conversion::create_temp_arrays): Likewise.
3341 (free_temp_arrays): Likewise.
3342 (gather_default_values): Likewise.
3343 (switch_conversion::gather_default_values): Likewise.
3344 (build_constructors): Likewise.
3345 (switch_conversion::build_constructors): Likewise.
3346 (constructor_contains_same_values_p): Likewise.
3347 (switch_conversion::contains_same_values_p): Likewise.
3348 (array_value_type): Likewise.
3349 (switch_conversion::array_value_type): Likewise.
3350 (build_one_array): Likewise.
3351 (switch_conversion::build_one_array): Likewise.
3352 (build_arrays): Likewise.
3353 (switch_conversion::build_arrays): Likewise.
3354 (gen_def_assigns): Likewise.
3355 (switch_conversion::gen_def_assigns): Likewise.
3356 (prune_bbs): Likewise.
3357 (switch_conversion::prune_bbs): Likewise.
3358 (fix_phi_nodes): Likewise.
3359 (switch_conversion::fix_phi_nodes): Likewise.
3360 (gen_inbound_check): Likewise.
3361 (switch_conversion::gen_inbound_check): Likewise.
3362 (process_switch): Use the newly created class.
3363 (switch_conversion::expand): New.
3364 (switch_conversion::~switch_conversion): New.
3365 * tree-switch-conversion.h: New file.
3367 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3369 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3370 tree-vect-patterns.c.
3371 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3372 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3373 (vect_recog_sad_pattern): Likewise.
3374 (vect_recog_widen_sum_pattern): Likewise.
3375 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3376 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3377 (vect_recog_rotate_pattern): Likewise.
3378 (vect_recog_mult_pattern): Likewise.
3379 (vect_recog_vector_vector_shift_pattern): Likewise.
3380 (vect_recog_divmod_pattern): Likewise.
3381 (vect_recog_mixed_size_cond_pattern): Likewise.
3382 (vect_recog_bool_pattern): Likewise.
3383 (vect_recog_mask_conversion_pattern): Likewise.
3384 (vect_try_gather_scatter_pattern): Likewise.
3385 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3386 (vect_recog_over_widening_pattern): Likewise.
3387 (vect_recog_gather_scatter_pattern): Likewise.
3388 (vect_recog_func_ptr): Move from tree-vectorizer.h
3389 (vect_vect_recog_func_ptrs): Move further down the file.
3390 (vect_recog_func): Likewise. Remove the third argument.
3391 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3392 (vect_pattern_recog_1): Expect the pattern function to do any
3393 necessary target tests. Also expect it to provide a vector type.
3394 Remove the type_in handling.
3396 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3398 * tree-vect-patterns.c (vect_pattern_detected): New function.
3399 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3400 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3401 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3402 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3403 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3404 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3405 (vect_recog_mask_conversion_pattern)
3406 (vect_try_gather_scatter_pattern): Likewise.
3408 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3410 * tree-vect-patterns.c (vect_get_internal_def): New function.
3411 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3412 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3413 (search_type_for_mask_1): Use it.
3415 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3417 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3418 redundant WIDEN_SUM_EXPR handling.
3419 (vect_recog_sad_pattern): Likewise.
3421 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3423 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3424 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3425 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3426 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3427 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3429 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3431 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3432 use the stmt_vec_info of the original bb statement for the
3433 new zero assignment, even if the call is part of a pattern.
3435 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3437 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3438 that the sequence is attached to the original statement rather
3439 than the pattern statement.
3440 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3441 PATTERN_DEF_SEQ from the original statement rather than
3442 the main pattern statement.
3443 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3444 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3445 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3447 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3449 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3450 definition statements before the early exit for statements that aren't
3452 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3454 (vect_transform_loop): ...here. Process pattern definition
3455 statements without first checking whether the main pattern
3456 statement is live or relevant.
3458 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3460 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3461 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3463 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3465 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3466 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3467 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3468 (expand_block_compare): Change select_block_compare_mode call.
3469 (expand_strncmp_align_check): Use new functions, fix comment.
3470 (emit_final_str_compare_gpr): New function.
3471 (expand_strn_compare): Refactor and clean up code.
3472 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3474 2018-06-19 Tony Reix <tony.reix@atos.com>
3475 Damien Bergamini <damien.bergamini@atos.com>
3476 David Edelsohn <dje.gcc@gmail.com>
3478 * collect2.c (static_obj): New variable.
3479 (static_libs): New variable.
3480 (is_in_list): Uncomment declaration.
3481 (main): Track AIX libraries linked statically.
3482 (is_in_list): Uncomment definition.
3483 (scan_prog_file): Don't add AIX shared libraries initializer
3484 to constructor list if linking statically.
3486 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3488 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3490 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3492 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3494 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3497 2018-06-19 Martin Liska <mliska@suse.cz>
3499 * config/i386/i386.c (ix86_can_inline_p): Do not use
3500 ipa_fn_summaries::get_create.
3501 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3503 (devirtualization_time_bonus): Likewise.
3504 (ipcp_propagate_stage): Likewise.
3505 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3506 (edge_set_predicate): Likewise.
3507 (evaluate_conditions_for_known_args): Likewise.
3508 (evaluate_properties_for_edge): Likewise.
3509 (ipa_call_summary::reset): Tranform to ...
3510 (ipa_call_summary::~ipa_call_summary): ... this.
3511 (ipa_fn_summary::reset): Transform to ...
3512 (ipa_fn_summary::~ipa_fn_summary): ... this.
3513 (ipa_fn_summary_t::remove): Rename to ...
3514 (ipa_fn_summary_t::remove_callees): ... this.
3515 (ipa_fn_summary_t::duplicate): Use placement new
3516 instead of memory copy.
3517 (ipa_call_summary_t::duplicate): Likewise.
3518 (ipa_call_summary_t::remove): Remove.
3519 (dump_ipa_call_summary): Change get_create to get.
3520 (ipa_dump_fn_summary): Dump only when summary exists.
3521 (analyze_function_body): Use symbol_summary::get instead
3523 (compute_fn_summary): Likewise.
3524 (estimate_edge_devirt_benefit): Likewise.
3525 (estimate_edge_size_and_time): Likewise.
3526 (inline_update_callee_summaries): Likewise.
3527 (remap_edge_change_prob): Likewise.
3528 (remap_edge_summaries): Likewise.
3529 (ipa_merge_fn_summary_after_inlining): Likewise.
3530 (write_ipa_call_summary): Likewise.
3531 (ipa_fn_summary_write): Likewise.
3532 (ipa_free_fn_summary): Likewise.
3533 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3534 (struct ipa_call_summary): Likewise.
3535 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3537 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3538 (estimate_size_after_inlining): Likewise.
3539 (estimate_growth): Likewise.
3540 (growth_likely_positive): Likewise.
3541 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3542 (inline_call): Likewise.
3543 * ipa-inline.c (caller_growth_limits): Likewise.
3544 (can_inline_edge_p): Likewise.
3545 (can_inline_edge_by_limits_p): Likewise.
3546 (compute_uninlined_call_time): Likewise.
3547 (compute_inlined_call_time): Likewise.
3548 (want_inline_small_function_p): Likewise.
3549 (edge_badness): Likewise.
3550 (update_caller_keys): Likewise.
3551 (update_callee_keys): Likewise.
3552 (inline_small_functions): Likewise.
3553 (inline_to_all_callers_1): Likewise.
3554 (dump_overall_stats): Likewise.
3555 (early_inline_small_functions): Likewise.
3556 (early_inliner): Likewise.
3557 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3558 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3559 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3560 * ipa-split.c (execute_split_functions): Likewise.
3561 * symbol-summary.h: Likewise.
3562 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3564 2018-06-19 Richard Biener <rguenther@suse.de>
3566 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3567 (vectorize_loops): ... here. Fix dbgcnt handling.
3568 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3570 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3573 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3574 ieee128 argument takes up only one (vector) register, not two (floating
3577 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3579 * gimplify.c (gimplify_init_constructor): Really never clear for an
3580 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3582 2018-06-19 Richard Biener <rguenther@suse.de>
3584 PR tree-optimization/86179
3585 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3586 after failed recognition.
3588 2018-06-18 Martin Sebor <msebor@redhat.com>
3591 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3592 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3593 Handle integer subtraction.
3594 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3595 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3597 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3599 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3600 param from rtx to rtx_insn *.
3601 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3603 (frv_ifcvt_modify_insn): Likwise.
3604 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3605 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3606 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3607 as_a <rtx_insn *> cast to local "unprotected_region" once
3608 it's been established that it's not NULL or pc_rtx.
3609 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3610 param "sethi" from rtx to rtx_insn *.
3611 (nds32_group_float_insns): Likewise for param "insn".
3612 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3614 (vax_output_int_subtract): Likewise.
3615 * config/vax/vax.c (vax_output_int_add): Likewise for param
3617 (vax_output_int_subtract): Likewise.
3618 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3619 (emit_pattern_after): Likewise for param "after".
3620 (emit_insn_after): Likewise.
3621 (emit_jump_insn_after): Likewise.
3622 (emit_call_insn_after): Likewise.
3623 (emit_debug_insn_after): Likewise.
3624 (emit_pattern_before): Likewise for param "before".
3625 (emit_insn_before): Likewise.
3626 (emit_jump_insn_before): Likewise.
3627 * final.c (get_insn_template): Likewise for param "insn", removing
3629 * output.h (get_insn_template): Likewise for 2nd param.
3630 * rtl.h (emit_insn_before): Likewise.
3631 (emit_jump_insn_before): Likewise.
3632 (emit_debug_insn_before_noloc): Likewise.
3633 (emit_insn_after): Likewise.
3634 (emit_jump_insn_after): Likewise.
3635 (emit_call_insn_after): Likewise.
3636 (emit_debug_insn_after): Likewise.
3637 (set_insn_deleted): Likewise for param.
3639 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3642 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3643 floating point modes, so that IFmode is numerically greater than
3644 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3645 to declare the ordering. This prevents IFmode from being
3646 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3647 machine. Include rs6000-modes.h to share the fractional values
3648 between genmodes* and the rest of the compiler.
3652 * config/rs6000/rs6000-modes.h: New file.
3653 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3654 meaning of rs6000_long_double_size so that 126..128 selects an
3655 appropriate 128-bit floating point type.
3656 (rs6000_option_override_internal): Likewise.
3657 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3658 (TARGET_LONG_DOUBLE_128): Change the meaning of
3659 rs6000_long_double_size so that 126..128 selects an appropriate
3660 128-bit floating point type.
3661 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3662 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3663 source and destination to match the standard usage.
3664 (truncifkf2): Likewise.
3665 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3666 ISA 2.07 to use an explicit clobber, instead of passing in a
3668 (copysign<mode>3_soft): Likewise.
3670 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3672 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3673 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3674 (vect_slp_analyze_instance_dependence): Likewise.
3675 (vect_enhance_data_refs_alignment): Likewise.
3676 (vect_analyze_data_refs_alignment): Likewise.
3677 (vect_slp_analyze_and_verify_instance_alignment
3678 (vect_analyze_data_ref_accesses): Likewise.
3679 (vect_prune_runtime_alias_test_list): Likewise.
3680 (vect_analyze_data_refs): Likewise.
3681 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3682 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3683 (vect_analyze_scalar_cycles_1): Likewise.
3684 (vect_get_loop_niters): Likewise.
3685 (vect_analyze_loop_form_1): Likewise.
3686 (vect_update_vf_for_slp): Likewise.
3687 (vect_analyze_loop_operations): Likewise.
3688 (vect_analyze_loop): Likewise.
3689 (vectorizable_induction): Likewise.
3690 (vect_transform_loop): Likewise.
3691 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3692 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3693 (vect_make_slp_decision): Likewise.
3694 (vect_detect_hybrid_slp): Likewise.
3695 (vect_slp_analyze_operations): Likewise.
3696 (vect_slp_bb): Likewise.
3697 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3698 (vectorizable_bswap): Likewise.
3699 (vectorizable_call): Likewise.
3700 (vectorizable_simd_clone_call): Likewise.
3701 (vectorizable_conversion): Likewise.
3702 (vectorizable_assignment): Likewise.
3703 (vectorizable_shift): Likewise.
3704 (vectorizable_operation): Likewise.
3705 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3707 2018-06-18 Martin Sebor <msebor@redhat.com>
3709 PR tree-optimization/81384
3710 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3711 * builtins.c (expand_builtin_strnlen): New function.
3712 (expand_builtin): Call it.
3713 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3714 * builtins.def (BUILT_IN_STRNLEN): New.
3715 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3716 Warn for bounds in excess of maximum object size.
3717 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3718 single-value ranges. Handle strnlen.
3719 (handle_builtin_strlen): Handle strnlen.
3720 (strlen_check_and_optimize_stmt): Same.
3721 * doc/extend.texi (Other Builtins): Document strnlen.
3723 2018-06-18 Maya Rashish <coypu@sdf.org>
3725 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3726 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3727 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3729 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3731 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3733 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3735 * tree.c (escaped_string::escape): Replace cast to char * by
3736 const_cast<char *> (unescaped).
3738 2018-06-18 Nick Clifton <nickc@redhat.com>
3741 * tree.c (escaped_string): New class. Converts an unescaped
3742 string into its escaped equivalent.
3743 (warn_deprecated_use): Use the new class to convert the
3744 deprecation message, if present.
3745 (test_escaped_strings): New self test.
3746 (test_c_tests): Add test_escaped_strings.
3747 * doc/extend.texi (deprecated): Add a note that the
3748 deprecation message is affected by the -fmessage-length
3749 option, and that control characters will be escaped.
3750 (#pragma GCC error): Document this pragma.
3751 (#pragma GCC warning): Likewise.
3752 * doc/invoke.texi (-fmessage-length): Document this option's
3753 effect on the #warning and #error preprocessor directives and
3754 the deprecated attribute.
3756 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3758 * tree.c (decl_value_expr_lookup): Revert latest change.
3759 (decl_value_expr_insert): Likewise.
3761 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3763 * gimplify.c (nonlocal_vlas): Delete.
3764 (nonlocal_vla_vars): Likewise.
3765 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3767 (gimplify_body): Do not create and destroy nonlocal_vlas.
3768 * tree-nested.c: Include diagnostic.h.
3769 (use_pointer_in_frame): Tweak.
3770 (lookup_field_for_decl): Add assertion and declare the transformation.
3771 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3772 internal error when the reference is in a wrong context. Do not
3773 create a debug decl by default.
3774 (note_nonlocal_block_vlas): Delete.
3775 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3776 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3777 create a debug decl by default.
3778 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3779 call is in a wrong context.
3780 (fixup_vla_decls): New function.
3781 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3782 debug variables were created.
3783 * tree.c (decl_value_expr_lookup): Add checking assertion.
3784 (decl_value_expr_insert): Likewise.
3786 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3789 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3790 * tree-scalar-evolution.c (interpret_expr): Likewise.
3791 (expression_expensive_p): Likewise.
3792 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3793 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3794 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3795 (ssa_defined_by_minus_one_stmt_p): New.
3797 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3800 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3801 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3802 * dojump.c (do_jump): Likewise.
3803 * expr.c (expand_expr_real_2): Check operand type's sign.
3804 * fold-const.c (const_unop): Handle ABSU_EXPR.
3805 (fold_abs_const): Likewise.
3806 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3807 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3808 (strip_sign_op_1): Likesise.
3809 * match.pd: Add new pattern to generate ABSU_EXPR.
3810 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3811 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3812 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3813 * tree-inline.c (estimate_operator_cost): Likewise.
3814 * tree-pretty-print.c (dump_generic_node): Likewise.
3815 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3816 * tree.def (ABSU_EXPR): New.
3818 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3821 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3822 documented as preserved for backward compatibility only.
3823 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3825 PR rtl-optimization/86108
3826 * bb-reorder.c (create_forwarder_block): Renamed to ...
3827 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3828 jump from new landing pad to the second part.
3829 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3832 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3835 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3836 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3837 Only call store_expr for halves if the mode is the same.
3840 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3841 Fix up comment formatting.
3843 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3845 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3846 (typed_splay_tree::closure,
3847 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3848 (typed_splay_tree::typed_splay_tree,
3849 typed_splay_tree::operator =): Declared private.
3850 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3851 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3852 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3853 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3854 typed_splay_tree::splay_tree_splay,
3855 typed_splay_tree::splay_tree_foreach_helper,
3856 typed_splay_tree::splay_tree_insert,
3857 typed_splay_tree::splay_tree_remove,
3858 typed_splay_tree::splay_tree_lookup,
3859 typed_splay_tree::splay_tree_predecessor,
3860 typed_splay_tree::splay_tree_successor,
3861 typed_splay_tree::splay_tree_min,
3862 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3863 (typed_splay_tree::root, typed_splay_tree::comp,
3864 typed_splay_tree::delete_key,
3865 typed_splay_tree::delete_value): New data members.
3866 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3867 typed_splay_tree::remove.
3869 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3871 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3872 -mginv and -mno-ginv to the assembler.
3873 * config/mips/mips.opt (-mcrc): New option.
3875 * doc/invoke.text (-mcrc): Document.
3878 2018-06-15 Nick Clifton <nickc@redhat.com>
3881 * tree.c (escaped_string): New class. Converts an unescaped
3882 string into its escaped equivalent.
3883 (warn_deprecated_use): Use the new class to convert the
3884 deprecation message, if present.
3885 (test_escaped_strings): New self test.
3886 (test_c_tests): Add test_escaped_strings.
3887 * doc/extend.texi (deprecated): Add a note that the
3888 deprecation message is affected by the -fmessage-length
3889 option, and that control characters will be escaped.
3890 (#pragma GCC error): Document this pragma.
3891 (#pragma GCC warning): Likewise.
3892 * doc/invoke.texi (-fmessage-length): Document this option's
3893 effect on the #warning and #error preprocessor directives and
3894 the deprecated attribute.
3896 2018-06-15 Richard Biener <rguenther@suse.de>
3898 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3899 here, also noting vector size used.
3900 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3901 size used in MSG_OPTIMIZED_LOCATIONS dump.
3902 (pass_slp_vectorize::execute): Adjust.
3904 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3907 * config/arc/arc.c (arc_return_address_register): Fix
3910 2018-06-15 Richard Biener <rguenther@suse.de>
3913 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3914 leave useless conversion stripping to force_gimple_operand_gsi.
3915 (gimplify_build2): Likewise.
3916 (gimplify_build1): Likewise.
3918 2018-06-15 Richard Biener <rguenther@suse.de>
3921 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3922 before adjusting their block.
3924 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3926 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3927 multilibs for *-*-rtems*.
3928 * config/riscv/t-rtems: New file.
3930 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3933 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3934 unsigned_type_for returns NULL.
3937 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3938 subregs of multi-word pseudos unless the float mode has word size.
3940 2018-06-14 Richard Biener <rguenther@suse.de>
3943 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3944 duplicate and harmful type_hash_canon.
3945 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3947 2018-06-14 Richard Biener <rguenther@suse.de>
3950 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3953 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3955 * config/rtems.h (STDINT_LONG32): Define.
3957 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3958 Prachi Godbole <prachi.godbole@imgtec.com>
3960 * config/mips/mips-cpus.def: Define P6600.
3961 * config/mips/mips-tables.opt: Regenerate.
3962 * config/mips/mips.c (mips_ucbranch_type): New enum.
3963 (mips_rtx_cost_data): Add support for P6600.
3964 (mips_issue_rate): Likewise.
3965 (mips_multipass_dfa_lookahead): Likewise.
3966 (mips_avoid_hazard): Likewise.
3967 (mips_reorg_process_insns): Likewise.
3968 (mips_classify_branch_p6600): New function.
3969 * config/mips/mips.h (TUNE_P6600): New define.
3970 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3971 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3972 * config/mips/mips.md: Include p6600.md.
3973 (processor): Add p6600.
3974 * config/mips/p6600.md: New file.
3975 * doc/invoke.texi: Add p6600 to supported architectures.
3977 2018-06-13 Martin Sebor <msebor@redhat.com>
3979 PR tree-optimization/86114
3980 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3982 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3984 2018-06-13 Richard Biener <rguenther@suse.de>
3986 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3987 Properly set vector type of the intermediate stmt.
3988 * tree-vect-stmts.c (vectorizable_operation): The destination
3989 var always has vectype_out type.
3991 2018-06-13 Jeff Law <law@redhat.com>
3993 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3994 integer 0 for argument to print_rtl_with_bb.
3995 (rl78_reorg): Likewise.
3997 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3999 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4000 from rtx to rtx_insn *.
4001 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4003 (add_sched_insns_for_speculation): Likewise for local "target",
4004 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4005 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4006 from rtx_insn ** to rtx_code_label **.
4007 (reorg_emit_nops): Likewise.
4008 (c6x_reorg): Likewise for local "call_labels".
4009 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4011 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4012 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4013 the loops over LABEL_REFS.
4014 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4016 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4017 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4018 (split_branches): Strengthen local "olabel" from rtx to
4019 rtx_insn *, adding a safe_as_a cast.
4020 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4022 (add_insn_after): Likewise for first two params.
4023 (add_insn_before): Likewise.
4024 (remove_insn): Likewise for param.
4025 (emit_pattern_before_noloc): Likewise for second and third params.
4026 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4027 (emit_call_insn_before_noloc): Likewise.
4028 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4030 (emit_barrier_before): Likewise.
4031 (emit_label_before): Strengthen "label" param from "rtx" to
4032 "rtx_code_label *". Strengthen "before" param from "rtx" to
4034 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4036 (emit_pattern_after_noloc): Likewise.
4037 (emit_insn_after_noloc): Likewise.
4038 (emit_jump_insn_after_noloc): Likewise.
4039 (emit_call_insn_after_noloc): Likewise.
4040 (emit_debug_insn_after_noloc): Likewise.
4041 (emit_barrier_after): Likewise.
4042 (emit_label_after): Likewise for both params.
4043 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4044 "loc" param from "int" to "location_t".
4045 (emit_insn_after_setloc): Likewise.
4046 (emit_jump_insn_after_setloc): Likewise.
4047 (emit_call_insn_after_setloc): Likewise.
4048 (emit_debug_insn_after_setloc): Likewise.
4049 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4050 "loc" param from "int" to "location_t".
4051 (emit_pattern_before): Convert NULL_RTX to NULL.
4052 (emit_insn_before_setloc): Convert "loc" param from "int" to
4054 (emit_jump_insn_before_setloc): Likewise.
4055 (emit_call_insn_before_setloc): Likewise.
4056 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4057 rtx_insn *. Convert "loc" param from "int" to "location_t".
4058 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4059 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4060 Convert 3rd param from "int" to "location_t".
4061 (emit_barrier_before, emit_barrier_after, next_real_insn):
4062 Strengthen param from rtx to rtx_insn *.
4063 (emit_label_before): Strengthen 1st param from "rtx" to
4064 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4066 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4067 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4068 Strengthen 2nd param from "rtx" to "rtx_insn *".
4069 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4070 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4071 Likewise. Convert 3rd param from "int" to "location_t".
4072 (emit_label_after): Strengthen 1st param from "rtx" to
4074 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4076 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4077 from "rtx" to "rtx_insn *".
4079 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4081 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4082 bodies streamed in with -Q.
4083 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4084 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4085 * gimple-streamer-out.c: Include gimple-pretty-print.h
4086 (output_bb): Dump stmts streamed.
4087 * lto-section-out.c: Include print-tree.h
4088 (lto_begin_section): Dump sections created.
4089 (lto_output_decl_index): Dump decl encoded.
4090 * lto-streamer-out.c: Include print-tree.h
4091 (create_output_block): Dump output block created.
4092 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4093 (output_function): Dump function output.
4094 (output_constructor): Dump constructor streamed.
4095 (write_global_stream): Output indexes encoded.
4096 (produce_asm_for_decls): Dump streams encoded.
4097 * lto-streamer.c (streamer_dump_file): New global var.
4098 * lto-streamer.h (streamer_dump_file): Declare.
4099 * passes.c (ipa_write_summaries): Initialize streamer dump.
4100 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4103 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4106 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4107 offsets for register save directives. Emit a second batch of save
4108 directives, if need be, when the function accesses prior frames.
4110 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4112 * config/arc/fpu.md (fmasf4): Force operand to register.
4113 (fnmasf4): Likewise.
4115 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4117 * config/arc/arc-protos.h (arc_pad_return): Remove.
4118 * config/arc/arc.c (machine_function): Remove force_short_suffix
4120 (arc_print_operand): Adjust printing of '&'.
4121 (arc_verify_short): Remove conditional printing of short suffix.
4122 (arc_final_prescan_insn): Remove reference to size_reason.
4123 (pad_return): New function.
4124 (arc_reorg): Call pad_return.
4125 (arc_pad_return): Remove.
4126 (arc_init_machine_status): Remove reference to force_short_suffix.
4127 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4128 (attr length): When attribute iscompact is true force to 2
4129 regardless; in the case of maybe check if we want to force the
4130 instruction to have 4 bytes length.
4131 (nopv): Change it to generate 4 byte long nop as well.
4132 (blockage): New pattern.
4133 (simple_return): Remove call to arc_pad_return.
4134 (p_return_i): Likewise.
4136 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4138 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4140 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4142 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4145 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4147 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4148 for ARC700 and ARCv2.
4150 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4153 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4154 operands[2] instead of operands[1].
4157 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4159 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4160 case, check whether the outer register overlaps an unallocatable
4161 register, not just whether it fits the required class.
4163 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4165 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4167 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4168 (memrefs_conflict_p): Likewise.
4169 (init_alias_analysis): Likewise.
4170 * cfgexpand.c (expand_debug_expr): Likewise.
4171 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4172 * cse.c (fold_rtx): Likewise.
4173 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4174 * expr.c (emit_block_move_hints): Likewise.
4175 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4176 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4177 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4178 (emit_group_store): Likewise.
4179 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4180 to read the PRE/POST_MODIFY increment.
4181 * calls.c (store_one_arg): Use strip_offset.
4182 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4184 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4186 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4187 (simplify_binary_operation_1): Extend CONST_INT handling to
4189 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4190 than a HOST_WIDE_INT.
4191 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4193 (adjust_mems, add_stores): Update accodingly.
4194 (vt_canonicalize_addr): Track polynomial offsets.
4195 (emit_note_insn_var_location): Likewise.
4196 (vt_add_function_parameter): Likewise.
4197 (vt_initialize): Likewise.
4199 2018-06-12 Jeff Law <law@redhat.com>
4201 * config.gcc (alpha*-*-freebsd*): Remove.
4202 * config/alpha/freebsd.h: Remove.
4204 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4207 * spellcheck-tree.c (levenshtein_distance): Rename to...
4208 (get_edit_distance): ...this, and update for underlying renaming.
4209 * spellcheck-tree.h (levenshtein_distance): Rename to...
4210 (get_edit_distance): ...this.
4211 * spellcheck.c (levenshtein_distance): Rename to...
4212 (get_edit_distance): ...this. Convert from Levenshtein distance
4213 to Damerau-Levenshtein distance by supporting transpositions of
4214 adjacent characters. Rename "v1" to "v_next" and "v0" to
4216 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4217 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4218 update for underlying renaming.
4219 (selftest::levenshtein_distance_unit_test): Rename to...
4220 (selftest::test_get_edit_distance_unit): ...this, and update for
4221 underlying renaming.
4222 (selftest::test_find_closest_string): Add example from PR 69968
4223 where transposition helps
4224 (selftest::test_metric_conditions): Update for renaming.
4225 (selftest::test_metric_conditions): Likewise.
4226 (selftest::spellcheck_c_tests): Likewise.
4227 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4228 (get_edit_distance): ...this.
4229 (best_match::consider): Update for renaming.
4231 2018-06-12 Martin Sebor <msebor@redhat.com>
4233 PR tree-optimization/85259
4234 * builtins.c (compute_objsize): Handle constant offsets.
4235 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4236 true iff a warning has been issued.
4237 * gimple.h (gimple_nonartificial_location): New function.
4238 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4239 gimple_nonartificial_location and handle -Wno-system-headers.
4240 (handle_builtin_stxncpy): Same.
4242 2018-06-12 Martin Sebor <msebor@redhat.com>
4245 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4247 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4249 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4250 BUILTIN_VEC_XST entries for pointer to double and long long.
4252 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4255 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4257 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4260 2018-06-12 Martin Liska <mliska@suse.cz>
4262 * doc/options.texi: Document IntegerRange.
4264 2018-06-12 Martin Liska <mliska@suse.cz>
4266 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4267 * opt-functions.awk: Handle Deprecated flag.
4268 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4270 (read_cmdline_option): Report warning for a deprecated option.
4271 * opts.h (struct cl_option): Add new field cl_deprecated.
4272 (CL_ERR_DEPRECATED): New.
4274 2018-06-12 Martin Liska <mliska@suse.cz>
4276 * doc/options.texi: Document Deprecated option flag.
4278 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4280 * config/arc/arc-arch.h (arc_extras): New enum.
4281 (arc_cpu_t):Add field extra.
4282 (arc_cpu_types): Consider the extras.
4283 * config/arc/arc-cpus.def: Add extras info.
4284 * config/arc/arc-opts.h (processor_type): Consider extra field.
4285 * config/arc/arc.c (arc_override_options): Handle extra field.
4287 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4289 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4290 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4292 * config/arc/arc.c (arc_init): Update pic warning.
4293 (irq_range): Update irq range parsing warnings.
4294 (arc_override_options): Update various warning messages.
4295 (arc_handle_aux_attribute): Likewise.
4297 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4299 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4301 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4303 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4304 compiler messages directives.
4306 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4308 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4309 * config/mips/mips-tables.opt: Regenerate.
4310 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4312 * doc/invoke.texi: Document -march=i6500.
4314 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4316 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4317 (i6400_gpmul): Add cpu_unit.
4318 (i6400_gpdiv): Likewise.
4319 (i6400_msa_add_d): Update reservations.
4320 (i6400_msa_int_add) Likewise.
4321 (i6400_msa_short_logic3) Likewise.
4322 (i6400_msa_short_logic2) Likewise.
4323 (i6400_msa_short_logic) Likewise.
4324 (i6400_msa_move) Likewise.
4325 (i6400_msa_cmp) Likewise.
4326 (i6400_msa_short_float2) Likewise.
4327 (i6400_msa_div_d) Likewise.
4328 (i6400_msa_long_logic1) Likewise.
4329 (i6400_msa_long_logic2) Likewise.
4330 (i6400_msa_mult) Likewise.
4331 (i6400_msa_long_float2) Likewise.
4332 (i6400_msa_long_float4) Likewise.
4333 (i6400_msa_long_float5) Likewise.
4334 (i6400_msa_long_float8) Likewise.
4335 (i6400_fpu_fadd): Include frint type.
4336 (i6400_fpu_store): New define_insn_reservation.
4337 (i6400_fpu_load): Likewise.
4338 (i6400_fpu_move): Likewise.
4339 (i6400_fpu_fcmp): Likewise.
4340 (i6400_fpu_fmadd): Likewise.
4341 (i6400_int_mult): Include imul3nc type and update reservation.
4342 (i6400_int_div): Include idiv3 type and update reservation.
4343 (i6400_int_load): Update to check type not move_type.
4344 (i6400_int_store): Likewise.
4345 (i6400_int_prefetch): Set zero latency.
4347 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4349 * gcc.c: Document new %@{...} sequence.
4350 (LINK_COMMAND_SPEC): Use it for the -L switches.
4351 (cpp_unique_options): Use it for the -I switches.
4352 (at_file_argbuf): New global variable.
4353 (in_at_file): Likewise.
4354 (alloc_args): Create at_file_argbuf.
4355 (clear_args): Truncate at_file_argbuf.
4356 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4357 (open_at_file): New function.
4358 (close_at_file): Likewise.
4359 (create_at_file): Delete.
4360 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4363 (validate_switches_from_spec): Deal with %@{...} sequence.
4364 (validate_switches): Likewise.
4365 (driver::finalize): Call clear_args.
4367 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4369 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4371 2018-06-11 Martin Sebor <msebor@redhat.com>
4373 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4374 (Wno-class-memaccess): Add @opindex.
4375 (Wno-templates, Wno-multiple-inheritance): Same.
4376 (Wno-virtual-inheritance, Wno-namespaces): Same.
4377 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4378 (Wno-format-overflow, Wno-format-truncation): Same.
4379 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4380 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4381 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4382 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4383 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4384 (Wno-misspelled-isr): Same.
4386 2018-06-11 Martin Sebor <msebor@redhat.com>
4388 * PR tree-optimization/86083
4389 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4391 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4393 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4395 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4398 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4399 on the correct operand.
4400 (*movdi_internal64): Ditto.
4402 2018-06-11 Martin Liska <mliska@suse.cz>
4404 PR tree-optimization/86089
4405 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4407 2018-06-11 Julia Koval <julia.koval@intel.com>
4409 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4410 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4411 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4413 2018-06-11 Olivier Hainque <hainque@adacore.com>
4415 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4416 for Ada with strict dwarf2.
4418 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4421 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4424 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4426 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4428 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4430 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4433 2018-06-08 Martin Liska <mliska@suse.cz>
4435 * tree-cfg.h (debug_function): Fix argument type to match
4438 2018-06-08 Martin Liska <mliska@suse.cz>
4440 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4441 Remove usage of MPX-related (and removed) fields.
4442 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4444 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4446 * cfg.c (debug): Use TDF_NONE rather than 0.
4447 * cfghooks.c (debug): Likewise.
4448 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4449 (struct dump_option_value_info): Convert to...
4450 (struct kv_pair): ...this template type.
4451 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4453 (optinfo_verbosity_options): Likewise.
4454 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4456 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4457 than int for "optgroup_flags" param.
4458 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4460 (dump_dec): Likewise.
4461 (dump_finish): Use TDF_NONE rather than 0.
4462 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4463 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4464 than 0. Update for change to option_ptr.
4465 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4466 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4467 0. Update for changes to optinfo_verbosity_options and
4469 (opt_info_switch_p): Convert optgroup_flags from int to
4471 (dump_basic_block): Use dump_flags_t rather than int
4472 for "dump_kind" param.
4473 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4474 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4475 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4476 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4477 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4478 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4479 TDF_NONE): Convert from macros to...
4480 (enum dump_flag): ...this new enum.
4481 (dump_flags_t): Update to use enum.
4482 (operator|, operator&, operator~, operator|=, operator&=):
4483 Implement for dump_flags_t.
4484 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4485 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4486 Convert from macros to...
4487 (enum optgroup_flag): ...this new enum.
4488 (optgroup_flags_t): New typedef.
4489 (operator|, operator|=): Implement for optgroup_flags_t.
4490 (struct dump_file_info): Convert field "alt_flags" to
4491 dump_flags_t. Convert field "optgroup_flags" to
4493 (dump_basic_block): Use dump_flags_t rather than int for param.
4494 (dump_generic_expr_loc): Likewise.
4495 (dump_dec): Likewise.
4496 (dump_register): Convert param "optgroup_flags" to
4498 (opt_info_enable_passes): Likewise.
4499 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4501 * gimple-pretty-print.c (debug): Likewise.
4502 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4503 (merged_store_group::apply_stores): Likewise.
4504 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4505 * gimple.c (verify_gimple_pp): Likewise.
4506 * graphite-poly.c (print_pbb_body): Likewise.
4507 * passes.c (pass_manager::register_one_dump_file): Convert
4508 local "optgroup_flags" to optgroup_flags_t.
4509 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4511 (debug_body): Likewise.
4512 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4513 to optgroup_flags_t.
4514 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4516 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4517 (convert_mult_to_fma): Likewise.
4518 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4519 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4520 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4521 "dump_kind" to dump_flags_t.
4523 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4525 * config/rs6000/rs6000.c (min, max): Delete.
4527 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4529 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4530 -mabi=spe and -mabi=no-spe.
4532 2018-06-08 Martin Liska <mliska@suse.cz>
4534 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4535 where we expect an existing summary.
4537 2018-06-08 Martin Liska <mliska@suse.cz>
4539 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4540 * ipa-inline.h (estimate_edge_growth): Likewise.
4542 2018-06-08 Martin Liska <mliska@suse.cz>
4544 * cgraph.c (function_version_hasher::hash): Use
4545 cgraph_node::get_uid ().
4546 (function_version_hasher::equal):
4547 * cgraph.h (cgraph_node::get_uid): New method.
4548 * ipa-inline.c (update_caller_keys): Use
4549 cgraph_node::get_uid ().
4550 (update_callee_keys): Likewise.
4551 * ipa-utils.c (searchc): Likewise.
4552 (ipa_reduced_postorder): Likewise.
4553 * lto-cgraph.c (input_node): Likewise.
4554 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4555 * symbol-summary.h (symtab_insertion): Likewise.
4556 (symtab_removal): Likewise.
4557 (symtab_duplication): Likewise.
4558 * tree-pretty-print.c (dump_function_header): Likewise.
4559 * tree-sra.c (convert_callers_for_node): Likewise.
4561 2018-06-08 Martin Liska <mliska@suse.cz>
4563 * cgraph.c (symbol_table::create_edge): Always assign a new
4565 (symbol_table::free_edge): Do not recycle numbers.
4566 * cgraph.h (cgraph_edge::get): New method.
4567 * symbol-summary.h (symtab_removal): Use it.
4568 (symtab_duplication): Likewise.
4569 (call_summary::hashable_uid): Remove.
4571 2018-06-08 Martin Liska <mliska@suse.cz>
4573 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4574 (initialize_growth_caches): Remove.
4575 (free_growth_caches): Likewise.
4576 (do_estimate_edge_time): Use edge_growth_cache.
4577 (do_estimate_edge_size): Likewise.
4578 (do_estimate_edge_hints): Likewise.
4579 * ipa-inline.c (reset_edge_caches): Likewise.
4580 (recursive_inlining): Likewise.
4581 (inline_small_functions): Likewise.
4582 * ipa-inline.h (initialize_growth_caches): Remove.
4583 (estimate_edge_size): Likewise.
4584 (estimate_edge_time): Likewise.
4585 (estimate_edge_hints): Likewise.
4586 (reset_edge_growth_cache): Likewise.
4587 * symbol-summary.h (call_summary::remove): New method.
4589 2018-06-08 Martin Liska <mliska@suse.cz>
4591 * ipa-cp.c (class edge_clone_summary): New summary.
4592 (grow_edge_clone_vectors): Remove.
4593 (ipcp_edge_duplication_hook): Remove.
4594 (class edge_clone_summary_t): New call_summary class.
4595 (ipcp_edge_removal_hook): Remove.
4596 (edge_clone_summary_t::duplicate): New function.
4597 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4598 (create_specialized_node): Likewise.
4599 (ipcp_driver): Initialize edge_clone_summaries and do not
4602 2018-06-08 Martin Liska <mliska@suse.cz>
4604 * symbol-summary.h (get): New function.
4605 (call_summary::m_initialize_when_cloning): New class member.
4607 2018-06-08 Martin Liska <mliska@suse.cz>
4609 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4610 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4611 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4612 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4613 (remove_cgraph_node_from_order): Use the removed_nodes set.
4614 (do_per_function_toporder): Likwise.
4615 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4616 instead of summary_uid.
4617 (symtab_removal): Likewise.
4618 (symtab_duplication): Likewise.
4620 2018-06-08 Martin Liska <mliska@suse.cz>
4622 * ipa-cp.c (ipcp_store_bits_results): Use
4623 ipcp_transformation_sum.
4624 (ipcp_store_vr_results): Likewise.
4625 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4627 (ipcp_transformation_initialize): ... this.
4628 (ipa_set_node_agg_value_chain):
4629 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4630 (write_ipcp_transformation_info): Likewise.
4631 (read_ipcp_transformation_info): Likewise.
4632 (ipcp_update_bits): Likewise.
4633 (ipcp_update_vr): Likewise.
4634 (ipcp_transform_function): Likewise.
4635 * ipa-prop.h: Rename ipcp_transformation_summary to
4636 ipcp_transformation.
4637 (class ipcp_transformation_t): New function summary.
4638 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4639 (ipa_get_agg_replacements_for_node): Likewise.
4641 2018-06-08 Martin Liska <mliska@suse.cz>
4643 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4645 (class funct_state_summary_t): New function_summary class.
4646 (has_function_state): Remove.
4647 (get_function_state): Likewise.
4648 (set_function_state): Likewise.
4649 (add_new_function): Likewise.
4650 (funct_state_summary_t::insert): New function.
4651 (duplicate_node_data): Remove.
4652 (remove_node_data): Remove.
4653 (funct_state_summary_t::duplicate): New function.
4654 (register_hooks): Create new funct_state_summaries.
4655 (pure_const_generate_summary): Use it.
4656 (pure_const_write_summary): Likewise.
4657 (pure_const_read_summary): Likewise.
4658 (propagate_pure_const): Likewise.
4659 (propagate_nothrow): Likewise.
4660 (dump_malloc_lattice): Likewise.
4661 (propagate_malloc): Likewise.
4662 (execute): Do not register hooks, just remove summary
4664 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4667 2018-06-08 Martin Liska <mliska@suse.cz>
4669 * ipa-reference.c (remove_node_data): Remove.
4670 (duplicate_node_data): Likewise.
4671 (class ipa_ref_var_info_summary_t): New class.
4672 (class ipa_ref_opt_summary_t): Likewise.
4673 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4674 (get_reference_optimization_summary): Use
4675 ipa_ref_opt_sum_summaries.
4676 (set_reference_vars_info): Remove.
4677 (set_reference_optimization_summary): Likewise.
4678 (ipa_init): Create summaries.
4679 (init_function_info): Use function summary.
4680 (ipa_ref_opt_summary_t::duplicate): New function.
4681 (ipa_ref_opt_summary_t::remove): New function.
4682 (get_read_write_all_from_node): Fix GNU coding style.
4683 (propagate): Use function summary.
4684 (write_node_summary_p): Fix GNU coding style.
4685 (stream_out_bitmap): Likewise.
4686 (ipa_reference_read_optimization_summary): Use function summary.
4687 (ipa_reference_c_finalize): Do not release hooks.
4689 2018-06-08 Martin Liska <mliska@suse.cz>
4691 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4692 (analyze_function_body): Extract multiple calls of get_create.
4693 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4694 * ipa-inline.c (recursive_inlining): Use ::get method.
4695 * ipa-inline.h (estimate_edge_growth): Likewise.
4697 2018-06-08 Martin Liska <mliska@suse.cz>
4699 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4701 (hsa_function_summary::hsa_function_summary): Use the new enum
4703 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4704 * hsa-gen.c (hsa_get_host_function): Likewise.
4705 (get_brig_function_name): Likewise.
4706 * ipa-hsa.c (process_hsa_functions): Likewise.
4707 (ipa_hsa_write_summary): Likewise.
4708 * symbol-summary.h (symtab_duplication): Use ::get function/
4709 (get): New function.
4711 2018-06-08 Martin Liska <mliska@suse.cz>
4713 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4715 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4716 (hsa_register_kernel): Likewise.
4717 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4718 * hsa-gen.c (hsa_get_host_function): Likewise.
4719 (get_brig_function_name): Likewise.
4720 (generate_hsa): Likewise.
4721 (pass_gen_hsail::execute): Likewise.
4722 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4723 (devirtualization_time_bonus): Likewise.
4724 (ipcp_propagate_stage): Likewise.
4725 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4726 (edge_set_predicate): Likewise.
4727 (evaluate_conditions_for_known_args): Likewise.
4728 (evaluate_properties_for_edge): Likewise.
4729 (ipa_fn_summary::reset): Likewise.
4730 (ipa_fn_summary_t::duplicate): Likewise.
4731 (dump_ipa_call_summary): Likewise.
4732 (ipa_dump_fn_summary): Likewise.
4733 (analyze_function_body): Likewise.
4734 (compute_fn_summary): Likewise.
4735 (estimate_edge_devirt_benefit): Likewise.
4736 (estimate_edge_size_and_time): Likewise.
4737 (estimate_calls_size_and_time): Likewise.
4738 (estimate_node_size_and_time): Likewise.
4739 (inline_update_callee_summaries): Likewise.
4740 (remap_edge_change_prob): Likewise.
4741 (remap_edge_summaries): Likewise.
4742 (ipa_merge_fn_summary_after_inlining): Likewise.
4743 (ipa_update_overall_fn_summary): Likewise.
4744 (read_ipa_call_summary): Likewise.
4745 (inline_read_section): Likewise.
4746 (write_ipa_call_summary): Likewise.
4747 (ipa_fn_summary_write): Likewise.
4748 (ipa_free_fn_summary): Likewise.
4749 * ipa-hsa.c (process_hsa_functions): Likewise.
4750 (ipa_hsa_write_summary): Likewise.
4751 (ipa_hsa_read_section): Likewise.
4752 * ipa-icf.c (sem_function::merge): Likewise.
4753 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4754 (do_estimate_edge_time): Likewise.
4755 (estimate_size_after_inlining): Likewise.
4756 (estimate_growth): Likewise.
4757 (growth_likely_positive): Likewise.
4758 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4759 (inline_call): Likewise.
4760 * ipa-inline.c (caller_growth_limits): Likewise.
4761 (can_inline_edge_p): Likewise.
4762 (can_inline_edge_by_limits_p): Likewise.
4763 (compute_uninlined_call_time): Likewise.
4764 (compute_inlined_call_time): Likewise.
4765 (want_inline_small_function_p): Likewise.
4766 (edge_badness): Likewise.
4767 (update_caller_keys): Likewise.
4768 (update_callee_keys): Likewise.
4769 (recursive_inlining): Likewise.
4770 (inline_small_functions): Likewise.
4771 (inline_to_all_callers_1): Likewise.
4772 (dump_overall_stats): Likewise.
4773 (early_inline_small_functions): Likewise.
4774 (early_inliner): Likewise.
4775 * ipa-inline.h (estimate_edge_growth): Likewise.
4776 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4777 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4778 * ipa-prop.h (IPA_NODE_REF): Likewise.
4779 (IPA_EDGE_REF): Likewise.
4780 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4781 (propagate_malloc): Likewise.
4782 * ipa-split.c (execute_split_functions): Likewise.
4783 * symbol-summary.h: Rename get to get_create.
4785 (get_create): Likewise.
4786 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4788 2018-06-08 Martin Liska <mliska@suse.cz>
4790 * symbol-summary.h (release): Move definition out of class
4792 (symtab_removal): Likewise.
4793 (symtab_duplication): Likewise.
4795 2018-06-08 Martin Liska <mliska@suse.cz>
4797 * symbol-summary.h (function_summary): Move constructor
4798 implementation out of class declaration.
4799 (release): Likewise.
4800 (symtab_insertion): Likewise.
4801 (symtab_removal): Likewise.
4802 (symtab_duplication): Likewise.
4805 2018-06-08 Martin Liska <mliska@suse.cz>
4807 * Makefile.in: Remove support for MPX (macros, related functions,
4808 fields in cgraph_node, ...).
4809 * builtin-types.def (BT_BND): Likewise.
4810 (BT_FN_BND_CONST_PTR): Likewise.
4811 (BT_FN_CONST_PTR_BND): Likewise.
4812 (BT_FN_VOID_PTR_BND): Likewise.
4813 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4814 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4815 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4816 (expand_builtin_mempcpy_with_bounds): Likewise.
4817 (expand_builtin_memset_with_bounds): Likewise.
4818 (expand_builtin_memset_args): Likewise.
4819 (std_expand_builtin_va_start): Likewise.
4820 (expand_builtin): Likewise.
4821 (expand_builtin_with_bounds): Likewise.
4822 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4823 (DEF_LIB_BUILTIN_CHKP): Likewise.
4824 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4825 (DEF_CHKP_BUILTIN): Likewise.
4826 (BUILT_IN_MEMCPY): Likewise.
4827 (BUILT_IN_MEMMOVE): Likewise.
4828 (BUILT_IN_MEMPCPY): Likewise.
4829 (BUILT_IN_MEMSET): Likewise.
4830 (BUILT_IN_STPCPY): Likewise.
4831 (BUILT_IN_STRCAT): Likewise.
4832 (BUILT_IN_STRCHR): Likewise.
4833 (BUILT_IN_STRCPY): Likewise.
4834 (BUILT_IN_STRLEN): Likewise.
4835 (BUILT_IN_MEMCPY_CHK): Likewise.
4836 (BUILT_IN_MEMMOVE_CHK): Likewise.
4837 (BUILT_IN_MEMPCPY_CHK): Likewise.
4838 (BUILT_IN_MEMSET_CHK): Likewise.
4839 (BUILT_IN_STPCPY_CHK): Likewise.
4840 (BUILT_IN_STRCAT_CHK): Likewise.
4841 (BUILT_IN_STRCPY_CHK): Likewise.
4842 * calls.c (store_bounds): Likewise.
4843 (emit_call_1): Likewise.
4844 (special_function_p): Likewise.
4845 (maybe_warn_nonstring_arg): Likewise.
4846 (initialize_argument_information): Likewise.
4847 (finalize_must_preallocate): Likewise.
4848 (compute_argument_addresses): Likewise.
4849 (expand_call): Likewise.
4850 * cfgexpand.c (expand_call_stmt): Likewise.
4851 (expand_return): Likewise.
4852 (expand_gimple_stmt_1): Likewise.
4853 (pass_expand::execute): Likewise.
4854 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4855 (cgraph_node::remove): Likewise.
4856 (cgraph_node::dump): Likewise.
4857 (cgraph_node::verify_node): Likewise.
4858 * cgraph.h (chkp_function_instrumented_p): Likewise.
4859 (symtab_node::get_alias_target): Likewise.
4860 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4861 (cgraph_local_p): Likewise.
4862 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4863 (cgraph_edge::rebuild_references): Likewise.
4864 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4865 (walk_polymorphic_call_targets): Likewise.
4866 (cgraph_node::expand_thunk): Likewise.
4867 (symbol_table::output_weakrefs): Likewise.
4868 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4869 (ix86_handle_option): Likewise.
4870 * config/i386/constraints.md: Likewise.
4871 * config/i386/i386-builtin-types.def (BND): Likewise.
4875 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4876 (BDESC_FIRST): Likewise.
4878 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4879 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4880 * config/i386/i386.c (enum reg_class): Likewise.
4881 (ix86_target_string): Likewise.
4882 (ix86_option_override_internal): Likewise.
4883 (ix86_conditional_register_usage): Likewise.
4884 (ix86_valid_target_attribute_inner_p): Likewise.
4885 (ix86_set_indirect_branch_type): Likewise.
4886 (ix86_set_current_function): Likewise.
4887 (ix86_function_arg_regno_p): Likewise.
4888 (init_cumulative_args): Likewise.
4889 (ix86_function_arg_advance): Likewise.
4890 (ix86_function_arg): Likewise.
4891 (ix86_pass_by_reference): Likewise.
4892 (ix86_function_value_regno_p): Likewise.
4893 (ix86_function_value_1): Likewise.
4894 (ix86_function_value_bounds): Likewise.
4895 (ix86_return_in_memory): Likewise.
4896 (ix86_setup_incoming_vararg_bounds): Likewise.
4897 (ix86_va_start): Likewise.
4898 (indirect_thunk_need_prefix): Likewise.
4899 (print_reg): Likewise.
4900 (ix86_print_operand): Likewise.
4901 (ix86_expand_call): Likewise.
4902 (ix86_output_function_return): Likewise.
4903 (reg_encoded_number): Likewise.
4904 (BDESC_VERIFYS): Likewise.
4905 (ix86_init_mpx_builtins): Likewise.
4906 (ix86_init_builtins): Likewise.
4907 (ix86_emit_cmove): Likewise.
4908 (ix86_emit_move_max): Likewise.
4909 (ix86_expand_builtin): Likewise.
4910 (ix86_builtin_mpx_function): Likewise.
4911 (ix86_get_arg_address_for_bt): Likewise.
4912 (ix86_load_bounds): Likewise.
4913 (ix86_store_bounds): Likewise.
4914 (ix86_load_returned_bounds): Likewise.
4915 (ix86_store_returned_bounds): Likewise.
4916 (ix86_class_likely_spilled_p): Likewise.
4917 (ix86_hard_regno_mode_ok): Likewise.
4918 (x86_order_regs_for_local_alloc): Likewise.
4919 (ix86_mitigate_rop): Likewise.
4920 (ix86_bnd_prefixed_insn_p): Likewise.
4921 (ix86_mpx_bound_mode): Likewise.
4922 (ix86_make_bounds_constant): Likewise.
4923 (ix86_initialize_bounds): Likewise.
4924 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4925 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4926 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4927 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4928 (TARGET_CHKP_BOUND_MODE): Likewise.
4929 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4930 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4931 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4932 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4933 * config/i386/i386.h (TARGET_MPX): Likewise.
4934 (TARGET_MPX_P): Likewise.
4935 (VALID_BND_REG_MODE): Likewise.
4936 (FIRST_BND_REG): Likewise.
4937 (LAST_BND_REG): Likewise.
4938 (enum reg_class): Likewise.
4939 (BND_REG_P): Likewise.
4940 (BND_REGNO_P): Likewise.
4941 (BNDmode): Likewise.
4942 (ADJUST_INSN_LENGTH): Likewise.
4943 * config/i386/i386.md: Likewise.
4944 * config/i386/i386.opt: Likewise.
4945 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4946 (defined): Likewise.
4947 (LINK_MPX): Likewise.
4948 (MPX_SPEC): Likewise.
4949 (LIBMPX_SPEC): Likewise.
4950 (LIBMPXWRAPPERS_SPEC): Likewise.
4951 (CHKP_SPEC): Likewise.
4952 * config/i386/predicates.md: Likewise.
4953 * dbxout.c (dbxout_type): Likewise.
4954 * doc/extend.texi: Likewise.
4955 * doc/invoke.texi: Likewise.
4956 * doc/md.texi: Likewise.
4957 * doc/tm.texi: Likewise.
4958 * doc/tm.texi.in: Likewise.
4959 * dwarf2out.c (is_base_type): Likewise.
4960 (gen_formal_types_die): Likewise.
4961 (gen_subprogram_die): Likewise.
4962 (gen_type_die_with_usage): Likewise.
4963 (gen_decl_die): Likewise.
4964 (dwarf2out_late_global_decl): Likewise.
4965 * expr.c (expand_assignment): Likewise.
4966 (emit_storent_insn): Likewise.
4967 (store_expr_with_bounds): Likewise.
4968 (store_expr): Likewise.
4969 (expand_expr_real_1): Likewise.
4970 * expr.h (store_expr_with_bounds): Likewise.
4971 * function.c (use_register_for_decl): Likewise.
4972 (struct bounds_parm_data): Likewise.
4973 (assign_parms_augmented_arg_list): Likewise.
4974 (assign_parm_find_entry_rtl): Likewise.
4975 (assign_parm_is_stack_parm): Likewise.
4976 (assign_parm_load_bounds): Likewise.
4977 (assign_bounds): Likewise.
4978 (assign_parms): Likewise.
4979 (expand_function_start): Likewise.
4980 * gcc.c (CHKP_SPEC): Likewise.
4981 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4982 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4983 (wrestrict_dom_walker::check_call): Likewise.
4984 * gimple.c (gimple_build_call_from_tree): Likewise.
4985 * gimple.h (enum gf_mask): Likewise.
4986 (gimple_call_with_bounds_p): Likewise.
4987 (gimple_call_set_with_bounds): Likewise.
4988 * gimplify.c (gimplify_init_constructor): Likewise.
4989 * ipa-cp.c (initialize_node_lattices): Likewise.
4990 (propagate_constants_across_call): Likewise.
4991 (find_more_scalar_values_for_callers_subset): Likewise.
4992 * ipa-hsa.c (process_hsa_functions): Likewise.
4993 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4994 * ipa-icf.c (sem_function::merge): Likewise.
4995 * ipa-inline.c (early_inliner): Likewise.
4996 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4997 (warn_function_cold): Likewise.
4998 (propagate_pure_const): Likewise.
4999 * ipa-ref.h (enum GTY): Likewise.
5000 * ipa-split.c (find_retbnd): Likewise.
5001 (consider_split): Likewise.
5002 (split_function): Likewise.
5003 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5004 * ipa.c (walk_polymorphic_call_targets): Likewise.
5005 (symbol_table::remove_unreachable_nodes): Likewise.
5006 (process_references): Likewise.
5007 (cgraph_build_static_cdtor_1): Likewise.
5008 * lto-cgraph.c (lto_output_node): Likewise.
5009 (output_refs): Likewise.
5010 (compute_ltrans_boundary): Likewise.
5011 (input_overwrite_node): Likewise.
5012 (input_node): Likewise.
5013 (input_cgraph_1): Likewise.
5014 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5015 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5016 (class pass_chkp_instrumentation_passes): Likewise.
5017 (make_pass_chkp_instrumentation_passes): Likewise.
5018 * passes.def: Likewise.
5019 * rtl.h (struct GTY): Likewise.
5020 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5021 * stor-layout.c (layout_type): Likewise.
5022 * symtab.c: Likewise.
5023 * target.def: Likewise.
5024 * targhooks.c (default_chkp_bound_type): Likewise.
5025 (default_chkp_bound_mode): Likewise.
5026 (default_builtin_chkp_function): Likewise.
5027 (default_chkp_function_value_bounds): Likewise.
5028 (default_chkp_make_bounds_constant): Likewise.
5029 (default_chkp_initialize_bounds): Likewise.
5030 * targhooks.h (default_chkp_bound_type): Likewise.
5031 (default_chkp_bound_mode): Likewise.
5032 (default_builtin_chkp_function): Likewise.
5033 (default_chkp_function_value_bounds): Likewise.
5034 (default_chkp_make_bounds_constant): Likewise.
5035 (default_chkp_initialize_bounds): Likewise.
5036 * toplev.c (compile_file): Likewise.
5037 (process_options): Likewise.
5038 * tree-core.h (DEF_BUILTIN): Likewise.
5039 (DEF_BUILTIN_CHKP): Likewise.
5040 * tree-inline.c (declare_return_variable): Likewise.
5041 (remap_gimple_stmt): Likewise.
5042 (copy_bb): Likewise.
5043 (initialize_inlined_parameters): Likewise.
5044 (expand_call_inline): Likewise.
5045 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5046 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5047 (make_pass_ipa_chkp_produce_thunks): Likewise.
5048 (make_pass_chkp): Likewise.
5049 (make_pass_chkp_opt): Likewise.
5050 (make_pass_chkp_instrumentation_passes): Likewise.
5051 * tree-pretty-print.c (dump_generic_node): Likewise.
5052 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5053 * tree-ssa-dce.c (propagate_necessity): Likewise.
5054 (eliminate_unnecessary_stmts): Likewise.
5055 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5056 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5057 * tree-ssa-sccvn.h: Likewise.
5058 * tree-ssa-strlen.c (get_string_length): Likewise.
5059 (valid_builtin_call): Likewise.
5060 (adjust_last_stmt): Likewise.
5061 (handle_builtin_strchr): Likewise.
5062 (handle_builtin_strcpy): Likewise.
5063 (handle_builtin_stxncpy): Likewise.
5064 (handle_builtin_memcpy): Likewise.
5065 (handle_builtin_strcat): Likewise.
5066 (strlen_check_and_optimize_stmt): Likewise.
5067 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5068 * tree-streamer-in.c: Likewise.
5069 * tree-streamer.c (record_common_node): Likewise.
5070 * tree.c (tree_code_size): Likewise.
5071 (wide_int_to_tree_1): Likewise.
5072 (type_contains_placeholder_1): Likewise.
5073 (build_common_tree_nodes): Likewise.
5074 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5075 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5076 (POINTER_BOUNDS_P): Likewise.
5077 (BOUNDED_TYPE_P): Likewise.
5078 (BOUNDED_P): Likewise.
5079 (CALL_WITH_BOUNDS_P): Likewise.
5080 (pointer_bounds_type_node): Likewise.
5081 * value-prof.c (gimple_ic): Likewise.
5082 * var-tracking.c (vt_add_function_parameters): Likewise.
5083 * varasm.c (make_decl_rtl): Likewise.
5084 (assemble_start_function): Likewise.
5085 (output_constant): Likewise.
5086 (maybe_assemble_visibility): Likewise.
5087 * varpool.c (ctor_for_folding): Likewise.
5088 * chkp-builtins.def: Remove.
5089 * ipa-chkp.c: Remove.
5090 * ipa-chkp.h: Remove.
5091 * rtl-chkp.c: Remove.
5092 * rtl-chkp.h: Remove.
5093 * tree-chkp-opt.c: Remove.
5094 * tree-chkp.c: Remove.
5095 * tree-chkp.h: Remove.
5097 2018-06-07 Carl Love <cel@us.ibm.com>
5099 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5100 vextract_fp_from_shortl): Add BE support.
5102 2018-06-07 Paul Koning <ni1d@arrl.net>
5104 * compare-elim.c (try_merge_compare): Don't merge compare if
5105 address contains a side effect.
5106 (try_eliminate_compare): Likewise.
5108 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5110 * config.gcc: Support "tremont".
5111 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5112 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5114 * config/i386/i386.c (m_TREMONT): Define.
5115 (processor_target_table): Add "tremont".
5116 (PTA_TREMONT): Define.
5117 (ix86_lea_outperforms): Add TARGET_TREMONT.
5118 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5119 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5120 and M_INTEL_GOLDMONT_PLUS.
5121 (fold_builtin_cpu): Add "tremont".
5122 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5123 (ix86_option_override_internal): Add "tremont".
5124 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5125 (processor_type): Add PROCESSOR_TREMONT.
5126 * config/i386/x86-tune.def: Add m_TREMONT.
5127 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5129 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5131 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5132 symbol defined for msp430i* devices to be lower case.
5134 2018-06-07 Richard Biener <rguenther@suse.de>
5136 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5137 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5138 Properly wrap signed arithmetic if overflow wraps.
5140 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5142 PR tree-optimization/69615
5143 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5144 of a cast from a same precision integral SSA_NAME in a bb dominated
5145 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5146 cast to utype if rhs2 has already a compatible type.
5148 2018-06-07 Richard Biener <rguenther@suse.de>
5150 PR tree-optimization/85935
5151 * graphite-scop-detection.c (find_params_in_bb): Analyze
5152 condition operands with respect to the correct loop. Assert
5153 the analysis doesn't fail.
5155 2018-06-04 Carl Love <cel@us.ibm.com>
5157 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5158 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5159 as it is slightly cheaper.
5160 (first_match_or_eos_index_<mode>):
5161 Calculate index using natural element order.
5162 (first_match_index_<mode>):
5163 Calculate index using natural element order.
5164 (first_match_or_eos_index_<mode>):
5165 Calculate index using natural order.
5166 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5167 for BE and LE modes.
5168 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5169 P9V_BUILTIN_VCLZLSBB_V16QI.
5170 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5173 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5175 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5176 indentation and line wrap for many prototypes. Add missing
5177 @smallexample directives around block of prototypes for vec_xl and
5180 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5182 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5183 track if we pass or return IEEE 128-bit floating point.
5184 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5185 C++ mangling that is compatible with GCC 8.1.
5186 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5187 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5188 floating point types.
5189 (rs6000_function_arg_advance_1): Likewise.
5190 (rs6000_mangle_type): Optionally generate mangled names that match
5191 what GCC 8.1 generated for IEEE 128-bit floating point types.
5192 (rs6000_globalize_decl_name): If we have an external function that
5193 passes or returns IEEE 128-bit types, generate a weak reference
5194 from the mangled name used in GCC 8.1 to the current mangled
5196 (rs6000_init_builtins): Make __ibm128 use the long double type if
5197 long double is IBM extended double. Make __float128 use the long
5198 double type if long double is IEEE 128-bit.
5201 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5202 macro for __ibm128 built-in functions.
5203 (PACK_IF): Add __ibm128 pack/unpack functions.
5204 (UNPACK_IF): Likewise.
5205 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5206 enable long double built-in functions if long double is IEEE
5207 128-bit floating point.
5208 (rs6000_invalid_builtin): Update long double built-in function
5210 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5211 functions, adjust the built-in function to use the long double
5212 built-in function if __ibm128 and long double are the same type.
5213 * doc/extend.texi (PowerPC builtins): Update documention for
5214 __builtin_{,un}pack_longdouble. Add documentation for
5215 __builtin_{,un}pack_ibm128.
5217 2018-06-06 Jim Wilson <jimw@sifive.com>
5219 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5220 (struct machine_function): New field interrupt_mode.
5221 (riscv_handle_type_attribute): New function. Add forward declaration.
5222 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5223 (riscv_expand_epilogue): Check interrupt_mode field.
5224 (riscv_set_current_function): Check interrupt attribute args and
5225 set interrupt_mode field.
5226 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5227 (riscv_sret, riscv_uret): New.
5228 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5229 new arguments to interrupt attribute.
5231 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5234 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5235 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5237 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5239 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5240 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5241 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5242 several redundant entries.
5244 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5246 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5247 type from "rtx" to "rtx_insn *".
5248 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5249 for local "call_insn", removing cast.
5250 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5252 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5254 PR tree-optimization/86066
5255 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5256 for BIT_INSERT_EXPR stores.
5258 2018-06-06 Richard Biener <rguenther@suse.de>
5260 PR tree-optimization/86062
5261 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5262 component refs ontop
5263 of to be offsetted base.
5265 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5267 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5268 to be static and remove check on interrupt attribute name.
5270 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5272 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5273 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5275 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5278 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5280 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5281 Remove second argument, change how error is called.
5282 (aarch64_layout_arg): Remove second argument from
5283 aarch64_err_no_fpadvsimd call.
5284 (aarch64_init_cumulative_args): Ditto.
5285 (aarch64_gimplify_va_arg_expr): Ditto.
5286 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5288 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5290 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5291 (*simple_return_indirect_internal<mode>): Rename from
5292 simple_return_indirect_internal. Use W mode iterator.
5293 (rstorssp): New expander.
5294 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5295 (clrssbsy): New expander.
5296 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5298 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5300 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5302 (cmse_check_pointed_object): Likewise.
5304 2018-06-05 Martin Liska <mliska@suse.cz>
5306 PR gcov-profile/47618
5307 * doc/invoke.texi: Document how -fprofile-dir format
5310 2018-06-05 Richard Biener <rguenther@suse.de>
5312 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5313 removal pretend DOM info isn't available so we do not update
5314 it and only remove edges, not dominated blocks. Actually free
5315 DOM info in case we removed something. Remove unreachable blocks.
5316 (mfb_keep_latches): Work with either DOM info or marked backedges.
5317 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5318 first. Mark backedges if DOM info isn't available.
5319 (Re-)compute DOM info after cleanup_control_flow_pre.
5321 2018-06-05 Richard Biener <rguenther@suse.de>
5323 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5324 (locus_discrim_hasher::hash): Adjust.
5325 (locus_discrim_hasher::equal): Likewise.
5326 (next_discriminator_for_locus): Work on line directly.
5327 (same_line_p): Pass in expanded locus1 as well.
5328 (assign_discriminators): Avoid redundant location expansions.
5330 2018-06-05 Richard Biener <rguenther@suse.de>
5332 PR tree-optimization/86046
5333 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5334 if required after clearing TREE_ADDRESSABLE.
5336 2018-06-05 Richard Biener <rguenther@suse.de>
5338 PR tree-optimization/86047
5339 * tree-ssa-loop.c (for_each_index): Glob handling of all
5340 decls and constants and really handle all of them.
5342 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5345 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5346 qualifier_void_pointer and qualifier_const_void_pointer.
5347 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5348 (arm_init_builtins): Handle the above.
5349 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5350 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5351 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5354 2018-06-05 Martin Liska <mliska@suse.cz>
5356 * auto-profile.c (read_autofdo_file): Do not use
5357 gcov_ctr_summary struct.
5358 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5359 * coverage.c (struct counts_entry): Likewise.
5360 (read_counts_file): Read just single summary entry.
5361 (get_coverage_counts): Use gcov_summary struct.
5362 * coverage.h (get_coverage_counts): Likewise.
5363 * gcov-dump.c (dump_working_sets): Likewise.
5364 (tag_summary): Dump just single summary.
5365 * gcov-io.c (gcov_write_summary): Write just histogram
5367 (gcov_read_summary): Read just single summary.
5368 (compute_working_sets): Use gcov_summary struct.
5369 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5370 of GCOV_COUNTERS_SUMMABLE.
5371 (GCOV_COUNTERS_SUMMABLE): Remove.
5372 (GCOV_FIRST_VALUE_COUNTER): Replace with
5373 GCOV_COUNTER_V_INTERVAL.
5374 (struct gcov_ctr_summary): Remove.
5375 (struct gcov_summary): Directly use fields of former
5377 (compute_working_sets): Use gcov_summary struct.
5378 * gcov.c (read_count_file): Do not use ctrs fields.
5379 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5381 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5383 * profile.c: Likewise.
5384 * profile.h: Likewise.
5386 2018-06-05 Martin Liska <mliska@suse.cz>
5388 PR gcov-profile/84846
5389 * gcov.c (output_lines): Print working directory only
5390 in intermediate format.
5392 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5394 * config/s390/s390-builtin-types.def: Add void function type.
5395 * config/s390/s390-builtins.def: Use the function type for the
5398 2018-06-04 Jim Wilson <jimw@sifive.com>
5400 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5402 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5403 and maybe_eh_return. Change regno to unsigned int. Use new args to
5404 handle EH_RETURN_DATA_REGNO registers properly.
5405 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5406 (riscv_expand_epilogue): Update comment. Change argument name and
5407 type. Update code to use new name and type. Pass new args to
5408 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5410 * config/riscv/riscv.md (NORMAL_RETURN): New.
5411 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5412 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5413 (eh_return): Call gen_eh_return_internal and emit barrier.
5414 (eh_return_internal): Call riscv_expand_epilogue.
5416 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5418 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5419 bit_insertion field and declare can_be_merged_into method.
5420 (merged_store_group::can_be_merged_into): New method.
5421 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5422 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5423 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5425 2018-06-04 Richard Biener <rguenther@suse.de>
5427 PR tree-optimization/85955
5428 * builtins.c (fold_builtin_sincos): Convert pointers to
5429 destination to appropriate type before dereferencing.
5431 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5433 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5435 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5437 * expr.c (expand_expr_real_1): Force the operand into memory if
5438 its TYPE_MODE is BLKmode and if there is no integer mode for
5439 the number of bits being extracted.
5441 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5445 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5446 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5448 2018-06-04 Richard Biener <rguenther@suse.de>
5450 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5451 (cleanup_tree_cfg_noloop): ... single caller. Do
5452 start_recording_case_labels later.
5454 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5456 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5457 to _IMMINTRIN_H_INCLUDED.
5458 * config/i386/pconfigintrin.h: Ditto.
5459 * config/i386/waitpkgintrin.h: Ditto.
5460 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5461 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5462 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5463 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5464 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5465 waitpkgintrin.h and cldemoteintrin.h.
5467 2018-06-04 Richard Biener <rguenther@suse.de>
5469 PR tree-optimization/86038
5470 * tracer.c (find_best_successor): Check probability for
5471 being initialized, bail out if not.
5473 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5476 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5477 of bits to ignore when comparing architectures.
5479 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5481 PR tree-optimization/69615
5482 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5483 maximum or minimum of the type, try to merge it also as if
5484 range1 is + [-, x - 1] or + [x + 1, -].
5487 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5489 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5491 PR tree-optimization/86034
5492 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5493 the unsigned bitfield type in a bit insertion sequence if it does not
5494 have a larger precision than the bitfield size.
5495 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5497 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5499 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5501 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5503 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5504 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5505 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5506 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5508 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5510 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5511 Disable -fdelete-null-pointer-checks for ELF toolchain.
5513 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5514 Kito Cheng <kito.cheng@gmail.com>
5516 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5517 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5518 (nds32*-*-*): Add glibc and uclibc conditions.
5519 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5520 (TARGET_EXCEPT_UNWIND_INFO): Define.
5521 * config/nds32/elf.h: New file.
5522 * config/nds32/linux.h: New file.
5523 * config/nds32/nds32-elf.opt: New file.
5524 * config/nds32/nds32-linux.opt: New file.
5525 * config/nds32/nds32-fp-as-gp.c
5526 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5527 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5529 (nds32_asm_file_end): Ditto.
5530 (nds32_print_operand): Ditto.
5531 (nds32_insert_attributes): Ditto.
5532 (nds32_init_libfuncs): New function.
5533 (TARGET_HAVE_TLS): Define.
5534 (TARGET_INIT_LIBFUNCS): Define.
5535 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5537 (TARGET_ELF): Apply different mcmodel setting.
5538 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5539 been migrated into elf.h and linux.h files.
5540 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5541 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5542 (mcmodel): The content has been migrated into nds32-elf.opt and
5543 nds32-linux.opt files.
5544 * config/nds32/t-elf: New file.
5545 * config/nds32/t-linux: New file.
5547 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5548 Shiva Chen <shiva0217@gmail.com>
5550 * config/nds32/constants.md (unspec_volatile_element): Add
5551 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5552 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5554 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5555 (make_pass_nds32_fp_as_gp): Declare.
5556 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5558 (nds32_asm_function_end_prologue): Remove unused asm output.
5559 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5560 (nds32_asm_file_start): Output necessary fp_as_gp information.
5561 (nds32_option_override): Adjust register usage.
5562 (nds32_expand_prologue): Consider fp_as_gp situation.
5563 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5564 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5567 (simple_return): Ditto.
5568 (omit_fp_begin): Output special directive for fp_as_gp.
5569 (omit_fp_end): Output special directive for fp_as_gp.
5570 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5571 mforbid-fp-as-gp): New options.
5573 2018-06-01 Mark Wielaard <mark@klomp.org>
5575 * dwarf2out.c (dwarf2out_finish): Remove generation of
5576 DW_AT_loclists_base.
5578 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5580 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5581 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5582 (struct merged_store_group): Add bit_insertion field.
5583 (dump_char_array): Use standard hexadecimal format.
5584 (merged_store_group::merged_store_group): Set bit_insertion to false.
5585 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5586 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5587 also print the mask in the dump file.
5588 (pass_store_merging::gate): Minor tweak.
5589 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5590 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5591 stores with INTEGER_CST stores.
5592 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5593 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5594 and use it throughout. Generate bit insertion sequences if need be.
5595 (pass_store_merging::process_store): Remove redundant condition.
5596 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5598 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5600 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5601 the 128-bit floating point types. Fix function comment.
5603 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5605 * config/aarch64/aarch64-simd.md
5606 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5608 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5611 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5613 PR tree-optimization/85989
5614 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5616 (backprop::intersect_uses): Check it when deciding whether this
5617 is a backedge reference.
5618 (backprop::process_block): Add each phi to m_visited_phis
5619 after visiting it, then clear it at the end.
5621 2018-06-01 Richard Biener <rguenther@suse.de>
5623 * tree-vectorizer.h (vect_dr_stmt): New function.
5624 (vect_get_load_cost): Adjust.
5625 (vect_get_store_cost): Likewise.
5626 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5627 Use vect_dr_stmt instead of DR_SMTT.
5628 (vect_record_base_alignments): Likewise.
5629 (vect_calculate_target_alignment): Likewise.
5630 (vect_compute_data_ref_alignment): Likewise and make static.
5631 (vect_update_misalignment_for_peel): Likewise.
5632 (vect_verify_datarefs_alignment): Likewise.
5633 (vector_alignment_reachable_p): Likewise.
5634 (vect_get_data_access_cost): Likewise. Pass down
5635 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5636 (vect_get_peeling_costs_all_drs): Likewise.
5637 (vect_peeling_hash_get_lowest_cost): Likewise.
5638 (vect_enhance_data_refs_alignment): Likewise.
5639 (vect_find_same_alignment_drs): Likewise.
5640 (vect_analyze_data_refs_alignment): Likewise.
5641 (vect_analyze_group_access_1): Likewise.
5642 (vect_analyze_group_access): Likewise.
5643 (vect_analyze_data_ref_access): Likewise.
5644 (vect_analyze_data_ref_accesses): Likewise.
5645 (vect_vfa_segment_size): Likewise.
5646 (vect_small_gap_p): Likewise.
5647 (vectorizable_with_step_bound_p): Likewise.
5648 (vect_prune_runtime_alias_test_list): Likewise.
5649 (vect_analyze_data_refs): Likewise.
5650 (vect_supportable_dr_alignment): Likewise.
5651 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5652 (vect_gen_prolog_loop_niters): Likewise.
5653 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5654 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5656 (vect_recog_mask_conversion_pattern): Likewise.
5657 (vect_try_gather_scatter_pattern): Likewise.
5658 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5659 to vect_get_store_cost.
5660 (vect_get_store_cost): Get stmt_info instead of DR.
5661 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5662 (vect_get_load_cost): Get stmt_info instead of DR.
5664 2018-06-01 Richard Biener <rguenther@suse.de>
5667 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5668 vars wrapped in MEM_REFs.
5670 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5672 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5673 Fix subreg tests so that we only return a choice between
5674 GENERAL_REGS and FP_REGS if the original classes included both.
5676 2018-06-01 Richard Biener <rguenther@suse.de>
5679 * tree-ssa-structalias.c (get_function_part_constraint):
5680 Handle NULL fi->decl.
5681 (find_func_aliases_for_call): Properly handle indirect
5682 fi from direct call.
5683 (find_func_clobbers): Likewise.
5684 (ipa_pta_execute): Likewise.
5685 (create_variable_info_for): For functions that are ifunc_resolver
5686 resolve to a varinfo that contains the result of the resolver call.
5687 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5690 2018-05-31 Michael Collison <michael.collison@arm.com>
5692 * config/aarch64/aarch64.md:
5693 (*fix_to_zero_extenddfdi2): New pattern.
5694 * gcc.target/aarch64/fix_extend1.c: New testcase.
5696 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5700 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5701 and BUILT_IN_STRNCMP_EQ.
5702 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5703 BUILT_IN_STRNCMP_EQ.
5704 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5705 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5706 (gimple_fold_builtin): Likewise.
5707 * tree-ssa-strlen.c (compute_string_length): New function.
5708 (determine_min_obsize): New function.
5709 (handle_builtin_string_cmp): New function to handle calls to
5710 string compare functions.
5711 (strlen_optimize_stmt): Add handling to builtin string compare
5713 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5714 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5715 * tree.c (build_common_builtin_nodes): Add new defines of
5716 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5718 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5721 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5722 functions with naked attribute.
5724 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5726 * config/i386/sse.md (avx_vec_concat<mode>):
5727 Substitute concat_tg_mode mode attribute with xtg_mode.
5728 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5729 (concat_tg_mode): Remove mode attribute.
5731 2018-05-31 Martin Sebor <msebor@redhat.com>
5734 * calls.c (alloc_max_size): Correct a logic error/typo.
5735 Treat excessive arguments as infinite. Warn for invalid arguments.
5736 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5738 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5741 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5742 and movx for Haswell.
5744 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5745 Cesar Philippidis <cesar@codesourcery.com>
5748 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5749 when emitting error on private/firstprivate reductions.
5750 * omp-low.c (lower_omp_target): Avoid reference-type processing
5751 on pointers for firstprivate clause.
5753 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5755 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5758 * config/aarch64/aarch64-simd.md
5759 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5760 (aarch64_ld1_x3_<mode>): Likewise
5761 (aarch64_st1x2<VALLDIF:mode>): Likewise
5762 (aarch64_st1_x2_<mode>): Likewise
5763 (aarch64_st1x3<VALLDIF:mode>): Likewise
5764 (aarch64_st1_x3_<mode>): Likewise
5765 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5766 (vld1_s8_x3): Likewise.
5767 (vld1_u16_x3): Likewise.
5768 (vld1_s16_x3): Likewise.
5769 (vld1_u32_x3): Likewise.
5770 (vld1_s32_x3): Likewise.
5771 (vld1_u64_x3): Likewise.
5772 (vld1_s64_x3): Likewise.
5773 (vld1_f16_x3): Likewise.
5774 (vld1_f32_x3): Likewise.
5775 (vld1_f64_x3): Likewise.
5776 (vld1_p8_x3): Likewise.
5777 (vld1_p16_x3): Likewise.
5778 (vld1_p64_x3): Likewise.
5779 (vld1q_u8_x3): Likewise.
5780 (vld1q_s8_x3): Likewise.
5781 (vld1q_u16_x3): Likewise.
5782 (vld1q_s16_x3): Likewise.
5783 (vld1q_u32_x3): Likewise.
5784 (vld1q_s32_x3): Likewise.
5785 (vld1q_u64_x3): Likewise.
5786 (vld1q_s64_x3): Likewise.
5787 (vld1q_f16_x3): Likewise.
5788 (vld1q_f32_x3): Likewise.
5789 (vld1q_f64_x3): Likewise.
5790 (vld1q_p8_x3): Likewise.
5791 (vld1q_p16_x3): Likewise.
5792 (vld1q_p64_x3): Likewise.
5793 (vst1_s64_x2): Likewise.
5794 (vst1_u64_x2): Likewise.
5795 (vst1_f64_x2): Likewise.
5796 (vst1_s8_x2): Likewise.
5797 (vst1_p8_x2): Likewise.
5798 (vst1_s16_x2): Likewise.
5799 (vst1_p16_x2): Likewise.
5800 (vst1_s32_x2): Likewise.
5801 (vst1_u8_x2): Likewise.
5802 (vst1_u16_x2): Likewise.
5803 (vst1_u32_x2): Likewise.
5804 (vst1_f16_x2): Likewise.
5805 (vst1_f32_x2): Likewise.
5806 (vst1_p64_x2): Likewise.
5807 (vst1q_s8_x2): Likewise.
5808 (vst1q_p8_x2): Likewise.
5809 (vst1q_s16_x2): Likewise.
5810 (vst1q_p16_x2): Likewise.
5811 (vst1q_s32_x2): Likewise.
5812 (vst1q_s64_x2): Likewise.
5813 (vst1q_u8_x2): Likewise.
5814 (vst1q_u16_x2): Likewise.
5815 (vst1q_u32_x2): Likewise.
5816 (vst1q_u64_x2): Likewise.
5817 (vst1q_f16_x2): Likewise.
5818 (vst1q_f32_x2): Likewise.
5819 (vst1q_f64_x2): Likewise.
5820 (vst1q_p64_x2): Likewise.
5821 (vst1_s64_x3): Likewise.
5822 (vst1_u64_x3): Likewise.
5823 (vst1_f64_x3): Likewise.
5824 (vst1_s8_x3): Likewise.
5825 (vst1_p8_x3): Likewise.
5826 (vst1_s16_x3): Likewise.
5827 (vst1_p16_x3): Likewise.
5828 (vst1_s32_x3): Likewise.
5829 (vst1_u8_x3): Likewise.
5830 (vst1_u16_x3): Likewise.
5831 (vst1_u32_x3): Likewise.
5832 (vst1_f16_x3): Likewise.
5833 (vst1_f32_x3): Likewise.
5834 (vst1_p64_x3): Likewise.
5835 (vst1q_s8_x3): Likewise.
5836 (vst1q_p8_x3): Likewise.
5837 (vst1q_s16_x3): Likewise.
5838 (vst1q_p16_x3): Likewise.
5839 (vst1q_s32_x3): Likewise.
5840 (vst1q_s64_x3): Likewise.
5841 (vst1q_u8_x3): Likewise.
5842 (vst1q_u16_x3): Likewise.
5843 (vst1q_u32_x3): Likewise.
5844 (vst1q_u64_x3): Likewise.
5845 (vst1q_f16_x3): Likewise.
5846 (vst1q_f32_x3): Likewise.
5847 (vst1q_f64_x3): Likewise.
5848 (vst1q_p64_x3): Likewise.
5850 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5852 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5853 user_label_prefix to name.
5855 * tree-core.h: Update comment about the format of NAME string
5856 passed to handler in attribute_spec.
5858 * config/msp430/msp430.md: Remove erroneous subreg expression from
5859 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5860 zero_extend{q,h}isi2.
5862 2018-05-30 Borislav Petkov <bp@suse.de>
5864 * doc/extend.texi: Document some architecture specific
5865 constraints and sort entries.
5867 2018-05-30 Martin Sebor <msebor@redhat.com>
5870 * builtins.c (expand_builtin_stpcpy_1): New function.
5871 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5872 only if the former succeeds.
5874 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5876 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5879 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5881 * doc/invoke.texi (-flinker-output): Document
5883 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5885 * passes.c (ipa_write_summaries): Only modify statements if body
5887 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5888 incrementally linking.
5889 (ipa_passes): Likewise.
5890 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5891 pass down resolution info.
5892 * common.opt (flag_incremental_link): Update info.
5893 * gcc.c (plugin specs): Turn flinker-output=* to
5894 -plugin-opt=-linker-output-known
5895 * toplev.c (compile_file): Also cut compilation when doing incremental
5897 * flag-types. (enum lto_partition_model): Add
5898 LTO_LINKER_OUTPUT_NOLTOREL.
5899 (invoke.texi): Add -flinker-output docs.
5900 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5901 link same way as WPA; do not stream in dead initializers.
5903 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5904 darf2out_register_external_decl): Support incremental link.
5906 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5908 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5910 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5912 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5913 it down to simple_object_copy_lto_debug_sections.
5914 (run_gcc): Determine incremental LTO link time and configure
5915 lto1 into non-wpa mode, disable renaming of debug sections.
5917 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5919 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5920 descriptions of various incorrectly documented functions.
5922 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5925 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5926 address check not strict.
5928 2018-05-30 Richard Biener <rguenther@suse.de>
5930 PR tree-optimization/85964
5931 * tracer.c (better_p): Drop initialized count check, we only
5932 call the function with initialized counts now.
5933 (find_best_successor): Do find a best edge if one
5934 has uninitialized count.
5935 (find_best_predecessor): Likewise. Do BB frequency check only
5936 if count is initialized.
5938 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5940 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5941 (aarch64_ldrstr_offset_compare): New.
5942 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5943 load/store orderings.
5944 (aarch64_gen_adjusted_ldpstp): Likewise.
5946 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5948 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5949 Check for subset of GENERAL_REGS and FP_REGS.
5950 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5953 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5955 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5956 and wi::to_poly_offset. Add the current offset and then check
5957 whether the sum fits, rather than using an unchecked addition of
5958 a checked term. Check for a shwi rather than a uhwi.
5959 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5960 (store_constructor): Use poly_int_tree_p.
5961 (expand_expr_real_1): Likewise.
5962 * function.c (assign_temp): Likewise.
5963 * fold-const.c (const_binop): Use poly_int_tree_p and
5965 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5967 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5968 to_poly_offset for MEM offsets.
5969 * ipa-icf.c (sem_variable::equals): Likewise.
5970 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5971 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5972 wi::to_poly_offset for BIT_FIELD_REF offsets.
5973 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5975 * var-tracking.c (emit_note_insn_var_location): Use
5976 tree_to_poly_uint64.
5978 2018-05-29 Jim Wilson <jimw@sifive.com>
5980 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5982 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5985 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5986 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5988 (sse4_1_round<mode>2): Use nonimmediate_operand
5989 for operand 1 predicate.
5991 2018-05-29 Martin Sebor <msebor@redhat.com>
5992 Richard Biener <rguenther@suse.de>
5995 * calls.c (get_size_range): Call determine_value_range instead
5996 of get_value_range..
5997 * tree-vrp.h (determine_value_range): Declared new function.
5998 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6000 2018-05-29 Richard Biener <rguenther@suse.de>
6002 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6003 sure to use non-pattern stmts for get_earlier_stmt arguments.
6004 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6005 called on pattern stmts.
6006 (get_later_stmt): Likewise.
6008 2018-05-29 Martin Liska <mliska@suse.cz>
6010 PR gcov-profile/85759
6011 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6014 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6016 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6018 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6022 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6023 VEC_PACK_FLOAT_EXPR): New tree codes.
6024 * tree-pretty-print.c (op_code_prio): Handle
6025 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6026 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6027 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6028 * tree-inline.c (estimate_operator_cost): Likewise.
6029 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6030 * fold-const.c (const_binop): Likewise.
6031 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6032 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6033 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6034 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6035 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6036 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6037 * expr.c (expand_expr_real_2): Likewise.
6038 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6039 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6040 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6042 * optabs.c (expand_widen_pattern_expr): For
6043 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6044 sign from result type rather than operand's type.
6045 (expand_binop_directly): For vec_packu_float_optab and
6046 vec_packs_float_optab allow result type to be different from operand's
6048 * optabs-tree.c (optab_for_tree_code): Handle
6049 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6050 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6051 * tree-vect-generic.c (expand_vector_operations_1): Handle
6052 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6053 VEC_PACK_FLOAT_EXPR.
6054 * tree-vect-stmts.c (supportable_widening_operation): Handle
6056 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6057 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6058 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6059 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6060 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6062 (vec_pack<floatprefix>_float_<mode>): New expander.
6063 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6065 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6066 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6067 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6068 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6069 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6071 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6072 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6073 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6074 VEC_PACK_FLOAT_EXPR): Document.
6076 2018-05-29 Richard Biener <rguenther@suse.de>
6078 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6080 (stmt_vec_info_vec): Make pointer.
6081 (init_stmt_vec_info_vec): Remove.
6082 (free_stmt_vec_info_vec): Likewise.
6083 (set_stmt_vec_info_vec): New function.
6084 (free_stmt_vec_infos): Likewise.
6085 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6086 (set_vinfo_for_stmt): Likewise.
6087 (get_earlier_stmt): Likewise.
6088 (get_later_stmt): Likewise.
6089 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6090 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6091 (vec_info::~vec_info): Free stmt_vec_infos.
6092 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6093 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6094 (pass_slp_vectorize::execute): Likewise.
6095 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6096 (free_stmt_vec_info_vec): Likewise.
6097 (set_stmt_vec_info_vec): New function.
6098 (free_stmt_vec_infos): Likewise.
6099 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6100 the global stmt_vec_info_vec.
6101 * tree-parloops.c (gather_scalar_reductions): Use
6102 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6105 2018-05-29 Richard Biener <rguenther@suse.de>
6107 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6109 2018-05-29 Martin Liska <mliska@suse.cz>
6110 David Malcolm <dmalcolm@redhat.com>
6112 * vec.c (test_reverse): New.
6113 (vec_c_tests): Add new test.
6114 * vec.h (vl_ptr>::reverse): New function.
6116 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6118 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6120 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6123 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6125 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6127 2018-05-28 Richard Biener <rguenther@suse.de>
6129 PR tree-optimization/85933
6130 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6131 look at stmts marked as vectorizable.
6133 2018-05-28 Richard Biener <rguenther@suse.de>
6135 PR tree-optimization/85934
6136 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6137 vector boolean check before scalar optimization.
6139 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6141 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6144 2018-05-28 Mark Wielaard <mark@klomp.org>
6146 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6147 if it is an expression containing a minus sign.
6149 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6151 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6153 2018-05-27 Paul Koning <ni1d@arrl.net>
6155 * config/pdp11/pdp11.md (truncsihi2): Remove.
6157 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6158 Chung-Ju Wu <jasonwucj@gmail.com>
6160 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6162 (unaligned_store_dw): Ditto.
6163 * config/nds32/nds32-memory-manipulation.c
6164 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6165 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6166 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6167 (emit_setmem_word_loop): Rename to ...
6168 (emit_setmem_doubleword_loop): ... this.
6169 (nds32_gen_dup_4_byte_to_word_value): New function.
6170 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6171 (nds32_expand_setmem_loop): Refine implementation.
6172 (nds32_expand_setmem_loop_v3m): Ditto.
6173 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6176 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6178 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6180 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6182 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6183 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6184 machine->attr_no_prologue_p.
6185 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6186 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6187 (nds32_expand_epilogue): Consider attr_naked_p.
6188 (nds32_expand_epilogue_v3pop): Likewise.
6189 (nds32_can_use_return_insn): Likewise.
6190 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6191 attr_no_prologue_p fields.
6192 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6194 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6197 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6199 * config/i386/sse.md
6200 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6202 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6204 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6206 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6208 (*<floatsuffix>floatv2div2sf2): Rename to ...
6209 (*float<floatunssuffix>v2div2sf2): ... this.
6210 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6211 (float<floatunssuffix>v2div2sf2_mask): ... this.
6212 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6213 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6214 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6216 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6218 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6220 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6222 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6224 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6226 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6227 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6228 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6229 gen_ufix_truncv8dfv8si2.
6230 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6231 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6232 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6233 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6234 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6235 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6236 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6237 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6239 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6243 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6245 2018-05-25 Jim Wilson <jimw@sifive.com>
6247 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6248 * config/riscv/riscv.c (struct machine_function): Add
6249 interrupt_handler_p and attribute_checked_p fields.
6250 (riscv_attribute_table): Add interrupt.
6251 (riscv_interrupt_type_p): New.
6252 (riscv_save_reg_p): Save extra regs for interrupt handler.
6253 (riscv_use_save_libcall): Return false for interrupt handler.
6254 (riscv_first_stack_step): Add forward declaration.
6255 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6256 for interrupt handler with large frame. Use it for saved reg list.
6257 (riscv_expand_prologue): Move flag_stack_usage_info support to
6258 eliminate duplication.
6259 (riscv_expand_epilogue): Generate mret for interrupt handler.
6260 (riscv_epilogue_uses): New.
6261 (riscv_can_use_return_insn): Return false for interrupt handler.
6262 (riscv_function_ok_for_sibcall): Likewise.
6263 (riscv_set_current_function): Add interrupt handler support.
6264 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6265 * config/riscv/riscv.md (UNSPECV_MRET): New.
6267 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6269 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6271 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6273 PR tree-optimization/85712
6274 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6275 this candidate has already been replaced in-situ by a copy.
6277 2018-05-25 Jason Merrill <jason@redhat.com>
6279 PR c++/80485 - inline function non-zero address.
6280 * symtab.c (nonzero_address): Check DECL_COMDAT.
6282 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6285 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6286 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6287 extension of SImode operation. Use const123_operand predicate.
6288 (*saddsi_1): Remove.
6289 (*saddl_se_1): Ditto.
6291 (*ssubl_se_1): Ditto.
6292 * config/alpha/predicates.md (const123_operand): New predicate.
6293 * config/alpha/constraints.md (P): Use IN_RANGE.
6295 2018-05-25 Richard Biener <rguenther@suse.de>
6297 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6299 (ref_maybe_used_by_stmt_p): Likewise.
6300 (stmt_may_clobber_ref_p): Likewise.
6301 (stmt_may_clobber_ref_p_1): Likewise.
6302 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6304 (ref_maybe_used_by_stmt_p): Likewise.
6305 (stmt_may_clobber_ref_p): Likewise.
6306 (stmt_may_clobber_ref_p_1): Likewise.
6307 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6308 the alias oracle to disambiguate DRs with stmts DR analysis
6310 (vect_analyze_data_refs): Do not give up on not analyzable
6311 DRs for BB vectorization. Remove code truncating the dataref
6314 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6317 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6318 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6319 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6321 2018-05-25 Richard Biener <rguenther@suse.de>
6323 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6324 function, combining stmt data ref gathering and fatal analysis
6326 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6327 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6328 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6329 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6330 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6332 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6334 PR tree-optimization/85720
6335 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6336 SCC if all partitions are builtins.
6337 (version_loop_by_alias_check): New parameter. Generate cancelable
6338 runtime alias check if all partitions are builtins.
6339 (distribute_loop): Update call to above function.
6341 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6343 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6344 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6345 (parm_default_def_partition_arg): Ditto.
6346 (set_parm_default_def_partition): Ditto.
6347 (get_parm_default_def_partitions): Ditto and make it static.
6348 (get_undefined_value_partitions): Ditto and make it static.
6349 (remove_ssa_form): Refactor call to init_var_map here.
6350 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6351 computation for loop region.
6352 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6353 (register_default_def): Delete.
6354 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6355 (parm_default_def_partition_arg): Ditto.
6356 (set_parm_default_def_partition): Ditto.
6357 (get_parm_default_def_partitions): Ditto and make it static.
6358 (get_undefined_value_partitions): Ditto and make it static.
6359 (coalesce_with_default, coalesce_with_default): Update comment.
6360 (create_coalesce_list_for_region): New func factored out from
6361 create_outofssa_var_map.
6362 (populate_coalesce_list_for_outofssa): New func factored out from
6363 create_outofssa_var_map and coalesce_ssa_name.
6364 (create_outofssa_var_map): Delete.
6365 (coalesce_ssa_name): Refactor to support live range computation.
6366 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6367 (get_parm_default_def_partitions): Delete.
6368 (get_undefined_value_partitions): Ditto.
6369 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6370 computation for loop region.
6371 (new_tree_live_info, loe_visit_block): Ditto.
6372 (live_worklist, set_var_live_on_entry): Ditto.
6373 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6374 * tree-ssa-live.h (struct _var_map): New fields.
6375 (init_var_map): Change decl.
6376 (region_contains_p): New.
6378 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6380 * tree-ssa-live.h (live_merge_and_clear): Delete.
6382 2018-05-25 Richard Biener <rguenther@suse.de>
6385 * tree-dump.c (dequeue_and_dump): Remove access to removed
6386 operand 2 of a SWITCH_EXPR.
6388 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6390 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6391 multiplication and division.
6392 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6393 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6394 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6395 (cond_udiv_optab, cond_umod_optab): New optabs.
6396 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6397 (IFN_COND_RDIV): New internal functions.
6398 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6399 TRUNC_MOD_EXPR and RDIV_EXPR.
6400 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6401 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6403 (SVE_INT_BINARY): Include mult.
6404 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6405 (optab, sve_int_op): Handle mult.
6406 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6408 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6409 for SVE_INT_BINARY_SD.
6411 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6413 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6414 (optab, sve_int_op): Handle div and udiv.
6415 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6416 for SVE_INT_BINARY_SD.
6417 (*<optab><mode>3): New insn for the same.
6419 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6421 * tree-vect-patterns.c: Include predict.h.
6422 (vect_recog_divmod_pattern): Restrict check for division support
6423 to when optimizing for size.
6425 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6427 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6428 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6429 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6430 (gimple_match_op::set_op): Likewise.
6431 (gimple_resimplify4): Declare.
6432 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6433 (expr::gen_transform): Likewise.
6434 (decision_tree::gen): Generate a simplification routine for 4 operands.
6435 * gimple-match-head.c (gimple_simplify): Add an overload for
6436 4 operands. In the top-level function, handle up to 4 call
6437 arguments and call gimple_resimplify4.
6438 (gimple_resimplify4): New function.
6439 (build_call_internal): Pass a fourth operand.
6440 (maybe_push_to_seq): Likewise.
6441 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6442 Fold VEC_COND_EXPRs of an operation and a default value into
6443 an IFN_COND_* function if possible.
6444 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6446 (SVE_COND_FP_BINARY): Include them.
6447 (optab, sve_fp_op): Handle them.
6448 (SVE_INT_BINARY_REV): New code iterator.
6449 (SVE_COND_FP_BINARY_REV): New int iterator.
6450 (commutative): New int attribute.
6451 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6453 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6455 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6456 (*cond_<optab><mode>): New patterns for reversed operands.
6458 2018-05-25 Richard Biener <rguenther@suse.de>
6460 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6461 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6462 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6463 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6464 * tree-vect-data-refs.c (everywhere): Adjust users.
6465 * tree-vect-loop.c (everywhere): Likewise.
6466 * tree-vect-slp.c (everywhere): Likewise.
6467 * tree-vect-stmts.c (everywhere): Likewise.
6468 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6470 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6472 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6474 (gcc_cv_as_section_exclude): ... this.
6475 Try Solaris as #exclude syntax.
6476 * configure: Regenerate.
6477 * config.in: Regenerate.
6478 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6480 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6481 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6483 * varasm.c (default_elf_asm_named_section): Don't check if
6484 HAVE_GAS_SECTION_EXCLUDE is defined.
6486 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6488 * doc/md.texi: Update the documentation of the cond_* optabs
6489 to mention the new final operand. Fix GET_MODE_NUNITS call.
6490 Describe the scalar case too.
6491 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6492 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6494 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6495 (get_conditional_internal_fn): Update comment.
6496 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6497 accumulator value as a final argument to conditional functions.
6498 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6499 a define_expand and add an "else" operand. Assert for now that
6500 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6501 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6502 (*cond_<optab><mode>): New patterns.
6503 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6504 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6505 (UNSPEC_COND_EOR): Delete.
6506 (optab): Remove associated mappings.
6507 (SVE_INT_BINARY): New code iterator.
6508 (sve_int_op): Remove int attribute and add "minus" to the code
6510 (SVE_COND_INT_OP): Delete.
6511 (SVE_COND_FP_OP): Rename to...
6512 (SVE_COND_FP_BINARY): ...this.
6514 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6516 * optabs.c (can_reuse_operands_p): New function.
6517 (maybe_legitimize_operands): Try to reuse the results for
6520 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6522 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6523 Add {q} suffix to insn mnemonic.
6525 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6527 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6528 (msp430_warn_func_return): New.
6530 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6532 * fold-const.c (tree_nonzero_bits): New function.
6533 * fold-const.h (tree_nonzero_bits): Likewise.
6534 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6535 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6537 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6541 * varasm.c (assemble_alias): Check ifunc_resolver only on
6544 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6547 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6548 when memory input operand is handled.
6550 2018-05-24 Luis Machado <luis.machado@linaro.org>
6552 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6554 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6556 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6558 * match.pd: Delay FMA folds until after vectorization.
6560 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6563 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6564 address check not strict.
6566 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6568 * gimple-match.h (gimple_match_op): New class.
6569 (mprts_hook): Replace parameters with a gimple_match_op *.
6570 (maybe_build_generic_op): Likewise.
6571 (gimple_simplified_result_is_gimple_val): Replace parameters with
6572 a const gimple_match_op *.
6573 (gimple_simplify): Replace code_helper * and tree * parameters with
6574 a gimple_match_op * parameter.
6575 (gimple_resimplify1): Replace code_helper *, tree and tree *
6576 parameters with a gimple_match_op * parameter.
6577 (gimple_resimplify2): Likewise.
6578 (gimple_resimplify3): Likewise.
6579 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6580 parameters with a gimple_match_op * parameter.
6581 * gimple-match-head.c (gimple_simplify): Change prototypes of
6582 auto-generated functions to take a gimple_match_op * instead of
6583 separate code_helper * and tree * parameters. Make the same
6584 change in the top-level overload and update calls to the
6585 gimple_resimplify routines. Update calls to the auto-generated
6586 functions and to maybe_push_res_to_seq in the publicly-facing
6587 operation-specific gimple_simplify overloads.
6588 (gimple_match_op::MAX_NUM_OPS): Define.
6589 (gimple_resimplify1): Replace rcode and ops with a single res_op
6590 parameter. Update call to gimple_simplify.
6591 (gimple_resimplify2): Likewise.
6592 (gimple_resimplify3): Likewise.
6593 (mprts_hook): Replace parameters with a gimple_match_op *.
6594 (maybe_build_generic_op): Likewise.
6595 (build_call_internal): Replace type, nargs and ops with
6596 a gimple_match_op *.
6597 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6598 with a single gimple_match_op *. Update calls to mprts_hook,
6599 build_call_internal and gimple_simplified_result_is_gimple_val.
6600 Factor out code that is common to the tree_code and combined_fn cases.
6601 * genmatch.c (expr::gen_transform): Replace tem_code and
6602 tem_ops with a gimple_match_op called tem_op. Update calls
6603 to the gimple_resimplify functions and maybe_push_res_to_seq.
6604 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6605 res_ops. Update call to the gimple_resimplify functions.
6606 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6607 (decision_tree::gen): Make the functions take a gimple_match_op *
6608 called res_op instead of separate res_code and res_ops parameters.
6609 Update call accordingly.
6610 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6611 and ops with a single res_op parameter. Update calls to
6612 maybe_build_generic_op and maybe_push_res_to_seq.
6613 (fold_stmt_1): Update calls to gimple_simplify and
6614 replace_stmt_with_simplification.
6615 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6616 and gimple_simplified_result_is_gimple_val.
6617 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6619 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6620 with a gimple_match_op *.
6621 (vn_nary_build_or_lookup): Likewise. Update call to
6622 vn_nary_build_or_lookup_1.
6623 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6624 gimple_match_op *. Update calls to the gimple_resimplify routines
6625 and to gimple_simplified_result_is_gimple_val.
6626 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6627 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6628 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6629 (visit_nary_op): Likewise.
6630 (visit_reference_op_load): Likewise.
6632 2018-05-23 Luis Machado <luis.machado@linaro.org>
6634 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6635 modifier for printing the step amount.
6637 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6640 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6643 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6645 * doc/sourcebuild.texi (Endianness): New subsubsection.
6647 2018-05-23 Luis Machado <luis.machado@linaro.org>
6649 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6650 <prefetch_dynamic_strides>: New const bool field.
6651 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6652 prefetch_dynamic_strides.
6653 (exynosm1_prefetch_tune): Likewise.
6654 (thunderxt88_prefetch_tune): Likewise.
6655 (thunderx_prefetch_tune): Likewise.
6656 (thunderx2t99_prefetch_tune): Likewise.
6657 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6659 (aarch64_override_options_internal): Update to set
6660 PARAM_PREFETCH_DYNAMIC_STRIDES.
6661 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6662 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6663 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6664 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6665 prefetch-dynamic-strides setting.
6667 2018-05-23 Luis Machado <luis.machado@linaro.org>
6669 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6670 <minimum_stride>: New const int field.
6671 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6672 minimum_stride field defaulting to -1.
6673 (exynosm1_prefetch_tune): Likewise.
6674 (thunderxt88_prefetch_tune): Likewise.
6675 (thunderx_prefetch_tune): Likewise.
6676 (thunderx2t99_prefetch_tune): Likewise.
6677 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6678 <default_opt_level>: Set to 3.
6679 (aarch64_override_options_internal): Update to set
6680 PARAM_PREFETCH_MINIMUM_STRIDE.
6681 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6682 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6683 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6684 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6685 stride is constant and is below the minimum stride threshold.
6687 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6689 * config/arm/arm-cpus.in (mode26): Delete.
6690 (armv4): Delete mode26 reference.
6691 * config/arm/arm.c (arm_configure_build_target): Delete use of
6694 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6696 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6698 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6699 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6701 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6702 (floatunsdidf2): Ditto.
6704 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6706 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6707 (fixuns_trunc<mode>si2_avx512f): Ditto.
6708 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6709 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6710 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6712 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6714 PR rtl-optimization/79985
6715 * df-scan.c (df_insn_refs_collect): Remove special case for
6716 global registers and asm statements.
6718 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6720 * extend.texi (Global Register Variables): Rewrite the bullet list.
6721 Note that the register is available for allocation. Note that access
6722 via inline asm must use constraints. Add note about async-signal
6723 handlers. Remove paragraph about automagic register selection.
6725 2018-05-23 Richard Biener <rguenther@suse.de>
6727 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6728 of fixed offset from memset VN.
6730 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6732 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6734 (alloc_cand_and_find_basis): Initialize first_interp field.
6735 (slsr_process_mul): Modify first_interp field.
6736 (slsr_process_add): Likewise.
6737 (slsr_process_cast): Modify first_interp field for each new
6739 (slsr_process_copy): Likewise.
6740 (dump_candidate): Dump first_interp field.
6741 (replace_mult_candidate): Process all interpretations, not just
6743 (replace_rhs_if_not_dup): Likewise.
6744 (replace_one_candidate): Likewise.
6746 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6748 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6750 (aarch64_needs_frame_chain): New function.
6751 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6753 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6756 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6757 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6758 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6759 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6760 as true for strict-align.
6761 (aarch64_can_inline_p): Perform checks even when callee has no
6762 attributes to check for strict alignment.
6763 * doc/extend.texi (AArch64 Function Attributes): Document
6765 * doc/invoke.texi: (AArch64 Options): Likewise.
6767 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6769 PR tree-optimization/85853
6770 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6771 the handling of the root of the node to...
6772 (vect_slp_analyze_node_operations_1): ...this new function,
6773 and run the whole thing with the child nodes' def types
6774 set according to their SLP node's def type.
6776 2018-05-23 Richard Biener <rguenther@suse.de>
6779 * tree-data-ref.c (create_runtime_alias_checks): Defer
6780 and ignore overflow warnings.
6782 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6784 PR tree-optimization/85822
6785 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6788 2018-05-23 Richard Biener <rguenther@suse.de>
6790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6791 memset constants via native_interpret_expr.
6793 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6796 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6798 (cgraph_node::create_alias): Likewise.
6799 (cgraph_node::get_availability): Check ifunc_resolver instead
6800 of looking up ifunc attribute.
6801 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6802 * varasm.c (do_assemble_alias): Likewise.
6803 (assemble_alias): Likewise.
6804 (default_binds_local_p_3): Likewise.
6805 * cgraph.h (cgraph_node): Add ifunc_resolver.
6806 (cgraph_node::only_called_directly_or_aliased_p): Return false
6808 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6810 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6811 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6812 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6813 instead of looking up ifunc attribute.
6815 2018-05-22 Luis Machado <luis.machado@linaro.org>
6817 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6819 2018-05-22 Martin Sebor <msebor@redhat.com>
6822 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6823 only when expasion succeeds.
6824 (expand_builtin_strcmp): Same.
6825 (expand_builtin_strncmp): Same.
6827 2018-05-22 Martin Sebor <msebor@redhat.com>
6829 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6831 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6832 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6834 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6835 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6837 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6838 Add check for legitimate_address.
6839 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6840 (aarch64_swap_ldrstr_operands): New.
6841 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6844 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6845 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6847 * config/aarch64/aarch64.md: New patterns to generate stp
6849 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6850 single words and double words.
6851 (load_pair_sw, load_pair_dw): Likewise.
6852 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6854 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6856 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6857 for different mode ldpstp and add peephole for merged zero stores.
6859 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6861 (aarch64_gen_store_pair): Rename calls to match new patterns.
6862 (aarch64_gen_load_pair): Rename calls to match new patterns.
6863 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6864 (load_pair<DREG:mode><DREG2:mode>): ... This.
6865 (store_pair<mode>): Rename to...
6866 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6867 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6869 (V_INT_EQUIV): Handle SImode.
6870 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6873 2018-05-22 Martin Sebor <msebor@redhat.com>
6876 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6877 or ajust the presumed bound on an operation to avoid unnecessary
6880 2018-05-22 Martin Sebor <msebor@redhat.com>
6882 PR tree-optimization/85826
6883 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6884 assuming that a DECL necesarily has a constant size.
6886 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6889 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6891 2018-05-22 Richard Biener <rguenther@suse.de>
6893 PR tree-optimization/85834
6894 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6895 non-constant and non-zero memset arguments.
6897 2018-05-22 Martin Liska <mliska@suse.cz>
6900 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6902 2018-05-22 Richard Biener <rguenther@suse.de>
6904 PR tree-optimization/85863
6905 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6906 comparisons when vectype is specified.
6907 (vectorizable_condition): Do not specify vectype for
6908 vect_is_simple_cond when SLP vectorizing.
6910 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6913 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6914 define __ibm128 as long double.
6915 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6916 as a distinct type when IEEE 128-bit support is enabled.
6917 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6918 128-bit types to use the correct functions.
6919 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6920 convert between 128-bit floating point types that have different
6921 modes but the same representation, instead of using gen_lowpart to
6923 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6925 (IFKF_reg): New attributes to give the register constraints for
6927 (extend<mode>tf2_internal): New insns to mark an explicit
6928 conversion between 128-bit floating point types that have a
6929 different mode but share the same representation.
6931 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6933 PR tree-optimization/85814
6934 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6935 a null return from get_strinfo when unsharing the next
6936 strinfo in the chain.
6938 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6941 * varasm.c (weak_finish): Clean up weak_decls.
6943 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6945 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6946 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6947 UNSPEC_UADALP values.
6948 * config/aarch64/iterators.md (ABAL): New int iterator.
6951 (sur): Add mappings for the above.
6952 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6954 (aarch64_<sur>abal<mode>_4): Likewise.
6955 (aarch64_<sur>adalp<mode>_3): Likewise.
6956 (<sur>sadv16qi): New define_expand.
6958 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6960 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6961 (*movdf_internal): Ditto.
6962 (*rcpsf2_sse): Ditto.
6963 (*rsqrtsf2_sse): Ditto.
6964 (*sqrt<mode>2_sse): Ditto.
6966 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6968 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6970 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6971 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6972 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6973 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6975 * config/aarch64/arm_neon.h: Likewise.
6976 * config/aarch64/iterators.md (VQ_I): New.
6978 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6980 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6982 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6985 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6987 * config/nds32/constraints.md (S): New constraint.
6988 * config/nds32/nds32.md (call_internal): Use constraint S.
6989 (call_value_internal): Likewise.
6990 (sibcall_internal): Likewise.
6991 (sibcall_value_internal): Likewise.
6993 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6994 Chung-Ju Wu <jasonwucj@gmail.com>
6996 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6999 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7000 Chung-Ju Wu <jasonwucj@gmail.com>
7002 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7003 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7004 (nds32_rtx_costs_impl): Simplify.
7005 (nds32_address_cost_impl): Simplify.
7006 (nds32_init_rtx_costs): New function.
7007 (nds32_rtx_costs_speed_prefer): Likewise.
7008 (nds32_rtx_costs_size_prefer): Likewise.
7009 (nds32_address_cost_speed_prefer): Likewise.
7010 (nds32_address_cost_speed_fwprop): Likewise.
7011 (nds32_address_cost_size_prefer): Likewise.
7012 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7013 * config/nds32/nds32.c (nds32_option_override): Use
7014 nds32_init_rtx_costs function.
7016 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7018 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7019 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7020 (TARGET_PIPELINE_N8): Likewise.
7021 (TARGET_PIPELINE_N10): Likewise.
7022 (TARGET_PIPELINE_N13): Likewise.
7023 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7025 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7027 * config/nds32/nds32-fpu.md: Update copyright year.
7029 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7031 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7033 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7035 * config/nds32/nds32.c
7036 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7037 * config/nds32/nds32.opt (minline-asm-r15): New option.
7039 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7041 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7043 * config/nds32/nds32.md (abssi2): New pattern.
7045 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7047 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7048 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7049 Merge insn pattern from sse_cvtsi2ss<round_name> and
7050 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7051 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7052 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7053 using SWI48 mode iterator.
7054 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7055 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7056 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7057 pattern from sse_cvttss2si<round_saeonly_name>
7058 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7059 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7060 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7061 using SWI48 mode iterator.
7062 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7063 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7064 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7065 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7066 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7067 using SWI48 mode iterator.
7068 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7069 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7070 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7071 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7072 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7073 SWI48 mode iterator.
7074 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7075 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7076 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7077 pattern from sse_cvttsd2si<round_saeonly_name>
7078 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7080 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7082 * config/nds32/nds32-md-auxiliary.c
7083 (nds32_valid_smw_lwm_base_p): Refine.
7084 (nds32_output_smw_single_word): Refine.
7085 (nds32_output_smw_double_word): New.
7086 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7088 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7090 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7091 (nds32_output_stack_pop): Refine.
7092 (nds32_expand_unaligned_load): Refine.
7093 (nds32_expand_unaligned_store): Refine.
7095 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7096 Chung-Ju Wu <jasonwucj@gmail.com>
7098 * config/nds32/constants.md: Add TP_REGNUM constant.
7099 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7100 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7102 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7103 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7104 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7105 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7106 and PIC code generation.
7107 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7109 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7111 * config/nds32/nds32.md: Support TLS and PIC.
7112 * config/nds32/nds32.c: Support TLS and PIC.
7113 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7114 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7117 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7119 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7120 mode with E_ prefix.
7122 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7123 Chung-Ju Wu <jasonwucj@gmail.com>
7125 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7126 * config/nds32/nds32-md-auxiliary.c
7127 (symbolic_reference_mentioned_p): New.
7128 (nds32_legitimize_ict_address): New.
7129 (nds32_expand_ict_move): New.
7130 (nds32_indirect_call_referenced_p): New.
7131 (nds32_symbol_binds_local_p): Delete.
7132 (nds32_long_call_p): Modify.
7133 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7134 * config/nds32/nds32-protos.h
7135 (symbolic_reference_mentioned_p): Declare.
7136 (nds32_legitimize_ict_address): Declare.
7137 (nds32_expand_ict_move): Declare.
7138 (nds32_indirect_call_referenced_p): Declare.
7139 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7140 (nds32_relax_group): Use nds32_ict_const_p as condition.
7141 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7142 (nds32_asm_file_start): Output ict_model directive in asm code.
7143 (nds32_legitimate_address_p): Consider indirect call.
7144 (nds32_print_operand): Consider indirect call.
7145 (nds32_print_operand_address): Consider indirect call.
7146 (nds32_insert_attributes): Handle "indirect_call" attribute.
7147 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7148 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7149 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7150 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7151 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7152 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7153 (TARGET_ICT_MODEL_SMALL): Define.
7154 (TARGET_ICT_MODEL_LARGE): Define.
7155 * config/nds32/nds32.md (movsi): Consider ict model.
7156 (call, call_value): Consider ict model.
7157 (sibcall, sibcall_value): Consider ict model.
7158 * config/nds32/nds32.opt (mict-model): New option.
7159 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7162 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7163 Monk Chiang <sh.chiang04@gmail.com>
7164 Jim Wilson <jimw@sifive.com>
7166 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7167 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7168 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7169 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7170 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7171 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7172 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7173 compute save_libcall_adjustment properly.
7174 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7175 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7176 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7177 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7178 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7179 (ABI_SPEC): Handle mabi=ilp32e.
7180 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7181 (RVE): Add RVE mask.
7182 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7183 <-march>: Add rv32e as an example.
7185 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7188 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7189 (intra_create_variable_infos): Handle C++ constructors.
7191 2018-05-18 Martin Liska <mliska@suse.cz>
7193 * passes.def: Remove a redundant pass.
7195 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7198 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7200 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7202 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7204 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7206 (armv2, armv2a, armv3, armv3m): Delete architectures.
7207 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7208 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7209 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7211 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7212 (*mulsidi3adddi): Likewise.
7213 (mulsidi3): Likewise.
7214 (*mulsidi3_nov6): Likewise.
7215 (umulsidi3): Likewise.
7216 (umulsidi3_nov6): Likewise.
7217 (umaddsidi4): Likewise.
7218 (*umulsidi3adddi): Likewise.
7219 (smulsi3_highpart): Likewise.
7220 (*smulsi3_highpart_nov6): Likewise.
7221 (umulsi3_highpart): Likewise.
7222 (*umulsi3_highpart_nov6): Likewise.
7223 * config/arm/arm.h (arm_arch3m): Delete.
7224 * config/arm/arm.c (arm_arch3m): Delete.
7225 (arm_option_override_internal): Update armv3-related comment.
7226 (arm_configure_build_target): Delete use of isa_bit_mode32.
7227 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7228 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7229 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7231 (mulusa3): Likewise.
7232 * config/arm/arm-protos.h (arm_arch3m): Delete.
7233 * config/arm/arm-tables.opt: Regenerate.
7234 * config/arm/arm-tune.md: Likewise.
7235 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7236 deleted architectures.
7238 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7240 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7241 (armv5t, armv5te): New features.
7242 (ARMv5, ARMv5e): Delete fgroups.
7243 (ARMv5t, ARMv5te): Adjust for above changes.
7245 (armv5, armv5e): Delete arches.
7246 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7248 (*call_reg_arm): Likewise.
7249 (*call_value_reg_armv5): Likewise.
7250 (*call_value_reg_arm): Likewise.
7251 (*call_symbol): Likewise.
7252 (*call_value_symbol): Likewise.
7253 (*sibcall_insn): Likewise.
7254 (*sibcall_value_insn): Likewise.
7256 (prefetch): Likewise.
7257 (define_split and define_peephole2 dependent on arm_arch5):
7259 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7261 (TARGET_ARM_QBIT): Likewise.
7262 (TARGET_DSP_MULTIPLY): Likewise.
7263 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7264 (arm_arch5, arm_arch5e): Delete.
7265 (arm_arch5t, arm_arch5te): Declare.
7266 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7267 (arm_arch5t): Declare.
7268 (arm_option_reconfigure_globals): Update for the above.
7269 (arm_options_perform_arch_sanity_checks): Update comment, replace
7270 use of arm_arch5 with arm_arch5t.
7271 (use_return_insn): Likewise.
7272 (arm_emit_call_insn): Likewise.
7273 (output_return_instruction): Likewise.
7274 (arm_final_prescan_insn): Likewise.
7275 (arm_coproc_builtin_available): Likewise.
7276 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7277 arm_arch5e with arm_arch5t and arm_arch5te.
7278 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7279 (arm_arch5t, arm_arch5te): Declare.
7280 * config/arm/arm-tables.opt: Regenerate.
7281 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7282 * config/arm/t-multilib: Likewise.
7283 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7284 instead of arm_arch5.
7285 (*call_reg_thumb1): Likewise.
7286 (*call_value_reg_thumb1_v5): Likewise.
7287 (*call_value_reg_thumb1): Likewise.
7288 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7290 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7292 2018-05-18 Martin Liska <mliska@suse.cz>
7294 PR gcov-profile/84846
7295 * doc/gcov.texi: Document -t option of gcov tool.
7297 2018-05-18 Martin Liska <mliska@suse.cz>
7299 PR gcov-profile/84846
7300 * gcov.c (print_usage): Add new -t option.
7301 (process_args): Handle the option.
7302 (generate_results): Use stdout as output when requested by
7305 2018-05-18 Martin Liska <mliska@suse.cz>
7307 PR gcov-profile/84846
7308 * coverage.c (coverage_init): Write PWD to .gcno file.
7309 * doc/gcov.texi: Document how working directory is printed.
7310 * gcov-dump.c (dump_gcov_file): Print PWD.
7311 * gcov.c (output_intermediate_file): Likewise.
7312 (read_graph_file): Read PWD string.
7313 (output_lines): Print PWD.
7315 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7318 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7319 for retval and return false if all args to phi are zero.
7321 2018-05-18 Richard Biener <rguenther@suse.de>
7323 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7325 (evrp_dom_walker::before_dom_children): Call it.
7327 2018-05-18 Richard Biener <rguenther@suse.de>
7329 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7330 results when processing array refs with variable index.
7332 2018-05-18 Toon Moene <toon@moene.org>
7334 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7335 directly after that of -floop-interchange. Indicate that both
7336 options are enabled by default when specifying -O3.
7338 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7340 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7341 iterator. Delete separate integer-mode vec_set<mode> expander.
7342 (aarch64_simd_vec_setv2di): Delete.
7343 (vec_setv2di): Delete.
7344 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7345 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7346 the "w, r" alternative.
7348 2018-05-18 Martin Liska <mliska@suse.cz>
7350 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7351 * tree-pass.h (make_pass_lower_switch_O0): New function.
7352 * tree-switch-conversion.c (node_has_low_bound): Remove.
7353 (node_has_high_bound): Likewise.
7354 (node_is_bounded): Likewise.
7355 (class pass_lower_switch): Make it a template type and create
7357 (pass_lower_switch::execute): Add template argument.
7358 (make_pass_lower_switch): New function.
7359 (make_pass_lower_switch_O0): New function.
7360 (do_jump_if_equal): Remove.
7361 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7362 all the hard work to tree optimization passes.
7364 2018-05-18 Martin Liska <mliska@suse.cz>
7366 * dbgcnt.c (limit_low): Renamed from limit.
7367 (limit_high): New variable.
7368 (dbg_cnt_is_enabled): Check for upper limit.
7369 (dbg_cnt): Adjust dumping.
7370 (dbg_cnt_set_limit_by_index): Add new argument for high
7372 (dbg_cnt_set_limit_by_name): Likewise.
7373 (dbg_cnt_process_single_pair): Parse new format.
7374 (dbg_cnt_process_opt): Use strtok.
7375 (dbg_cnt_list_all_counters): Remove 'value' and add
7377 * doc/invoke.texi: Document changes.
7379 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7381 * doc/sourcebuild.texi (scalar_all_fma): Document.
7382 * tree.def (FMA_EXPR): Delete.
7383 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7384 * internal-fn.c (ternary_direct): New macro.
7385 (expand_ternary_optab_fn): Likewise.
7386 (direct_ternary_optab_supported_p): Likewise.
7387 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7388 * builtins.c (fold_builtin_fma): Delete.
7389 (fold_builtin_3): Don't call it.
7390 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7391 * expr.c (expand_expr_real_2): Likewise.
7392 * fold-const.c (operand_equal_p): Likewise.
7393 (fold_ternary_loc): Likewise.
7394 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7395 * gimple.c (DEFTREECODE): Likewise.
7396 * gimplify.c (gimplify_expr): Likewise.
7397 * optabs-tree.c (optab_for_tree_code): Likewise.
7398 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7399 * tree-eh.c (operation_could_trap_p): Likewise.
7400 (stmt_could_throw_1_p): Likewise.
7401 * tree-inline.c (estimate_operator_cost): Likewise.
7402 * tree-pretty-print.c (dump_generic_node): Likewise.
7403 (op_code_prio): Likewise.
7404 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7405 * tree-ssa-operands.c (get_expr_operands): Likewise.
7406 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7407 * fold-const-call.h (fold_fma): Delete.
7408 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7409 CFN_FNMA and CFN_FNMS.
7411 * genmatch.c (combined_fn): New enum.
7412 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7413 (commutative_op): New function.
7414 (commutate): Use it. Handle more than 2 operands.
7415 (dt_operand::gen_gimple_expr): Use commutative_op.
7416 (parser::parse_expr): Allow :c to be used with non-binary
7417 operators if the commutative operand is known.
7418 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7419 CFN_FMS, CFN_FNMA and CFN_FNMS.
7420 (backprop::process_assign_use): Remove FMA_EXPR handling.
7421 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7422 (gen_hsa_fma): New function.
7423 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7424 IFN_FNMA and IFN_FNMS.
7425 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7426 * gimple-fold.h (follow_all_ssa_edges): Declare.
7427 * gimple-fold.c (follow_all_ssa_edges): New function.
7428 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7429 gimple_build interface and use follow_all_ssa_edges to fold the result.
7430 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7431 instead of checking for optabs directly.
7432 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7433 rather than FMA_EXPRs.
7434 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7435 call to IFN_FMA instead of an FMA_EXPR.
7437 2018-05-17 Jim Wilson <jimw@sifive.com>
7439 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7440 sign extended promoted subreg, and the range does not have the sign bit
7441 set, then do a sign extend.
7443 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7444 test, check for sign extended subreg and/or constant operands, and
7445 do a sign extend in that case.
7447 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7449 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7450 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7452 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7453 Change logics_shift_reg to logics_shift_imm.
7454 (thunderx2t99_fp_loadpair_basic): Delete.
7455 (thunderx2t99_fp_storepair_basic): Delete.
7456 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7457 (thunderx2t99_asimd_polynomial): Delete.
7458 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7459 and neon_fp_mul_d_scalar_q.
7460 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7461 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7462 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7463 (thunderx2t99_asimd_lut): Add missing tbl types.
7464 (thunderx2t99_asimd_ext): Delete.
7465 (thunderx2t99_asimd_load1_1_mult): Delete.
7466 (thunderx2t99_asimd_load1_2_mult): Delete.
7467 (thunderx2t99_asimd_load1_ldp): New.
7468 (thunderx2t99_asimd_load1): New.
7469 (thunderx2t99_asimd_load2): Add missing *load2* types.
7470 (thunderx2t99_asimd_load3): New.
7471 (thunderx2t99_asimd_load4): New.
7472 (thunderx2t99_asimd_store1_1_mult): Delete.
7473 (thunderx2t99_asimd_store1_2_mult): Delete.
7474 (thunderx2t99_asimd_store2_mult): Delete.
7475 (thunderx2t99_asimd_store2_onelane): Delete.
7476 (thunderx2t99_asimd_store_stp): New.
7477 (thunderx2t99_asimd_store1): New.
7478 (thunderx2t99_asimd_store2): New.
7479 (thunderx2t99_asimd_store3): New.
7480 (thunderx2t99_asimd_store4): New.
7482 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7484 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7485 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7487 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7488 Segher Boessenkool <segher@kernel.crashing.org>
7491 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7494 2018-05-17 Richard Biener <rguenther@suse.de>
7496 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7497 for pruning loop and prune defs feeding only already visited PHIs.
7499 2018-05-17 Richard Biener <rguenther@suse.de>
7501 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7503 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7504 Richard Biener <rguenther@suse.de>
7506 PR tree-optimization/85793
7507 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7508 for VMAT_ELEMENTWISE.
7510 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7512 * internal-fn.h (lookup_internal_fn): Declare
7513 * internal-fn.c (lookup_internal_fn): New function.
7514 * gimple.c (gimple_build_call_from_tree): Handle calls to
7516 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7517 internal function names.
7518 * tree-pretty-print.c (dump_generic_node): Likewise.
7519 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7521 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7523 * gimple-fold.h (gimple_build): Make the function forms take
7524 combined_fn rather than built_in_function.
7525 (gimple_simplify): Likewise.
7526 * gimple-match-head.c (gimple_simplify): Likewise.
7527 * gimple-fold.c (gimple_build): Likewise.
7528 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7529 rather than gimple_build_call_internal.
7530 (get_initial_defs_for_reduction): Likewise.
7531 (vect_create_epilog_for_reduction): Likewise.
7532 (vectorizable_live_operation): Likewise.
7534 2018-05-17 Martin Liska <mliska@suse.cz>
7536 * gimple-ssa-sprintf.c (format_directive): Do not use
7537 space in between 'G_' and '('.
7539 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7542 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7543 even if the mask is not all ones.
7546 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7548 (ix86_gimple_fold_builtin): Likewise.
7551 * config/i386/i386.c: Include tree-vector-builder.h.
7552 (ix86_vector_shift_count): New function.
7553 (ix86_fold_builtin): Fold shift builtins by scalar count.
7554 (ix86_gimple_fold_builtin): Likewise.
7556 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7557 _mm512_setzero): New intrinsics.
7559 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7560 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7562 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7563 code generation for cases where splatting a value is not useful.
7564 * simplify-rtx.c (simplify_ternary_operation): Simplify
7565 vec_merge across a vec_duplicate and a paradoxical subreg forming
7566 a vector mode to a vec_concat.
7568 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7570 * config.gcc: Support "goldmont-plus".
7571 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7573 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7574 PROCESSOR_GOLDMONT_PLUS.
7575 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7576 (processor_target_table): Add "goldmont-plus".
7577 (PTA_GOLDMONT_PLUS): Define.
7578 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7579 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7580 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7581 (fold_builtin_cpu): Add "goldmont-plus".
7582 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7583 (ix86_option_override_internal): Add "goldmont-plus".
7584 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7585 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7586 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7587 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7589 2018-05-17 Richard Biener <rguenther@suse.de>
7591 PR tree-optimization/85757
7592 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7593 remove defs that only feed that PHI from further processing.
7595 2018-05-16 Jim Wilson <jimw@sifive.com>
7597 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7599 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7601 2018-05-16 Mark Wielaard <mark@klomp.org>
7603 * dwarf2out.c (count_index_strings): New function.
7604 (output_indirect_strings): Call count_index_strings and generate
7605 header for dwarf_version >= 5.
7607 2018-05-16 Mark Wielaard <mark@klomp.org>
7609 * dwarf2out.c (dwarf_FORM): New function.
7610 (set_indirect_string): Use dwarf_FORM.
7611 (reset_indirect_string): Likewise.
7612 (size_of_die): Likewise.
7613 (value_format): Likewise.
7614 (output_die): Likewise.
7615 (add_skeleton_AT_string): Likewise.
7616 (output_macinfo_op): Likewise.
7617 (index_string): Likewise.
7618 (output_index_string_offset): Likewise.
7619 (output_index_string): Likewise.
7620 (count_index_strings): Likewise.
7622 2018-05-16 Carl Love <cel@us.ibm.com>
7624 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7625 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7627 2018-05-16 Martin Jambor <mjambor@suse.cz>
7629 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7630 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7632 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7634 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7635 (fnma<mode>4): Likewise.
7636 (fms<mode>4): Likewise.
7637 (fnms<mode>4): Likewise.
7638 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7639 (aarch64_fnma<mode>4): Likewise.
7640 (aarch64_fms<mode>4): Likewise.
7641 (aarch64_fnms<mode>4): Likewise.
7642 (aarch64_fnmadd<mode>4): Likewise.
7644 2018-05-16 Jason Merrill <jason@redhat.com>
7646 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7648 2018-05-16 Richard Biener <rguenther@suse.de>
7650 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7651 (dump_stmt_cost): Declare.
7652 (add_stmt_cost): Dump cost we add.
7653 (add_stmt_costs): New function.
7654 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7656 (vect_analyze_stmt): Adjust prototype.
7657 (vectorizable_condition): Likewise.
7658 (vectorizable_live_operation): Likewise.
7659 (vectorizable_reduction): Likewise.
7660 (vectorizable_induction): Likewise.
7661 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7662 cost vector to pass to vectorizable_ and record afterwards.
7663 (vect_model_reduction_cost): Take cost vector argument and adjust.
7664 (vect_model_induction_cost): Likewise.
7665 (vectorizable_reduction): Likewise.
7666 (vectorizable_induction): Likewise.
7667 (vectorizable_live_operation): Likewise.
7668 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7669 SLP_TREE_NUMBER_OF_VEC_STMTS.
7670 (vect_analyze_slp_cost_1): Remove.
7671 (vect_analyze_slp_cost): Likewise.
7672 (vect_slp_analyze_node_operations): Take visited args and
7673 a target cost vector. Avoid processing already visited stmt sets.
7674 (vect_slp_analyze_operations): Use a local cost vector to gather
7675 costs and register those of non-discarded instances.
7676 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7677 (vect_schedule_slp_instance): Remove copying of
7678 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7680 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7681 adding cost. Record cost entry location.
7682 (vect_prologue_cost_for_slp_op): Function to compute cost of
7683 a constant or invariant generated for SLP vect in the prologue,
7684 split out from vect_analyze_slp_cost_1.
7685 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7686 (vect_model_promotion_demotion_cost): Likewise.
7687 (vect_model_store_cost): Likewise, make static.
7688 (vect_model_load_cost): Likewise.
7689 (vectorizable_bswap): Add cost vector arg and adjust.
7690 (vectorizable_call): Likewise.
7691 (vectorizable_simd_clone_call): Likewise.
7692 (vectorizable_conversion): Likewise.
7693 (vectorizable_assignment): Likewise.
7694 (vectorizable_shift): Likewise.
7695 (vectorizable_operation): Likewise.
7696 (vectorizable_store): Likewise.
7697 (vectorizable_load): Likewise.
7698 (vectorizable_condition): Likewise.
7699 (vectorizable_comparison): Likewise.
7700 (can_vectorize_live_stmts): Likewise.
7701 (vect_analyze_stmt): Likewise.
7702 (vect_transform_stmt): Adjust calls to vectorizable_*.
7703 * tree-vectorizer.c: Include gimple-pretty-print.h.
7704 (dump_stmt_cost): New function.
7706 2018-05-16 Richard Biener <rguenther@suse.de>
7708 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7709 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7710 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7711 (check_name): New callback.
7712 (dse_classify_store): Track cycles via a visited bitmap of PHI
7713 defs and simplify handling of in-loop and across loop dead stores
7714 and properly fail for loop-variant refs. Handle byte-tracking with
7715 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7718 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7720 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7721 (vect_get_mask_type_for_stmt): Likewise.
7722 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7724 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7725 to determine the statement's vector type and the vector type that
7726 should be used for calculating nunits. Deal with cases in which
7727 the type has to be deferred.
7728 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7729 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7730 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7731 (vect_determine_vf_for_stmt): New functions, split out from...
7732 (vect_determine_vectorization_factor): ...here.
7733 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7734 (vect_get_mask_type_for_stmt): New functions, split out from
7735 vect_determine_vectorization_factor.
7737 2018-05-16 Richard Biener <rguenther@suse.de>
7739 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7740 verify the [VEC_]COND_EXPR embedded comparison.
7742 2018-05-15 Martin Sebor <msebor@redhat.com>
7744 PR tree-optimization/85753
7745 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7746 RECORD_TYPE in addition to ARRAY_TYPE.
7748 2018-05-15 Martin Sebor <msebor@redhat.com>
7751 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7753 2018-05-15 Richard Biener <rguenther@suse.de>
7755 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7756 add by_clobber_p one. Change algorithm to collect all defs
7757 representing uses we need to walk and try reducing them to
7758 a single one before failing.
7759 (dse_dom_walker::dse_optimize_stmt): Adjust.
7761 2018-05-13 Mark Wielaard <mark@klomp.org>
7763 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7764 (size_of_loc_descr): Likewise.
7765 (output_loc_operands): Likewise.
7766 (output_loc_operands_raw): Likewise.
7767 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7768 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7769 (hash_loc_operands): Likewise.
7770 (compare_loc_operands): Likewise.
7772 2018-05-14 Mark Wielaard <mark@klomp.org>
7774 * dwarf2out.c (count_index_addrs): New function.
7775 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7777 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7779 PR tree-optimization/83648
7780 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7781 return value as malloc candidate.
7783 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7786 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7787 param as true in call to suggest_attribute.
7789 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7791 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7792 -mreadonly-in-sdata.
7794 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7796 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7798 (aarch64_crypto_aesd_fused): Likewise.
7800 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7802 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7803 (movsi_aarch64): Likewise.
7804 (load_pairsi): Likewise.
7805 (load_pairdi): Likewise.
7806 (store_pairsi): Likewise.
7807 (store_pairdi): Likewise.
7808 (load_pairsf): Likewise.
7809 (load_pairdf): Likewise.
7810 (store_pairsf): Likewise.
7811 (store_pairdf): Likewise.
7812 (zero_extend): Likewise.
7813 (trunc): Swap alternatives.
7814 (fcvt_target): Add '?' to prefer w over r.
7816 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7819 * config/i386/i386.md: Disallow non-commutative arithmetics in
7820 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7821 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7822 in the peephole2 before it.
7824 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7826 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7827 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7828 (ix86_handle_option): Handle -mcldemote.
7829 * config.gcc: New header.
7830 * config/i386/cldemoteintrin.h: New file.
7831 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7832 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7834 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7835 OPTION_MASK_ISA_CLDEMOTE.
7836 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7837 (ix86_valid_target_attribute_inner_p): Ditto.
7838 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7839 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7840 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7841 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7842 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7844 * config/i386/i386.opt: Add -mcldemote.
7845 * config/i386/x86intrin.h: New header.
7846 * doc/invoke.texi: Add -mcldemote.
7848 2018-05-14 Richard Biener <rguenther@suse.de>
7850 * doc/match-and-simplify.texi: Adjust :s documentation.
7852 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7854 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7855 intended memcpy size.
7856 (REORDER_45): Likewise.
7858 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7860 * sort.cc: New file.
7861 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7862 * vec.c (qsort_chk): Use gcc_qsort.
7863 * Makefile.in (OBJS-libcommon): Add sort.o.
7864 (build/sort.o): New target. Use it...
7865 (BUILD_RTL): ... here, and...
7866 (build/gencfn-macros): ... here, and...
7867 (build/genmatch): ... here.
7869 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7870 Chung-Ju Wu <jasonwucj@gmail.com>
7872 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7873 * config/nds32/nds32-graywolf.md: New file.
7874 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7875 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7877 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7878 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7879 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7880 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7881 * config/nds32/pipelines.md: Include n15 settings.
7883 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7884 Chung-Ju Wu <jasonwucj@gmail.com>
7886 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7887 * config/nds32/nds32-n13.md: New file.
7888 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7889 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7891 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7892 * config/nds32/nds32.md (pipeline_model): Add n13.
7893 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7894 * config/nds32/pipelines.md: Include n13 settings.
7896 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7897 Chung-Ju Wu <jasonwucj@gmail.com>
7899 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7900 * config/nds32/nds32-n10.md: New file.
7901 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7902 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7904 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7905 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7906 * config/nds32/nds32.md (pipeline_model): Add n10.
7907 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7908 * config/nds32/pipelines.md: Include n10 settings.
7910 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7911 Kito Cheng <kito.cheng@gmail.com>
7912 Chung-Ju Wu <jasonwucj@gmail.com>
7914 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7915 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7916 Add enum values for DSP extension instructions.
7917 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7919 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7920 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7922 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7923 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7924 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7925 * config/nds32/nds32-intrinsic.md: Likewise.
7926 * config/nds32/nds32_intrinsic.h: Likewise.
7927 * config/nds32/nds32-md-auxiliary.c: Likewise.
7928 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7929 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7930 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7931 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7932 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7933 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7934 TYPE_DMAC in switch statement.
7935 * config/nds32/nds32.c: New checking and implementation for DSP
7936 extension instructions.
7937 * config/nds32/nds32.h: Likewise.
7938 * config/nds32/nds32.md: Likewise.
7939 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7940 * config/nds32/predicates.md: Implement new predicates for DSP
7943 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7945 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7946 Reformat alternatives and attributes so it is easier to identify
7947 which constraints/attributes go with which instruction.
7948 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7949 (mov<mode>_softfloat32, FMOVE64): Likewise.
7950 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7951 (mov<mode>_softfloat64, FMOVE64): Likewise.
7953 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7955 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7957 (Basic PowerPC Built-in Functions): The new name of the
7958 subsection previously known as "PowerPC Built-in Functions".
7959 (Basic PowerPC Built-in Functions Available on all Configurations):
7961 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7962 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7963 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7964 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7966 2018-05-11 Martin Jambor <mjambor@suse.cz>
7969 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7972 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7975 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7977 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7979 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7980 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7981 (ix86_handle_option): Handle -mwaitpkg.
7982 * config.gcc: New header.
7983 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7984 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7985 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7987 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7988 OPTION_MASK_ISA_WAITPKG.
7989 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7990 (ix86_option_override_internal): Add PTA_WAITPKG.
7991 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7992 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7993 IX86_BUILTIN_TPAUSE.
7994 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7995 __builtin_ia32_umwait and __builtin_ia32_tpause.
7996 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7997 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7998 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7999 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8000 UNSPECV_TPAUSE): New.
8001 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8002 * config/i386/i386.opt: Add -mwaitpkg.
8003 * config/i386/waitpkgintrin.h: New file.
8004 * config/i386/x86intrin.h: New header.
8005 * doc/invoke.texi: Add -mwaitpkg.
8007 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8010 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8012 (cortex-m0): Use armv6s-m isa.
8013 (cortex-m0plus): Likewise.
8014 (cortex-m1): Likewise.
8015 (cortex-m0.small-multiply): Likewise.
8016 (cortex-m0plus.small-multiply): Likewise.
8017 (cortex-m1.small-multiply): Likewise.
8019 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8020 Jakub Jelinek <jakub@redhat.com>
8022 PR tree-optimization/85692
8023 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8024 source permute as well.
8026 2018-05-11 Martin Liska <mliska@suse.cz>
8029 * doc/extend.texi: Document LLVM style format for no_sanitize
8032 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8034 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8035 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8036 (mode_supports_vsx_dform_quad): Likewise.
8037 (mode_supports_vmx_dform): Move these functions to be next to the
8038 other mode_supports functions.
8039 (mode_supports_dq_form): Likewise.
8040 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8041 mode_supports_dq_form.
8042 (reg_offset_addressing_ok_p): Likewise.
8043 (offsettable_ok_by_alignment): Likewise.
8044 (rs6000_legitimate_offset_address_p): Likewise.
8045 (legitimate_lo_sum_address_p): Likewise.
8046 (rs6000_legitimize_address): Likewise.
8047 (rs6000_legitimize_reload_address): Likewise.
8048 (rs6000_secondary_reload_inner): Likewise.
8049 (rs6000_preferred_reload_class): Likewise.
8050 (rs6000_output_move_128bit): Likewise.
8052 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8054 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8055 Generate SImode target register for null target.
8056 <case IX86_BUILTIN_XGETBV>: Ditto.
8057 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8058 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8060 2018-05-10 Carl Love <cel@us.ibm.com>
8062 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8063 dcbtt and dcbtstt if operands[2] is 0.
8065 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8068 * config/i386/sse.md (usadv64qi): New expander.
8070 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8072 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8073 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8074 -maltivec=be support.
8075 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8076 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8077 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8078 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8079 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8080 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8081 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8082 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8083 altivec_vsumsws): Adjust.
8084 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8085 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8086 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8088 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8089 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8090 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8091 (altivec_lve<VI_char>x): Delete expand.
8092 (*altivec_lve<VI_char>x_internal): Rename to...
8093 (altivec_lve<VI_char>x): ... this.
8094 (altivec_lvxl_<mode>): Delete expand.
8095 (*altivec_lvxl_<mode>_internal): Rename to ...
8096 (altivec_lvxl_<mode>): ... this.
8097 (altivec_stvxl_<mode>): Delete expand.
8098 (*altivec_stvxl_<mode>_internal): Rename to ...
8099 (altivec_stvxl_<mode>): ... this.
8100 (altivec_stve<VI_char>x): Delete expand.
8101 (*altivec_stve<VI_char>x_internal): Rename to ...
8102 (altivec_stve<VI_char>x): ... this.
8103 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8104 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8105 reduc_plus_scal_<mode>): Adjust.
8106 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8108 (rs6000_cpu_cpp_builtins): Adjust.
8109 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8110 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8111 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8112 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8113 -maltivec=be support.
8114 (rs6000_split_vec_extract_var): Adjust.
8115 (rs6000_split_v4si_init): Adjust.
8116 (swap_selector_for_mode): Delete.
8117 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8118 altivec_expand_stvex_be): Delete.
8119 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8120 -maltivec=be support.
8121 (rs6000_gimple_fold_builtin): Ditto.
8122 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8124 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8125 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8126 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8127 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8128 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8129 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8130 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8131 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8132 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8133 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8134 anonymous split): Adjust.
8135 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8136 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8138 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8140 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8141 when --with-gxx-include-dir is also specified.
8142 * configure: Regenerate.
8144 2018-05-09 Jim Wilson <jimw@sifive.com>
8147 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8148 * config/riscv/t-withmultilib: New.
8149 * config/riscv/withmultilib.h: New.
8150 * doc/install.texi: Document RISC-V --with-multilib-list support.
8152 2018-05-09 Richard Biener <rguenther@suse.de>
8154 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8156 (vect_bb_vectorization_profitable_p): Adjust. Compute
8157 actual scalar cost using the cost vector and the add_stmt_cost
8160 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8162 PR rtl-optimization/85645
8163 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8164 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8166 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8168 PR rtl-optimization/85645
8169 * shrink-wrap.c (spread_components): Return a boolean saying if
8170 anything was changed.
8171 (try_shrink_wrapping_separate): Iterate spread_components until
8172 nothing changes anymore.
8174 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8176 PR rtl-optimization/85645
8177 * regrename.c (build_def_use): Also kill the chains that include the
8178 destination of a REG_CFA_REGISTER note.
8180 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8182 PR rtl-optimization/85645
8183 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8184 insn that has a REG_CFA_REGISTER note.
8186 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8188 * cfgexpand.c (expand_clobber): New function.
8189 (expand_gimple_stmt_1): Use it.
8190 * tree-vect-stmts.c (vect_clobber_variable): New function,
8192 (vectorizable_simd_clone_call): ...here.
8193 (vectorizable_store): Emit a clobber either side of an
8194 IFN_STORE_LANES sequence.
8195 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8197 2018-05-09 Tom de Vries <tom@codesourcery.com>
8200 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8201 (define_insn "trap_if_false"): Add exit after trap.
8203 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8205 PR rtl-optimization/85638
8206 * bb-reorder.c: Include common/common-target.h.
8207 (create_forwarder_block): New function extracted from...
8208 (fix_up_crossing_landing_pad): ...here. Rename into...
8209 (dw2_fix_up_crossing_landing_pad): ...this.
8210 (sjlj_fix_up_crossing_landing_pad): New function.
8211 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8212 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8213 from both partitions and exit the loop after one iteration.
8215 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8218 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8220 (Basic PowerPC Built-in Functions): The new name of the
8221 subsection previously known as "PowerPC Built-in Functions".
8222 (Basic PowerPC Built-in Functions Available on all Configurations):
8224 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8226 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8228 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8230 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8233 2018-05-08 Jim Wilson <jimw@sifive.com>
8235 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8236 (LD_EMUL_SUFFIX): New.
8237 (LINK_SPEC): Use it.
8239 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8241 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8243 (Basic PowerPC Built-in Functions): The new name of the
8244 subsection previously known as "PowerPC Built-in Functions".
8245 (Basic PowerPC Built-in Functions Available on all Configurations):
8247 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8249 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8251 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8253 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8256 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8259 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8260 after cmpelim optimization.
8262 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8264 * config.gcc: Support "goldmont".
8265 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8266 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8268 * config/i386/i386.c (m_GOLDMONT): Define.
8269 (processor_target_table): Add "goldmont".
8270 (PTA_GOLDMONT): Define.
8271 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8272 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8273 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8274 (fold_builtin_cpu): Add "goldmont".
8275 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8276 (ix86_option_override_internal): Add "goldmont".
8277 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8278 (processor_type): Add PROCESSOR_GOLDMONT.
8279 * config/i386/i386.md: Add CPU "glm".
8280 * config/i386/glm.md: New file.
8281 * config/i386/x86-tune.def: Add m_GOLDMONT.
8282 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8284 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8287 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8289 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8290 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8291 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8294 * config/i386/i386.c (ix86_fold_builtin): Handle
8295 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8298 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8299 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8301 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8304 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8305 (check_arch): Likewise.
8306 (check_fpu): Return the result rather than printing it.
8307 (end arch): Fix operator precedence.
8308 (end cpu): Likewise.
8309 (END): Print the result from check_fpu.
8311 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8312 Alan Hayward <alan.hayward@arm.com>
8313 David Sherwood <david.sherwood@arm.com>
8315 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8316 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8317 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8318 (*fcmuo<mode>_and): New patterns.
8320 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8322 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8323 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8324 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8325 (cmp_op, sve_imm_con): New code attributes.
8326 (SVE_COND_INT_CMP, imm_con): Delete.
8327 (cmp_op): Remove above unspecs from int attribute.
8328 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8330 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8331 comparison-specific unspecs.
8332 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8333 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8334 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8335 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8336 (*vec_fcm<cmp_op><mode>): Rename to...
8337 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8338 (*vec_fcmuo<mode>): Rename to...
8339 (*fcmuo<mode>): ...this and adjust likewise.
8340 (*pred_fcm<cmp_op><mode>): New pattern.
8341 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8342 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8344 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8346 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8347 (aarch64_emit_sve_predicated_cond): New function.
8348 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8349 (aarch64_emit_unspec_cond_or): Replace with...
8350 (aarch64_emit_sve_or_conds): ...this new function. Use
8351 aarch64_emit_sve_ptrue_op for the individual comparisons and
8352 aarch64_emit_binop to OR them together.
8353 (aarch64_emit_inverted_unspec_cond): Replace with...
8354 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8355 aarch64_emit_sve_ptrue_op for the comparison and
8356 aarch64_emit_unop to invert the result.
8357 (aarch64_expand_sve_vec_cmp_float): Update after the above
8358 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8360 2018-05-07 Nathan Sidwell <nathan@acm.org>
8362 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8363 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8364 (Backwards Compatibility): Likewise.
8366 2018-05-07 Luis Machado <luis.machado@linaro.org>
8370 2018-05-07 Luis Machado <luis.machado@linaro.org>
8372 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8373 <prefetch_dynamic_strides>: New const bool field.
8374 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8375 prefetch_dynamic_strides.
8376 (exynosm1_prefetch_tune): Likewise.
8377 (thunderxt88_prefetch_tune): Likewise.
8378 (thunderx_prefetch_tune): Likewise.
8379 (thunderx2t99_prefetch_tune): Likewise.
8380 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8382 (aarch64_override_options_internal): Update to set
8383 PARAM_PREFETCH_DYNAMIC_STRIDES.
8384 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8385 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8386 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8387 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8388 prefetch-dynamic-strides setting.
8390 2018-05-07 Luis Machado <luis.machado@linaro.org>
8392 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8393 <minimum_stride>: New const int field.
8394 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8395 minimum_stride field.
8396 (exynosm1_prefetch_tune): Likewise.
8397 (thunderxt88_prefetch_tune): Likewise.
8398 (thunderx_prefetch_tune): Likewise.
8399 (thunderx2t99_prefetch_tune): Likewise.
8400 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8401 (aarch64_override_options_internal): Update to set
8402 PARAM_PREFETCH_MINIMUM_STRIDE.
8403 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8404 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8405 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8406 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8407 stride is constant and is below the minimum stride threshold.
8409 2018-05-07 Luis Machado <luis.machado@linaro.org>
8411 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8414 2018-05-07 Luis Machado <luis.machado@linaro.org>
8416 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8417 <prefetch_dynamic_strides>: New const bool field.
8418 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8419 prefetch_dynamic_strides.
8420 (exynosm1_prefetch_tune): Likewise.
8421 (thunderxt88_prefetch_tune): Likewise.
8422 (thunderx_prefetch_tune): Likewise.
8423 (thunderx2t99_prefetch_tune): Likewise.
8424 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8426 (aarch64_override_options_internal): Update to set
8427 PARAM_PREFETCH_DYNAMIC_STRIDES.
8428 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8429 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8430 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8431 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8432 prefetch-dynamic-strides setting.
8434 2018-05-07 Luis Machado <luis.machado@linaro.org>
8436 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8437 <minimum_stride>: New const int field.
8438 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8439 minimum_stride field.
8440 (exynosm1_prefetch_tune): Likewise.
8441 (thunderxt88_prefetch_tune): Likewise.
8442 (thunderx_prefetch_tune): Likewise.
8443 (thunderx2t99_prefetch_tune): Likewise.
8444 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8445 (aarch64_override_options_internal): Update to set
8446 PARAM_PREFETCH_MINIMUM_STRIDE.
8447 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8448 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8449 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8450 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8451 stride is constant and is below the minimum stride threshold.
8453 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8456 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8457 the type is addressable. Don't force op into register if it has
8460 2018-05-05 Roland McGrath <mcgrathr@google.com>
8463 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8464 any section for which we don't know a specific type it should have,
8465 regardless of name. Previously this was done only for the exact
8466 names ".init_array", ".fini_array", and ".preinit_array".
8467 (default_elf_asm_named_section): Add comment about
8468 relationship with default_section_type_flags and SECTION_NOTYPE.
8469 (get_section): Don't consider it a type conflict if one side has
8470 SECTION_NOTYPE and the other doesn't, as long as neither has the
8471 SECTION_BSS et al used in the default_section_type_flags logic.
8473 2018-05-05 Tom de Vries <tom@codesourcery.com>
8476 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8477 (workaround_barsyncs): New function.
8478 (nvptx_reorg): Use workaround_barsyncs.
8479 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8480 (define_expand "nvptx_membar_cta"): New define_expand.
8481 (define_insn "*nvptx_membar_cta"): New insn.
8483 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8485 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8486 To improve optimization opportunities.
8487 * builtin-types.def: The new needed builtin types for the above.
8489 2018-05-04 Richard Biener <rguenther@suse.de>
8491 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8492 * gimple-ssa-store-merging.c
8493 (imm_store_chain_info::output_merged_store): Remove redundant create,
8494 release split_store vector contents on failure.
8495 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8496 scalar stmt vector on cache hit.
8498 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8500 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8502 * config.gcc (powerpc-xilinx-eabi*): Remove.
8503 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8505 (fusion_addis_mem_combo_load): Ditto.
8506 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8508 (rs6000_cpu_cpp_builtins): Ditto.
8509 * config/rs6000/rs6000-linux.c
8510 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8511 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8512 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8514 (rs6000_setup_reg_addr_masks): Ditto.
8515 (rs6000_init_hard_regno_mode_ok): Ditto.
8516 (rs6000_option_override_internal): Ditto.
8517 (legitimate_lo_sum_address_p): Ditto.
8518 (rs6000_legitimize_address): Ditto.
8519 (rs6000_legitimize_reload_address): Ditto.
8520 (rs6000_legitimate_address_p): Ditto.
8521 (abi_v4_pass_in_fpr): Ditto.
8522 (setup_incoming_varargs): Ditto.
8523 (rs6000_gimplify_va_arg): Ditto.
8524 (rs6000_split_multireg_move): Ditto.
8525 (rs6000_savres_strategy): Ditto.
8526 (rs6000_emit_prologue_components): Ditto.
8527 (rs6000_emit_epilogue_components): Ditto.
8528 (rs6000_emit_prologue): Ditto.
8529 (rs6000_emit_epilogue): Ditto.
8530 (rs6000_elf_file_end): Ditto.
8531 (rs6000_function_value): Ditto.
8532 (rs6000_libcall_value): Ditto.
8533 * config/rs6000/rs6000.h: Ditto.
8534 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8535 (TARGET_MINMAX): ... this. New.
8536 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8537 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8538 (*movsi_internal1_single): Delete.
8539 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8540 mfpu=, mxilinx-fpu): Delete.
8541 * config/rs6000/singlefp.h: Delete.
8542 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8543 * config/rs6000/t-rs6000: Ditto.
8544 * config/rs6000/t-xilinx: Delete.
8545 * config/rs6000/titan.md: Adjust for fp_type removal.
8546 * config/rs6000/vsx.md: Remove Xilinx FP support.
8547 (VStype_simple): Delete.
8548 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8549 * config/rs6000/xfpu.h: Delete.
8550 * config/rs6000/xfpu.md: Delete.
8551 * config/rs6000/xilinx.h: Delete.
8552 * config/rs6000/xilinx.opt: Delete.
8553 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8554 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8556 2018-05-04 Tom de Vries <tom@codesourcery.com>
8559 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8562 2018-05-04 Richard Biener <rguenther@suse.de>
8565 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8566 (expand_complex_div_wide): Likewise.
8567 (expand_complex_operations_1): Likewise.
8568 (expand_complex_libcall): Preserve EH info of the original stmt.
8569 (tree_lower_complex): Handle removed blocks.
8570 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8571 on complex multiplication and division libcall builtins.
8573 2018-05-04 Richard Biener <rguenther@suse.de>
8576 * fold-const.c (negate_expr_p): Restrict negation of operand
8577 zero of a division to when we know that can happen without
8579 (fold_negate_expr_1): Likewise.
8581 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8584 * real.h (real_nextafter): Declare.
8585 * real.c (real_nextafter): New function.
8586 * fold-const-call.c (fold_const_nextafter): New function.
8587 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8588 CASE_CFN_NEXTTOWARD.
8589 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8590 even when arg1_mode is different from arg0_mode.
8592 2018-05-03 Nathan Sidwell <nathan@acm.org>
8594 * doc/extend.texi (Deprecated Features): Remove
8596 (Backwards Compatibility): Likewise.
8597 * doc/invoke.texi (C++ Language Options): Likewise.
8598 (C++ Dialect Options): Likewise.
8600 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8603 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8604 _mm512_mask_mullox_epi64): New intrinsics.
8606 2018-05-03 Tom de Vries <tom@codesourcery.com>
8609 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8610 dump files): Add offload-tree.
8612 2018-05-03 Richard Biener <rguenther@suse.de>
8614 PR tree-optimization/85615
8615 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8616 to loops not nested in BBs loop father to avoid creating multi-entry
8619 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8621 PR tree-optimization/70291
8622 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8623 arguments. Change return type to tree. Emit libcall as a new
8624 statement rather than replacing existing one when inplace_p is true.
8625 (expand_complex_multiplication_components): New function.
8626 (expand_complex_multiplication): Expand floating-point complex
8627 multiplication using the above.
8628 (expand_complex_division): Rename inner_type parameter to type.
8629 Update expand_complex_libcall call-site.
8630 (expand_complex_operations_1): Update expand_complex_multiplication
8631 and expand_complex_division call-sites.
8633 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8636 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8637 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8638 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8639 the highest significant bit of the shift count mask is clear. In
8640 check whether and[sq]i3 is needed verify that all significant bits
8641 of the shift count other than the highest are set.
8643 2018-05-02 Tom de Vries <tom@codesourcery.com>
8646 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8647 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8648 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8649 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8650 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8651 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8652 __builtin_goacc_parlevel_size.
8654 2018-05-02 Richard Biener <rguenther@suse.de>
8656 PR tree-optimization/85597
8657 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8658 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8661 2018-05-02 Tom de Vries <tom@codesourcery.com>
8664 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8665 dump files): Add ltrans-tree.
8667 2018-05-02 Tom de Vries <tom@codesourcery.com>
8670 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8671 dump files): Add wpa-ipa.
8673 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8675 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8676 powerpc*-*-linux*paired* target.
8677 * config/rs6000/750cl.h: Delete.
8678 * config/rs6000/paired.h: Delete.
8679 * config/rs6000/paired.md: Delete.
8680 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8682 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8683 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8684 comment. Remove paired float support.
8685 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8686 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8688 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8689 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8691 * config/rs6000/rs6000.c: Remove paired float support.
8692 (paired_expand_vector_init, paired_expand_vector_move,
8693 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8694 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8695 paired_expand_builtin, paired_expand_predicate_builtin,
8696 paired_init_builtins): Delete.
8697 * config/rs6000/rs6000.h: Remove paired float support.
8698 * config/rs6000/rs6000.md: Remove paired float support.
8699 (move_from_CR_ov_bit): Delete.
8700 * config/rs6000/rs6000.opt (mpaired): Delete.
8701 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8702 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8704 2018-05-02 Richard Biener <rguenther@suse.de>
8707 * gimplify.c (gimplify_save_expr): When in SSA form allow
8708 SAVE_EXPRs to compute to SSA vars.
8710 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8713 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8714 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8715 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8716 clobber operands[2], instead use a new pseudo. Formatting fixes.
8718 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8720 PR tree-optimization/85586
8721 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8722 exit early for statements in the same group if the accesses are
8725 2018-05-02 Tom de Vries <tom@codesourcery.com>
8728 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8731 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8733 PR tree-optimization/85143
8734 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8736 2018-05-01 Tom de Vries <tom@codesourcery.com>
8739 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8740 not found" error message.
8742 2018-05-01 Tom de Vries <tom@codesourcery.com>
8745 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8746 * vec.c (test_ordered_remove_if): New function.
8747 (vec_c_tests): Call test_ordered_remove_if.
8748 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8749 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8750 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8751 VEC_ORDERED_REMOVE_IF.
8753 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8755 PR tree-optimization/82665
8756 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8757 pointer subtraction where arguments come from a memchr call.
8759 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8761 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8762 --push-state --as-needed and --pop-state instead of --as-needed and
8763 --no-as-needed if ld supports it.
8764 * configure: Regenerated.
8767 * doc/install.texi2html: Replace _002d with - and _002a with * in
8768 generated html files using sed.
8770 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8773 * gcc-rich-location.c (blank_line_before_p): New function.
8774 (use_new_line): New function.
8775 (gcc_rich_location::add_fixit_insert_formatted): New function.
8776 * gcc-rich-location.h
8777 (gcc_rich_location::add_fixit_insert_formatted): New function.
8779 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8781 * selftest.c (assert_streq): Rename "expected" and "actual" to
8782 "val1" and "val2". Extend NULL-handling to cover both inputs
8783 symmetrically, while still requiring both to be non-NULL for a pass.
8784 * selftest.h (assert_streq): Rename "expected" and "actual" to
8786 (ASSERT_EQ): Likewise.
8787 (ASSERT_EQ_AT): Likewise.
8788 (ASSERT_KNOWN_EQ): Likewise.
8789 (ASSERT_KNOWN_EQ_AT): Likewise.
8790 (ASSERT_NE): Likewise.
8791 (ASSERT_MAYBE_NE): Likewise.
8792 (ASSERT_MAYBE_NE_AT): Likewise.
8793 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8794 the assertion to pass.
8795 (ASSERT_STREQ_AT): Likewise.
8797 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8799 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8800 interaction with -pie.
8802 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8804 * selftest.h: Fix alphabetization of per-source-file selftest
8807 2018-04-30 Jason Merrill <jason@redhat.com>
8809 PR c++/61982 - dead stores to destroyed objects.
8810 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8813 2018-04-30 Jason Merrill <jason@redhat.com>
8815 * tree.c (build_clobber): New.
8816 * tree.h: Declare it.
8817 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8819 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8821 * diagnostic-show-locus.c (layout::layout): Update for
8822 location_get_source_line returning a char_span.
8823 (struct char_span): Move to input.h.
8824 (struct correction): Update for fields in char_span becoming
8826 (struct source_line): Update for location_get_source_line
8827 returning a char_span.
8828 (layout::print_line): Likewise.
8829 * edit-context.c (edited_file::print_content): Likewise.
8830 (edited_file::print_diff_hunk): Likewise.
8831 (edited_file::print_run_of_changed_lines): Likewise.
8832 (edited_file::get_num_lines): Likewise.
8833 (edited_line::edited_line): Likewise.
8834 * final.c (asm_show_source): Likewise.
8835 * input.c (location_get_source_line): Convert return type
8836 from const char * to char_span, losing the final "line_len"
8838 (dump_location_info): Update for the above.
8839 (get_substring_ranges_for_loc): Likewise. Use a char_span
8840 when handling the literal within the line.
8841 (test_reading_source_line): Update for location_get_source_line
8842 returning a char_span.
8843 * input.h (class char_span): Move here from
8844 diagnostic-show-locus.c, converting from a struct to a class.
8845 Make data members private.
8846 (char_span::operator bool): New.
8847 (char_span::length): New.
8848 (char_span::get_buffer): New.
8849 (char_span::operator[]): New.
8850 (char_span::subspan): Make const.
8851 (char_span::xstrdup): New.
8852 (location_get_source_line): Convert return type from const char *
8853 to char_span, losing the final "line_size" param.
8855 2018-04-30 Jan Hubicka <jh@suse.cz>
8857 * lto-wrapper.c (ltrans_priorities): New static var.
8858 (cmp_priority): New.
8859 (run_gcc): Read priorities and if doing parallel build order
8860 the Makefile by them.
8862 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8864 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8866 2018-04-30 Richard Biener <rguenther@suse.de>
8868 * tree-cfg.c (verify_address): Remove base argument, add
8869 flag whether to check TREE_ADDRESSABLE and do that.
8870 (verify_expr): Remove.
8871 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8872 (verify_gimple_assign_single): Likewise.
8873 (verify_gimple_switch): Likewise.
8874 (verify_expr_location_1): Dereference tp once. Add (disabled)
8875 piece from verify_expr.
8876 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8878 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8880 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8882 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8884 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8885 (small_data_pattern): Likewise.
8886 (arc_rewrite_small_data): Likewise.
8887 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8888 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8889 (get_symbol_alignment): New function.
8890 (legitimate_small_data_address_p): Likewise.
8891 (legitimate_scaled_address): Update, call
8892 legitimate_small_data_address_p.
8893 (output_sdata): New static variable.
8894 (arc_print_operand): Update how we handle small data operands.
8895 (arc_print_operand_address): Likewise.
8896 (arc_legitimate_address_p): Update, use
8897 legitimate_small_data_address_p.
8898 (arc_rewrite_small_data_p): Remove.
8899 (arc_rewrite_small_data_1): Likewise.
8900 (arc_rewrite_small_data): Likewise.
8901 (small_data_pattern): Likewise.
8902 (compact_sda_memory_operand): Update to use
8903 legitimate_small_data_address_p and get_symbol_alignment.
8904 (prepare_move_operands): Don't rewite sdata pattern.
8905 (prepare_extend_operands): Remove.
8906 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8908 (zero_extendqisi2): Likewise.
8909 (zero_extendhisi2): Likewise.
8910 (extendqihi2): Likewise.
8911 (extendqisi2): Likewise.
8912 (extendhisi2): Likewise.
8916 * config/arc/constraints.md (Usd): Change it to memory constraint.
8918 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8920 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8921 as source of std instructions.
8922 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8923 allow 6-bit constants as source for store instructions.
8924 (movdi_insn): Update instruction pattern to allow 6-bit constants
8925 as source for store instructions.
8927 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8929 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8931 2018-04-30 Nathan Sidwell <nathan@acm.org>
8932 Sandra Loosemore <sandra@codesourcery.com>
8934 * dumpfile.c (dump_open): Allow '-' for stdout.
8935 * doc/invoke.texi (Developer Options): Document dump filename
8936 determination early. Document stdin/stdout selection.
8938 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8940 Microblaze Target: PIC data text relative
8942 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8943 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8945 * config/microblaze/microblaze.h (microblaze_constant_address_p):
8946 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8947 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8948 New addressing mode for data-text relative position indepenedent code.
8949 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8950 'ADDRESS_SYMBOLIC_TXT_REL'.
8951 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8952 (microblaze_legitimate_pic_operand): Exclude function calls from
8953 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8954 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8956 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8957 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8958 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8959 for 'address + offset'.
8960 (microblaze_expand_prologue): Add new function prologue call for
8962 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8963 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8964 table in case of TARGET_PIC_DATA_TEXT_REL.
8965 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8966 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8967 Add new macros 'UNSPEC_TEXT',
8968 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8969 + exclude function calls from 'UNSPEC_PLT' in case of data text
8971 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8972 new target hook for generating address diff vector tables in case of
8974 * doc/tm.texi : Regenerate.
8975 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8976 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8977 of addr diff vector generation.
8978 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8979 target hook definition.
8980 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8981 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8982 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8983 option for data text relative.
8985 2018-04-30 Richard Biener <rguenther@suse.de>
8987 * tree-chrec.h (evolution_function_is_constant_p): Remove
8989 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8991 2018-04-30 Richard Biener <rguenther@suse.de>
8994 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8996 2018-04-30 Richard Biener <rguenther@suse.de>
8998 PR tree-optimization/28364
8999 PR tree-optimization/85275
9000 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9001 copying first exit test.
9003 2018-04-28 Mark Wielaard <mark@klomp.org>
9005 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9007 (dwarf_AT): Handle DW_AT_addr_base.
9008 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9010 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9013 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9014 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9015 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9016 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9018 2018-04-28 Richard Biener <rguenther@suse.de>
9020 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9021 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9022 to reflect use. Only add interesting stmts.
9024 2018-04-27 Martin Jambor <mjambor@suse.cz>
9027 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9028 the jump function allows for passing through aggregate values.
9030 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9032 * input.h (in_system_header_at): Convert from macro to inline
9034 (from_macro_expansion_at): Likewise.
9035 (from_macro_definition_at): Likewise.
9037 2018-04-27 Jeff Law <law@redhat.com>
9039 * config.gcc: Mark tile* targets as deprecated/obsolete.
9041 2018-04-27 Richard Biener <rguenther@suse.de>
9043 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9046 2018-04-27 Richard Biener <rguenther@suse.de>
9048 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9050 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9052 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9053 with Yd constraint. Set "preferred_for_speed" attribute from
9054 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9056 (*movdi_internal): Ditto.
9057 (movti_interunit splitters): Remove
9058 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9059 (movdi_interunit splitters): Ditto.
9060 * config/i386/constraints.md (Ye): Remove.
9061 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9063 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9066 * config/aarch64/constraints.md (Usg): Limit to 31.
9069 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9071 PR tree-optimization/85529
9072 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9073 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9074 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9075 zero extension or masking of the MSB bit.
9076 (optimize_range_tests): Add FIRST_BB argument, pass it through
9077 to optimize_range_tests_var_bound.
9078 (maybe_optimize_range_tests, reassociate_bb): Adjust
9079 optimize_range_tests callers.
9081 2018-04-26 Richard Biener <rguenther@suse.de>
9082 Jakub Jelinek <jakub@redhat.com>
9084 * cgraph.h (symbol_table): Just declare debug method here.
9085 * symtab.c (symbol_table::debug): Define.
9087 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9089 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9091 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9093 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9094 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9095 (*movdi_internal): Substitute Yi and Yj constraint with x
9096 and Ym and Yn constraint with y constraint. Update "isa"
9097 attribute and set "preferred_for_speed" attribute from
9098 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9099 (*movsi_internal): Ditto.
9100 (*movdf_internal): Ditto.
9101 (*movsf_internal): Ditto.
9102 (*zero_extendsidi2): Ditto.
9103 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9104 (sse2_loadld): Ditto.
9105 (*vec_extract<ssevecmodelower>_0): Ditto.
9106 (*vec_extractv4si_0_zext_sse4): Ditto.
9107 (vec_concatv2di): Ditto.
9108 (*vec_dup<mode>): Ditto.
9109 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9110 * config/i386/constraints.md (Yi): Remove.
9115 2018-04-26 Nathan Sidwell <nathan@acm.org>
9117 * dumpfile.c (dump_open): New.
9118 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9119 (dump_finish): Detect stdio/stderr by value not name.
9121 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9123 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9125 2018-04-26 Tom de Vries <tom@codesourcery.com>
9128 * config/nvptx/nvptx.c (verify_neutering_jumps)
9129 (verify_neutering_labels): New function
9130 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9132 2018-04-26 Tom de Vries <tom@codesourcery.com>
9135 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9136 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9139 2018-04-26 Richard Biener <rguenther@suse.de>
9140 Tom de Vries <tom@codesourcery.com>
9143 * lto-streamer-out.c (output_function): Fixup loops if required to match
9144 discovery done in the reader.
9146 2018-04-26 Richard Biener <rguenther@suse.de>
9148 PR tree-optimization/85116
9149 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9150 have a loop exit from the single latch predecessor. Remove
9151 case of header with just condition.
9152 (ch_base::copy_headers): Exclude infinite loops from any
9154 (pass_ch::execute): Record exits.
9156 2018-04-26 Richard Biener <rguenther@suse.de>
9158 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9159 prologue cost vector and pass it to vect_get_load_cost.
9160 (vect_get_peeling_costs_all_drs): Likewise.
9161 (vect_peeling_hash_get_lowest_cost): Likewise.
9162 (vect_enhance_data_refs_alignment): Likewise.
9164 2018-04-26 Richard Biener <rguenther@suse.de>
9167 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9168 checking of integer<->pointer conversions.
9169 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9170 sign-/zero-extending pointer types.
9171 (expand_omp_for_static_chunk): Likewise.
9173 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9174 Jean Lee <xiaoyur347@gmail.com>
9176 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9177 (TARGET_ASAN_SHADOW_OFFSET): Define.
9178 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9179 true for -fsanitize=address.
9181 2018-04-25 Mark Wielaard <mark@klomp.org>
9183 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9186 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9188 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9189 than "alu", remove explicit "memory" and "imm_disp" attributes.
9190 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9193 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9194 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9195 gen_lowpart_no_emit.
9197 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9200 * config/i386/i386.c (ix86_expand_builtin): Change memory
9201 operand to XI, extend p0 to Pmode.
9202 * config/i386/i386.md: Change unspec volatile and operand
9203 1 mode to XI, change operand 0 mode to P.
9205 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9207 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9208 GET_MODE_MASK before any checking.
9209 (nds32_can_use_bset_p): Likewise.
9210 (nds32_can_use_btgl_p): Likewise.
9212 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9214 * config/nds32/nds32-doubleword.md: New define_split pattern for
9215 illegal register number.
9217 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9219 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9221 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9223 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9225 2018-04-25 Richard Biener <rguenther@suse.de>
9227 * lto-streamer.h (LTO_major_version): Bump to 8.
9229 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9231 * BASE-VER: Set to 9.0.0.
9233 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9235 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9236 in __abskf2 and __powikf2.
9238 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9241 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9242 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9243 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9244 Use the above on operand 2. Reindent.
9245 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9247 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9250 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9252 * config/i386/i386.opt (mcet): Removed.
9253 * doc/install.texi: Remove -mcet documentation.
9254 * doc/invoke.texi: Likewise.
9256 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9259 * doc/install.texi: Remove -mcet from bootstrap-cet.
9261 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9264 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9265 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9269 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9270 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9271 containing a CONST_VECTOR.
9273 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9275 * doc/install.texi: Update newlib dependency for nvptx.
9277 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9280 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9281 instead of INTVAL when shifting x left.
9283 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9285 PR tree-optimization/85478
9286 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9287 vect_grouped_store_supported for single element vectors.
9289 2018-04-24 Richard Biener <rguenther@suse.de>
9292 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9293 load cost increase to the case of non-constant step.
9295 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9298 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9299 destination if any_malformed_asm.
9301 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9304 * expr.c (store_field): In the bitfield case, if the value comes from
9305 a function call and is returned in registers by means of a PARALLEL,
9306 do not change the mode of the temporary unless BLKmode and VOIDmode.
9308 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9310 PR rtl-optimization/85423
9311 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9312 dependencies to debug insns when the previous insn is non-debug.
9314 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9316 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9317 enums into a single definition.
9318 (fls): Fix predicates and printing.
9321 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9323 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9324 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9325 and short u6 immediate.
9326 (check_if_valid_sleep_operand): Remove.
9327 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9329 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9331 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9332 flag_always_save_lp condition.
9333 * config/nds32/nds32.opt (malways-save-lp): New option.
9335 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9337 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9338 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9339 * config/nds32/nds32.h
9340 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9341 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9343 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9345 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9346 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9348 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9349 Chung-Ju Wu <jasonwucj@gmail.com>
9351 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9353 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9354 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9356 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9358 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9360 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9362 * config/nds32/nds32-protos.h (nds32_data_alignment,
9363 nds32_local_alignment): Declare.
9364 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9365 nds32_local_alignment): New functions.
9366 (TARGET_CONSTANT_ALIGNMENT): Define.
9367 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9369 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9371 * config/nds32/nds32.c
9372 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9373 (TARGET_MODES_TIEABLE_P): Likewise.
9375 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9377 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9380 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9381 Chung-Ju Wu <jasonwucj@gmail.com>
9383 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9384 for unaligned access.
9385 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9387 * config/nds32/nds32-intrinsic.md: Likewise.
9388 * config/nds32/nds32_intrinsic.h: Likewise.
9389 * config/nds32/nds32.h (nds32_builtins): Likewise.
9390 * config/nds32/nds32.opt (munaligned-access): New option.
9391 * config/nds32/nds32.c (nds32_asm_file_start): Display
9392 flag_unaligned_access status.
9394 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9396 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9397 -mno-relax is present.
9398 * config/riscv/linux.h (LINK_SPEC): Ditto.
9400 2018-04-20 Martin Sebor <msebor@redhat.com>
9403 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9405 (gimple_fold_builtin_stxcpy_chk): Same.
9406 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9408 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9411 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9412 __powikf2 when long double is IEEE 128-bit.
9414 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9416 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9417 step to make sure stack always aligned.
9419 2018-04-20 Carl Love <cel@us.ibm.com>
9422 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9423 size check for arg0.
9425 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9426 Tom de Vries <tom@codesourcery.com>
9429 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9430 Emit insns for calls too.
9431 (nvptx_find_par): Always look for worker-level predecessor insn.
9432 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9434 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9435 (nvptx_process_pars): Propagate frames for calls.
9437 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9440 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9442 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9443 (ix86_handle_option): Don't handle OPT_mibt.
9444 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9446 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9448 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9449 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9450 (ix86_target_macros): Define __CET__ with flag_cf_protection
9451 for -fcf-protection.
9452 * config/i386/i386.c (isa2_opts): Remove -mibt.
9453 * config/i386/i386.h (TARGET_IBT): Removed.
9454 (TARGET_IBT_P): Likewise.
9455 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9456 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9457 * config/i386/i386.opt (mcet): Update help message.
9460 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9461 -mcet as an alias for -mshstk.
9463 2018-04-20 Richard Biener <rguenther@suse.de>
9466 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9467 complexity by forcing a single use of the multiply operand.
9469 2018-04-20 Martin Jambor <mjambor@suse.cz>
9472 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9473 recursion dependency to only apply to non-clones.
9475 2018-04-20 Martin Jambor <mjambor@suse.cz>
9478 * ipa-cp.c (create_specialized_node): Check that clones of
9479 self-recursive edges exist during IPA-CP.
9481 2018-04-19 Toon Moene <toon@moene.org>
9483 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9486 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9488 PR tree-optimization/85467
9489 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9490 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9491 VECTOR_CST element to type.
9493 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9496 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9497 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9498 (builtin_longjmp): Likewise.
9499 (save_stack_nonlocal): New pattern.
9500 (restore_stack_nonlocal): Likewise.
9502 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9505 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9506 Replace ASM_OUTPUT_LABEL with fprintf.
9508 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9511 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9512 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9513 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9514 define __IBT__ and __SHSTK__ for -fcf-protection.
9515 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9517 (ix86_trampoline_init): Likewise.
9518 (x86_output_mi_thunk): Likewise.
9519 (ix86_notrack_prefixed_insn_p): Likewise.
9520 (ix86_option_override_internal): Don't disallow -fcf-protection.
9521 * config/i386/i386.md (rdssp<mode>): Also enable for
9523 (incssp<mode>): Likewise.
9524 (nop_endbr): Likewise.
9525 * config/i386/i386.opt (mcet): Change help message to built-in
9529 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9530 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9531 enable CET built-in functions.
9533 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9535 * common/config/i386/i386-common.c
9536 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9537 OPTION_MASK_ISA_MOVDIRI_UNSET,
9538 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9539 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9540 * config.gcc (movdirintrin.h): New header.
9541 * config/i386/cpuid.h (bit_MOVDIRI,
9542 bit_MOVDIR64B): New bits.
9543 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9545 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9546 (VOID, PVOID, PCVOID)): New function types.
9547 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9548 __builtin_ia32_directstoreu_u64,
9549 __builtin_ia32_movdir64b): New builtins.
9550 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9551 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9553 (ix86_valid_target_attribute_inner_p): Ditto.
9554 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9555 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9556 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9557 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9558 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9559 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9560 (movdiri<mode>, movdir64b_<mode>): New.
9561 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9562 * config/i386/immintrin.h: Include movdirintrin.h.
9563 * config/i386/movdirintrin.h: New file.
9564 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9566 2018-04-19 Richard Biener <rguenther@suse.de>
9569 * cfg.c (clear_bb_flags): When loop state says we have
9570 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9572 2018-04-19 Richard Biener <rguenther@suse.de>
9574 PR tree-optimization/84737
9575 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9576 copying restrict info.
9577 (vect_setup_realignment): Use it.
9578 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9579 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9580 the first DR to all generated stores.
9581 (vectorizable_load): Likewise for loads.
9583 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9585 PR tree-optimization/85446
9586 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9587 the integral and pointer types to have the same precision.
9589 * doc/install.texi: Document --disable-cet being the default and
9592 2018-04-18 Martin Liska <mliska@suse.cz>
9594 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9597 2018-04-18 Martin Liska <mliska@suse.cz>
9600 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9603 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9604 arguments if they are comparable.
9606 2018-04-18 Martin Liska <mliska@suse.cz>
9609 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9612 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9615 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9618 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9619 ENDBR after calling __morestack.
9621 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9624 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9625 by using gcc_base_ver to generate a gcc_driver_version, and use
9626 it when generating GCC_DRIVER_NAME.
9627 * configure: Regenerate.
9629 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9632 * config.gcc: Obsolete powerpc*-*-*spe*.
9634 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9637 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9638 (stabstr_D): Change type of unum from unsigned int to
9639 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9642 2018-04-17 Jim Wilson <jimw@sifive.com>
9645 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9646 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9647 Set arg_pointer_offset after using pretend_args_size.
9649 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9651 PR rtl-optimization/85431
9652 * dse.c (record_store): Ignore zero width stores.
9655 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9656 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9657 __builtin_stack_restore rather than after it.
9658 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9659 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9660 argument instead of virtual_dynamic_stack_rtx.
9662 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9664 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9666 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9667 Add note to error message to explain internal mapping of overloaded
9668 built-in function name to non-overloaded built-in function name.
9669 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9672 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9675 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9676 where the inputs overlap with the output.
9678 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9680 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9681 (=v, v) alternative and explicit "memory" attribute.
9682 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9683 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9685 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9686 "sselog1" type instead of "sselog".
9687 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9688 "sselog". Remove explicit "memory" attribute.
9689 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9690 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9692 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9693 "sselog1" type instead of "sselog". Remove explicit "memory"
9695 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9696 use "sselog1" type instead of "sselog". Remove explicit "memory"
9698 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9699 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9701 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9702 "sselog1" type instead of "sselog". Remove explicit "memory"
9704 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9705 use "sselog1" type instead of "sselog". Remove explicit "memory"
9709 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9712 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9715 2018-04-17 Martin Jambor <mjambor@suse.cz>
9718 * ipa-cp.c (create_specialized_node): Call
9719 expand_all_artificial_thunks if necessary.
9721 2018-04-17 Martin Liska <mliska@suse.cz>
9724 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9725 in message, remote space in between '_G' and '('.
9727 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9730 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9731 avx512f_vmcmp<mode>3<round_saeonly_name>,
9732 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9733 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9734 avx512f_rndscale<mode><round_saeonly_name>,
9735 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9736 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9737 Use %<iptr>2 instead of %2 for -masm=intel.
9738 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9739 avx512f_vcvttss2usi<round_saeonly_name>,
9740 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9742 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9743 avx512f_vcvttsd2usi<round_saeonly_name>,
9744 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9745 Use %q1 instead of %1 for -masm=intel.
9746 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9747 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9748 of %3 for -masm=intel.
9749 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9751 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9753 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9755 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9756 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9758 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9760 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9761 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9762 %g1 and one with %0 and %1.
9763 (avx512er_vmrcp28<mode><round_saeonly_name>,
9764 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9766 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9767 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9768 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9769 of %0 and %{%4%} for -masm=intel.
9770 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9771 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9772 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9773 order of %0 and %{%5%}%{z%} for -masm=intel.
9775 2018-04-17 Jan Hubicka <jh@suse.cz>
9778 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9780 2018-04-17 Martin Liska <mliska@suse.cz>
9783 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9784 for target_clone error message. Make default implementation
9785 clone to be a local declaration.
9786 (separate_attrs): Add new argument and check for an empty
9788 (expand_target_clones): Handle it.
9789 (ipa_target_clone): Make redirection just for target_clones
9792 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9793 Tom de Vries <tom@codesourcery.com>
9796 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9797 tiled basic blocks without omp continue statements.
9799 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9802 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9803 vec_extract expression as having side effects to make sure it gets
9806 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9809 * config/i386/i386.c (get_builtin_code_for_version): Check
9812 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9815 * config.gcc: Support "skylake".
9816 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9818 * config/i386/i386.c (m_SKYLAKE): Define.
9819 (processor_target_table): Add "skylake".
9820 (ix86_option_override_internal): Add "skylake".
9821 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9822 PROCESSOR_CANNONLAKE.
9823 (get_builtin_code_for_version): Fix priority for
9824 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9825 PROCESSOR_SKYLAKE-AVX512.
9826 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9827 (processor_type): Add PROCESSOR_SKYLAKE.
9829 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9830 Jason Merrill <jason@redhat.com>
9833 * convert.c (convert_to_integer_1): Use direct recursion for
9834 enumeral types and types with a precision less than the number
9835 of bits in their mode.
9837 2018-04-16 Julia Koval <julia.koval@intel.com>
9840 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9841 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9843 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9846 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9847 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9848 and -mno-direct-move.
9850 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9853 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9854 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9855 Ensure negative shifts result in {0}.
9857 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9859 PR rtl-optimization/79916
9860 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9861 regs (if any) to define how to gnerate SD moves when LRA is in
9864 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9866 PR rtl-optimization/85393
9867 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9868 * except.c (expand_dw2_landing_pad_for_region): Make static.
9869 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9870 a label and unconditional jump to old_bb, rather than
9871 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9874 PR rtl-optimization/85376
9875 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9876 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9877 instead of a specific value.
9879 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9880 Bin Cheng <bin.cheng@arm.com>
9882 PR tree-optimization/82965
9883 PR tree-optimization/83991
9884 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9885 by_profile_only parameter.
9886 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9887 information if the loop was predicted to iterate too many times.
9888 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9890 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9893 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9896 2018-04-13 Martin Liska <mliska@suse.cz>
9897 Jakub Jelinek <jakub@redhat.com>
9900 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9901 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9902 * builtins.c (expand_builtin_memory_copy_args): Use
9903 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9904 handle dest_addr == pc_rtx.
9906 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9909 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9910 asked to not generate direct moves.
9911 (fix_trunc<mode>si2_stfiwx): Similar.
9912 (fix_trunc<mode>si2_internal): Similar.
9914 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9917 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9918 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9919 lookup if dest in some wider mode is known to be const0_rtx and
9920 if so, record permanent equivalence for it to be ZERO_EXTEND of
9921 the narrower mode destination.
9923 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9925 * lto-streamer-out.c (output_function): Revert 259346.
9926 * omp-expand.c (expand_oacc_for): Likewise.
9928 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9930 PR rtl-optimization/85354
9931 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9932 * sel-sched.c (sel_global_init): ... here.
9934 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9937 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9938 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9939 mode for PE-COFF targets.
9940 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9941 (i386_pe_asm_lto_end): Likewise.
9942 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9943 (TARGET_ASM_LTO_END): Likewise.
9944 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9945 (i386_pe_asm_lto_start): New function.
9946 (i386_pe_asm_lto_end): Likewise.
9948 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9949 Richard Biener <rguenther@suse.de>
9952 * lto-streamer-out.c (output_function): Fix CFG loop state before
9954 * omp-expand.c (expand_oacc_for): Handle calls to internal
9955 functions like regular functions.
9957 2018-04-12 Richard Biener <rguenther@suse.de>
9960 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9961 for the early LTO debug to properly generate references to it
9962 during DIE emission. Do not re-use that for the skeleton for
9964 (dwarf2out_early_finish): Likewise.
9966 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9969 * config/i386/sse.md
9970 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9971 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9972 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9973 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9974 and output is a reg, avoid creating invalid lowpart subreg, but
9975 instead split into a 512-bit move. Don't split if not AVX512VL,
9976 input is xmm16+ reg and output is a mem.
9977 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9978 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9979 xmm16+ reg and output is a mem.
9981 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9983 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9984 also for flag_dwarf2_cfi_asm.
9986 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9988 PR rtl-optimization/85342
9989 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9990 a bool scalar var inside of the loop instead. Don't try to update
9991 recog_data.operand after failed apply_change_group.
9993 2018-04-12 Tom de Vries <tom@codesourcery.com>
9996 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9997 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9998 array with flexible array member as array without given dimension.
9999 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10000 argument for undefined param to true.
10002 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10005 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10006 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10007 from PowerPC section.
10008 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10009 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10010 help text that is too long.
10011 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10012 help text that is too long.
10013 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10014 help text that is too long.
10016 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10018 * config/alpha/alpha.md (stack_probe_internal): Rename
10019 from "probe_stack". Update all callers.
10021 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10023 PR rtl-optimization/84566
10024 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10025 sched_macro_fuse_insns.
10027 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10030 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10031 (compute_block_dependences): ... from here.
10033 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10035 PR tree-optimization/85331
10036 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10037 from int to HOST_WIDE_INT.
10039 2018-04-11 Martin Jambor <mjambor@suse.cz>
10042 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10043 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10044 not the same as the source val.
10045 (cgraph_edge_brings_value_p): New parameter.
10046 (gather_edges_for_value): Pass destination value to
10047 cgraph_edge_brings_value_p.
10048 (perhaps_add_new_callers): Likewise.
10049 (get_info_about_necessary_edges): Likewise and exclude values brought
10050 only by self-recursive edges.
10051 (create_specialized_node): Redirect only clones of self-calling edges.
10052 (+self_recursive_pass_through_p): New function.
10053 (find_more_scalar_values_for_callers_subset): Use it.
10054 (find_aggregate_values_for_callers_subset): Likewise.
10055 (known_aggs_to_agg_replacement_list): Removed.
10056 (decide_whether_version_node): Re-calculate known constants for all
10057 remaining context clones.
10059 2018-04-11 Richard Biener <rguenther@suse.de>
10062 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10063 from early DWARF output.
10064 (dwarf2out_early_finish): Output line info unconditionally into
10065 early DWARF and add reference to it.
10067 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10070 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10071 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10072 other than V2DFmode using iptr mode attribute.
10073 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10075 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10077 PR rtl-optimization/84659
10078 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10080 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10083 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10085 (output_loc_list): Pass address of a dummy size variable even in the
10086 locview handling loop.
10087 (index_location_lists): Add comment on why skip_loc_list_entry can't
10090 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10093 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10096 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10099 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10100 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10101 and -mstring-compare-inline-limit.
10103 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10106 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10107 for stack clash protection in a register whenever we need it to be in
10110 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10112 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10113 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10115 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10118 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10120 (mlong-double-): Ditto.
10121 * config/rs6000/sysv4.opt (msdata=): Ditto.
10122 (mtls-size=): Ditto.
10124 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10126 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10127 erroneous entries for
10128 "vector int vec_ldl (int, long int *)", and
10129 "vector unsigned int vec_ldl (int, unsigned long int *)".
10130 Add comments and entries for
10131 "vector bool char vec_ldl (int, bool char *)",
10132 "vector bool short vec_ldl (int, bool short *)",
10133 "vector bool int vec_ldl (int, bool int *)",
10134 "vector bool long long vec_ldl (int, bool long long *)",
10135 "vector pixel vec_ldl (int, pixel *)",
10136 "vector long long vec_ldl (int, long long *)",
10137 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10138 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10139 type tree bool_long_long_type_node and correct definition of
10140 bool_V2DI_type_node to make reference to this new type tree.
10141 (rs6000_mangle_type): Replace erroneous reference to
10142 bool_long_type_node with bool_long_long_type_node.
10143 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10144 comments to emphasize sign distinctions for char and int types and
10145 replace RS6000_BTI_bool_long constant with
10146 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10147 use of RS6000_BTI_pixel.
10148 (bool_long_type_node): Remove this macro definition.
10149 (bool_long_long_type_node): New macro definition
10151 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10153 PR rtl-optimization/85300
10154 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10155 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10156 simplify_unary_operation fails.
10158 2018-04-10 Martin Liska <mliska@suse.cz>
10160 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10161 cgraph_edge and ipa_ref.
10163 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10167 * config/i386/sse.md
10168 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10169 computation of the VEC_MERGE selector from mask.
10170 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10171 Fix decoding of the VEC_MERGE selector into mask.
10173 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10175 PR tree-optimization/85286
10176 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10178 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10180 * final.c (final_1): Set insn_last_address as well as
10181 insn_current_address.
10183 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10186 * explow.c (emit_stack_probe): Call validize_mem on memory location
10187 before passing it to gen_probe_stack. Create address operand and
10188 legitimize it for the probe_stack_address case.
10190 2018-04-09 Jan Hubicka <jh@suse.cz>
10193 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10194 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10195 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10197 (free_lang_data): Rebuild type inheritance graph.
10199 2018-04-09 Martin Sebor <msebor@redhat.com>
10201 * invoke.texi (-finline-small-functions): Mention other optimization
10203 (-findirect-inlining, -fpartial-inlining): Same.
10204 (-finline-functions-called-once): Same.
10205 (-freorder-blocks-and-partition): Same.
10207 2018-04-09 Jan Hubicka <jh@suse.cz>
10210 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10211 jumps; choose last target that matches the criteria (i.e.
10212 no partition changes for non-crossing jumps).
10213 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10214 support for redirecting crossing jumps to non-crossing.
10216 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10218 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10219 also for naked functions.
10221 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10223 * config/arc/arc.md (add_shift): New pattern.
10224 (add_shift2): Likewise.
10225 (sub_shift): Likewise.
10226 (sub_shift_cmp0_noout): Likewise.
10227 (compare_si_ashiftsi): Likewise.
10228 (xbfu_cmp0_noout): New combine pattern.
10229 (xbfu_cmp0"): Likewise.
10230 (movsi_set_cc_insn): Place the predicable variant first.
10231 (commutative_binary_cmp0_noout): Remove clobber.
10232 (commutative_binary_cmp0): New pattern.
10233 (noncommutative_binary_cmp0): Likewise.
10234 (noncommutative_binary_cmp0_noout): Likewise.
10235 (noncommutative_binary_comparison_result_used): Removed.
10236 (rsub_cmp0): New pattern.
10237 (rsub_cmp0_noout): Likewise.
10238 (extzvsi): Changed, keep only meaningful variants.
10239 (SQH, SEZ): New iterators.
10240 (SQH_postfix): New mode attribute.
10241 (SEZ_prefix): New code attribute.
10242 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10243 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10244 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10245 of numerical value.
10246 (noncommutative_operator): Check the availability of barrel
10249 2018-04-09 Richard Biener <rguenther@suse.de>
10251 PR tree-optimization/85284
10252 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10253 Only use the niter constraining form of simple_iv when the exit
10254 is always executed.
10256 2018-04-09 Tom de Vries <tom@codesourcery.com>
10259 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10260 (define_expand "*memory_barrier"): New define_expand.
10261 (define_insn "memory_barrier"): New insn.
10263 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10265 PR rtl-optimization/80463
10266 PR rtl-optimization/83972
10267 PR rtl-optimization/83480
10269 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10270 correct producer for the insn.
10271 (tidy_control_flow): Fixup seqnos in case of debug insns.
10273 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10275 PR rtl-optimization/83913
10277 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10278 different sched-times when merging exprs.
10280 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10282 PR rtl-optimization/83962
10284 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10285 tidy_fallthru_edge and tidy_control_flow.
10287 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10289 PR rtl-optimization/83530
10291 * sel-sched.c (force_next_insn): New global variable.
10292 (remove_insn_for_debug): When force_next_insn is true, also leave only
10293 next insn in the ready list.
10294 (sel_sched_region): When the region wasn't scheduled, make another pass
10295 over it with force_next_insn set to 1.
10297 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10299 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10301 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10302 for interrupt control.
10303 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10304 functions for interrupt control.
10305 * config/nds32/nds32-intrinsic.md: Likewise.
10306 * config/nds32/nds32_intrinsic.h: Likewise.
10307 * config/nds32/nds32.h (nds32_builtins): Likewise.
10309 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10311 * config/nds32/nds32.c (nds32_init_machine_status,
10312 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10313 strict_aligned_p field.
10314 (nds32_expand_to_rtl_hook): New function.
10315 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10316 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10318 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10319 Chung-Ju Wu <jasonwucj@gmail.com>
10321 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10322 * config/nds32/nds32-n7.md: New file.
10323 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10324 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10326 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10327 * config/nds32/nds32.md (pipeline_model): Add n7.
10328 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10329 * config/nds32/pipelines.md: Include n7 settings.
10331 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10332 Chung-Ju Wu <jasonwucj@gmail.com>
10334 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10335 * config/nds32/nds32-e8.md: New file.
10336 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10337 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10339 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10340 * config/nds32/nds32.md (pipeline_model): Add e8.
10341 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10342 * config/nds32/pipelines.md: Include e8 settings.
10344 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10345 Chung-Ju Wu <jasonwucj@gmail.com>
10347 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10348 * config/nds32/nds32-n8.md: New file.
10349 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10350 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10352 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10353 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10354 * config/nds32/nds32.md (pipeline_model): Add n8.
10355 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10356 * config/nds32/pipelines.md: Include n8 settings.
10358 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10359 Chung-Ju Wu <jasonwucj@gmail.com>
10361 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10362 * config/nds32/nds32-n9-2r1w.md: New file.
10363 * config/nds32/nds32-n9-3r2w.md: New file.
10364 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10365 nds32_register_ports): New or modify for cpu n9.
10366 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10368 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10369 * config/nds32/nds32-utils.c: New file.
10370 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10371 TARGET_MUL_SLOW): Define.
10372 * config/nds32/nds32.md (pipeline_model): New attribute.
10373 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10374 New options that support cpu n9.
10375 * config/nds32/pipelines.md: Include n9 settings.
10376 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10378 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10380 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10381 information if necessary.
10382 (output_cond_branch_compare_zero): Likewise.
10383 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10384 (nds32_target_alignment): Refine for alignment.
10385 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10386 (FUNCTION_BOUNDARY): Modify.
10387 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10389 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10391 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10393 * config/nds32/constants.md (unspec_volatile_element): Add values for
10394 TLB operation and data prefetch.
10395 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10396 functions for TLB operation and data prefetch.
10397 * config/nds32/nds32-intrinsic.md: Likewise.
10398 * config/nds32/nds32_intrinsic.h: Likewise.
10399 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10400 (nds32_print_operand): Likewise.
10401 * config/nds32/nds32.h (nds32_builtins): Likewise.
10403 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10404 Andrew Pinski <pinsika@gcc.gnu.org>
10406 PR middle-end/82976
10407 * match.pd: Use constant_boolean_node of correct type instead of
10408 boolean_true_node or boolean_false_node for simplifying
10409 pointer comparisons to zero.
10411 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10413 PR tree-optimization/80021
10414 * tree.c (verify_type_variant): Make error call in verify_variant_match
10415 translatable and remove final full stop.
10417 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10419 * config/nds32/constants.md (unspec_volatile_element): Add
10420 UNSPEC_VOLATILE_EH_RETURN.
10421 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10422 nds32_output_stack_pop): Support dwarf exception handling process.
10423 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10424 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10425 exception handling process.
10426 (nds32_compute_stack_frame): Likewise.
10427 (nds32_return_addr_rtx): Likewise.
10428 (nds32_initial_elimination_offset): Likewise.
10429 (nds32_expand_prologue): Likewise.
10430 (nds32_expand_epilogue): Likewise.
10431 (nds32_dynamic_chain_address): New function.
10432 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10433 exception handling.
10434 (DYNAMIC_CHAIN_ADDRESS): Define.
10435 (EH_RETURN_DATA_REGNO): Define.
10436 (EH_RETURN_STACKADJ_RTX): Define.
10437 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10438 patterns for dwarf exception handling.
10440 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10442 * config/nds32/nds32.h: Clean up obsolete macros.
10444 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10446 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10447 Add enum values for particular instructions.
10448 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10449 particular intrinsic functions.
10450 * config/nds32/nds32-intrinsic.md: Likewise.
10451 * config/nds32/nds32_intrinsic.h: Likewise.
10452 * config/nds32/nds32.h (nds32_builtins): Likewise.
10453 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10454 (btst, ave): New patterns for particular instructions.
10456 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10458 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10459 Add enum values for atomic load/store and memory sync.
10460 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10462 * config/nds32/nds32-intrinsic.md: Likewise.
10463 * config/nds32/nds32_intrinsic.h: Likewise.
10464 * config/nds32/nds32.h (nds32_builtins): Likewise.
10466 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10468 PR tree-optimization/85257
10469 * fold-const.c (native_encode_vector): If not all elts could fit
10470 and off is -1, return 0 rather than offset.
10471 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10472 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10473 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10474 adjust buffer in native_interpret_expr call.
10476 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10478 * config/nds32/constants.md (unspec_volatile_element): Add cache
10479 control enum values.
10480 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10481 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10482 * config/nds32/nds32.c (nds32_cctl_names): New.
10483 (nds32_print_operand): Handle cache control register names.
10484 * config/nds32/nds32.h (nds32_builtins): New enum values.
10485 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10487 * config/nds32/nds32.md (type): Add mmu.
10488 * config/nds32/pipelines.md (simple_insn): Add mmu.
10490 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10492 * config/nds32/nds32.md (type): Remove call.
10493 * config/nds32/pipelines.md (simple_insn): Likewise.
10495 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10497 * config/nds32/constants.md (unspec_volatile_element): Add
10498 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10499 UNSPEC_VOLATILE_FMFCFG.
10500 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10501 description for fmfcfg and fmfcsr.
10502 (bdesc_1arg): Add fmtcsr.
10503 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10504 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10505 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10506 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10507 unspec_fmfcfg): New patterns.
10508 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10509 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10510 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10511 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10512 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10513 __nds32__fmfcfg): Define.
10515 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10517 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10518 intrinsic register names.
10519 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10520 intrinsic register enum values and macros.
10522 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10524 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10525 for load/store addressing form.
10526 (nds32_print_operand_address): Likewise.
10528 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10531 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10532 based on LABEL_REF. Remove useless assertion.
10533 (pic_address_needs_scratch): Fix formatting.
10534 (sparc_legitimize_pic_address): Minor tweaks.
10535 (sparc_delegitimize_address): Adjust assertion accordingly.
10536 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10537 into symbolic_operand.
10538 (movsi_high_pic_label_ref): Likewise.
10539 (movsi_lo_sum_pic_label_ref): Likewise.
10540 (movdi_pic_label_ref): Likewise.
10541 (movdi_high_pic_label_ref): Likewise.
10542 (movdi_lo_sum_pic_label_ref): Likewise.
10544 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10546 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10547 custom LIB_SPEC setup.
10549 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10550 Kito Cheng <kito.cheng@gmail.com>
10552 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10553 * config/riscv/freebsd.h: New.
10555 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10557 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10558 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10561 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10562 Kito Cheng <kito.cheng@gmail.com>
10564 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10565 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10566 * config/nds32/nds32-protos.h (nds32_output_call,
10567 nds32_output_return): Declare.
10568 * config/nds32/nds32.md: Refine all the call and return patterns.
10570 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10573 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10574 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10576 PR rtl-optimization/84872
10577 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10578 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10579 EDGE_CROSSING edge.
10581 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10583 * expr.c (copy_blkmode_to_reg): Revert 254862.
10584 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10586 2018-04-06 Richard Biener <rguenther@suse.de>
10588 PR middle-end/85244
10589 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10590 after seeing a component reference with an adjacent field. Treat
10591 refs to arrays at struct end of external decls similar to
10592 refs to unconstrained commons.
10594 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10597 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10598 look through SAVE_EXPRs with non-side-effects argument. Adjust
10600 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10603 2018-04-06 Richard Biener <rguenther@suse.de>
10605 PR middle-end/85180
10606 * alias.c (find_base_term): New wrapper around find_base_term
10607 unwinding CSELIB_VAL_PTR changes.
10608 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10611 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10613 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10615 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10616 constant definitions.
10617 ("nop"): lr 0,0 -> nopr r0
10618 ("nop_lr0", "nop_lr1"): New insn definitions.
10620 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10622 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10623 NDS32_V3PUSH_AVAILABLE_P macro.
10625 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10626 Chung-Ju Wu <jasonwucj@gmail.com>
10628 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10629 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10630 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10631 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10632 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10633 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10634 * config/nds32/constraints.md: New constraints and checking for hard
10635 float configuration.
10636 * config/nds32/iterators.md: New mode iterator and attribute for hard
10637 float configuration.
10638 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10640 * config/nds32/nds32-fpu.md: New file.
10641 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10642 deal with hard float code generation.
10643 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10645 (abi_type, float_reg_number): New enum type.
10646 * config/nds32/nds32-predicates.c: New predicates for hard float.
10647 * config/nds32/nds32-protos.h: Declare functions for hard float.
10648 * config/nds32/nds32.c: Implementation for hard float configuration.
10649 * config/nds32/nds32.h: Definitions for hard float configuration.
10650 * config/nds32/nds32.md: Include hard float machine description and
10651 modify patterns for hard float configuration.
10652 * config/nds32/nds32.opt: New options for hard float configuration.
10653 * config/nds32/predicates.md: New predicates for hard float
10656 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10658 * common/config/nds32/nds32-common.c
10659 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10661 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10663 PR middle-end/85195
10664 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10665 CONSTRUCTOR_ELT (ctor, ...)->value.
10667 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10670 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10672 2018-04-05 Tom de Vries <tom@codesourcery.com>
10675 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10678 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10679 Kito Cheng <kito.cheng@gmail.com>
10681 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10682 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10683 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10684 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10686 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10687 Kito Cheng <kito.cheng@gmail.com>
10689 * config/nds32/constraints.md (Ufe): New memory constraint.
10690 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10691 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10692 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10694 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10695 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10697 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10699 * config/nds32/nds32.md: Use optimize_size in the condition for
10700 alu-shift instructions.
10702 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10704 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10706 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10708 * config/nds32/nds32.md (negsi2): Refine pattern.
10710 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10711 Chung-Ju Wu <jasonwucj@gmail.com>
10713 * config/nds32/iterators.md (shift_rotate): New code iterator.
10714 (shift): New code attribute.
10715 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10716 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10717 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10718 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10719 bit-wise operations.
10720 (andsi3, *andsi3): Ditto.
10721 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10722 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10723 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10724 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10725 nds32_ior_operand, nds32_xor_operand): New predicates.
10727 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10729 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10730 (addsi3, subsi3): ... this.
10732 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10734 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10736 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10738 * config/nds32/nds32.md: Adjust indention.
10740 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10742 * config/nds32/nds32.md (feature): New attribute.
10744 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10746 * config/nds32/nds32.md (subtype): New attribute.
10748 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10751 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10752 expansion to perform a bitwise AND of the argument followed by a
10753 boolean negation of the result.
10755 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10757 PR rtl-optimization/84878
10758 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10759 the basic block. Assert the use reference is not artificial and that
10760 it has an associated insn.
10762 2018-04-04 Michael Matz <matz@suse.de>
10764 * builtins.c (compute_objsize): Pass correct operand
10765 to array_at_struct_end_p.
10767 2018-04-04 Richard Biener <rguenther@suse.de>
10770 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10771 from contexts for DINFO_LEVEL_TERSE and below.
10773 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10775 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10776 resiter_operand condition.
10777 * config/nds32/nds32.md (*move<mode>): Ditto.
10779 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10780 Monk Chiang <sh.chiang04@gmail.com>
10782 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10784 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10786 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10788 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10789 Kito Cheng <kito.cheng@gmail.com>
10791 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10792 nds32_cond_code_str, output_cond_branch,
10793 output_cond_branch_compare_zero, nds32_expand_cbranch,
10794 nds32_expand_cstore, nds32_expand_movcc,
10795 nds32_output_cbranchsi4_equality_zero,
10796 nds32_output_cbranchsi4_equality_reg,
10797 nds32_output_cbranchsi4_equality_reg_or_const_int,
10798 nds32_output_cbranchsi4_greater_less_zero: New functions.
10799 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10800 nds32_expand_cstore, nds32_expand_movcc,
10801 nds32_output_cbranchsi4_equality_zero,
10802 nds32_output_cbranchsi4_equality_reg,
10803 nds32_output_cbranchsi4_equality_reg_or_const_int,
10804 nds32_output_cbranchsi4_greater_less_zero): Declare.
10805 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10806 nds32_rimm11s_operand): New predicates.
10807 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10808 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10811 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10813 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10814 * config/nds32/nds32.md: Ditto.
10815 * config/nds32/pipelines.md: Ditto.
10817 2018-04-04 Richard Biener <rguenther@suse.de>
10819 PR tree-optimization/85168
10820 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10821 propagating abnormals.
10823 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10825 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10827 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10828 Kito Cheng <kito.cheng@gmail.com>
10830 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10831 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10832 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10833 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10834 * config/nds32/nds32.md (sibcall_internal): New.
10835 (sibcall_register): Remove.
10836 (sibcall_immediate): Remove.
10837 (sibcall_value_internal): New.
10838 (sibcall_value_register): Remove.
10839 (sibcall_value_immediate): Remove.
10840 * config/nds32/predicates.md (nds32_general_register_operand): New.
10841 (nds32_call_address_operand): New.
10843 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10845 PR rtl-optimization/85167
10846 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10847 bb_defs if *split_p, instead preinitialize it to NULL.
10849 PR tree-optimization/85156
10850 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10851 evaluating the argument multiple times.
10853 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10855 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10857 (_mm_cvtpd_ps): Likewise.
10858 (_mm_cvttpd_epi32): Likewise.
10859 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10860 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10861 vector, pixel, and bool following altivec.h include.
10863 2018-04-03 Martin Sebor <msebor@redhat.com>
10865 * doc/extend.texi (Common Function Attributes): Clarify.
10866 (const attribute): Likewise.
10867 (pure attribute): Likewise.
10869 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10872 * config/i386/i386.c (ix86_expand_vector_set): Use
10873 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10875 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10877 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10878 instructions when changing rounding bits to preserve precision bits
10879 in the x87 control word.
10881 2018-04-03 Martin Liska <mliska@suse.cz>
10883 PR tree-optimization/82491
10884 * rtl.h (strip_offset_and_add): Replace += suboffset with
10885 poly_uint64 () + suboffset.
10887 2018-03-29 Martin Liska <mliska@suse.cz>
10888 Martin Jambor <mjambor@suse.cz>
10891 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10892 param_type is not an integral or pointer type.
10894 2018-04-03 Richard Biener <rguenther@suse.de>
10896 * sese.h (recompute_all_dominators): Remove.
10898 2018-04-02 Martin Sebor <msebor@redhat.com>
10900 * doc/invoke.texi (-Wrestrict): Fix typos.
10902 2018-04-02 Jim Wilson <jimw@sifive.com>
10904 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10905 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10906 (<optab>di3, <optab>si3_extend): Likewise.
10907 (<optab>si3_mask, <optab>si3_mask_1): New.
10908 (<optab>di3_mask, <optab>di3_mask_1): New.
10909 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10910 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10911 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10913 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10915 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10918 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10920 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10921 (nds32_canonicalize_comparison): New function.
10923 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10924 Kito Cheng <kito.cheng@gmail.com>
10925 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10927 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10928 * config/nds32/constants.md (unspec_volatile_element): Add
10929 UNSPEC_VOLATILE_RELAX_GROUP.
10930 * config/nds32/nds32-relax-opt.c: New file.
10931 * config/nds32/nds32-predicates.c
10932 (nds32_symbol_load_store_p): New function.
10933 * config/nds32/nds32-protos.h
10934 (nds32_symbol_load_store_p): Declare function.
10935 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10936 * config/nds32/nds32.c
10937 (nds32_register_pass): New function to register pass.
10938 (nds32_register_passes): New function to register passes.
10939 * config/nds32/nds32.md (relax_group): New pattern.
10940 * config/nds32/nds32.opt (mrelax-hint): New option.
10941 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10943 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10945 * config/nds32/t-nds32: Modify files dependency.
10947 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10949 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10950 (PROFILE_HOOK): Define its implementation.
10952 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10954 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10955 type and 32-bit size.
10957 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10959 PR middle-end/85090
10960 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10961 (V_128_256): New mode iterator.
10962 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10963 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10964 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10966 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10967 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10969 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10972 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10973 NaN inputs correctly.
10975 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10978 * config/rs6000/vsx.md (??r): New mode attribute.
10979 (*vsx_mov<mode>_64bit): Use it.
10980 (*vsx_mov<mode>_32bit): Likewise.
10982 2018-03-30 Martin Sebor <msebor@redhat.com>
10984 PR tree-optimization/84818
10985 * builtins.c (check_access): Use warning_n.
10987 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10990 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10992 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10995 2018-03-30 Julia Koval <julia.koval@intel.com>
10998 * x86-tune.def (movx, partial_reg_dependency): Enable for
11001 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11003 PR inline-asm/84985
11004 * lra-constraints.c (process_alt_operands): Move setting
11005 this_alternative_matches below.
11007 2018-03-29 Martin Liska <mliska@suse.cz>
11010 * doc/invoke.texi: Document how LTO works with debug info.
11011 Describe auto-load support of binutils. Mention 'x86-64'
11012 as valid option value of -march option.
11014 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11016 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11019 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11020 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11021 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11024 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11027 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11028 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11029 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11030 for RS6000_BTM_POWERPC64.
11031 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11032 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11033 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11036 (DIVDEU): Likewise.
11038 2018-03-28 Carl Love <cel@us.ibm.com>
11041 2017-09-27 Carl Love <cel@us.ibm.com>
11043 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11044 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11045 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11048 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11050 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11051 instead of __vector bool.
11052 (_mm_max_pu8): Likewise.
11053 (_mm_min_pi16): Likewise.
11055 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11058 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11059 (DIVWEUO): Likewise.
11060 (DIVDEO): Likewise.
11061 (DIVDEUO): Likewise.
11062 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11063 DIVWEUO and DIVDEUO.
11064 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11065 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11066 (div_extend): Likewise.
11067 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11069 (__builtin_divweuo): Likewise.
11070 (__builtin_divdeo): Likewise.
11071 (__builtin_divdeuo): Likewise.
11073 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11076 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11077 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11079 PR tree-optimization/82004
11080 * gimple-match-head.c (optimize_pow_to_exp): New function.
11081 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11082 Don't fold to exp if optimize_pow_to_exp is false.
11084 2018-03-28 Martin Liska <mliska@suse.cz>
11087 * calls.c (initialize_argument_information): Fix trailing space.
11088 * common.opt: Fix typo and provide better explanation for
11089 -fsanitize-coverage option.
11090 * config/i386/i386.opt: Fix typo.
11092 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11093 Martin Liska <mliska@suse.cz>
11096 * gimplify.c (asan_poison_variable): Don't do the check for
11097 gimplify_omp_ctxp here.
11098 (gimplify_decl_expr): Do it here.
11099 (gimplify_target_expr): Likewise.
11101 2018-03-28 Martin Liska <mliska@suse.cz>
11104 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11105 chkp_type_bounds_count if MPX is not enabled.
11107 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11109 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11111 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11114 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11115 function to create the function decl for complex long double
11116 multiply and divide for -mabi=ieeelongdouble.
11117 (init_float128_ieee): Call it.
11119 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11122 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11123 -fcf-protection=branch -mibt.
11124 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11126 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11129 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11130 arm_word_relocations.
11132 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11135 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11136 extern array declarations.
11138 2018-03-27 Richard Biener <rguenther@suse.de>
11140 PR middle-end/84067
11141 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11142 explicit single_use checks.
11144 2018-03-27 Richard Biener <rguenther@suse.de>
11146 PR tree-optimization/85082
11147 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11150 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11152 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11153 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11154 Turn on fasynchronous-unwind-tables and funwind-tables.
11156 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11159 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11160 (*bmi_blsr_<mode>_ccz): Ditto.
11162 2018-03-26 Tom de Vries <tom@codesourcery.com>
11164 PR tree-optimization/85063
11165 * omp-general.c (offloading_function_p): New function. Factor out
11167 * omp-offload.c (pass_omp_target_link::gate): ... here.
11168 * omp-general.h (offloading_function_p): Declare.
11169 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11170 with attribute omp declare target for offloading functions.
11172 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11174 PR tree-optimization/84005
11175 * tree-data-ref.h (get_base_for_alignment): Declare.
11176 * tree-data-ref.c (get_base_for_alignment_1): New function.
11177 (get_base_for_alignment): Likewise.
11178 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11179 get_base_for_alignment to find a suitable base object, instead
11180 of always using drb->base_address.
11182 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11184 PR inline-asm/85022
11185 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11186 known size by default.
11188 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11190 PR inline-asm/85030
11191 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11192 and non BLKmode operands.
11194 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11197 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11198 Clean up attributes.
11200 2018-03-23 Richard Biener <rguenther@suse.de>
11203 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11204 we are going to emit early debug for LTO.
11206 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11208 PR inline-asm/85034
11209 * function.c (match_asm_constraints_1): Don't optimize if input
11210 doesn't satisfy general_operand predicate for output's mode.
11212 PR inline-asm/85022
11213 * alias.c (write_dependence_p): Don't require for x_canonicalized
11214 non-VOIDmode if x has VOIDmode.
11217 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11218 just don't try to optimize it rather than assert it never happens.
11220 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11222 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11223 macro expansions for definition of ST_INTERNAL_<mode> and
11224 LD_INTERNAL_<mode> builtins.
11225 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11227 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11229 (altivec_expand_st_builtin): Likewise.
11230 (altivec_expand_builtin): Remove calls to deleted functions.
11231 (rs6000_address_for_altivec): Delete this function.
11232 * config/rs6000/vector.md: Remove expands for
11233 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11235 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11238 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11239 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11240 re-computing once computed.
11241 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11242 (arm_init_machine_status): Initialize
11243 machine->static_chain_stack_bytes.
11245 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11248 * doc/extend.texi: Add four new prototypes for vec_ld.
11249 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11250 definitions for more logical presentation.
11251 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11252 entries for V1TI variants of __builtin_altivec_ld builtin.
11253 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11254 handling of V1TI variant of LVX icode pattern.
11255 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11256 (rs6000_gimple_fold_builtin): Likewise.
11257 (altivec_init_builtins): Add code to define
11258 __builtin_altivec_lvx_v1ti function.
11260 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11262 PR inline-asm/84941
11263 * function.c (match_asm_constraints_1): Don't do the optimization
11264 if input isn't a REG, SUBREG, MEM or constant.
11266 2018-03-22 Tom de Vries <tom@codesourcery.com>
11268 PR tree-optimization/84956
11269 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11270 bb_has_abnormal_pred.
11272 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11275 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11276 DECL_INITIAL (decl) to decl at the end.
11277 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11278 adjust the comment.
11280 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11282 * doc/extend.texi (__builtin_tgmath): Document when complex
11283 integer types are treated as _Complex _Float64.
11285 2018-03-21 Tom de Vries <tom@codesourcery.com>
11287 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11289 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11291 PR tree-optimization/84960
11292 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11293 if it is ENTRY block, move them into single succ of ENTRY in that case.
11295 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11297 PR tree-optimization/84811
11298 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11299 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11300 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11301 (known_subrange_p): Update accordingly. Cast each value involved
11302 in the size comparison, rather than casting the result of the
11305 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11307 PR tree-optimization/84982
11308 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11309 by flipping the least significant bit rather than all bits from
11310 bitpos to bitpos + bitsize - 1.
11312 2018-03-21 Nathan Sidwell <nathan@acm.org>
11314 * doc/extend.texi (Deprecated Features): Remove mention of
11315 long-deleted deprecations.
11317 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11320 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11321 * configure: Regenerate.
11323 2018-03-21 Tom de Vries <tom@codesourcery.com>
11325 PR tree-optimization/83126
11326 * tree-parloops.c (num_phis): New function.
11327 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11329 2018-03-21 Nathan Sidwell <nathan@acm.org>
11331 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11332 mention anon-struct/union members and trailing attributes.
11334 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11336 PR tree-optimization/84969
11337 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11338 builtin memset partitions if they set different rhs values.
11340 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11342 PR rtl-optimization/84989
11343 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11344 VEC_DUPLICATE with scalar result mode.
11346 2018-03-21 Martin Liska <mliska@suse.cz>
11349 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11350 not intended return statement.
11352 2018-03-21 Martin Liska <mliska@suse.cz>
11355 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11356 (chkp_find_bound_slots_1): Limit number of iterations.
11358 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11361 * Minor grammar fixes for x86 options.
11363 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11366 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11367 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11369 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11372 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11373 (altivec_lvx_<mode>_1op): Likewise.
11374 (altivec_stvx_<mode>_2op): Likewise.
11375 (altivec_stvx_<mode>_1op): Likewise.
11376 (altivec_lvx_<VM2:mode>): New define_expand.
11377 (altivec_stvx_<VM2:mode>): Likewise.
11378 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11379 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11380 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11381 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11382 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11383 (rs6000_gen_lvx): Likewise.
11384 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11385 (altivec_expand_stv_builtin): Likewise.
11386 (altivec_expand_builtin): Likewise.
11387 * config/rs6000/vector.md: Likewise.
11389 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11392 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11395 2018-03-20 Richard Biener <rguenther@suse.de>
11398 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11399 sign-conversions as zero, fall back to standard scalar_stmt
11402 2018-03-20 Martin Liska <mliska@suse.cz>
11405 * predict.c (rebuild_frequencies): Handle case when we have
11406 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11408 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11411 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11412 flag_section_anchors.
11413 * varasm.c (use_blocks_for_decl_p): Remove hack for
11414 dw2_force_const_mem.
11417 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11419 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11420 be created, use lowpart_subreg of operands[0] rather than operands[0]
11422 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11423 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11424 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11425 and n constraint instead of aarch64_shift_imm_di and Usd.
11426 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11427 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11429 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11432 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11433 to favor GPR over NEON registers.
11434 (<shift>di3_neon): Likewise.
11436 2018-03-20 Tom de Vries <tom@codesourcery.com>
11439 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11440 (nvptx_process_pars): Emit bar.sync asap and alap.
11442 2018-03-20 Tom de Vries <tom@codesourcery.com>
11445 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11446 seen_label if seen_label is already set.
11448 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11451 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11452 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11453 Use 1U instead of 1. Formatting fixes.
11456 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11457 instead of TREE_TYPE (s1) for the return value.
11459 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11461 PR tree-optimization/84946
11462 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11463 bitsize + bitsize in poly_uint64 rather than poly_int64.
11466 * dwarf2asm.c: Include fold-const.c.
11467 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11468 of decl rather than decl itself.
11470 PR rtl-optimization/84643
11471 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11473 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11476 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11477 calling assemble_variable.
11479 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11482 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11483 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11485 2018-03-19 Jim Wilson <jimw@sifive.com>
11488 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11489 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11490 (riscv_first_stack_step): Likewise.
11491 (riscv_option_override): Use STACK_BOUNDARY instead of
11492 MIN_STACK_BOUNDARY.
11493 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11494 MIN_STACK_BOUNDARY.
11495 (BIGGEST_ALIGNMENT): Set to 128.
11496 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11497 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11500 2018-03-19 Richard Biener <rguenther@suse.de>
11502 PR tree-optimization/84933
11503 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11504 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11506 2018-03-19 Richard Biener <rguenther@suse.de>
11508 PR tree-optimization/84859
11509 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11510 (cond_if_else_store_replacement): Perform sinking operation on
11511 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11512 Generalize what a BB with a single eligible store is.
11514 2018-03-19 Richard Biener <rguenther@suse.de>
11516 PR tree-optimization/84929
11517 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11518 chrec_is_positive against non-chrec arg.
11520 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11523 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11525 2018-03-18 Martin Liska <mliska@suse.cz>
11527 PR rtl-optimization/84635
11528 * regrename.c (build_def_use): Use matches_mode only when
11531 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11533 PR tree-optimization/84913
11534 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11535 vectorize chains of COND_EXPRs.
11537 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11539 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11541 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11543 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11545 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11547 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11549 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11550 Kito Cheng <kito.cheng@gmail.com>
11552 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11553 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11554 (nds32_adjust_reg_alloc_order): New function.
11555 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11557 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11559 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11560 nds32_print_operand, nds32_print_operand_address): Use
11561 HOST_WIDE_INT_PRINT_DEC instead.
11563 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11565 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11567 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11570 * config/i386/i386.c (initial_ix86_tune_features,
11571 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11572 unsigned long long.
11573 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11574 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11575 rather than 1u << ix86_tune. Formatting fix.
11576 (ix86_option_override_internal): Change ix86_arch_mask from
11577 unsigned int to unsigned HOST_WIDE_INT, initialize to
11578 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11579 (ix86_function_specific_restore): Likewise.
11581 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11584 * postreload.c (reload_combine_recognize_pattern): Perform
11585 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11586 truncate_int_for_mode the result for the destination's mode.
11589 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11590 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11592 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11595 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11598 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11599 regno_allocno_class_array and sorted_pseudos.
11600 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11601 insns where regno is used.
11603 2018-03-16 Martin Liska <mliska@suse.cz>
11606 * multiple_target.c (create_dispatcher_calls): Redirect
11607 reference in the symbol table.
11609 2018-03-16 Martin Liska <mliska@suse.cz>
11612 * multiple_target.c (create_dispatcher_calls): Redirect also
11615 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11619 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11620 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11621 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11623 2018-03-16 Julia Koval <julia.koval@intel.com>
11625 * doc/invoke.texi (Skylake Server): Add CLWB.
11626 Cannonlake): Remove CLWB.
11628 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11630 PR tree-optimization/84841
11631 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11633 (FLOAT_ONE_CONST_TYPE): Define.
11634 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11635 (sort_by_operand_rank): Put entries with higher constant_type last
11636 rather than first to match comments.
11638 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11640 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11643 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11646 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11647 check and use address of target always.
11649 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11652 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11653 (indirect_thunk_bnd_needed): Likewise.
11654 (indirect_thunks_used): Likewise.
11655 (indirect_thunks_bnd_used): Likewise.
11656 (indirect_return_needed): New.
11657 (indirect_return_bnd_needed): Likewise.
11658 (output_indirect_thunk_function): Add a bool argument for
11660 (output_indirect_thunk_function): Don't generate alias for
11661 function return thunk.
11662 (ix86_code_end): Call output_indirect_thunk_function to generate
11663 function return thunks.
11664 (ix86_output_function_return): Set indirect_return_bnd_needed
11665 and indirect_return_needed instead of indirect_thunk_bnd_needed
11666 and indirect_thunk_needed.
11668 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11670 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11671 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11672 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11674 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11675 Paul Hua <paul.hua.gm@gmail.com>
11678 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11680 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11682 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11683 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11684 resp. SFmode cases.
11686 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11689 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11690 instead of GET_MODE_SIZE when comparing Units.
11692 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11695 * varasm.c (hash_section): Return an unchangeble hash value
11696 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11697 Return !aarch64_can_use_per_function_literal_pools_p ().
11699 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11702 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11703 rather than address of cmode as last argument to prepare_cmp_insn.
11705 2018-03-15 Julia Koval <julia.koval@intel.com>
11707 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11708 F_AVX512VNNI, F_AVX512BITALG): New.
11710 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11713 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11714 insn for floating-point loads and stores.
11716 2018-03-14 Carl Love <cel@us.ibm.com>
11718 * config/rs6000/rs6000-c.c: Add macro definitions for
11719 ALTIVEC_BUILTIN_VEC_PERMXOR.
11720 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11721 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11722 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11723 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11725 * config/doc/extend.texi: Add prototypes for vec_permxor.
11727 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11730 * diagnostic-show-locus.c (class layout_point): Convert m_line
11731 from int to linenum_type.
11732 (line_span::comparator): Use linenum "compare" function when
11733 comparing line numbers.
11734 (test_line_span): New function.
11735 (layout_range::contains_point): Convert param "row" from int to
11737 (layout_range::intersects_line_p): Likewise.
11738 (layout::will_show_line_p): Likewise.
11739 (layout::print_source_line): Likewise.
11740 (layout::should_print_annotation_line_p): Likewise.
11741 (layout::print_annotation_line): Likewise.
11742 (layout::print_leading_fixits): Likewise.
11743 (layout::annotation_line_showed_range_p): Likewise.
11744 (struct line_corrections): Likewise for field m_row.
11745 (line_corrections::line_corrections): Likewise for param "row".
11746 (layout::print_trailing_fixits): Likewise.
11747 (layout::get_state_at_point): Likewise.
11748 (layout::get_x_bound_for_row): Likewise.
11749 (layout::print_line): Likewise.
11750 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11751 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11752 * input.c (selftest::test_linenum_comparisons): New function.
11753 (selftest::input_c_tests): Call it.
11754 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11755 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11756 * selftest.h (ASSERT_GT): New macro.
11757 (ASSERT_GT_AT): New macro.
11758 (ASSERT_LT): New macro.
11759 (ASSERT_LT_AT): New macro.
11761 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11763 PR rtl-optimization/84780
11764 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11766 2018-03-14 Martin Liska <mliska@suse.cz>
11768 * tree.c (record_node_allocation_statistics): Use
11769 get_stats_node_kind.
11770 (get_stats_node_kind): New function extracted from
11771 record_node_allocation_statistics.
11772 (free_node): Use get_stats_node_kind.
11774 2018-03-14 Richard Biener <rguenther@suse.de>
11776 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11777 that the value-set of ANTIC_IN doesn't grow.
11780 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11782 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11783 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11785 2018-03-14 Julia Koval <julia.koval@intel.com>
11787 * config.gcc (icelake-client, icelake-server): New.
11789 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11790 (initial_ix86_arch_features): Ditto.
11791 (PTA_SKYLAKE): Add SGX.
11792 (PTA_ICELAKE): Remove.
11793 (PTA_ICELAKE_CLIENT): New.
11794 (PTA_ICELAKE_SERVER): New.
11795 (ix86_option_override_internal): Split up icelake on icelake client and
11797 (get_builtin_code_for_version): Ditto.
11798 (fold_builtin_cpu): Ditto.
11799 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11800 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11801 * config/i386/i386.h (processor_type): Ditto.
11802 * doc/invoke.texi: Ditto.
11804 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11807 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11808 INTEGER_CST offset, add it together with bitpos / 8 and
11809 sign extend based on POINTER_SIZE.
11813 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11816 * config/i386/constraints.md (Yc): New register constraint.
11817 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11818 Use Yc constraint for alternative 2 of operand 0. Remove
11819 preferred_for_speed attribute.
11821 2018-03-14 Richard Biener <rguenther@suse.de>
11823 PR tree-optimization/84830
11824 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11825 with the old one to avoid oscillations.
11827 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11830 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11832 (assign_by_spills): Return a flag of reload assignment failure.
11833 Do not process the reload assignment failures. Do not spill other
11834 reload pseudos if they has the same reg class. Update n if
11836 (lra_assign): Add a return arg. Set up from the result of
11837 assign_by_spills call.
11838 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11839 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11840 usage_insns if it is not NULL.
11841 (spill_hard_reg_in_range): New function.
11842 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11843 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11844 function prototypes.
11845 (lra_assign): Change prototype.
11846 * lra.c (lra): Add code to deal with fails by splitting hard reg
11849 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11851 * config/riscv/riscv.opt (mrelax): New option.
11852 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11853 "norelax" when riscv_mrelax is disabled.
11854 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11856 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11859 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11860 reassociation for int modes.
11862 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11864 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11865 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11867 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11868 * config/aarch64/aarch64-sve.md
11869 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11870 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11871 (*extend<mode><Vwide>2): Rename to...
11872 (aarch64_sve_extend<mode><Vwide>2): ...this.
11873 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11874 renaming the old pattern to...
11875 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11877 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11878 define_expand, renaming the old pattern to...
11879 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11880 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11881 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11882 account when deciding which SVE instruction the optab should use.
11883 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11885 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11887 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11888 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11889 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11890 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11891 (tlsdesc_small_<mode>): Turn a define_expand and use
11892 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11893 (tlsdesc_small_advsimd_<mode>): ...this.
11894 (tlsdesc_small_sve_<mode>): New pattern.
11896 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11898 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11899 (UNSPEC_UMUL_HIGHPART): New constants.
11900 (MUL_HIGHPART): New int iteraor.
11901 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11902 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11904 (*<su>mul<mode>3_highpart): New define_insn.
11906 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11909 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11912 2018-03-13 Martin Liska <mliska@suse.cz>
11915 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11917 (sem_item_optimizer::~sem_item_optimizer): Release it.
11918 (sem_item_optimizer::merge_classes): Register variable aliases.
11919 (sem_item_optimizer::fixup_pt_set): New function.
11920 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11921 * ipa-icf.h: Declare new variables and functions.
11923 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11925 PR middle-end/84834
11926 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11927 integer_pow2p@2 and test integer_pow2p in condition.
11928 (A < 0 ? C : 0): Similarly for @1.
11930 PR middle-end/84831
11931 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11932 characters starting at p contain '\0' character, don't look beyond
11936 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11937 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11940 * reg-stack.c (change_stack): Change update_end var from int to
11941 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11942 also call set_block_for_insn on the newly added insns and rescan.
11945 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11946 on the last operand.
11949 * tree.c (stabilize_reference_1): Return save_expr (e) for
11950 STATEMENT_LIST even if it doesn't have side-effects.
11952 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11954 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11956 2018-03-12 Renlin Li <renlin.li@arm.com>
11958 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11959 aarch64_output_scalar_simd_mov_immediate.
11961 2018-03-12 Martin Sebor <msebor@redhat.com>
11963 PR tree-optimization/83456
11964 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11965 for perfectly overlapping calls to memcpy.
11966 (gimple_fold_builtin_memory_chk): Same.
11967 (gimple_fold_builtin_strcpy): Handle no-warning.
11968 (gimple_fold_builtin_stxcpy_chk): Same.
11969 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11971 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11973 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11974 parameter. Use it for SFmode.
11975 (rs6000_function_arg_advance_1): Adjust.
11976 (rs6000_function_arg): Adjust.
11977 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11979 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11981 PR rtl-optimization/84169
11982 PR rtl-optimization/84780
11983 * combine.c (can_combine_p): Check for a 2-insn combination whether
11984 the destination register is used between the two insns, too.
11986 2018-03-12 Richard Biener <rguenther@suse.de>
11988 PR tree-optimization/84803
11989 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11990 for refs DR analysis didn't process.
11992 2018-03-12 Richard Biener <rguenther@suse.de>
11994 PR tree-optimization/84777
11995 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11996 force-vectorize loops ignore whether we are optimizing for size.
11998 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12000 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12001 (TARGET_MD_ASM_ADJUST): Define.
12003 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12004 Kito Cheng <kito.cheng@gmail.com>
12005 Chung-Ju Wu <jasonwucj@gmail.com>
12007 * config/nds32/nds32.c (nds32_compute_stack_frame,
12008 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12009 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12010 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12011 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12012 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12013 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12014 * config/nds32/nds32.md (prologue, epilogue): Use macro
12015 NDS32_V3PUSH_AVAILABLE_P to do checking.
12017 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12020 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12021 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12022 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12023 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12024 addition of most attributes on !orig_type_die or the attribute not
12025 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12027 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12028 Chung-Ju Wu <jasonwucj@gmail.com>
12030 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12031 __NDS32_VH__ macro.
12032 * config/nds32/nds32.opt (mvh): New option.
12034 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12035 Chung-Ju Wu <jasonwucj@gmail.com>
12037 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12039 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12040 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12043 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12044 Chung-Ju Wu <jasonwucj@gmail.com>
12046 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12048 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12049 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12051 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12052 Kito Cheng <kito.cheng@gmail.com>
12053 Chung-Ju Wu <jasonwucj@gmail.com>
12055 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12056 UNSPEC_FFMISM and UNSPEC_FLMISM.
12057 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12058 for ffb, ffmism and flmism.
12059 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12060 (unspec_ffmism): Ditto.
12061 (unspec_flmism): Ditto.
12062 (nds32_expand_builtin_impl): Check if string extension is available.
12063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12064 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12066 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12069 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12072 * lra-assigns.c (assign_by_spills): Return a flag of reload
12073 assignment failure. Do not process the reload assignment
12074 failures. Do not spill other reload pseudos if they has the same
12076 (lra_assign): Add a return arg. Set up from the result of
12077 assign_by_spills call.
12078 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12079 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12080 usage_insns if it is not NULL.
12081 (spill_hard_reg_in_range): New function.
12082 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12083 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12084 function prototypes.
12085 (lra_assign): Change prototype.
12086 * lra.c (lra): Add code to deal with fails by splitting hard reg
12089 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12092 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12094 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12097 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12098 (dw_val_node): Add val_symbolic_view.
12099 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12100 (symview_upper_bound): New.
12101 (new_line_info_table): Initialize symviews_since_reset.
12102 (dwarf2out_source_line): Count symviews_since_reset and set
12103 symview_upper_bound.
12104 (dw_val_equal_p): Handle symview.
12105 (add_AT_symview): New.
12106 (print_dw_val): Handle symview.
12107 (attr_checksum, attr_checksum_ordered): Likewise.
12108 (same_dw_val_p, size_of_die): Likewise.
12109 (value_format, output_die): Likewise.
12110 (add_high_low_attributes): Use add_AT_symview for entry_view.
12111 (dwarf2out_finish): Reset symview_upper_bound, clear
12114 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12117 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12118 Add strict argument and use it.
12119 (rs6000_split_multireg_move): Update for new strict argument.
12120 (mem_operand_gpr): Disallow all non-offsettable addresses.
12121 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12123 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12126 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12127 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12128 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12131 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12132 decl, use remap_type if we want to use the type.
12134 2018-03-09 Martin Sebor <msebor@redhat.com>
12136 PR tree-optimization/84526
12137 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12139 (builtin_access::generic_overlap): Be prepared to handle non-array
12142 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12144 PR rtl-optimization/84682
12145 * lra-constraints.c (process_address_1): Check is_address flag
12146 for address constraints.
12147 (process_alt_operands): Likewise.
12148 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12149 preprocess_constraints.
12150 * recog.h (preprocess_constraints): Add oploc parameter.
12152 * recog.c (preprocess_constraints): Test address_operand for
12153 CT_ADDRESS constraints.
12155 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12158 * lra-assigns.c (assign_by_spills): Return a flag of reload
12159 assignment failure. Do not process the reload assignment
12160 failures. Do not spill other reload pseudos if they has the same
12162 (lra_assign): Add a return arg. Set up from the result of
12163 assign_by_spills call.
12164 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12165 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12166 usage_insns if it is not NULL.
12167 (spill_hard_reg_in_range): New function.
12168 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12169 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12170 function prototypes.
12171 (lra_assign): Change prototype.
12172 * lra.c (lra): Add code to deal with fails by splitting hard reg
12175 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12178 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12179 Accept complain bool parameter. Only emit errors if it is true.
12180 (arm_parse_cpu_option_name): Likewise.
12181 (arm_target_thumb_only): Adjust callers of the above.
12182 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12183 prototype to take a default true bool parameter.
12184 (arm_parse_arch_option_name): Likewise.
12186 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12187 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12191 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12192 * configure: Regenerate.
12193 * configure.ac ("linker --version-script option"): New.
12194 ("linker soname option"): New.
12196 2018-03-09 Richard Biener <rguenther@suse.de>
12198 PR tree-optimization/84775
12199 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12200 immediate uses of predicate stmts and mark them modified.
12203 PR tree-optimization/84178
12204 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12206 (version_loop_for_if_conversion): Delay update_ssa call.
12207 (tree_if_conversion): Delay update_ssa until after predicate
12210 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12213 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12214 when the function accesses prior frames.
12216 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12219 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12220 gen_llsym, otherwise call maybe_gen_llsym.
12222 PR inline-asm/84742
12223 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12224 has ',' character inside of it.
12226 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12229 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12230 as clobbering CC_REGNUM.
12232 2018-03-08 Richard Biener <rguenther@suse.de>
12234 PR middle-end/84552
12235 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12236 (follow_copies_to_constant): Do not follow SSA names registered
12239 2018-03-08 Richard Biener <rguenther@suse.de>
12241 PR tree-optimization/84178
12242 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12244 (version_loop_for_if_conversion): Delay update_ssa call.
12245 (tree_if_conversion): Delay update_ssa until after predicate
12248 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12250 PR tree-optimization/84178
12251 * tree-if-conv.c (release_bb_predicate): Remove the
12252 the assertion that the stmts have NULL use_ops.
12253 Discard the statements, asserting that they haven't
12254 yet been added to a BB.
12256 2018-03-08 Richard Biener <rguenther@suse.de>
12258 PR tree-optimization/84746
12259 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12260 (phi_translate): Pass in destination ANTIC_OUT set.
12261 (phi_translate_1): Likewise. For a simplified result lookup
12262 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12263 (phi_translate_set): Adjust.
12264 (do_pre_regular_insertion): Likewise.
12265 (do_pre_partial_partial_insertion): Likewise.
12267 2018-03-08 Martin Liska <mliska@suse.cz>
12269 PR gcov-profile/84735
12270 * doc/gcov.texi: Document usage of profile files.
12271 * gcov-io.h: Document changes in the format.
12273 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12277 * dwarf2out.c (struct dw_line_info_table): Update comments for
12279 (FORCE_RESET_NEXT_VIEW): New.
12280 (FORCE_RESETTING_VIEW_P): New.
12281 (RESETTING_VIEW_P): Check for -1 too.
12282 (ZERO_VIEW_P): Likewise.
12283 (new_line_info_table): Force-reset next view.
12284 (dwarf2out_begin_function): Likewise.
12285 (dwarf2out_source_line): Simplify zero_view_p initialization.
12286 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12287 view directly. Omit view when omitting .loc at line 0.
12289 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12291 PR tree-optimization/84740
12292 * tree-switch-conversion.c (process_switch): Call build_constructors
12293 only if info.phi_count is non-zero.
12295 PR tree-optimization/84739
12296 * tree-tailcall.c (find_tail_calls): Check call arguments against
12297 DECL_ARGUMENTS (current_function_decl) rather than
12298 DECL_ARGUMENTS (func) when checking for tail recursion.
12300 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12302 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12303 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12304 Volker Reichelt's entry and add entries for people that perform
12305 GCC fuzzy testing and report numerous bugs.
12307 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12310 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12311 readonly data in sdata, if that is disabled.
12312 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12313 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12314 -mreadonly-in-sdata option.
12316 2018-03-07 Martin Sebor <msebor@redhat.com>
12318 PR tree-optimization/84468
12319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12320 basic block when looking for nul assignment.
12322 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12325 * except.h (output_function_exception_table): Adjust prototype.
12326 * except.c (output_function_exception_table): Remove FNNAME parameter
12327 and add SECTION parameter. Ouput one part of the table at a time.
12328 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12329 the first part of the exception table and emit unwind directives.
12330 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12331 (i386_pe_seh_cold_init): Likewise.
12332 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12333 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12334 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12335 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12336 * config/i386/winnt.c: Include except.h.
12337 (struct seh_frame_state): Add reg_offset, after_prologue and
12338 in_cold_section fields.
12339 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12340 (i386_pe_seh_cold_init): New function.
12341 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12342 to seh->in_cold_section.
12343 (seh_emit_push): Record the offset of the push.
12344 (seh_emit_save): Record the offet of the save.
12345 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12346 Test seh->after_prologue to disregard the epilogue.
12347 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12348 (i386_pe_end_cold_function): New function.
12350 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12353 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12354 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12357 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12359 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12360 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12362 PR middle-end/84723
12363 * multiple_target.c: Include tree-inline.h and intl.h.
12364 (expand_target_clones): Diagnose and fail if node->definition and
12365 !tree_versionable_function_p (node->decl).
12367 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12369 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12371 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12372 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12373 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12375 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12378 * combine.c (try_combine): Use reg_or_subregno instead of handling
12379 just paradoxical SUBREGs and REGs.
12381 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12383 * config/arc/arc.c (arc_finalize_pic): Remove function.
12384 (arc_must_save_register): We use single base PIC register, remove
12385 checks to save/restore the PIC register.
12386 (arc_expand_prologue): Likewise.
12387 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12389 (arc_verify_short): Likewise.
12390 (arc_attr_type): Likewise.
12391 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12392 (walk_stores): Likewise.
12393 (arc_address_cost): Make it static.
12394 (arc_verify_short): Likewise.
12395 (branch_dest): Likewise.
12396 (arc_attr_type): Likewise.
12397 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12398 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12399 (arc_final_prescan_insn): Remove inserting the nops due to
12400 hardware hazards. It is done in reorg step.
12401 (insn_length_variant_t): Remove.
12402 (insn_length_parameters_t): Likewise.
12403 (arc_insn_length_parameters): Likewise.
12404 (arc_get_insn_variants): Likewise.
12405 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12407 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12409 PR inline-asm/84683
12410 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12413 PR tree-optimization/84687
12414 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12416 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12418 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12420 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12421 Rename to ppc_speculation_barrier.
12422 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12423 __builtin_ppc_speculation_barrier.
12425 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12428 * combine.c (combine_simplify_rtx): Don't try to simplify if
12429 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12432 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12434 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12435 to 32 bytes when compiling for POWER9.
12437 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12440 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12441 regparm >= 3 with no arg reg available also for calls with
12442 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12445 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12447 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12449 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12452 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12454 2018-03-05 Richard Biener <rguenther@suse.de>
12456 PR tree-optimization/84486
12457 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12458 When inserting a __builtin_assume_aligned call set the LHS
12459 SSA name alignment info accordingly.
12461 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12463 PR tree-optimization/84114
12464 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12465 Avoid reassociation of FLOAT_MODE addition.
12467 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12469 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12470 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12471 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12472 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12473 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12474 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12475 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12477 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12478 __builtin_ia32_wbinvd): New builtins.
12479 (SPECIAL_ARGS2): New.
12480 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12481 (SPECIAL_ARGS2): New.
12482 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12483 (ix86_valid_target_attribute_inner_p): Ditto.
12484 (ix86_init_mmx_sse_builtins): Add special_args2.
12485 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12486 TARGET_WBNOINVD_P): New.
12487 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12488 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12489 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12490 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12491 * config/i386/pconfigintrin.h: New file.
12492 * config/i386/wbnoinvdintrin.h: Ditto.
12493 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12495 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12497 2018-03-05 Richard Biener <rguenther@suse.de>
12499 PR tree-optimization/84670
12500 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12502 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12503 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12504 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12505 doesn't grow if all successors (recursively) were visited at least
12508 2018-03-05 Richard Biener <rguenther@suse.de>
12510 PR tree-optimization/84650
12511 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12512 if executed in the loop pipeline.
12514 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12516 * doc/configfiles.texi (Configuration Files): Move info about
12517 conditionalizing $target-protos.h to...
12518 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12519 differs from $target-protos.h.
12521 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12522 Chung-Ju Wu <jasonwucj@gmail.com>
12524 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12525 * config/nds32/nds32-multiple.md (setmemsi): Define.
12526 * config/nds32/nds32-memory-manipulation.c
12527 (nds32_gen_dup_4_byte_to_word_value): New.
12528 (emit_setmem_word_loop): New.
12529 (emit_setmem_byte_loop): New.
12530 (nds32_expand_setmem_loop): New.
12531 (nds32_expand_setmem_loop_v3m): New.
12532 (nds32_expand_setmem_unroll): New.
12533 (nds32_expand_setmem): New.
12535 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12536 Chung-Ju Wu <jasonwucj@gmail.com>
12538 * config/nds32/nds32-memory-manipulation.c
12539 (nds32_emit_load_store): New.
12540 (nds32_emit_post_inc_load_store): New.
12541 (nds32_emit_mem_move): New.
12542 (nds32_emit_mem_move_block): New.
12543 (nds32_expand_movmemsi_loop_unknown_size): New.
12544 (nds32_expand_movmemsi_loop_known_size): New.
12545 (nds32_expand_movmemsi_loop): New.
12546 (nds32_expand_movmemsi_unroll): New.
12547 (nds32_expand_movmemqi): Rename ...
12548 (nds32_expand_movmemsi): ... to this.
12549 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12550 (movmemsi): ... to this.
12551 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12552 (nds32_expand_movmemsi): ... to this.
12554 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12555 Monk Chiang <sh.chiang04@gmail.com>
12556 Chung-Ju Wu <jasonwucj@gmail.com>
12558 * config/nds32/nds32-protos.h
12559 (nds32_expand_load_multiple): New arguments.
12560 (nds32_expand_store_multiple): Ditto.
12561 (nds32_valid_multiple_load_store): Rename ...
12562 (nds32_valid_multiple_load_store_p): ... to this.
12563 * config/nds32/nds32-memory-manipulation.c
12564 (nds32_expand_load_multiple): Refine implementation.
12565 (nds32_expand_store_multiple): Ditto.
12566 * config/nds32/nds32-multiple.md
12567 (load_multiple): Update nds32_expand_load_multiple interface.
12568 (store_multiple): Update nds32_expand_store_multiple interface.
12569 * config/nds32/nds32-predicates.c
12570 (nds32_valid_multiple_load_store): Rename ...
12571 (nds32_valid_multiple_load_store_p): ... to this and refine
12573 * config/nds32/predicates.md
12574 (nds32_load_multiple_and_update_address_operation): New predicate.
12575 (nds32_store_multiple_and_update_address_operation): New predicate.
12577 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12578 Chung-Ju Wu <jasonwucj@gmail.com>
12580 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12581 (combo): New attribute.
12582 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12584 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12586 * config/nds32/nds32.opt: Change -mcmodel= default value.
12588 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12589 Monk Chiang <sh.chiang04@gmail.com>
12590 Chung-Ju Wu <jasonwucj@gmail.com>
12592 * config/nds32/constants.md (unspec_element): New enum.
12593 * config/nds32/constraints.md (Umw): New constraint.
12594 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12595 * config/nds32/nds32-intrinsic.md: Likewise.
12596 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12597 (nds32_valid_smw_lwm_base_p): New.
12598 (nds32_output_smw_single_word): New.
12599 (nds32_output_lmw_single_word): New.
12600 (nds32_expand_unaligned_load): New.
12601 (nds32_expand_unaligned_store): New.
12602 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12603 (nds32_output_smw_single_word): Declare.
12604 (nds32_output_lmw_single_word): Declare.
12605 (nds32_expand_unaligned_load): Declare.
12606 (nds32_expand_unaligned_store): Declare.
12607 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12608 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12609 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12610 NDS32_BUILTIN_UASTORE_DW.
12611 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12614 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12615 Kito Cheng <kito.cheng@gmail.com>
12616 Chung-Ju Wu <jasonwucj@gmail.com>
12618 * config/nds32/nds32-intrinsic.c
12619 (nds32_expand_builtin_null_ftype_reg): Delete.
12620 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12621 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12622 (nds32_read_argument): New.
12623 (nds32_legitimize_target): Ditto.
12624 (nds32_legitimize_argument): Ditto.
12625 (nds32_check_constant_argument): Ditto.
12626 (nds32_expand_unop_builtin): Ditto.
12627 (nds32_expand_unopimm_builtin): Ditto.
12628 (nds32_expand_binop_builtin): Ditto.
12629 (nds32_builtin_decl_impl): Ditto.
12630 (builtin_description): Ditto.
12631 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12632 (nds32_init_builtins_impl): Ditto.
12633 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12634 (nds32_builtin_decl): New.
12635 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12636 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12638 2018-03-02 Jeff Law <law@redhat.com>
12640 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12641 (redundant_insn, fill_simple_delay_slots): Likewise.
12642 (fill_slots_from_thread): Likewise.
12643 * resource.c (mark_referenced_resources): Likewise.
12644 (mark_set_resources, find_dead_or_set_registers): Likewise.
12646 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12648 * substring-locations.h (format_warning_va): Formatting fix for
12649 ATTRIBUTE_GCC_DIAG.
12650 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12652 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12653 * substring-locations.c: Include intl.h.
12654 (format_warning_va): Turned into small wrapper around
12655 format_warning_n_va, renamed to ...
12656 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12657 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12659 (format_warning_at_substring_n): New function.
12660 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12661 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12662 format_warning_at_substring with just a shorter name instead of
12663 const function pointer.
12664 (fmtwarn_n): New function.
12665 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12666 appropriate, get rid of all the fmtstr temporaries, move conditionals
12667 with G_() wrapped string literals directly into fmtwarn arguments,
12668 cast dir.len to (int), formatting fixes.
12670 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12672 * doc/invoke.texi: Remove "Cilk Plus" references.
12674 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12675 Richard Biener <rguenther@suse.de>
12678 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12679 for error or warning attributes if CALL_FROM_THUNK_P is set.
12682 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12685 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12686 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12689 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12690 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12692 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12693 instead of -1U in last predictors element's probability member.
12695 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12698 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12699 arguments if they are comparable.
12701 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12703 PR tree-optimization/84634
12704 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12705 masks and masked_loop_p with a single loop_masks, making sure it's
12706 null for bb vectorization.
12708 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12710 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12711 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12712 loop->force_vectorize to check whether there is no alias.
12714 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12717 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12719 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12721 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12722 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12723 instead of a loop around prev_real_insn.
12724 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12727 PR inline-asm/84625
12728 * config/i386/i386.c (ix86_print_operand): Use conditional
12729 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12732 2018-03-02 Richard Biener <rguenther@suse.de>
12734 PR tree-optimization/84427
12735 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12736 (bitmap_set_subtract_values): Rewrite to handle multiple
12739 (prune_clobbered_mems): Likewise.
12740 (phi_translate): Take edge instead of pred/phiblock.
12741 (phi_translate_1): Likewise.
12742 (phi_translate_set): Likewise. Insert all translated
12743 exprs for a value into the set, keeping possibly multiple
12744 expressions per value.
12745 (compute_antic_aux): Adjust for phi_translate changes.
12746 When intersecting union the expressions and prune those
12747 not in the final value set, keeping possibly multiple
12748 expressions per value. Do not use value-insertion
12749 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12750 all expressions. Add verification that the value-sets
12751 only shrink during iteration.
12752 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12753 (do_pre_regular_insertion): Likewise.
12754 (do_pre_partial_partial_insertion): Likewise.
12756 2018-03-02 Richard Biener <rguenther@suse.de>
12759 * config/darwin.c (saved_debug_info_level): New static global.
12760 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12761 (darwin_asm_lto_end): Restore debug info generation settings.
12763 2018-03-01 Martin Liska <mliska@suse.cz>
12766 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12767 volatile arguments.
12769 2018-03-01 Richard Biener <rguenther@suse.de>
12772 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12773 type annotation with LTO when debug was disabled at compile-time.
12775 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12777 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12779 (mips_final_postscan_insn): Likewise.
12781 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12783 PR rtl-optimization/84528
12784 * alias.c (init_alias_target): Add commentary.
12785 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12786 a unique base value if the frame pointer is not eliminated
12787 to the stack pointer.
12789 2018-03-01 Tom de Vries <tom@codesourcery.com>
12791 PR rtl-optimization/83327
12792 * lra-int.h (hard_regs_spilled_into): Declare.
12793 * lra.c (hard_regs_spilled_into): Define.
12794 (init_reg_info): Init hard_regs_spilled_into.
12795 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12796 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12797 (process_bb_lives): Handle hard_regs_spilled_into.
12798 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12799 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12801 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12803 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12804 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12805 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12806 * config/rs6000/aix72.h: New file.
12808 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12810 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12811 instead of warning_at with conditional singular and plural messages
12815 * stor-layout.c (update_alignment_for_field): For
12816 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12817 && !DECL_PACKED (field), do the alignment update, just use
12818 only desired_align instead of MAX (type_align, desired_align)
12820 (place_field): Don't do known_align < desired_align handling
12821 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12822 is non-NULL, instead do it after rli->prev_field handling and
12823 only if not within a bitfield word. For DECL_PACKED (field)
12824 use type_align of BITS_PER_UNIT.
12826 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12828 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12829 superfluous parentheses and trailing spaces.
12831 2018-02-28 Richard Biener <rguenther@suse.de>
12833 PR tree-optimization/84584
12834 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12835 SCoPs with fake exit edge.
12837 2018-02-28 Martin Liska <mliska@suse.cz>
12840 * timevar.c (timer::print): Fix format to properly print 100%
12843 2018-02-28 Richard Biener <rguenther@suse.de>
12845 PR middle-end/84607
12846 * genmatch.c (capture_info::walk_match): Do not mark
12847 captured expressions without operands as expr_p given
12848 they act more like predicates and should be subject to
12849 "lost tail" side-effect preserving.
12851 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12853 PR rtl-optimization/81611
12854 * auto-inc-dec.c (attempt_change): Move dead note from
12855 mem_insn if it's the next use of regno
12856 (find_address): Take address use of reg holding
12857 non-incremented value. Add parm to limit search to the named
12859 (merge_in_block): Attempt to use a mem insn that is the next
12860 use of the original regno.
12862 2018-02-27 Martin Sebor <msebor@redhat.com>
12865 * doc/invoke.texi (-Wmissing-attributes): New option.
12866 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
12868 2018-02-27 Martin Sebor <msebor@redhat.com>
12870 PR translation/84207
12871 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12872 n argument to unsigned HOST_WIDE_INT.
12873 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12874 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12875 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12876 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12878 2018-02-27 Richard Biener <rguenther@suse.de>
12880 PR tree-optimization/84512
12881 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12882 Do not use the estimate returned from record_stmt_cost for
12883 the scalar iteration cost but sum properly using add_stmt_cost.
12885 2018-02-27 Richard Biener <rguenther@suse.de>
12887 PR tree-optimization/84466
12888 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12889 Adjust last change to less strictly validate use operands.
12891 2018-02-27 Martin Liska <mliska@suse.cz>
12893 PR gcov-profile/84548
12894 * gcov.c (process_file): Allow partial overlap and consider it
12895 also as group functions.
12896 (output_lines): Properly calculate range of lines for a group.
12898 2018-02-27 Martin Liska <mliska@suse.cz>
12900 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12901 'ggc' suffixes. Change first column width.
12902 (timer::print): Fix formatting of the column.
12904 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12906 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12907 preserve inline entry blocks for the sake of debug inline
12908 entry point markers alone.
12909 (remove_unused_locals): Suggest in comments a better place to
12910 force the preservation of inline entry blocks that are
12911 otherwise unused, but do not preserve them.
12913 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12915 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12917 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12920 * config/i386/constraints.md (Bs): Replace
12921 ix86_indirect_branch_register with
12922 TARGET_INDIRECT_BRANCH_REGISTER.
12924 * config/i386/i386.md (indirect_jump): Likewise.
12925 (tablejump): Likewise.
12926 (*sibcall_memory): Likewise.
12927 (*sibcall_value_memory): Likewise.
12928 Peepholes of indirect call and jump via memory: Likewise.
12929 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12930 (*sibcall_value_GOT_32): Likewise.
12931 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12932 (GOT_memory_operand): Likewise.
12933 (call_insn_operand): Likewise.
12934 (sibcall_insn_operand): Likewise.
12935 (GOT32_symbol_operand): Likewise.
12936 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12938 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12940 PR rtl-optimization/83496
12941 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12942 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12943 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12944 redundant insn, if any.
12945 (relax_delay_slots): Likewise.
12946 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12948 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12950 PR tree-optimization/83965
12951 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12952 that grouped statements are part of a reduction chain. Return
12953 true if the statement is not marked as a reduction itself but
12954 is part of a group.
12955 (vect_recog_dot_prod_pattern): Don't check whether the statement
12956 is part of a group here.
12957 (vect_recog_sad_pattern): Likewise.
12958 (vect_recog_widen_sum_pattern): Likewise.
12960 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12963 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12965 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12968 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12970 (ix86_output_indirect_function_return): New prototype.
12971 (ix86_split_simple_return_pop_internal): Likewise.
12972 * config/i386/i386.c (indirect_return_via_cx): New.
12973 (indirect_return_via_cx_bnd): Likewise.
12974 (indirect_thunk_name): Handle return va CX_REG.
12975 (output_indirect_thunk_function): Create alias for
12976 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12977 (ix86_output_indirect_jmp): Remove the bool argument.
12978 (ix86_output_indirect_function_return): New function.
12979 (ix86_split_simple_return_pop_internal): Likewise.
12980 * config/i386/i386.md (*indirect_jump): Don't pass false
12981 to ix86_output_indirect_jmp.
12982 (*tablejump_1): Likewise.
12983 (simple_return_pop_internal): Change it to define_insn_and_split.
12984 Call ix86_split_simple_return_pop_internal to split it for
12985 -mfunction-return=.
12986 (simple_return_indirect_internal): Call
12987 ix86_output_indirect_function_return instead of
12988 ix86_output_indirect_jmp.
12990 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12993 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12994 memset and value initialization afterwards.
12996 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12998 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13000 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13003 * common/config/aarch64/aarch64-common.c
13004 (aarch_option_optimization_table[]): Switch
13005 off fomit-frame-pointer
13007 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13008 Chung-Ju Wu <jasonwucj@gmail.com>
13010 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13012 (store_multiple): Ditto.
13014 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13016 * config.gcc: Add --with-cpu support for nds32 target.
13017 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13018 * config/nds32/nds32.opt: Add -mcpu= option.
13020 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13022 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13023 isel=yes): Warn for these deprecated options.
13025 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13027 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13028 ISA_2_5_MASKS_EMBEDDED.
13030 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13032 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13033 p->max as pointers rather than using iterative_hash_expr.
13035 2018-02-23 Carl Love <cel@us.ibm.com>
13037 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13038 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13040 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13041 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13042 P8V_BUILTIN_VEC_VUNSIGNED2.
13044 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13047 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13048 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13050 * lra-constraints.c (curr_insn_transform): Set up
13051 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13053 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13054 * lra-eliminations.c (spill_pseudos): Ditto.
13055 (process_insn_for_elimination): Ditto.
13056 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13057 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13058 LRA_NON_CLOBBERED_ALT.
13060 2018-02-22 Martin Sebor <msebor@redhat.com>
13062 PR tree-optimization/84480
13063 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13064 to maybe_diag_stxncpy_trunc. Call it.
13065 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13066 from gimple_fold_builtin_strcpy. Print inlining stack.
13067 (handle_builtin_stxncpy): Print inlining stack.
13068 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13070 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13073 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13074 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13075 and -fcheck-pointer-bounds are used together.
13076 (indirect_thunk_prefix): New enum.
13077 (indirect_thunk_need_prefix): New function.
13078 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13079 "_nt" instead of "_bnd" for NOTRACK prefix.
13080 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13081 (output_indirect_thunk_function): Likewise.
13083 (ix86_code_end): Update output_indirect_thunk_function calls.
13084 (ix86_output_indirect_branch_via_reg): Replace
13085 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13086 (ix86_output_indirect_branch_via_push): Likewise.
13087 (ix86_output_function_return): Likewise.
13088 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13089 incompatible with -fcf-protection=branch and
13090 -fcheck-pointer-bounds.
13092 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13095 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13096 Change gcc_assert call to output_operand_lossage.
13098 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13100 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13102 2018-02-22 DJ Delorie <dj@redhat.com>
13103 Sebastian Perta <sebastian.perta@renesas.com>
13104 Oleg Endo <olegendo@gcc.gnu.org>
13106 * config/rx/rx.c (rx_rtx_costs): New function.
13107 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13109 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13111 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13113 2018-02-22 Martin Liska <mliska@suse.cz>
13116 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13117 Add "native" as a possible value.
13119 2018-02-22 Martin Liska <mliska@suse.cz>
13122 * config/i386/i386.c (ix86_option_override_internal):
13123 Add "native" as a possible value for -march and -mtune.
13125 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13128 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13129 to all type variants.
13131 PR tree-optimization/84503
13132 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13133 width as info->bitpos + info->bitsize - start.
13134 (merged_store_group::merge_overlapping): Simplify width computation.
13135 (check_no_overlap): New function.
13136 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13137 start + width and last_order of the group, fail if check_no_overlap
13139 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13140 to group if check_no_overlap fails.
13142 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13144 * config/rs6000/altivec.md: Delete contraint arguments to
13145 define_expand, define_split, and define_peephole2, and in
13146 define_insn_and_split if always unused.
13147 * config/rs6000/darwin.md: Ditto.
13148 * config/rs6000/dfp.md: Ditto.
13149 * config/rs6000/rs6000.md: Ditto.
13150 * config/rs6000/sync.md: Ditto.
13151 * config/rs6000/vector.md: Ditto.
13152 * config/rs6000/vsx.md: Ditto.
13154 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13156 * config/rs6000/altivec.md: Write output control strings as braced
13157 blocks instead of double-quoted strings.
13158 * config/rs6000/darwin.md: Ditto.
13159 * config/rs6000/rs6000.md: Ditto.
13160 * config/rs6000/vector.md: Ditto.
13161 * config/rs6000/vsx.md: Ditto.
13163 2018-02-21 Jason Merrill <jason@redhat.com>
13165 PR c++/84314 - ICE with templates and fastcall attribute.
13166 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13168 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13170 * ipa-cp.c (determine_versionability): Fix comment typos.
13172 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13175 * ipa-cp.c (determine_versionability): Do not version functions caling
13178 2018-02-21 Martin Liska <mliska@suse.cz>
13181 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13182 Add "native" as a possible value.
13183 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13184 the macro when native cpu detection is available.
13186 2018-02-21 Martin Liska <mliska@suse.cz>
13189 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13190 Add "native" as a possible value.
13191 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13192 when native cpu detection is available.
13194 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13195 Martin Sebor <msebor@redhat.com>
13197 PR tree-optimization/84478
13198 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13200 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13201 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13202 support which is conservatively correct, for 2 only stay conservative
13203 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13204 argument to the 2 argument get_range_strlen, adjust 6 arg
13205 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13207 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13208 (gimple_fold_builtin_strlen): Pass true as last argument to
13211 2018-02-20 Martin Sebor <msebor@redhat.com>
13213 PR middle-end/84095
13214 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13215 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13216 (builtin_memref::builtin_memref): Factor out parts into
13217 set_base_and_offset and call it.
13219 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13221 PR middle-end/84406
13222 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13223 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13224 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13225 search at the associated MODE_INT.
13227 2018-02-20 Jeff Law <law@redhat.com>
13229 PR middle-end/82123
13230 PR tree-optimization/81592
13231 PR middle-end/79257
13232 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13233 for range data rather than using global data.
13234 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13235 range data rather than using global data.
13236 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13237 pass it to children as needed.
13238 (struct directive::fmtresult): Similarly.
13239 (struct directive::set_width): Similarly.
13240 (struct directive::set_precision): Similarly.
13241 (format_integer, format_directive, parse_directive): Similarly.
13242 (format_none): Accept unnamed vr_values parameter.
13243 (format_percent, format_floating, format_character): Similarly.
13244 (format_string, format_plain): Similarly.
13245 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13246 the EVRP range analyzer for range data rather than using global data.
13247 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13248 gimple-ssa-evrp-analyze.h
13249 (class sprintf_dom_walker): Add after_dom_children member function.
13250 Add evrp_range_analyzer member.
13251 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13252 range analyzer as needed.
13253 (sprintf_dom_walker::after_dom_children): New member function.
13254 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13256 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13257 (evrp_range_analyzer::pop_to_marker): Likewise.
13259 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13261 PR tree-optimization/84419
13262 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13263 with the required type if its current type is compatible but
13266 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13268 PR middle-end/82004
13269 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13270 after vectorization.
13272 2018-02-20 Martin Liska <mliska@suse.cz>
13275 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13276 possible values if we don't have a hint.
13278 2018-02-20 Martin Liska <mliska@suse.cz>
13282 * final.c (shorten_branches): Build align_tab array with one
13284 * opts.c (finish_options): Add alignment option limit check.
13285 (MAX_CODE_ALIGN): Likewise.
13286 (MAX_CODE_ALIGN_VALUE): Likewise.
13287 * doc/invoke.texi: Document maximum allowed option value for
13288 all -falign-* options.
13290 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13293 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13294 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13295 * var-tracking.c (emit_note_insn_var_location): Remove all references
13296 to NOTE_INSN_CALL_ARG_LOCATION.
13297 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13298 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13299 Use copy_rtx_if_shared.
13300 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13301 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13302 (dwarf2out_var_location): Remove handling of
13303 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13305 * final.c (final_scan_insn): Remove all references to
13306 NOTE_INSN_CALL_ARG_LOCATION.
13307 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13308 before dumping final insns.
13309 * except.c (emit_note_eh_region_end): Remove all references to
13310 NOTE_INSN_CALL_ARG_LOCATION.
13311 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13312 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13313 * config/arc/arc.c (hwloop_optimize): Likewise.
13314 * config/arm/arm.c (create_fix_barrier): Likewise.
13315 * config/s390/s390.c (s390_chunkify_start): Likewise.
13316 * config/sh/sh.c (find_barrier): Likewise.
13317 * config/i386/i386.c (rest_of_insert_endbranch,
13318 ix86_seh_fixup_eh_fallthru): Likewise.
13319 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13320 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13321 * config/frv/frv.c (frv_function_prologue): Likewise.
13322 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13324 (note_outside_basic_block_p): Remove all references to
13325 NOTE_INSN_CALL_ARG_LOCATION.
13326 * gengtype.c (adjust_field_rtx_def): Likewise.
13327 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13329 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13330 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13333 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13336 PR tree-optimization/84452
13337 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13338 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13341 2018-02-19 Martin Liska <mliska@suse.cz>
13344 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13346 2018-02-19 Martin Liska <mliska@suse.cz>
13347 Richard Sandiford <richard.sandiford@linaro.org>
13349 PR tree-optimization/82491
13350 * gimple-fold.c (get_base_constructor): Make earlier bail out
13353 2018-02-19 Carl Love <cel@us.ibm.com>
13355 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13356 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13358 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13359 P8V_BUILTIN_VEC_NEG.
13361 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13363 * config/rl78/rl78.md (movdf): New define expand.
13365 2018-02-19 Martin Liska <mliska@suse.cz>
13368 * doc/invoke.texi: Fix typo.
13369 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13371 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13373 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13374 handle rs6000_single_float and rs6000_double_float specially for
13377 2018-02-16 Jeff Law <law@redhat.com>
13379 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13380 (pop_regs): Use it.
13382 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13385 * ipa-inline.c (inline_small_functions): Fix a typo.
13387 2018-02-16 Nathan Sidwell <nathan@acm.org>
13389 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13391 2018-02-16 Carl Love <cel@us.ibm.com>
13393 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13394 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13395 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13396 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13397 expansion to P8V_BUILTIN_VEC_FLOAT2.
13399 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13401 PR rtl-optimization/70023
13402 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13403 src_regno into account.
13405 2018-02-16 Carl Love <cel@us.ibm.com>
13407 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13408 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13409 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13410 * config/rs6000/rs6000.c: Remove case statements for
13411 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13412 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13413 and P9V_BUILTIN_VEC_VINSERT4B.
13414 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13415 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13416 * config/rs6000/vsx.md:
13417 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13420 2018-02-16 Carl Love <cel@us.ibm.com>
13422 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13424 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13426 * config/rs6000/rs6000-c.c: Add the definitions for
13427 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13428 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13429 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13430 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13431 definition for insert4b and define insn *insert3b_internal.
13432 * doc/extend.texi: Add documentation for vec_extract4b.
13434 2018-02-16 Nathan Sidwell <nathan@acm.org>
13436 * doc/extend.texi (Backwards Compatibility): Mention friend
13437 injection. Note for-scope is deprecated.
13438 * doc/invoke.texi (-ffriend-injection): Deprecate.
13440 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13442 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13443 that moved to I2, also allow destinations that are a paradoxical
13444 subreg (instead of a normal reg).
13446 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13449 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13452 2018-02-16 Richard Biener <rguenther@suse.de>
13454 PR tree-optimization/84037
13455 PR tree-optimization/84016
13457 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13458 Adjust vec_construct for the fact we need additional higher latency
13459 128bit inserts for AVX256 and AVX512 vector builds.
13460 (ix86_add_stmt_cost): Scale vector construction cost for
13463 2018-02-16 Richard Biener <rguenther@suse.de>
13465 PR tree-optimization/84417
13466 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13467 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13468 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13470 2018-02-16 Martin Liska <mliska@suse.cz>
13473 * internal-fn.def (ASAN_CHECK): Set proper flags.
13474 (ASAN_MARK): Likewise.
13476 2018-02-16 Julia Koval <julia.koval@intel.com>
13478 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13479 from PTA_CANNONLAKE.
13481 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13484 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13485 Use ++iter rather than iter++ for std::list iterators.
13486 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13487 defer deleting them until all nodes in the forest are processed. Do
13488 free even leaf nodes. Change to_process into auto_vec.
13491 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13492 * vec.h (vec_default_construct): Use memset instead of placement new
13493 if BROKEN_VALUE_INITIALIZATION is defined.
13494 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13495 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13498 PR rtl-optimization/83723
13499 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13500 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13501 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13503 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13505 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13507 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13509 PR rtl-optimization/81443
13510 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13511 from inner REGs to paradoxical SUBREGs.
13513 2018-02-16 Richard Biener <rguenther@suse.de>
13515 PR tree-optimization/84399
13516 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13517 For operands we can analyze at their definition make sure we can
13518 analyze them at each use as well.
13520 2018-02-16 Richard Biener <rguenther@suse.de>
13522 PR tree-optimization/84190
13523 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13524 volatile accesses if the decl isn't volatile.
13526 2018-02-15 Jason Merrill <jason@redhat.com>
13528 PR c++/84314 - ICE with templates and fastcall attribute.
13529 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13530 TYPE_CANONICAL on an existing type.
13532 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13534 PR tree-optimization/84383
13535 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13536 dstoff nor call operand_equal_p if dstbase is NULL.
13538 PR tree-optimization/84334
13539 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13540 also a CONSTANT_CLASS_P, punt.
13542 2018-02-14 Jim Wilson <jimw@sifive.com>
13544 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13545 first SMALL_OPERAND check. New local min_second_step. Move assert
13546 to where locals are set. Add TARGET_RVC support.
13547 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13549 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13551 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13553 2018-02-14 Martin Sebor <msebor@redhat.com>
13555 PR tree-optimization/83698
13556 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13557 arrays constrain the offset range to their bounds.
13558 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13559 (builtin_access::overlap): Avoid setting the size of overlap if it's
13561 (maybe_diag_overlap): Also consider arrays when deciding what values
13562 of offsets to include in diagnostics.
13564 2018-02-14 Martin Sebor <msebor@redhat.com>
13567 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13568 that correspond to the kind of a declaration.
13570 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13573 * config/pa/pa.md: Load address of PIC label using the linkage table
13574 if the label is nonlocal.
13576 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13578 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13579 warning message if user requests -maltivec=be.
13580 * doc/invoke.texi: Document deprecation of -maltivec=be.
13582 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13585 * config/rs6000/rs6000-c.c: Update definitions for
13586 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13587 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13589 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13592 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13593 add _get_ssp intrinsics. Remove argument from
13594 __builtin_ia32_rdssp[d|q].
13595 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13596 * config/i386/i386-builtin.def: Remove argument from
13597 __builtin_ia32_rdssp[d|q].
13598 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13599 ix86_expand_special_args_builtin for _rdssp[d|q].
13600 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13601 Clear register before usage.
13602 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13603 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13605 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13607 PR tree-optimization/84357
13608 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13609 operand 1 of an ARRAY_REF too.
13611 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13614 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13615 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13617 (set_of_reg): New struct.
13618 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13619 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13620 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13622 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13623 Split into bitclr, bitset, bitinvert patterns if appropriate.
13624 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13625 use rx_fuse_in_memory_bitop.
13626 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13627 to named insn, correct maximum insn length.
13629 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13632 * machmode.def: Define a complex mode for PARTIAL_INT.
13633 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13635 * doc/rtl.texi: Document CSPImode.
13636 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13638 (msp430_hard_regno_nregs_with_padding): Likewise.
13640 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13643 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13645 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13647 PR rtl-optimization/84169
13648 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13649 we generated a parallel as new i3 and we split that to new i2 and i3
13650 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13651 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13652 those to i2, not i1. Partially rewrite this scan code.
13654 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13657 * stor-layout.c (place_field): For variable length fields, adjust
13658 offset_align afterwards not just based on the field's alignment,
13659 but also on the size.
13661 PR middle-end/84309
13662 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13663 of exps and logs in the use_exp2 case.
13665 2018-02-13 Jeff Law <law@redhat.com>
13667 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13668 entry for "vector".
13670 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13673 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13677 * common.opt (gas-loc-support, gas-locview-support): New.
13678 (ginline-points, ginternal-reset-location-views): New.
13679 * doc/invoke.texi: Document them. Use @itemx where intended.
13680 (gvariable-location-views): Adjust.
13681 * target.def (reset_location_view): New.
13682 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13683 (TARGET_RESET_LOCATION_VIEW): New.
13684 * doc/tm.texi: Rebuilt.
13685 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13686 (dwarf2out_default_as_locview_support): New.
13687 (output_asm_line_debug_info): Use option variables.
13688 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13689 (output_loc_list): Likewise.
13690 (add_high_low_attributes): Check option variables.
13691 Don't output entry view attribute in strict mode.
13692 (gen_inlined_subroutine_die): Check option variables.
13693 (dwarf2out_inline_entry): Likewise.
13694 (init_sections_and_labels): Likewise.
13695 (dwarf2out_early_finish): Likewise.
13696 (maybe_reset_location_view): New, from...
13697 (dwarf2out_var_location): ... here. Call it.
13698 * debug.h (dwarf2out_default_as_loc_support): Declare.
13699 (dwarf2out_default_as_locview_support): Declare.
13700 * hooks.c (hook_int_rtx_insn_0): New.
13701 * hooks.h (hook_int_rtx_insn_0): Declare.
13702 * toplev.c (process_options): Take -gas-loc-support and
13703 -gas-locview-support from dwarf2out. Enable
13704 -gvariable-location-views by default only with locview
13705 assembler support. Enable -ginternal-reset-location-views by
13706 default only if the target defines the corresponding hook.
13707 Enable -ginline-points by default if location views are
13708 enabled; force it disabled if statement frontiers are
13710 * tree-inline.c (expand_call_inline): Check option variables.
13711 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13713 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13715 PR tree-optimization/84321
13716 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13717 handling. Also check whether the anti-range contains any values
13718 that satisfy the mask; switch to a VR_RANGE if not.
13720 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13723 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13725 2018-02-13 Martin Jambor <mjambor@suse.cz>
13728 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13729 of call statements, also set location of a load to a temporary.
13731 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13733 * config/rl78/rl78.c (add_vector_labels): New function.
13734 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13735 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13736 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13737 which checks that no arguments are passed.
13738 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13739 * doc/extend.texi: Documentation for the new attribute.
13741 2018-02-13 Andreas Schwab <schwab@suse.de>
13743 * config/riscv/linux.h (CPP_SPEC): Define.
13745 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13748 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13749 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13750 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13751 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13752 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13753 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13754 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13755 temporarily for AES and PCLMUL builtins.
13757 PR tree-optimization/84339
13758 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13759 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13762 PR middle-end/84309
13763 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13764 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13765 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13767 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13769 * omp-simd-clone.h: New file.
13770 * omp-simd-clone.c: Include omp-simd-clone.h.
13771 (expand_simd_clones): No longer static.
13772 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13773 cgraph.h and omp-simd-clone.h.
13774 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13775 (vect_recog_widen_shift_pattern): Formatting fix.
13776 (vect_pattern_recog_1): Don't check optab for calls.
13779 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13780 operands[2] into a REG before using gen_lowpart on it.
13782 2018-02-12 Jeff Law <law@redhat.com>
13785 * config/sh/sh.c (find_barrier): Consider a sibling call
13788 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13789 successfully back substituting a reg.
13791 2018-02-12 Richard Biener <rguenther@suse.de>
13793 PR tree-optimization/84037
13794 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13795 parameter, move visited init to caller.
13796 (vect_slp_analyze_operations): Separate cost from validity
13797 check, initialize visited once for all instances.
13798 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13800 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13802 (vect_model_promotion_demotion_cost): Likewise.
13803 (vectorizable_bswap): Guard cost modeling with !slp_node
13804 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13806 (vectorizable_call): Likewise.
13807 (vectorizable_conversion): Likewise.
13808 (vectorizable_assignment): Likewise.
13809 (vectorizable_shift): Likewise.
13810 (vectorizable_operation): Likewise.
13811 (vectorizable_store): Likewise.
13812 (vectorizable_load): Likewise.
13813 (vectorizable_condition): Likewise.
13814 (vectorizable_comparison): Likewise.
13816 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13819 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13820 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13821 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13823 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13825 PR middle-end/83665
13826 * params.def (inline-min-speedup): Increase from 8 to 15.
13827 (max-inline-insns-auto): Decrease from 40 to 30.
13828 * ipa-split.c (consider_split): Add some buffer for function to
13829 be considered inlining candidate.
13830 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13833 2018-02-12 Richard Biener <rguenther@suse.de>
13835 PR tree-optimization/84037
13836 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13837 matched stmts if we cannot swap the non-matched ones.
13839 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13841 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13842 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13843 _mm_maskz_scalef_round_ss): New intrinsics.
13844 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13845 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13846 __builtin_ia32_scalefss_round): Remove.
13847 (__builtin_ia32_scalefsd_mask_round,
13848 __builtin_ia32_scalefss_mask_round): New intrinsics.
13849 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13850 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13851 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13852 "<round_constraint>")): Changed to ...
13853 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13854 "<round_scalar_constraint>")): ... this.
13855 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13856 %0, %1, %2<round_op3>}"): Changed to ...
13857 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13858 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13859 %2<round_scalar_mask_op3>}"): ... this.
13860 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13862 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13864 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13865 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13866 (_mm_maskz_sqrt_round_ss): New intrinsics.
13867 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13868 (__builtin_ia32_sqrtsd_mask_round)
13869 (__builtin_ia32_sqrtss_mask_round): New builtins.
13870 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13871 (__builtin_ia32_sqrtss_round): Remove.
13872 (__builtin_ia32_sqrtsd_mask_round)
13873 (__builtin_ia32_sqrtss_mask_round): New builtins.
13874 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13875 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13876 ((match_operand:VF_128 1 "vector_operand"
13877 "xBm,<round_constraint>")): Changed to ...
13878 ((match_operand:VF_128 1 "vector_operand"
13879 "xBm,<round_scalar_constraint>")): ... this.
13880 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13881 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13882 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13883 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13884 %<iptr>1<round_scalar_mask_op3>}): ... this.
13885 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13886 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13888 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13891 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13892 Cast vec_cmpeq result to correct type.
13893 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13894 Cast vec_cmpgt result to correct type.
13896 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13898 * final.c (final_scan_insn_1): Renamed from...
13899 (final_scan_insn): ... this. New wrapper, to recover
13900 seen from the outermost call in recursive ones.
13901 * config/sparc/sparc.c (output_return): Drop seen from call.
13902 (output_sibcall): Likewise.
13903 * config/visium/visium.c (output_branch): Likewise.
13905 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13907 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13910 2018-02-10 Alan Modra <amodra@gmail.com>
13913 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13914 Specify LR as an input.
13916 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13919 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13920 remove_member_access_dummy_vars): New functions.
13921 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13922 lower_omp_1, execute_lower_omp): Use them.
13924 PR rtl-optimization/84308
13925 * shrink-wrap.c (spread_components): Release todo vector.
13927 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13929 PR rtl-optimization/57193
13930 * ira-color.c (struct allocno_color_data): Add member
13931 conflict_allocno_hard_prefs.
13932 (update_conflict_allocno_hard_prefs): New.
13933 (bucket_allocno_compare_func): Add a preference based on
13934 conflict_allocno_hard_prefs.
13935 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13936 (color_allocnos): Remove a dead code. Initiate
13937 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13939 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13942 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13943 constraint from =wa to wa. Avoid a subreg on the output operand,
13944 instead use a pseudo and subreg it in a move.
13945 (p9_xxbrd_<mode>): Changed to ...
13946 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13947 (p9_xxbrd_v2df): New expander.
13948 (p9_xxbrw_<mode>): Changed to ...
13949 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13950 (p9_xxbrw_v4sf): New expander.
13952 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13954 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13956 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13959 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13960 multiply in 32-bit mode.
13961 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13962 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13965 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13967 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13968 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13969 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13970 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13972 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13975 * dwarf2out.c (is_trivial_indirect_ref): New function.
13976 (dwarf2out_late_global_decl): Do not generate a location
13977 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13978 and that are not defined in the current unit.
13980 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13982 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13983 instead of a libcall for UNORDERED.
13985 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13988 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13989 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13991 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13994 * config/s390/s390.c (s390_set_current_function): Invoke
13995 s390_indirect_branch_settings also if fndecl didn't change.
13997 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13999 * config/rs6000/rs6000.md (blockage): Set length to zero.
14001 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14003 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14005 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14008 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14009 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14013 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14014 PARALLEL incoming that failed vt_get_decl_and_offset check.
14016 PR middle-end/84237
14017 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14018 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14020 (get_variable_section): For decls in named .bss* sections pass true as
14021 second argument to bss_initializer_p.
14023 2018-02-09 Marek Polacek <polacek@redhat.com>
14024 Jakub Jelinek <jakub@redhat.com>
14027 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14028 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14029 Sync some changes from cxx_fold_indirect_ref.
14031 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14033 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14035 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14036 (BLOCK_INLINE_ENTRY_LABEL): New.
14037 (dwarf2out_var_location): Disregard inline entry markers.
14038 (inline_entry_data): New struct.
14039 (inline_entry_data_hasher): New hashtable type.
14040 (inline_entry_data_hasher::hash): New.
14041 (inline_entry_data_hasher::equal): New.
14042 (inline_entry_data_table): New variable.
14043 (add_high_low_attributes): Add DW_AT_entry_pc and
14044 DW_AT_GNU_entry_view attributes if a pending entry is found
14045 in inline_entry_data_table. Add old entry_pc attribute only
14046 if debug nonbinding markers are disabled.
14047 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14048 markers are enabled.
14049 (block_within_block_p, dwarf2out_inline_entry): New.
14050 (dwarf2out_finish): Check that no entries remained in
14051 inline_entry_data_table.
14052 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14053 (final_scan_insn, notice_source_line): Likewise.
14054 (rest_of_clean_state): Skip inline entry markers.
14055 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14057 * gimple.c (gimple_build_debug_inline_entry): New.
14058 * gimple.h (enum gimple_debug_subcode): Add
14059 GIMPLE_DEBUG_INLINE_ENTRY.
14060 (gimple_build_debug_inline_entry): Declare.
14061 (gimple_debug_inline_entry_p): New.
14062 (gimple_debug_nonbind_marker_p): Adjust.
14063 * insn-notes.def (INLINE_ENTRY): New.
14064 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14065 inline entry marker notes.
14066 (print_insn): Likewise.
14067 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14068 (INSN_DEBUG_MARKER_KIND): Likewise.
14069 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14070 * tree-inline.c (expand_call_inline): Build and insert
14071 debug_inline_entry stmt.
14072 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14073 inline entry blocks early, if nonbind markers are enabled.
14074 (dump_scope_block): Dump fragment info.
14075 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14076 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14077 (gimple_build_debug_inline_entry): New.
14078 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14079 Enable/disable inline entry points too.
14080 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14081 (DEBUG_INSN): Describe inline entry markers.
14083 * common.opt (gvariable-location-views): New.
14084 (gvariable-location-views=incompat5): New.
14085 * config.in: Rebuilt.
14086 * configure: Rebuilt.
14087 * configure.ac: Test assembler for view support.
14088 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14089 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14090 * dwarf2out.c (var_loc_view): New typedef.
14091 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14092 (dwarf2out_locviews_in_attribute): New.
14093 (dwarf2out_locviews_in_loclist): New.
14094 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14095 (enum dw_line_info_opcode): Add LI_adv_address.
14096 (struct dw_line_info_table): Add view.
14097 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14098 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14099 (zero_view_p): New variable.
14100 (ZERO_VIEW_P): New macro.
14101 (output_asm_line_debug_info): New.
14102 (struct var_loc_node): Add view.
14103 (add_AT_view_list, AT_loc_list): New.
14104 (add_var_loc_to_decl): Add view param. Test it against last.
14105 (new_loc_list): Add view params. Record them.
14106 (AT_loc_list_ptr): Handle loc and view lists.
14107 (view_list_to_loc_list_val_node): New.
14108 (print_dw_val): Handle dw_val_class_view_list.
14109 (size_of_die): Likewise.
14110 (value_format): Likewise.
14111 (loc_list_has_views): New.
14112 (gen_llsym): Set vl_symbol too.
14113 (maybe_gen_llsym, skip_loc_list_entry): New.
14114 (dwarf2out_maybe_output_loclist_view_pair): New.
14115 (output_loc_list): Output view list or entries too.
14116 (output_view_list_offset): New.
14117 (output_die): Handle dw_val_class_view_list.
14118 (output_dwarf_version): New.
14119 (output_compilation_unit_header): Use it.
14120 (output_skeleton_debug_sections): Likewise.
14121 (output_rnglists, output_line_info): Likewise.
14122 (output_pubnames, output_aranges): Update version comments.
14123 (output_one_line_info_table): Output view numbers in asm comments.
14124 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14125 Call maybe_gen_llsym.
14126 (loc_list_from_tree_1): Adjust.
14127 (add_AT_location_description): Create view list attribute if
14128 needed, check it's absent otherwise.
14129 (convert_cfa_to_fb_loc_list): Adjust.
14130 (maybe_emit_file): Call output_asm_line_debug_info for test.
14131 (dwarf2out_var_location): Reset views as needed. Precompute
14132 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14133 attribute. Set view.
14134 (new_line_info_table): Reset next view.
14135 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14136 (dwarf2out_source_line): Likewise. Output view resets and labels to
14137 the assembler, or select appropriate line info opcodes.
14138 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14139 (optimize_string_length): Catch it. Adjust.
14140 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14141 dw_val_class_view_list, and remove it if no longer needed.
14142 (hash_loc_list): Hash view numbers.
14143 (loc_list_hasher::equal): Compare them.
14144 (optimize_location_lists): Check whether a view list symbol is
14145 needed, and whether the locview attribute is present, and
14146 whether they match. Remove the locview attribute if no longer
14148 (index_location_lists): Call skip_loc_list_entry for test.
14149 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14150 Use output_dwarf_version.
14151 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14152 (struct dw_val_node): Add val_view_list.
14153 * final.c (SEEN_NEXT_VIEW): New.
14154 (set_next_view_needed): New.
14155 (clear_next_view_needed): New.
14156 (maybe_output_next_view): New.
14157 (final_start_function): Rename to...
14158 (final_start_function_1): ... this. Take pointer to FIRST,
14159 add SEEN parameter. Emit param bindings in the initial view.
14160 (final_start_function): Reintroduce SEEN-less interface.
14161 (final): Rename to...
14162 (final_1): ... this. Take SEEN parameter. Output final pending
14163 next view at the end.
14164 (final): Reintroduce seen-less interface.
14165 (final_scan_insn): Output pending next view before switching
14166 sections or ending a block. Mark the next view as needed when
14167 outputting variable locations. Notify debug backend of section
14168 changes, and of location view changes.
14169 (rest_of_handle_final): Adjust.
14170 * toplev.c (process_options): Autodetect value for debug variable
14171 location views option. Warn on incompat5 without -gdwarf-5.
14172 * doc/invoke.texi (gvariable-location-views): New.
14173 (gvariable-location-views=incompat5): New.
14174 (gno-variable-location-views): New.
14176 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14178 PR tree-optimization/84136
14179 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14180 that the result of find_edge is non-NULL.
14182 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14185 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14186 storing integer register in SImode. Fix cost of 256 and 512
14187 byte aligned SSE register store.
14189 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14191 * config/i386/i386.c (ix86_multiplication_cost): Fix
14192 multiplication cost for TARGET_AVX512DQ.
14194 2018-02-08 Marek Polacek <polacek@redhat.com>
14196 PR tree-optimization/84238
14197 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14200 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14202 PR tree-optimization/84265
14203 * tree-vect-stmts.c (vectorizable_store): Don't treat
14204 VMAT_CONTIGUOUS accesses as grouped.
14205 (vectorizable_load): Likewise.
14207 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14209 PR tree-optimization/81635
14210 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14211 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14212 (test_round_for_mask): New functions.
14213 (wide_int_cc_tests): Call test_round_for_mask.
14214 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14215 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14216 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14217 range returned by get_range_info.
14219 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14222 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14223 * symtab.c: Include builtins.h
14224 (symtab_node::output_to_lto_symbol_table_p): Move here
14225 from lto-streamer-out.c:output_symbol_p.
14226 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14227 (output_symbol_p): Move all logic to symtab.c
14228 (produce_symtab): Update.
14230 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14232 * config/s390/s390-opts.h (enum indirect_branch): Define.
14233 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14234 (s390_indirect_branch_via_thunk)
14235 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14236 (enum s390_indirect_branch_type): Define.
14237 * config/s390/s390.c (struct s390_frame_layout, struct
14238 machine_function): Remove.
14239 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14240 (indirect_branch_table_label_no, indirect_branch_table_name):
14242 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14243 (enum s390_indirect_branch_option): Define.
14244 (s390_return_addr_from_memory): New function.
14245 (s390_handle_string_attribute): New function.
14246 (s390_attribute_table): Add new attribute handler.
14247 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14248 (s390_indirect_branch_via_thunk): New function.
14249 (s390_indirect_branch_via_inline_thunk): New function.
14250 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14251 sibling call optimization for non z10 compiles.
14252 (s390_emit_call): Force indirect branch target to be a single
14253 register. Add r1 clobber for non-z10 compiles.
14254 (s390_emit_epilogue): Emit return jump via return_use expander.
14255 (s390_reorg): Handle JUMP_INSNs as execute targets.
14256 (s390_option_override_internal): Perform validity checks for the
14257 new command line options.
14258 (s390_indirect_branch_attrvalue): New function.
14259 (s390_indirect_branch_settings): New function.
14260 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14261 (s390_output_indirect_thunk_function): New function.
14262 (s390_code_end): Implement target hook.
14263 (s390_case_values_threshold): Implement target hook.
14264 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14266 * config/s390/s390.h (struct s390_frame_layout)
14267 (struct machine_function): Move here from s390.c.
14268 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14269 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14270 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14271 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14272 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14273 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14274 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14275 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14276 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14277 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14278 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14279 (mnemonic attribute): Add values which aren't recognized
14281 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14282 pattern for branch conversion. Fix mnemonic attribute.
14283 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14284 indirect branch via thunk if requested.
14285 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14286 ("*indirect_jump"): Disable for branch conversion using out of
14288 ("indirect_jump_via_thunk<mode>_z10")
14289 ("indirect_jump_via_thunk<mode>")
14290 ("indirect_jump_via_inlinethunk<mode>_z10")
14291 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14292 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14293 ("casesi_jump_via_inlinethunk<mode>_z10")
14294 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14295 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14296 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14297 ("*indirect2_jump"): Disable for branch conversion.
14298 ("casesi_jump"): Turn into expander and expand patterns for branch
14300 ("return_use"): New expander.
14301 ("*return"): Emit return via thunk and rename it to ...
14302 ("*return<mode>"): ... this one.
14303 * config/s390/s390.opt: Add new options and and enum for the
14306 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14308 * lra-constraints.c (match_reload): Unconditionally use
14309 gen_lowpart_SUBREG, rather than selecting between that
14310 and equivalent gen_rtx_SUBREG code.
14312 2018-02-08 Richard Biener <rguenther@suse.de>
14314 PR tree-optimization/84233
14315 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14316 changed flag instead of boguously re-using phi_inserted.
14318 2018-02-08 Martin Jambor <mjambor@suse.cz>
14320 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14321 static local variables.
14323 2018-02-08 Richard Biener <rguenther@suse.de>
14325 PR tree-optimization/84278
14326 * tree-vect-stmts.c (vectorizable_store): When looking for
14327 smaller vector types to perform grouped strided loads/stores
14328 make sure the mode is supported by the target.
14329 (vectorizable_load): Likewise.
14331 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14333 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14334 Increase LDP/STP opportunities by adding adjacent callee-saves.
14336 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14338 PR rtl-optimization/84068
14339 PR rtl-optimization/83459
14340 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14342 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14344 PR tree-optimization/84224
14345 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14346 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14347 non-zero arguments.
14349 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14352 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14353 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14354 count, remove one USE.
14356 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14358 * doc/install.texi (Configuration): Document the
14359 --with-long-double-format={ibm,ieee} PowerPC configuration
14363 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14364 Convert from define_expand to be define_insn_and_split. Rework
14365 float/double/_Float128 conversions to QI/HI/SImode to work with
14366 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14367 conversions to QI/HImode types did a store and then a load to
14368 truncate the value. For conversions to VSX registers, don't split
14369 the insn, instead emit the code directly. Use the code iterator
14370 any_fix to combine signed and unsigned conversions.
14371 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14372 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14373 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14374 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14375 (fix_<mode>di2_hw): Likewise.
14376 (fixuns_<mode>di2_hw): Likewise.
14377 (fix_<mode>si2_hw): Likewise.
14378 (fixuns_<mode>si2_hw): Likewise.
14379 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14380 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14381 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14382 fix<uns>_trunc<SFDF:mode>si2_p8.
14383 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14384 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14385 (fix<uns>_<mode>_mem): Likewise.
14386 (fctiw<u>z_<mode>_mem): Likewise.
14387 (fix<uns>_<mode>_mem): Likewise.
14388 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14389 the register allocator from doing a direct move to the GPRs to do
14390 a store, and instead use the ISA 3.0 store byte/half-word from
14391 vector register instruction. For IEEE 128-bit floating point,
14392 also optimize stores of 32-bit ints.
14393 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14395 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14397 * genextract.c (push_pathstr_operand): New function to support
14399 (walk_rtx): Call push_pathstr_operand.
14400 (print_path): Support [a-zA-Z].
14402 2018-02-07 Richard Biener <rguenther@suse.de>
14404 PR tree-optimization/84037
14405 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14406 (cse_and_gimplify_to_preheader): Declare.
14407 (vect_get_place_in_interleaving_chain): Likewise.
14408 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14410 (_loop_vec_info::~_loop_vec_info): Delete it.
14411 (cse_and_gimplify_to_preheader): New function.
14412 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14413 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14414 (vectorizable_load): Likewise. For grouped stores always base
14415 the IV on the first element.
14416 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14417 condition before gimplifying.
14419 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14421 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14422 *DIV_EXPR and *MOD_EXPR.
14424 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14427 * config/i386/i386.c (ix86_option_override_internal): Mask out
14428 the CF_SET bit when checking -fcf-protection.
14430 2018-02-07 Tom de Vries <tom@codesourcery.com>
14433 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14436 2018-02-07 Richard Biener <rguenther@suse.de>
14438 PR tree-optimization/84204
14439 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14442 PR tree-optimization/84205
14443 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14444 special-case isl_ast_op_zdiv_r.
14446 PR tree-optimization/84223
14447 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14448 Only add conditions from within the region.
14449 (gather_bbs::after_dom_children): Adjust.
14451 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14454 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14455 * config/avr/avr.md: Only post-reload split REG-REG moves if
14456 either register is GENERAL_REG_P.
14458 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14460 PR tree-optimization/84235
14461 * tree-ssa-scopedtables.c
14462 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14463 if the subtraction is performed in floating point type where NaNs are
14464 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14465 build 1. Formatting fix.
14467 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14470 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14471 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14472 and skip it regardless of bb boundaries. Use CALL_P macro,
14473 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14474 unnecessarily, formatting fix.
14476 2018-02-06 Michael Collison <michael.collison@arm.com>
14478 * config/arm/thumb2.md:
14479 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14480 (*thumb_mov_notscc): Ditto.
14482 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14485 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14486 unsigned_fix, not "s".
14488 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14490 * configure.ac (gcc_fn_eh_frame_ro): New function.
14491 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14492 correct .eh_frame permissions.
14493 * configure: Regenerate.
14495 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14497 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14498 irrelevant options.
14500 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14502 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14503 Display warning message for -mno-speculate-indirect-jumps.
14505 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14507 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14509 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14511 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14513 PR tree-optimization/84225
14514 * tree-eh.c (find_trapping_overflow): Only call
14515 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14517 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14520 * config/i386/i386.c: Reimplement the check of possible options
14521 -mibt/-mshstk conbination. Change error messages.
14522 * doc/invoke.texi: Fix a typo: remove extra '='.
14524 2018-02-06 Marek Polacek <polacek@redhat.com>
14526 PR tree-optimization/84228
14527 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14529 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14532 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14533 emitted arch directives.
14534 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14535 __ARM_FEATURE_COPROC before changing architectures.
14537 2018-02-06 Richard Biener <rguenther@suse.de>
14539 * config/i386/i386.c (print_reg): Fix typo.
14540 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14542 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14544 * configure: Regenerate.
14546 2018-02-05 Martin Sebor <msebor@redhat.com>
14548 PR tree-optimization/83369
14549 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14552 2018-02-05 Martin Liska <mliska@suse.cz>
14554 * doc/invoke.texi: Cherry-pick upstream r323995.
14556 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14558 * ira.c (ira_init_register_move_cost): Adjust comment.
14560 2018-02-05 Martin Liska <mliska@suse.cz>
14562 PR gcov-profile/84137
14563 * doc/gcov.texi: Fix typo in documentation.
14565 2018-02-05 Martin Liska <mliska@suse.cz>
14567 PR gcov-profile/83879
14568 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14569 using dlopen functionality.
14571 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14573 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14574 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14575 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14576 _mm_maskz_range_round_ss): New intrinsics.
14577 (__builtin_ia32_rangesd128_round)
14578 (__builtin_ia32_rangess128_round): Remove.
14579 (__builtin_ia32_rangesd128_mask_round,
14580 __builtin_ia32_rangess128_mask_round): New builtins.
14581 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14582 __builtin_ia32_rangess128_round): Remove.
14583 (__builtin_ia32_rangesd128_mask_round,
14584 __builtin_ia32_rangess128_mask_round): New builtins.
14585 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14586 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14587 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14588 "<round_saeonly_constraint>")): Changed to ...
14589 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14590 "<round_saeonly_scalar_constraint>")): ... this.
14591 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14592 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14593 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14594 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14595 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14597 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14599 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14601 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14602 Remove all values except native, 8540 and 8548.
14604 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14606 * config/i386/i386.c (ix86_output_function_return): Pass
14607 INVALID_REGNUM, instead of -1, as invalid register number to
14608 indirect_thunk_name and output_indirect_thunk.
14610 2018-02-02 Julia Koval <julia.koval@intel.com>
14612 * config.gcc: Add -march=icelake.
14613 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14614 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14615 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14616 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14617 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14618 (processor_target_table): Add icelake.
14619 (ix86_option_override_internal): Handle new PTAs.
14620 (get_builtin_code_for_version): Handle icelake.
14621 (M_INTEL_COREI7_ICELAKE): New.
14622 (fold_builtin_cpu): Handle icelake.
14623 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14624 * doc/invoke.texi: Add -march=icelake.
14626 2018-02-02 Julia Koval <julia.koval@intel.com>
14628 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14629 to wide_int_bitmask.
14630 * wide-int-bitmask.h: New.
14632 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14635 * config/i386/i386.md: Replace Pmode with word_mode in
14636 builtin_setjmp_setup and builtin_longjmp to support x32.
14638 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14642 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14644 (rs6000_supported_cpu_names): New static variable.
14645 (linux_cpu_translation_table): Likewise.
14646 (elf_platform) <cpu>: Define new static variable and use it.
14647 Translate kernel AT_PLATFORM name to canonical name if needed.
14648 Error if platform name is unknown.
14650 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14653 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14655 2018-02-01 Jeff Law <law@redhat.com>
14658 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14659 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14660 the scratch if RELEASE_VIA_POP is false.
14661 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14662 If we have to save a temporary register, decrement SIZE appropriately.
14663 Pass new arguments to release_scratch_register_on_entry.
14664 (ix86_adjust_stack_and_probe): Likewise.
14665 (ix86_emit_probe_stack_range): Pass new arguments to
14666 release_scratch_register_on_entry.
14668 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14670 PR rtl-optimization/84157
14671 * combine.c (change_zero_ext): Use REG_P predicate in
14672 front of HARD_REGISTER_P predicate.
14674 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14676 * config/avr/avr.c (avr_option_override): Move disabling of
14677 -fdelete-null-pointer-checks to...
14678 * common/config/avr/avr-common.c (avr_option_optimization_table):
14681 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14683 PR tree-optimization/81635
14684 * tree-data-ref.c (split_constant_offset_1): For types that
14685 wrap on overflow, try to use range info to prove that wrapping
14688 2018-02-01 Renlin Li <renlin.li@arm.com>
14691 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14692 TAILCALL_ADDR_REGS.
14693 (aarch64_register_move_cost): Likewise.
14694 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14695 TAILCALL_ADDR_REGS.
14696 (REG_CLASS_NAMES): Likewise.
14697 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14698 TAILCALL_ADDR_REGS. Remove IP registers.
14699 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14701 2018-02-01 Richard Biener <rguenther@suse.de>
14703 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14704 for specifying RPO order and allow NULL for that.
14705 * domwalk.c (dom_walker::dom_walker): Likewise.
14706 (dom_walker::walk): Handle NULL RPO order.
14707 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14709 (rewrite_update_dom_walker): Likewise.
14710 (mark_def_dom_walker): Likewise.
14712 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14714 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14715 (aarch64_maybe_expand_sve_subreg_move): Declare.
14716 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14717 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14719 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14720 that are semantically a reverse operation.
14721 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14722 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14723 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14725 (aarch64_can_change_mode_class): For big-endian, forbid changes
14726 between two SVE modes if they have different element sizes.
14728 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14730 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14731 the TImode handling for big-endian targets.
14733 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14735 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14736 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14738 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14739 Remove BSWAP handing for big-endian targets and use the form of
14740 LD1RQ appropariate for the mode.
14742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14744 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14745 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14746 duplicated element.
14748 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14751 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14752 check for operands that need to go through aarch64_sve_reload_be.
14754 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14756 PR tree-optimization/81661
14757 PR tree-optimization/84117
14758 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14759 * tree-eh.c: Include gimplify.h.
14760 (find_trapping_overflow, replace_trapping_overflow,
14761 rewrite_to_non_trapping_overflow): New functions.
14762 * tree-vect-loop.c: Include tree-eh.h.
14763 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14764 * tree-data-ref.c: Include tree-eh.h.
14765 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14767 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14769 PR rtl-optimization/84123
14770 * combine.c (change_zero_ext): Check if hard register satisfies
14771 can_change_dest_mode before calling gen_lowpart_SUBREG.
14773 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14776 * ira.c (ira_init_register_move_cost): Remove assert.
14778 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14780 PR rtl-optimization/84071
14781 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14782 * doc/tm.texi: Regenerate.
14784 2018-01-31 Richard Biener <rguenther@suse.de>
14786 PR tree-optimization/84132
14787 * tree-data-ref.c (analyze_miv_subscript): Properly
14788 check whether evolution_function_is_affine_multivariate_p
14789 before calling gcd_of_steps_may_divide_p.
14791 2018-01-31 Julia Koval <julia.koval@intel.com>
14794 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14795 * config/i386/i386.md (rdpid_rex64) New.
14796 (rdpid): Make 32bit only.
14798 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14801 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14802 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14804 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14807 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14809 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14811 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14813 PR rtl-optimization/84071
14814 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14815 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14817 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14819 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14820 (arc_attribute_table): Add 'aux' attribute.
14821 (arc_in_small_data_p): Consider aux like variables.
14822 (arc_is_aux_reg_p): New function.
14823 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14824 (arc_get_aux_arg): New function.
14825 (prepare_move_operands): Handle aux-register access.
14826 (arc_handle_aux_attribute): New function.
14827 * doc/extend.texi (ARC Variable attributes): Add subsection.
14829 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14831 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14832 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14833 (arc_attribute_table): Add 'uncached' attribute.
14834 (arc_print_operand): Print '.di' flag for uncached memory
14836 (arc_in_small_data_p): Do not consider for small data the uncached
14838 (arc_is_uncached_mem_p): New function.
14839 * config/arc/predicates.md (compact_store_memory_operand): Check
14840 for uncached memory accesses.
14841 (nonvol_nonimm_operand): Likewise.
14842 * doc/extend.texi (ARC Type Attribute): New subsection.
14844 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14847 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14848 falign-loops=): Add Optimization flag.
14850 2018-01-30 Jeff Law <law@redhat.com>
14853 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14854 INT_REGISTERS_SAVED. Check it prior to calling
14855 get_scratch_register_on_entry.
14856 (ix86_adjust_stack_and_probe): Similarly.
14857 (ix86_emit_probe_stack_range): Similarly.
14858 (ix86_expand_prologue): Corresponding changes.
14860 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14863 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14864 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14866 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14869 * lra-constraints.c (curr_insn_transform): Process AND in the
14872 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14874 PR rtl-optimization/83986
14875 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14876 dependence against last_pending_memory_flush in addition to
14877 pending_jump_insns.
14879 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14881 PR tree-optimization/81611
14882 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14885 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14888 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14891 2018-01-30 Richard Biener <rguenther@suse.de>
14892 Jakub Jelinek <jakub@redhat.com>
14894 PR tree-optimization/84111
14895 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14896 inner loops added during recursion, as they don't have up-to-date
14899 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14902 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14903 (can_inline_edge_by_limits_p): ... here.
14904 (can_early_inline_edge_p, check_callers,
14905 update_caller_keys, update_callee_keys, recursive_inlining,
14906 add_new_edges_to_heap, speculation_useful_p,
14907 inline_small_functions,
14908 inline_small_functions, flatten_function,
14909 inline_to_all_callers_1): Update.
14911 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14913 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14916 2018-01-30 Richard Biener <rguenther@suse.de>
14918 PR tree-optimization/83008
14919 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14920 invariant and constant vector uses in stmts when they need
14921 more than one stmt.
14923 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14926 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14927 * configure: Regenerate.
14929 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14931 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14933 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14934 Use gen_rtx_REG rather than gen_lowpart.
14936 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14938 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14939 rather than 0 when creating partial subregs.
14941 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14943 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14946 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14949 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14950 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14951 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14952 flags. This restores the settings used before the 2017-07-24.
14953 Turning off pre increment/decrement/modify allows IVOPTS to
14954 optimize DF/SF loops where the index is an int.
14956 2018-01-29 Richard Biener <rguenther@suse.de>
14957 Kelvin Nilsen <kelvin@gcc.gnu.org>
14960 * tree-vect-stmts.c (vectorizable_call): Don't call
14961 targetm.vectorize_builtin_md_vectorized_function if callee is
14964 2018-01-22 Carl Love <cel@us.ibm.com>
14966 * doc/extend.tex: Fix typo in second arg in
14967 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14969 2018-01-29 Richard Biener <rguenther@suse.de>
14971 PR tree-optimization/84086
14972 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14973 (flush_ssaname_freelist): When SSA names were released reset
14974 the SCEV hash table.
14976 2018-01-29 Richard Biener <rguenther@suse.de>
14978 PR tree-optimization/84057
14979 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14980 removed paths when removing edges.
14982 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14984 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14985 -mfunction-return=@var{choice}.
14987 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14989 PR diagnostic/84034
14990 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14991 Handle CR like TAB.
14992 (layout::print_source_line): Likewise.
14993 (test_get_line_width_without_trailing_whitespace): Add test cases.
14995 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14997 PR middle-end/84040
14998 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15001 2018-01-26 Jim Wilson <jimw@sifive.com>
15003 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15005 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15008 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15010 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15011 and CMP + SUB-immediate -> SUBS.
15013 2018-01-26 Martin Sebor <msebor@redhat.com>
15015 PR tree-optimization/83896
15016 * tree-ssa-strlen.c (get_string_len): Rename...
15017 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15018 Avoid assuming length is constant.
15019 (handle_char_store): Use HOST_WIDE_INT for string length.
15021 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15024 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15025 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15027 2018-01-26 Richard Biener <rguenther@suse.de>
15029 PR rtl-optimization/84003
15030 * dse.c (record_store): Only record redundant stores when
15031 the earlier store aliases at least all accesses the later one does.
15033 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15035 PR rtl-optimization/83985
15036 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15037 REG_CFA_RESTORE insns.
15038 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15039 REG_CFA_RESTORE insns here.
15042 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15043 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15045 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15047 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15048 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15049 (arc_init): Likewise.
15050 (arc_override_options): Likewise.
15051 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15053 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15055 * config/arc/arc.h (TARGET_DBNZ): Define.
15056 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15057 properly set the tune attribute.
15058 (dbnz): Use TARGET_DBNZ guard.
15059 * config/arc/arc.opt (mtune): Add core3 option.
15061 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15063 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15064 recognize new pic like addresses.
15065 (arc_delegitimize_address): Clean up.
15067 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15069 * config/arc/arc-arches.def: Option mrf16 valid for all
15071 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15072 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15073 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15074 * config/arc/arc-tables.opt: Regenerate.
15075 * config/arc/arc.c (arc_conditional_register_usage): Handle
15076 reduced register file case.
15077 (arc_file_start): Set must have build attributes.
15078 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15079 mrf16 option value.
15080 * config/arc/arc.opt (mrf16): Add new option.
15081 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15082 * config/arc/genmultilib.awk: Handle new mrf16 option.
15083 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15084 * config/arc/t-multilib: Regenerate.
15085 * doc/invoke.texi (ARC Options): Document mrf16 option.
15087 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15089 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15090 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15091 (arc_attribute_table): Add 'secure_call' attribute.
15092 (arc_print_operand): Print secure call operand.
15093 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15095 (arc_is_secure_call_p): New function. * config/arc/arc.md
15096 (call_i): Add support for sjli instruction.
15097 (call_value_i): Likewise.
15098 * config/arc/constraints.md (Csc): New constraint.
15100 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15101 John Eric Martin <John.Martin@emmicro-us.com>
15103 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15104 * config/arc/arc.c (_arc_jli_section): New struct.
15105 (arc_jli_section): New type.
15106 (rc_jli_sections): New static variable.
15107 (arc_handle_jli_attribute): New function.
15108 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15109 (arc_file_end): New function.
15110 (TARGET_ASM_FILE_END): Define.
15111 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15112 (arc_add_jli_section): New function.
15113 (jli_call_scan): Likewise.
15114 (arc_reorg): Call jli_call_scan.
15115 (arc_output_addsi): Remove 'S' from printing asm operand.
15116 (arc_is_jli_call_p): New function.
15117 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15119 (movhi_insn): Likewise.
15120 (movsi_insn): Likewise.
15121 (movsi_set_cc_insn): Likewise.
15122 (loadqi_update): Likewise.
15123 (load_zeroextendqisi_update): Likewise.
15124 (load_signextendqisi_update): Likewise.
15125 (loadhi_update): Likewise.
15126 (load_zeroextendhisi_update): Likewise.
15127 (load_signextendhisi_update): Likewise.
15128 (loadsi_update): Likewise.
15129 (loadsf_update): Likewise.
15130 (movsicc_insn): Likewise.
15131 (bset_insn): Likewise.
15132 (bxor_insn): Likewise.
15133 (bclr_insn): Likewise.
15134 (bmsk_insn): Likewise.
15135 (bicsi3_insn): Likewise.
15136 (cmpsi_cc_c_insn): Likewise.
15137 (movsi_ne): Likewise.
15138 (movsi_cond_exec): Likewise.
15139 (clrsbsi2): Likewise.
15140 (norm_f): Likewise.
15150 (call_i): Remove 'S' asm letter, add jli instruction.
15151 (call_value_i): Likewise.
15152 * config/arc/arc.op (mjli-always): New option.
15153 * config/arc/constraints.md (Cji): New constraint.
15154 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15156 (subsf3_fpx): Likewise.
15157 (mulsf3_fpx): Likewise.
15158 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15160 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15161 function attrbutes.
15162 * doc/invoke.texi (ARC): Document mjli-always option.
15164 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15166 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15167 avoid addition with 0 and use incw and decw where possible.
15169 2018-01-26 Richard Biener <rguenther@suse.de>
15171 PR tree-optimization/81082
15172 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15173 association if it requires casting to unsigned.
15174 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15175 from fold_plusminus_mult_expr to catch important cases late when
15176 range info is available.
15178 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15180 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15181 * configure.ac (hidden_linkonce): New test.
15182 * configure: Regenerate.
15183 * config.in: Regenerate.
15185 2018-01-26 Julia Koval <julia.koval@intel.com>
15187 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15188 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15189 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15190 _mm_mask_bitshuffle_epi64_mask): Fix type.
15191 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15192 USI_FTYPE_V4DI_V4DI_USI): Remove.
15193 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15194 __builtin_ia32_vpshufbitqmb256_mask,
15195 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15196 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15197 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15199 2018-01-26 Alan Modra <amodra@gmail.com>
15202 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15203 UNSPEC_VBPERMQ. Sort other unspecs.
15205 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15207 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15209 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15211 PR middle-end/83055
15212 * predict.c (drop_profile): Do not push/pop cfun; update also
15214 (handle_missing_profiles): Fix logic looking for zero profiles.
15216 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15218 PR middle-end/83977
15219 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15220 on functions with #pragma omp declare simd or functions with simd
15222 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15223 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15224 Remove trailing \n from warning_at calls.
15226 2018-01-25 Tom de Vries <tom@codesourcery.com>
15229 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15230 for neutered workers.
15232 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15235 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15236 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15238 2018-01-24 Jeff Law <law@redhat.com>
15241 * i386.c (get_probe_interval): Move to earlier point.
15242 (ix86_compute_frame_layout): If -fstack-clash-protection and
15243 the frame is larger than the probe interval, then use pushes
15244 to save registers rather than reg->mem moves.
15245 (ix86_expand_prologue): Remove conditional for int_registers_saved
15248 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15251 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15252 min/max for never referenced object.
15254 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15256 PR middle-end/83977
15257 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15259 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15260 attributes from DECL_ATTRIBUTES (decl) without affecting
15261 DECL_ATTRIBUTES (current_function_decl).
15262 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15263 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15265 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15267 PR tree-optimization/83979
15268 * fold-const.c (fold_comparison): Use constant_boolean_node
15269 instead of boolean_{true,false}_node.
15271 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15273 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15276 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15278 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15279 Simplify the clause that sets the length attribute.
15280 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15281 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15282 clause that sets the length attribute.
15283 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15285 2018-01-24 Tom de Vries <tom@codesourcery.com>
15288 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15289 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15290 Add strict parameter.
15291 (prevent_branch_around_nothing): Insert dummy insn between branch to
15292 label and label with no ptx insn inbetween.
15293 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15295 2018-01-24 Tom de Vries <tom@codesourcery.com>
15298 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15299 for neutered threads in warp.
15300 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15302 2018-01-24 Richard Biener <rguenther@suse.de>
15304 PR tree-optimization/83176
15305 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15308 2018-01-24 Richard Biener <rguenther@suse.de>
15310 PR tree-optimization/82819
15311 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15312 code generating pluses that are no-ops in the target precision.
15314 2018-01-24 Richard Biener <rguenther@suse.de>
15316 PR middle-end/84000
15317 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15319 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15321 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15322 to merge probabilities.
15323 * predict.c (probably_never_executed): Also mark as cold functions
15324 with global 0 profile and guessed local profile.
15325 * profile-count.c (profile_probability::combine_with_count): New
15327 * profile-count.h (profile_probability::operator*,
15328 profile_probability::operator*=, profile_probability::operator/,
15329 profile_probability::operator/=): Reduce precision to adjusted
15330 and set value to guessed on contradictory divisions.
15331 (profile_probability::combine_with_freq): Remove.
15332 (profile_probability::combine_wiht_count): Declare.
15333 (profile_count::force_nonzero):: Set to adjusted.
15334 (profile_count::probability_in):: Set quality to adjusted.
15335 * tree-ssa-tail-merge.c (replace_block_by): Use
15336 combine_with_count.
15338 2018-01-23 Andrew Waterman <andrew@sifive.com>
15339 Jim Wilson <jimw@sifive.com>
15341 * config/riscv/riscv.c (riscv_stack_boundary): New.
15342 (riscv_option_override): Set riscv_stack_boundary. Handle
15343 riscv_preferred_stack_boundary_arg.
15344 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15345 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15346 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15347 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15348 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15349 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15351 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15354 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15355 of struct ix86_frame.
15356 (ix86_expand_epilogue): Likewise. Add a local variable for
15357 the reg_save_offset field in struct ix86_frame.
15359 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15361 PR tree-optimization/82604
15362 * tree-loop-distribution.c (enum partition_kind): New enum item
15363 PKIND_PARTIAL_MEMSET.
15364 (partition_builtin_p): Support above new enum item.
15365 (generate_code_for_partition): Ditto.
15366 (compute_access_range): Differentiate cases that equality can be
15367 proven at all loops, the innermost loops or no loops.
15368 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15369 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15370 (finalize_partitions, distribute_loop): Don't fuse partition of
15371 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15372 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15373 parloop is enabled.
15375 2018-01-23 Martin Liska <mliska@suse.cz>
15377 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15378 order to ignore the predictor.
15379 (PRED_POLYMORPHIC_CALL): Likewise.
15380 (PRED_RECURSIVE_CALL): Likewise.
15382 2018-01-23 Martin Liska <mliska@suse.cz>
15384 * tree-profile.c (tree_profiling): Print function header to
15385 aware reader which function we are working on.
15386 * value-prof.c (gimple_find_values_to_profile): Do not print
15387 not interesting value histograms.
15389 2018-01-23 Martin Liska <mliska@suse.cz>
15391 * profile-count.h (enum profile_quality): Add
15392 profile_uninitialized as the first value. Do not number values
15393 as they are zero based.
15394 (profile_count::verify): Update sanity check.
15395 (profile_probability::verify): Likewise.
15397 2018-01-23 Nathan Sidwell <nathan@acm.org>
15399 * doc/invoke.texi (ffor-scope): Deprecate.
15401 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15403 PR tree-optimization/83510
15404 * domwalk.c (set_all_edges_as_executable): New function.
15405 (dom_walker::dom_walker): Convert bool param
15406 "skip_unreachable_blocks" to enum reachability. Move setup of
15407 edge flags to set_all_edges_as_executable and only do it when
15408 reachability is REACHABLE_BLOCKS.
15409 * domwalk.h (enum dom_walker::reachability): New enum.
15410 (dom_walker::dom_walker): Convert bool param
15411 "skip_unreachable_blocks" to enum reachability.
15412 (set_all_edges_as_executable): New decl.
15413 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15414 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15416 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15417 but converting true to REACHABLE_BLOCKS.
15418 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15420 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15421 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15422 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15424 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15425 if check_all_array_refs will be called.
15427 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15429 * tree.c (selftest::test_location_wrappers): Add more test
15432 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15434 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15435 (selftest::test_bit_in_range): Likewise.
15437 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15440 * doc/sourcebuild.texi (vect_float): Say that the selector
15441 only describes the situation when -funsafe-math-optimizations is on.
15442 (vect_float_strict): Document.
15444 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15446 PR tree-optimization/83965
15447 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15448 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15449 instead of checking only for a reduction.
15450 (vect_recog_widen_sum_pattern): Likewise.
15452 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15454 * predict.c (probably_never_executed): Only use precise profile info.
15455 (compute_function_frequency): Skip after inlining hack since we now
15456 have quality checking.
15458 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15460 * profile-count.h (profile_probability::very_unlikely,
15461 profile_probability::unlikely, profile_probability::even): Set
15462 precision to guessed.
15464 2018-01-23 Richard Biener <rguenther@suse.de>
15466 PR tree-optimization/83963
15467 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15468 Properly terminate dominator walk when crossing the exit edge not
15469 when visiting its source block.
15471 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15474 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15475 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15477 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15479 PR tree-optimization/83957
15480 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15481 semicolon after for body surrounded by braces.
15483 PR tree-optimization/83081
15484 * profile-count.h (profile_probability::split): New method.
15485 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15486 Use profile_probability::split.
15487 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15488 when splitting a single conditional jump into 2.
15490 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15492 PR tree-optimization/69452
15493 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15496 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15498 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15499 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15500 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15502 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15504 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15506 * config/rl78/rl78.md (movdi): New define_expand.
15507 * config/rl78/rl78.c (rl78_split_movdi): New function.
15509 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15512 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15514 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15515 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15516 128-bit to produce an UNSPEC move to get the double word with the
15517 signbit and then a shift directly to do signbit.
15518 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15519 implementation with a new version that just does either a direct
15520 move or a regular move. Move memory interface to separate insns.
15521 Move insns so they are next to the expander.
15522 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15523 with signbit move. Split big and little endian case.
15524 (signbit<mode>2_dm_mem_le): Likewise.
15525 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15526 (signbit<mode>2_dm2): Likewise.
15528 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15530 * config/rl78/rl78.md (anddi3): New define_expand.
15532 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15534 * config/rl78/rl78.md (umindi3): New define_expand.
15536 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15538 * config/rl78/rl78.md (smindi3): New define_expand.
15540 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15542 * config/rl78/rl78.md (smaxdi3): New define_expand.
15544 2018-01-22 Carl Love <cel@us.ibm.com>
15546 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15547 LVX_V1TI): Add macro expansion.
15548 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15549 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15550 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15551 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15552 Change check to determine if the instruction is a byte reversing
15553 entry. Fix typo in comment.
15554 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15555 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15556 Add def_builtin calls for new builtins.
15557 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15558 Add define_insn expansion.
15560 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15562 * config/rl78/rl78.md (umaxdi3): New define_expand.
15564 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15566 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15567 for non-QImode registers.
15569 2018-01-22 Richard Biener <rguenther@suse.de>
15571 PR tree-optimization/83963
15572 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15573 including the loop exit block.
15574 (scop_detection::merge_sese): Likewise.
15575 (scop_detection::add_scop): Do it here instead.
15577 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15579 * doc/sourcebuild.texi (arm_softfloat): Document.
15581 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15584 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15585 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15586 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15588 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15589 David Edelsohn <dje.gcc@gmail.com>
15592 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15593 Change "crset eq" to "crset 2".
15594 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15595 (*call_indirect_aix<mode>_nospec): Likewise.
15596 (*call_value_indirect_aix<mode>_nospec): Likewise.
15597 (*call_indirect_elfv2<mode>_nospec): Likewise.
15598 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15599 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15600 change assembly output from . to $.
15601 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15602 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15603 (*tablejump<mode>_internal1_nospec): Likewise.
15605 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15608 * config/sh/sh_optimize_sett_clrt.cc:
15609 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15611 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15613 PR tree-optimization/83940
15614 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15615 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15616 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15617 use it to pass back the definition type.
15618 (vect_check_store_rhs): Likewise rhs_dt_out.
15619 (vect_build_gather_load_calls): Add a mask_dt argument and use
15620 it instead of a call to vect_is_simple_use.
15621 (vectorizable_store): Update calls to vect_check_load_store_mask
15622 and vect_check_store_rhs. Use the dt returned by the latter instead
15623 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15624 instead of calls to vect_is_simple_use. Pass the scalar rather
15625 than the vector operand to vect_is_simple_use when handling
15626 second and subsequent copies of an rhs value.
15627 (vectorizable_load): Update calls to vect_check_load_store_mask
15628 and vect_build_gather_load_calls. Use the cached mask_dt and
15629 gs_info.offset_dt instead of calls to vect_is_simple_use.
15631 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15633 PR middle-end/83945
15634 * tree-emutls.c: Include gimplify.h.
15635 (lower_emutls_2): New function.
15636 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15637 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15638 it before further processing.
15641 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15642 UINTVAL (trueop1) instead of INTVAL (op1).
15644 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15648 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15649 INCOMING_FRAME_SP_OFFSET if not defined.
15650 (scan_trace): Add ENTRY argument. If true and
15651 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15652 emit a note to adjust the CFA offset.
15653 (create_cfi_notes): Adjust scan_trace callers.
15654 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15655 INCOMING_FRAME_SP_OFFSET in the CIE.
15656 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15657 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15659 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15660 * doc/tm.texi: Regenerated.
15662 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15664 PR rtl-optimization/83147
15665 * lra-constraints.c (remove_inheritance_pseudos): Use
15666 lra_substitute_pseudo_within_insn.
15668 2018-01-19 Tom de Vries <tom@codesourcery.com>
15669 Cesar Philippidis <cesar@codesourcery.com>
15672 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15674 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15677 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15678 spaces for function labels.
15680 2018-01-19 Martin Liska <mliska@suse.cz>
15682 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15683 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15684 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15685 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15686 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15687 (PRED_CONST_RETURN): Change from 69 to 65.
15688 (PRED_NULL_RETURN): Change from 91 to 71.
15689 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15690 (PRED_LOOP_GUARD): Change from 66 to 73.
15692 2018-01-19 Martin Liska <mliska@suse.cz>
15694 * predict.c (predict_insn_def): Add new assert.
15695 (struct branch_predictor): Change type to signed integer.
15696 (test_prediction_value_range): Amend test to cover
15697 PROB_UNINITIALIZED.
15698 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15699 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15700 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15701 (PRED_LOOP_IV_COMPARE): Likewise.
15702 * predict.h (PROB_UNINITIALIZED): Define new constant.
15704 2018-01-19 Martin Liska <mliska@suse.cz>
15706 * predict.c (dump_prediction): Add new format for
15707 analyze_brprob.py script which is enabled with -details
15709 * profile-count.h (precise_p): New function.
15711 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15713 PR tree-optimization/83922
15714 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15715 there are no statements that need masking.
15716 (vect_active_double_reduction_p): New function.
15717 (vect_analyze_loop_operations): Use it when handling phis that
15718 are not in the loop header.
15720 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15722 PR tree-optimization/83914
15723 * tree-vect-loop.c (vectorizable_induction): Don't convert
15724 init_expr or apply the peeling adjustment for inductions
15725 that are nested within the vectorized loop.
15727 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15729 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15732 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15736 * function.h (gimplify_parameters): Add gimple_seq * argument.
15737 * function.c: Include gimple.h and options.h.
15738 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15739 for the added local temporaries if needed.
15740 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15741 if there are any parameter cleanups, wrap whole body into a
15742 try/finally with the cleanups.
15744 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15747 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15748 Use GET_MODE_CLASS for scalar floating point.
15750 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15754 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15755 Fix call of call_cgraph_insertion_hooks.
15757 2018-01-18 Martin Sebor <msebor@redhat.com>
15759 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15761 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15764 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15767 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15770 * common.opt: (-ffile-prefix-map): New option.
15771 * opts.c (common_handle_option): Defer it.
15772 * opts-global.c (handle_common_deferred_options): Handle it.
15773 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15774 * file-prefix-map.h: New file.
15775 (remap_debug_filename, add_debug_prefix_map): ...here.
15776 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15777 * final.c (debug_prefix_map, add_debug_prefix_map
15778 remap_debug_filename): Move to...
15779 * file-prefix-map.c: New file.
15780 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15781 generalize, get rid of alloca(), use strrchr() instead of strchr().
15782 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15783 Implement in terms of add_prefix_map().
15784 (remap_macro_filename, remap_debug_filename): Implement in term of
15786 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15787 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15788 * dbxout.c: Include file-prefix-map.h.
15789 * varasm.c: Likewise.
15790 * vmsdbgout.c: Likewise.
15791 * xcoffout.c: Likewise.
15792 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15793 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15794 * doc/invoke.texi (-ffile-prefix-map): Document.
15795 (-fdebug-prefix-map): Update description.
15797 2018-01-18 Martin Liska <mliska@suse.cz>
15799 * config/i386/i386.c (indirect_thunk_name): Document that also
15801 (output_indirect_thunk): Document why both instructions
15802 (pause and lfence) are generated.
15804 2018-01-18 Richard Biener <rguenther@suse.de>
15806 PR tree-optimization/83887
15807 * graphite-scop-detection.c
15808 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15809 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15810 (scop_detection::merge_sese): Re-implement with a flood-fill
15811 algorithm that properly finds a SESE region if it exists.
15813 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15816 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15817 pointer_diff optimizations use view_convert instead of convert.
15819 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15821 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15822 Generate different code for -mno-speculate-indirect-jumps.
15823 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15824 (*call_indirect_aix<mode>): Disable for
15825 -mno-speculate-indirect-jumps.
15826 (*call_indirect_aix<mode>_nospec): New define_insn.
15827 (*call_value_indirect_aix<mode>): Disable for
15828 -mno-speculate-indirect-jumps.
15829 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15830 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15831 -mno-speculate-indirect-jumps.
15832 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15834 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15836 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15837 long double type, set the flags for noting the default long double
15838 type, even if we don't pass or return a long double type.
15840 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15843 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15846 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15848 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15849 support for merge[hl].
15850 (fold_mergehl_helper): New helper function.
15851 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15852 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15853 (altivec_vmrglw_direct): Add xxmrglw insn.
15855 2018-01-17 Andrew Waterman <andrew@sifive.com>
15857 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15858 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15860 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15863 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15864 call the lto_location_cache before reading the
15865 DECL_SOURCE_LOCATION of the types.
15867 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15868 Richard Sandiford <richard.sandiford@linaro.org>
15870 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15871 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15872 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15873 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15874 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15876 * config/aarch64/constraints.md (aarch64_movti_operand):
15878 * config/aarch64/predicates.md (Uti): Add new constraint.
15880 2018-01-17 Carl Love <cel@us.ibm.com>
15882 * config/rs6000/vsx.md (define_expand xl_len_r,
15883 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15884 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15886 (define_expand, define_insn): Move the shift left from the
15887 define_insn to the define_expand for lxvl and stxvl instructions.
15888 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15889 and XL_LEN_R definitions to PURE.
15891 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15893 * config/i386/i386.c (indirect_thunk_name): Declare regno
15894 as unsigned int. Compare regno with INVALID_REGNUM.
15895 (output_indirect_thunk): Ditto.
15896 (output_indirect_thunk_function): Ditto.
15897 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15898 in the call to output_indirect_thunk_function.
15900 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15902 PR middle-end/83884
15903 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15904 rather than the size of inner_type to determine the stack slot size
15905 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15907 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15910 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15913 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15915 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15916 endian Linux systems to optionally enable multilibs for selecting
15917 the long double type if the user configured an explicit type.
15918 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15919 have no long double multilibs if not defined.
15920 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15921 warn if the user used -mabi={ieee,ibm}longdouble and we built
15922 multilibs for long double.
15923 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15924 appropriate multilib option.
15925 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15927 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15928 for building long double multilibs.
15929 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15931 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15933 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15936 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15938 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15941 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15944 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15947 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15949 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15950 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15952 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15954 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15955 different rtl trees depending on TARGET_64BIT.
15956 (rs6000_gen_lvx): Likewise.
15958 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15960 * config/visium/visium.md (nop): Tweak comment.
15961 (hazard_nop): Likewise.
15963 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15965 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15966 -mspeculate-indirect-jumps.
15967 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15968 for -mno-speculate-indirect-jumps.
15969 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15970 (*call_value_indirect_elfv2<mode>): Disable for
15971 -mno-speculate-indirect-jumps.
15972 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15973 (indirect_jump): Emit different RTL for
15974 -mno-speculate-indirect-jumps.
15975 (*indirect_jump<mode>): Disable for
15976 -mno-speculate-indirect-jumps.
15977 (*indirect_jump<mode>_nospec): New define_insn.
15978 (tablejump): Emit different RTL for
15979 -mno-speculate-indirect-jumps.
15980 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15981 (tablejumpsi_nospec): New define_expand.
15982 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15983 (tablejumpdi_nospec): New define_expand.
15984 (*tablejump<mode>_internal1): Disable for
15985 -mno-speculate-indirect-jumps.
15986 (*tablejump<mode>_internal1_nospec): New define_insn.
15987 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15990 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15992 * caller-save.c (insert_save): Drop unnecessary parameter. All
15995 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15996 Richard Biener <rguenth@suse.de>
15999 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16000 return early, inline manually is_gimple_sizepos. Make sure if we
16001 call gimplify_expr we don't end up with a gimple constant.
16002 * tree.c (variably_modified_type_p): Don't return true for
16003 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16004 * gimplify.h (is_gimple_sizepos): Remove.
16006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16008 PR tree-optimization/83857
16009 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16010 vectorizable_live_operation for pure SLP statements.
16011 (vectorizable_live_operation): Handle PHIs.
16013 2018-01-16 Richard Biener <rguenther@suse.de>
16015 PR tree-optimization/83867
16016 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16017 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16019 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16022 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16023 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16024 If off is not INTEGER_CST, issue a may not be aligned warning
16025 rather than isn't aligned. Use isn%'t rather than isn't.
16026 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16028 <case MULT_EXPR>: Improve the case when bottom and one of the
16029 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16030 operand, in that case check if the other operand is multiple of
16031 bottom divided by the INTEGER_CST operand.
16033 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16036 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16037 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16038 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16039 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16040 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16041 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16042 (pa_function_arg_size): New function.
16044 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16046 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16047 in a separate statement.
16049 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16051 PR tree-optimization/83847
16052 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16053 group gathers and scatters.
16055 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16057 PR rtl-optimization/86620
16058 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16060 PR rtl-optimization/83213
16061 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16062 to last if both are JUMP_INSNs.
16064 PR tree-optimization/83843
16065 * gimple-ssa-store-merging.c
16066 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16067 store_immediate_info for bswap/nop orig_stores.
16069 2018-01-15 Andrew Waterman <andrew@sifive.com>
16071 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16073 <UDIV>: Increase cost if !TARGET_DIV.
16075 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16077 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16078 (define_attr "cr_logical_3op"): New.
16079 (cceq_ior_compare): Adjust.
16080 (cceq_ior_compare_complement): Adjust.
16081 (*cceq_rev_compare): Adjust.
16082 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16083 (is_cracked_insn): Adjust.
16084 (insn_must_be_first_in_group): Adjust.
16085 * config/rs6000/40x.md: Adjust.
16086 * config/rs6000/440.md: Adjust.
16087 * config/rs6000/476.md: Adjust.
16088 * config/rs6000/601.md: Adjust.
16089 * config/rs6000/603.md: Adjust.
16090 * config/rs6000/6xx.md: Adjust.
16091 * config/rs6000/7450.md: Adjust.
16092 * config/rs6000/7xx.md: Adjust.
16093 * config/rs6000/8540.md: Adjust.
16094 * config/rs6000/cell.md: Adjust.
16095 * config/rs6000/e300c2c3.md: Adjust.
16096 * config/rs6000/e500mc.md: Adjust.
16097 * config/rs6000/e500mc64.md: Adjust.
16098 * config/rs6000/e5500.md: Adjust.
16099 * config/rs6000/e6500.md: Adjust.
16100 * config/rs6000/mpc.md: Adjust.
16101 * config/rs6000/power4.md: Adjust.
16102 * config/rs6000/power5.md: Adjust.
16103 * config/rs6000/power6.md: Adjust.
16104 * config/rs6000/power7.md: Adjust.
16105 * config/rs6000/power8.md: Adjust.
16106 * config/rs6000/power9.md: Adjust.
16107 * config/rs6000/rs64.md: Adjust.
16108 * config/rs6000/titan.md: Adjust.
16110 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16112 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16113 ix86_indirect_branch_register logic.
16115 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16117 * config/i386/constraints.md (Bs): Update
16118 ix86_indirect_branch_register check. Don't check
16119 ix86_indirect_branch_register with GOT_memory_operand.
16121 * config/i386/predicates.md (GOT_memory_operand): Don't check
16122 ix86_indirect_branch_register here.
16123 (GOT32_symbol_operand): Likewise.
16125 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16127 * config/i386/predicates.md (constant_call_address_operand):
16128 Rewrite ix86_indirect_branch_register logic.
16129 (sibcall_insn_operand): Likewise.
16131 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16133 * config/i386/constraints.md (Bs): Replace
16134 ix86_indirect_branch_thunk_register with
16135 ix86_indirect_branch_register.
16137 * config/i386/i386.md (indirect_jump): Likewise.
16138 (tablejump): Likewise.
16139 (*sibcall_memory): Likewise.
16140 (*sibcall_value_memory): Likewise.
16141 Peepholes of indirect call and jump via memory: Likewise.
16142 * config/i386/i386.opt: Likewise.
16143 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16144 (GOT_memory_operand): Likewise.
16145 (call_insn_operand): Likewise.
16146 (sibcall_insn_operand): Likewise.
16147 (GOT32_symbol_operand): Likewise.
16149 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16151 PR middle-end/83837
16152 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16153 type rather than type addr's type points to.
16154 (expand_omp_atomic_mutex): Likewise.
16155 (expand_omp_atomic): Likewise.
16157 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16160 * config/i386/i386.c (output_indirect_thunk_function): Use
16161 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16162 for __x86_return_thunk.
16164 2018-01-15 Richard Biener <rguenther@suse.de>
16166 PR middle-end/83850
16167 * expmed.c (extract_bit_field_1): Fix typo.
16169 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16172 * config/arm/iterators.md (VF): New mode iterator.
16173 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16174 Remove integer-related logic from pattern.
16175 (neon_vabd<mode>_3): Likewise.
16177 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16179 PR middle-end/82694
16180 * common.opt (fstrict-overflow): No longer an alias.
16181 (fwrapv-pointer): New option.
16182 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16183 also for pointer types based on flag_wrapv_pointer.
16184 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16185 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16186 opts->x_flag_wrapv got set.
16187 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16188 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16189 POINTER_TYPE_OVERFLOW_UNDEFINED.
16190 * match.pd: Likewise in address comparison pattern.
16191 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16193 2018-01-15 Richard Biener <rguenther@suse.de>
16196 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16197 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16198 Reset type names to their identifier if their TYPE_DECL doesn't
16199 have linkage (and thus is used for ODR and devirt).
16200 (save_debug_info_for_decl): Remove.
16201 (save_debug_info_for_type): Likewise.
16202 (add_tree_to_fld_list): Adjust.
16203 * tree-pretty-print.c (dump_generic_node): Make dumping of
16204 type names more robust.
16206 2018-01-15 Richard Biener <rguenther@suse.de>
16208 * BASE-VER: Bump to 8.0.1.
16210 2018-01-14 Martin Sebor <msebor@redhat.com>
16213 * builtins.c (check_access): Avoid warning when the no-warning bit
16216 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16218 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16219 * ira-color (allocno_hard_regs_compare): Likewise.
16221 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16224 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16225 Use .pushsection/.popsection.
16227 2018-01-14 Martin Sebor <msebor@redhat.com>
16230 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16232 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16234 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16235 entry from extra_headers.
16236 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16237 extra_headers, make the list bitwise identical to the i?86-*-* one.
16239 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16241 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16242 -mcmodel=large with -mindirect-branch=thunk,
16243 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16244 -mfunction-return=thunk-extern.
16245 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16246 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16247 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16249 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16251 * config/i386/i386.c (print_reg): Print the name of the full
16252 integer register without '%'.
16253 (ix86_print_operand): Handle 'V'.
16254 * doc/extend.texi: Document 'V' modifier.
16256 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16258 * config/i386/constraints.md (Bs): Disallow memory operand for
16259 -mindirect-branch-register.
16261 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16262 (GOT_memory_operand): Likewise.
16263 (call_insn_operand): Likewise.
16264 (sibcall_insn_operand): Likewise.
16265 (GOT32_symbol_operand): Likewise.
16266 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16267 for -mindirect-branch-register.
16268 (tablejump): Likewise.
16269 (*sibcall_memory): Likewise.
16270 (*sibcall_value_memory): Likewise.
16271 Disallow peepholes of indirect call and jump via memory for
16272 -mindirect-branch-register.
16273 (*call_pop): Replace m with Bw.
16274 (*call_value_pop): Likewise.
16275 (*sibcall_pop_memory): Replace m with Bs.
16276 * config/i386/i386.opt (mindirect-branch-register): New option.
16277 * doc/invoke.texi: Document -mindirect-branch-register option.
16279 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16281 * config/i386/i386-protos.h (ix86_output_function_return): New.
16282 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16283 set function_return_type.
16284 (indirect_thunk_name): Add ret_p to indicate thunk for function
16286 (output_indirect_thunk_function): Pass false to
16287 indirect_thunk_name.
16288 (ix86_output_indirect_branch_via_reg): Likewise.
16289 (ix86_output_indirect_branch_via_push): Likewise.
16290 (output_indirect_thunk_function): Create alias for function
16291 return thunk if regno < 0.
16292 (ix86_output_function_return): New function.
16293 (ix86_handle_fndecl_attribute): Handle function_return.
16294 (ix86_attribute_table): Add function_return.
16295 * config/i386/i386.h (machine_function): Add
16296 function_return_type.
16297 * config/i386/i386.md (simple_return_internal): Use
16298 ix86_output_function_return.
16299 (simple_return_internal_long): Likewise.
16300 * config/i386/i386.opt (mfunction-return=): New option.
16301 (indirect_branch): Mention -mfunction-return=.
16302 * doc/extend.texi: Document function_return function attribute.
16303 * doc/invoke.texi: Document -mfunction-return= option.
16305 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16307 * config/i386/i386-opts.h (indirect_branch): New.
16308 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16309 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16310 with local indirect jump when converting indirect call and jump.
16311 (ix86_set_indirect_branch_type): New.
16312 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16313 (indirectlabelno): New.
16314 (indirect_thunk_needed): Likewise.
16315 (indirect_thunk_bnd_needed): Likewise.
16316 (indirect_thunks_used): Likewise.
16317 (indirect_thunks_bnd_used): Likewise.
16318 (INDIRECT_LABEL): Likewise.
16319 (indirect_thunk_name): Likewise.
16320 (output_indirect_thunk): Likewise.
16321 (output_indirect_thunk_function): Likewise.
16322 (ix86_output_indirect_branch_via_reg): Likewise.
16323 (ix86_output_indirect_branch_via_push): Likewise.
16324 (ix86_output_indirect_branch): Likewise.
16325 (ix86_output_indirect_jmp): Likewise.
16326 (ix86_code_end): Call output_indirect_thunk_function if needed.
16327 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16329 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16330 (ix86_attribute_table): Add indirect_branch.
16331 * config/i386/i386.h (machine_function): Add indirect_branch_type
16332 and has_local_indirect_jump.
16333 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16335 (tablejump): Likewise.
16336 (*indirect_jump): Use ix86_output_indirect_jmp.
16337 (*tablejump_1): Likewise.
16338 (simple_return_indirect_internal): Likewise.
16339 * config/i386/i386.opt (mindirect-branch=): New option.
16340 (indirect_branch): New.
16343 (thunk-inline): Likewise.
16344 (thunk-extern): Likewise.
16345 * doc/extend.texi: Document indirect_branch function attribute.
16346 * doc/invoke.texi: Document -mindirect-branch= option.
16348 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16351 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16353 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16355 * ipa-inline.c (want_inline_small_function_p): Return false if
16356 inlining has already failed with CIF_FINAL_ERROR.
16357 (update_caller_keys): Call want_inline_small_function_p before
16359 (update_callee_keys): Likewise.
16361 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16363 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16365 (rs6000_quadword_masked_address_p): Likewise.
16366 (quad_aligned_load_p): Likewise.
16367 (quad_aligned_store_p): Likewise.
16368 (const_load_sequence_p): Add comment to describe the outer-most loop.
16369 (mimic_memory_attributes_and_flags): New function.
16370 (rs6000_gen_stvx): Likewise.
16371 (replace_swapped_aligned_store): Likewise.
16372 (rs6000_gen_lvx): Likewise.
16373 (replace_swapped_aligned_load): Likewise.
16374 (replace_swapped_load_constant): Capitalize argument name in
16375 comment describing this function.
16376 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16377 and stores that access quad-word aligned addresses and replace
16378 with stvx or lvx instructions when appropriate.
16379 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16380 New function prototype.
16381 (rs6000_quadword_masked_address_p): Likewise.
16382 (rs6000_gen_lvx): Likewise.
16383 (rs6000_gen_stvx): Likewise.
16384 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16385 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16386 when memory address is aligned.
16387 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16388 this split to select lvx instruction when memory address is aligned.
16389 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16390 instruction when memory address is aligned.
16391 (*vsx_le_perm_load_v16qi): Likewise.
16392 (four unnamed splitters): Modify to select the stvx instruction
16393 when memory is aligned.
16395 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16397 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16398 which appears in the queue multiple times.
16400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16401 Alan Hayward <alan.hayward@arm.com>
16402 David Sherwood <david.sherwood@arm.com>
16404 * tree-vectorizer.h (vec_lower_bound): New structure.
16405 (_loop_vec_info): Add check_nonzero and lower_bounds.
16406 (LOOP_VINFO_CHECK_NONZERO): New macro.
16407 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16408 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16409 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16410 fields. Make seg_len the distance travelled, not including the
16412 (dr_direction_indicator): Declare.
16413 (dr_zero_step_indicator): Likewise.
16414 (dr_known_forward_stride_p): Likewise.
16415 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16417 (runtime_alias_check_p): Allow runtime alias checks with
16419 (operator ==): Compare access_size and align.
16420 (prune_runtime_alias_test_list): Rework for new distinction between
16421 the access_size and seg_len.
16422 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16424 (get_segment_min_max): New function.
16425 (create_intersect_range_checks): Use it.
16426 (dr_step_indicator): New function.
16427 (dr_direction_indicator): Likewise.
16428 (dr_zero_step_indicator): Likewise.
16429 (dr_known_forward_stride_p): Likewise.
16430 * tree-loop-distribution.c (data_ref_segment_size): Return
16431 DR_STEP * (niters - 1).
16432 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16434 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16435 (vect_preserves_scalar_order_p): New function, split out from...
16436 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16437 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16438 (vect_vfa_access_size): New function.
16439 (vect_vfa_align): Likewise.
16440 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16441 (dump_lower_bound): New function.
16442 (vect_check_lower_bound): Likewise.
16443 (vect_small_gap_p): Likewise.
16444 (vectorizable_with_step_bound_p): Likewise.
16445 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16446 depencies if the vectorization factor is 1. Convert the checks
16447 for nonzero steps into checks on the bounds of DR_STEP. Try using
16448 a bunds check for variable steps if the minimum required step is
16449 relatively small. Update calls to the dr_with_seg_len
16450 constructor and to vect_compile_time_alias.
16451 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16453 (vect_loop_versioning): Call it.
16454 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16456 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16459 Alan Hayward <alan.hayward@arm.com>
16460 David Sherwood <david.sherwood@arm.com>
16462 * doc/sourcebuild.texi (vect_scatter_store): Document.
16463 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16465 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16467 * genopinit.c (main): Add supports_vec_scatter_store and
16468 supports_vec_scatter_store_cached to target_optabs.
16469 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16470 IFN_MASK_SCATTER_STORE.
16471 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16473 * internal-fn.h (internal_store_fn_p): Declare.
16474 (internal_fn_stored_value_index): Likewise.
16475 * internal-fn.c (scatter_store_direct): New macro.
16476 (expand_scatter_store_optab_fn): New function.
16477 (direct_scatter_store_optab_supported_p): New macro.
16478 (internal_store_fn_p): New function.
16479 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16480 IFN_MASK_SCATTER_STORE.
16481 (internal_fn_mask_index): Likewise.
16482 (internal_fn_stored_value_index): New function.
16483 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16484 for scatter stores.
16485 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16486 * optabs-query.c (supports_vec_scatter_store_p): New function.
16487 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16488 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16489 true for scatter stores.
16490 (vect_gather_scatter_fn_p): Handle scatter stores too.
16491 (vect_check_gather_scatter): Consider using scatter stores if
16492 supports_vec_scatter_store_p.
16493 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16494 scatter stores too.
16495 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16496 internal_fn_stored_value_index.
16497 (check_load_store_masking): Handle scatter stores too.
16498 (vect_get_store_rhs): Make public.
16499 (vectorizable_call): Use internal_store_fn_p.
16500 (vectorizable_store): Handle scatter store internal functions.
16501 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16502 when deciding whether the end of the group has been reached.
16503 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16504 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16505 (mask_scatter_store<mode>): New insns.
16507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16508 Alan Hayward <alan.hayward@arm.com>
16509 David Sherwood <david.sherwood@arm.com>
16511 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16512 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16513 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16515 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16516 Use vect_truncate_gather_scatter_offset if we can't treat the
16517 operation as a normal gather load or scatter store.
16518 (get_group_load_store_type): Take the gather_scatter_info
16519 as argument. Try using a gather load or scatter store for
16520 single-element groups.
16521 (get_load_store_type): Update calls to get_group_load_store_type
16522 and vect_use_strided_gather_scatters_p.
16524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16525 Alan Hayward <alan.hayward@arm.com>
16526 David Sherwood <david.sherwood@arm.com>
16528 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16529 optional tree argument.
16530 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16532 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16533 but continue to use the current value as a fallback.
16534 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16535 to compare the updates.
16536 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16537 (get_load_store_type): Use it when handling a strided access.
16538 (vect_get_strided_load_store_ops): New function.
16539 (vect_get_data_ptr_increment): Likewise.
16540 (vectorizable_load): Handle strided gather loads. Always pass
16541 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
16548 (mask_gather_load@var{m}): Likewise.
16549 * genopinit.c (main): Add supports_vec_gather_load and
16550 supports_vec_gather_load_cached to target_optabs.
16551 * optabs-tree.c (init_tree_optimization_optabs): Use
16552 ggc_cleared_alloc to allocate target_optabs.
16553 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16554 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16556 * internal-fn.h (internal_load_fn_p): Declare.
16557 (internal_gather_scatter_fn_p): Likewise.
16558 (internal_fn_mask_index): Likewise.
16559 (internal_gather_scatter_fn_supported_p): Likewise.
16560 * internal-fn.c (gather_load_direct): New macro.
16561 (expand_gather_load_optab_fn): New function.
16562 (direct_gather_load_optab_supported_p): New macro.
16563 (direct_internal_fn_optab): New function.
16564 (internal_load_fn_p): Likewise.
16565 (internal_gather_scatter_fn_p): Likewise.
16566 (internal_fn_mask_index): Likewise.
16567 (internal_gather_scatter_fn_supported_p): Likewise.
16568 * optabs-query.c (supports_at_least_one_mode_p): New function.
16569 (supports_vec_gather_load_p): Likewise.
16570 * optabs-query.h (supports_vec_gather_load_p): Declare.
16571 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16572 and memory_type field.
16573 (NUM_PATTERNS): Bump to 15.
16574 * tree-vect-data-refs.c: Include internal-fn.h.
16575 (vect_gather_scatter_fn_p): New function.
16576 (vect_describe_gather_scatter_call): Likewise.
16577 (vect_check_gather_scatter): Try using internal functions for
16578 gather loads. Recognize existing calls to a gather load function.
16579 (vect_analyze_data_refs): Consider using gather loads if
16580 supports_vec_gather_load_p.
16581 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16582 (vect_get_gather_scatter_offset_type): Likewise.
16583 (vect_convert_mask_for_vectype): Likewise.
16584 (vect_add_conversion_to_patterm): Likewise.
16585 (vect_try_gather_scatter_pattern): Likewise.
16586 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16587 (vect_vect_recog_func_ptrs): Add it.
16588 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16589 internal_fn_mask_index and internal_gather_scatter_fn_p.
16590 (check_load_store_masking): Take the gather_scatter_info as an
16591 argument and handle gather loads.
16592 (vect_get_gather_scatter_ops): New function.
16593 (vectorizable_call): Check internal_load_fn_p.
16594 (vectorizable_load): Likewise. Handle gather load internal
16596 (vectorizable_store): Update call to check_load_store_masking.
16597 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16598 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16599 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16600 (aarch64_gather_scale_operand_d): New predicates.
16601 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16602 (mask_gather_load<mode>): New insns.
16604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16605 Alan Hayward <alan.hayward@arm.com>
16606 David Sherwood <david.sherwood@arm.com>
16608 * optabs.def (fold_left_plus_optab): New optab.
16609 * doc/md.texi (fold_left_plus_@var{m}): Document.
16610 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16611 * internal-fn.c (fold_left_direct): Define.
16612 (expand_fold_left_optab_fn): Likewise.
16613 (direct_fold_left_optab_supported_p): Likewise.
16614 * fold-const-call.c (fold_const_fold_left): New function.
16615 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16616 * tree-parloops.c (valid_reduction_p): New function.
16617 (gather_scalar_reductions): Use it.
16618 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16619 (vect_finish_replace_stmt): Declare.
16620 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16621 (needs_fold_left_reduction_p): New function, split out from...
16622 (vect_is_simple_reduction): ...here. Accept reductions that
16623 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16624 (vect_force_simple_reduction): Also store the reduction type in
16625 the assignment's STMT_VINFO_REDUC_TYPE.
16626 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16627 (merge_with_identity): New function.
16628 (vect_expand_fold_left): Likewise.
16629 (vectorize_fold_left_reduction): Likewise.
16630 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16631 scalar phi in place for it. Check for target support and reject
16632 cases that would reassociate the operation. Defer the transform
16633 phase to vectorize_fold_left_reduction.
16634 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16635 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16636 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16640 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16641 call to ifc_temp_var.
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 * target.def (legitimize_address_displacement): Take the original
16648 offset as a poly_int.
16649 * targhooks.h (default_legitimize_address_displacement): Update
16651 * targhooks.c (default_legitimize_address_displacement): Likewise.
16652 * doc/tm.texi: Regenerate.
16653 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16654 as an argument, moving assert of ad->disp == ad->disp_term to...
16655 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16656 Try calling targetm.legitimize_address_displacement before expanding
16657 the address rather than afterwards, and adjust for the new interface.
16658 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16659 Match the new hook interface. Handle SVE addresses.
16660 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16661 new hook interface.
16663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16665 * Makefile.in (OBJS): Add early-remat.o.
16666 * target.def (select_early_remat_modes): New hook.
16667 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16668 * doc/tm.texi: Regenerate.
16669 * targhooks.h (default_select_early_remat_modes): Declare.
16670 * targhooks.c (default_select_early_remat_modes): New function.
16671 * timevar.def (TV_EARLY_REMAT): New timevar.
16672 * passes.def (pass_early_remat): New pass.
16673 * tree-pass.h (make_pass_early_remat): Declare.
16674 * early-remat.c: New file.
16675 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16677 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16680 Alan Hayward <alan.hayward@arm.com>
16681 David Sherwood <david.sherwood@arm.com>
16683 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16684 vfm1 with a bound_epilog parameter.
16685 (vect_do_peeling): Update calls accordingly, and move the prologue
16686 call earlier in the function. Treat the base bound_epilog as 0 for
16687 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16688 this base when peeling for gaps.
16689 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16690 with fully-masked loops.
16691 (vect_estimate_min_profitable_iters): Handle the single peeled
16692 iteration in that case.
16694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16695 Alan Hayward <alan.hayward@arm.com>
16696 David Sherwood <david.sherwood@arm.com>
16698 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16699 single-element interleaving even if the size is not a power of 2.
16700 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16701 accesses for single-element interleaving if the group size is
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
16709 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16710 * optabs.def (fold_extract_last_optab): New optab.
16711 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16712 * internal-fn.c (fold_extract_direct): New macro.
16713 (expand_fold_extract_optab_fn): Likewise.
16714 (direct_fold_extract_optab_supported_p): Likewise.
16715 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16716 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16717 EXTRACT_LAST_REDUCTION.
16718 (get_initial_def_for_reduction): Do not create an initial vector
16719 for EXTRACT_LAST_REDUCTION reductions.
16720 (vectorizable_reduction): Leave the scalar phi in place for
16721 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16722 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16723 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16724 transform phase to vectorizable_condition.
16725 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16727 (vect_finish_stmt_generation): ...here.
16728 (vect_finish_replace_stmt): New function.
16729 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16730 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16732 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16735 Alan Hayward <alan.hayward@arm.com>
16736 David Sherwood <david.sherwood@arm.com>
16738 * doc/md.texi (extract_last_@var{m}): Document.
16739 * optabs.def (extract_last_optab): New optab.
16740 * internal-fn.def (EXTRACT_LAST): New internal function.
16741 * internal-fn.c (cond_unary_direct): New macro.
16742 (expand_cond_unary_optab_fn): Likewise.
16743 (direct_cond_unary_optab_supported_p): Likewise.
16744 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16745 loops using EXTRACT_LAST.
16746 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16747 (extract_last_<mode>): ...this optab.
16748 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
16755 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16756 * doc/tm.texi: Regenerate.
16757 * targhooks.h (default_empty_mask_is_expensive): Declare.
16758 * targhooks.c (default_empty_mask_is_expensive): New function.
16759 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16760 if the target says that empty masks are expensive.
16761 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16763 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16766 Alan Hayward <alan.hayward@arm.com>
16767 David Sherwood <david.sherwood@arm.com>
16769 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16770 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16771 (vect_use_loop_mask_for_alignment_p): New function.
16772 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16773 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16774 niters_skip argument. Make sure that the first niters_skip elements
16775 of the first iteration are inactive.
16776 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16777 Update call to vect_set_loop_masks_directly.
16778 (get_misalign_in_elems): New function, split out from...
16779 (vect_gen_prolog_loop_niters): ...here.
16780 (vect_update_init_of_dr): Take a code argument that specifies whether
16781 the adjustment should be added or subtracted.
16782 (vect_update_init_of_drs): Likewise.
16783 (vect_prepare_for_masked_peels): New function.
16784 (vect_do_peeling): Skip prologue peeling if we're using a mask
16785 instead. Update call to vect_update_inits_of_drs.
16786 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16788 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16789 alignment. Do not include the number of peeled iterations in
16790 the minimum threshold in that case.
16791 (vectorizable_induction): Adjust the start value down by
16792 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16793 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16794 Take the number of skipped iterations into account when calculating
16796 * tree-vect-stmts.c (vect_gen_while_not): New function.
16798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16799 Alan Hayward <alan.hayward@arm.com>
16800 David Sherwood <david.sherwood@arm.com>
16802 * doc/sourcebuild.texi (vect_fully_masked): Document.
16803 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16804 default value to 0.
16805 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16807 (vect_analyze_loop_2): ...here. Don't check the vectorization
16808 factor against the number of loop iterations if the loop is
16811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16812 Alan Hayward <alan.hayward@arm.com>
16813 David Sherwood <david.sherwood@arm.com>
16815 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16816 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16817 (dump_groups): Update accordingly.
16818 (iv_use::mem_type): New member variable.
16819 (address_p): New function.
16820 (record_use): Add a mem_type argument and initialize the new
16822 (record_group_use): Add a mem_type argument. Use address_p.
16823 Remove obsolete null checks of base_object. Update call to record_use.
16824 (find_interesting_uses_op): Update call to record_group_use.
16825 (find_interesting_uses_cond): Likewise.
16826 (find_interesting_uses_address): Likewise.
16827 (get_mem_type_for_internal_fn): New function.
16828 (find_address_like_use): Likewise.
16829 (find_interesting_uses_stmt): Try find_address_like_use before
16830 calling find_interesting_uses_op.
16831 (addr_offset_valid_p): Use the iv mem_type field as the type
16832 of the addressed memory.
16833 (add_autoinc_candidates): Likewise.
16834 (get_address_cost): Likewise.
16835 (split_small_address_groups_p): Use address_p.
16836 (split_address_groups): Likewise.
16837 (add_iv_candidate_for_use): Likewise.
16838 (autoinc_possible_for_pair): Likewise.
16839 (rewrite_groups): Likewise.
16840 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16841 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16842 (get_alias_ptr_type_for_ptr_address): New function.
16843 (rewrite_use_address): Rewrite address uses in calls that were
16844 identified by find_address_like_use.
16846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16847 Alan Hayward <alan.hayward@arm.com>
16848 David Sherwood <david.sherwood@arm.com>
16850 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16852 * gimple-expr.h (is_gimple_addressable: Likewise.
16853 * gimple-expr.c (is_gimple_address): Likewise.
16854 * internal-fn.c (expand_call_mem_ref): New function.
16855 (expand_mask_load_optab_fn): Use it.
16856 (expand_mask_store_optab_fn): Likewise.
16858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16859 Alan Hayward <alan.hayward@arm.com>
16860 David Sherwood <david.sherwood@arm.com>
16862 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16863 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16864 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16865 (cond_umax@var{mode}): Document.
16866 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16867 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16868 (cond_umin_optab, cond_umax_optab): New optabs.
16869 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16870 (COND_IOR, COND_XOR): New internal functions.
16871 * internal-fn.h (get_conditional_internal_fn): Declare.
16872 * internal-fn.c (cond_binary_direct): New macro.
16873 (expand_cond_binary_optab_fn): Likewise.
16874 (direct_cond_binary_optab_supported_p): Likewise.
16875 (get_conditional_internal_fn): New function.
16876 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16877 Cope with reduction statements that are vectorized as calls rather
16879 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16880 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16881 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16882 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16883 (UNSPEC_COND_EOR): New unspecs.
16884 (optab): Add mappings for them.
16885 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16886 (sve_int_op, sve_fp_op): New int attributes.
16888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16889 Alan Hayward <alan.hayward@arm.com>
16890 David Sherwood <david.sherwood@arm.com>
16892 * optabs.def (while_ult_optab): New optab.
16893 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16894 * internal-fn.def (WHILE_ULT): New internal function.
16895 * internal-fn.h (direct_internal_fn_supported_p): New override
16896 that takes two types as argument.
16897 * internal-fn.c (while_direct): New macro.
16898 (expand_while_optab_fn): New function.
16899 (convert_optab_supported_p): Likewise.
16900 (direct_while_optab_supported_p): New macro.
16901 * wide-int.h (wi::udiv_ceil): New function.
16902 * tree-vectorizer.h (rgroup_masks): New structure.
16903 (vec_loop_masks): New typedef.
16904 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16905 and fully_masked_p.
16906 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16907 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16908 (vect_max_vf): New function.
16909 (slpeel_make_loop_iterate_ntimes): Delete.
16910 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16911 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16912 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16913 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16914 internal-fn.h, stor-layout.h and optabs-query.h.
16915 (vect_set_loop_mask): New function.
16916 (add_preheader_seq): Likewise.
16917 (add_header_seq): Likewise.
16918 (interleave_supported_p): Likewise.
16919 (vect_maybe_permute_loop_masks): Likewise.
16920 (vect_set_loop_masks_directly): Likewise.
16921 (vect_set_loop_condition_masked): Likewise.
16922 (vect_set_loop_condition_unmasked): New function, split out from
16923 slpeel_make_loop_iterate_ntimes.
16924 (slpeel_make_loop_iterate_ntimes): Rename to..
16925 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16926 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16927 (vect_do_peeling): Update call accordingly.
16928 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16930 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16931 mask_compare_type, can_fully_mask_p and fully_masked_p.
16932 (release_vec_loop_masks): New function.
16933 (_loop_vec_info): Use it to free the loop masks.
16934 (can_produce_all_loop_masks_p): New function.
16935 (vect_get_max_nscalars_per_iter): Likewise.
16936 (vect_verify_full_masking): Likewise.
16937 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16938 retries, and free the mask rgroups before retrying. Check loop-wide
16939 reasons for disallowing fully-masked loops. Make the final decision
16940 about whether use a fully-masked loop or not.
16941 (vect_estimate_min_profitable_iters): Do not assume that peeling
16942 for the number of iterations will be needed for fully-masked loops.
16943 (vectorizable_reduction): Disable fully-masked loops.
16944 (vectorizable_live_operation): Likewise.
16945 (vect_halve_mask_nunits): New function.
16946 (vect_double_mask_nunits): Likewise.
16947 (vect_record_loop_mask): Likewise.
16948 (vect_get_loop_mask): Likewise.
16949 (vect_transform_loop): Handle the case in which the final loop
16950 iteration might handle a partial vector. Call vect_set_loop_condition
16951 instead of slpeel_make_loop_iterate_ntimes.
16952 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16953 (check_load_store_masking): New function.
16954 (prepare_load_store_mask): Likewise.
16955 (vectorizable_store): Handle fully-masked loops.
16956 (vectorizable_load): Likewise.
16957 (supportable_widening_operation): Use vect_halve_mask_nunits for
16959 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16960 (vect_gen_while): New function.
16961 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16962 (aarch64_uqdec<mode>): New insn.
16964 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16965 Alan Hayward <alan.hayward@arm.com>
16966 David Sherwood <david.sherwood@arm.com>
16968 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16969 (reduc_xor_scal_optab): New optabs.
16970 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16971 (reduc_xor_scal_@var{m}): Document.
16972 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16973 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16974 internal functions.
16975 * fold-const-call.c (fold_const_call): Handle them.
16976 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16977 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16978 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16979 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16980 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16981 (UNSPEC_XORV): New unspecs.
16982 (optab): Add entries for them.
16983 (BITWISEV): New int iterator.
16984 (bit_reduc_op): New int attributes.
16986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16987 Alan Hayward <alan.hayward@arm.com>
16988 David Sherwood <david.sherwood@arm.com>
16990 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16991 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16992 * optabs.def (vec_shl_insert_optab): New optab.
16993 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16994 (duplicate_and_interleave): Likewise.
16995 * tree-vect-loop.c: Include internal-fn.h.
16996 (neutral_op_for_slp_reduction): New function, split out from
16997 get_initial_defs_for_reduction.
16998 (get_initial_def_for_reduction): Handle option 2 for variable-length
16999 vectors by loading the neutral value into a vector and then shifting
17000 the initial value into element 0.
17001 (get_initial_defs_for_reduction): Replace the code argument with
17002 the neutral value calculated by neutral_op_for_slp_reduction.
17003 Use gimple_build_vector for constant-length vectors.
17004 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17005 but the first group_size elements have a neutral value.
17006 Use duplicate_and_interleave otherwise.
17007 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17008 Update call to get_initial_defs_for_reduction. Handle SLP
17009 reductions for variable-length vectors by creating one vector
17010 result for each scalar result, with the elements associated
17011 with other scalar results stubbed out with the neutral value.
17012 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17013 Require IFN_VEC_SHL_INSERT for double reductions on
17014 variable-length vectors, or SLP reductions that have
17015 a neutral value. Require can_duplicate_and_interleave_p
17016 support for variable-length unchained SLP reductions if there
17017 is no neutral value, such as for MIN/MAX reductions. Also require
17018 the number of vector elements to be a multiple of the number of
17019 SLP statements when doing variable-length unchained SLP reductions.
17020 Update call to vect_create_epilog_for_reduction.
17021 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17022 and remove initial values.
17023 (duplicate_and_interleave): Make public.
17024 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17025 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17028 Alan Hayward <alan.hayward@arm.com>
17029 David Sherwood <david.sherwood@arm.com>
17031 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17032 (can_duplicate_and_interleave_p): New function.
17033 (vect_get_and_check_slp_defs): Take the vector of statements
17034 rather than just the current one. Remove excess parentheses.
17035 Restriction rejectinon of vect_constant_def and vect_external_def
17036 for variable-length vectors to boolean types, or types for which
17037 can_duplicate_and_interleave_p is false.
17038 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17039 (duplicate_and_interleave): New function.
17040 (vect_get_constant_vectors): Use gimple_build_vector for
17041 constant-length vectors and suitable variable-length constant
17042 vectors. Use duplicate_and_interleave for other variable-length
17043 vectors. Don't defer the update when inserting new statements.
17045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17046 Alan Hayward <alan.hayward@arm.com>
17047 David Sherwood <david.sherwood@arm.com>
17049 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17050 min_profitable_iters doesn't go negative.
17052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17053 Alan Hayward <alan.hayward@arm.com>
17054 David Sherwood <david.sherwood@arm.com>
17056 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17057 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17058 * optabs.def (vec_mask_load_lanes_optab): New optab.
17059 (vec_mask_store_lanes_optab): Likewise.
17060 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17061 (MASK_STORE_LANES): Likewise.
17062 * internal-fn.c (mask_load_lanes_direct): New macro.
17063 (mask_store_lanes_direct): Likewise.
17064 (expand_mask_load_optab_fn): Handle masked operations.
17065 (expand_mask_load_lanes_optab_fn): New macro.
17066 (expand_mask_store_optab_fn): Handle masked operations.
17067 (expand_mask_store_lanes_optab_fn): New macro.
17068 (direct_mask_load_lanes_optab_supported_p): Likewise.
17069 (direct_mask_store_lanes_optab_supported_p): Likewise.
17070 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17072 (vect_load_lanes_supported): Likewise.
17073 * tree-vect-data-refs.c (strip_conversion): New function.
17074 (can_group_stmts_p): Likewise.
17075 (vect_analyze_data_ref_accesses): Use it instead of checking
17076 for a pair of assignments.
17077 (vect_store_lanes_supported): Take a masked_p parameter.
17078 (vect_load_lanes_supported): Likewise.
17079 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17080 vect_store_lanes_supported and vect_load_lanes_supported.
17081 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17082 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17083 parameter. Don't allow gaps for masked accesses.
17084 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17085 and vect_load_lanes_supported.
17086 (get_load_store_type): Take a masked_p parameter and update
17087 call to get_group_load_store_type.
17088 (vectorizable_store): Update call to get_load_store_type.
17089 Handle IFN_MASK_STORE_LANES.
17090 (vectorizable_load): Update call to get_load_store_type.
17091 Handle IFN_MASK_LOAD_LANES.
17093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17094 Alan Hayward <alan.hayward@arm.com>
17095 David Sherwood <david.sherwood@arm.com>
17097 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17099 * config/aarch64/aarch64-protos.h
17100 (aarch64_sve_struct_memory_operand_p): Declare.
17101 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17102 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17103 (VPRED, vpred): Handle SVE structure modes.
17104 * config/aarch64/constraints.md (Utx): New constraint.
17105 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17106 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17107 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17108 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17109 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17110 structure modes. Split into pieces after RA.
17111 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17112 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17114 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17115 SVE structure modes.
17116 (aarch64_classify_address): Likewise.
17117 (sizetochar): Move earlier in file.
17118 (aarch64_print_operand): Handle SVE register lists.
17119 (aarch64_array_mode): New function.
17120 (aarch64_sve_struct_memory_operand_p): Likewise.
17121 (TARGET_ARRAY_MODE): Redefine.
17123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17124 Alan Hayward <alan.hayward@arm.com>
17125 David Sherwood <david.sherwood@arm.com>
17127 * target.def (array_mode): New target hook.
17128 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17129 * doc/tm.texi: Regenerate.
17130 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17131 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17132 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17133 targetm.array_mode.
17134 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17138 Alan Hayward <alan.hayward@arm.com>
17139 David Sherwood <david.sherwood@arm.com>
17141 * fold-const.c (fold_binary_loc): Check the argument types
17142 rather than the result type when testing for a vector operation.
17144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17146 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17147 * doc/tm.texi: Regenerate.
17149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17150 Alan Hayward <alan.hayward@arm.com>
17151 David Sherwood <david.sherwood@arm.com>
17153 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17154 (sve): Document new AArch64 extension.
17155 * doc/md.texi (w): Extend the description of the AArch64
17156 constraint to include SVE vectors.
17157 (Upl, Upa): Document new AArch64 predicate constraints.
17158 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17160 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17161 (msve-vector-bits=): New option.
17162 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17163 SVE when these are disabled.
17164 (sve): New extension.
17165 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17166 modes. Adjust their number of units based on aarch64_sve_vg.
17167 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17168 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17169 aarch64_addr_query_type.
17170 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17171 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17172 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17173 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17174 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17175 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17176 (aarch64_simd_imm_zero_p): Delete.
17177 (aarch64_check_zero_based_sve_index_immediate): Declare.
17178 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17179 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17180 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17181 (aarch64_sve_float_mul_immediate_p): Likewise.
17182 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17183 rather than an rtx.
17184 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17185 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17186 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17187 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17188 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17189 (aarch64_regmode_natural_size): Likewise.
17190 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17191 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17193 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17194 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17195 for VG and the SVE predicate registers.
17196 (V_ALIASES): Add a "z"-prefixed alias.
17197 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17198 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17199 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17200 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17201 (REG_CLASS_NAMES): Add entries for them.
17202 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17203 and the predicate registers.
17204 (aarch64_sve_vg): Declare.
17205 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17206 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17207 (REGMODE_NATURAL_SIZE): Define.
17208 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17210 * config/aarch64/aarch64.c: Include cfgrtl.h.
17211 (simd_immediate_info): Add a constructor for series vectors,
17212 and an associated step field.
17213 (aarch64_sve_vg): New variable.
17214 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17215 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17216 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17217 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17218 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17219 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17220 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17221 (aarch64_get_mask_mode): New functions.
17222 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17223 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17224 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17225 predicate modes and predicate registers. Explicitly restrict
17226 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17227 to store a vector mode if it is recognized by
17228 aarch64_classify_vector_mode.
17229 (aarch64_regmode_natural_size): New function.
17230 (aarch64_hard_regno_caller_save_mode): Return the original mode
17232 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17233 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17234 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17235 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17237 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17238 does not overlap dest if the function is frame-related. Handle
17240 (aarch64_split_add_offset): New function.
17241 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17242 them aarch64_add_offset.
17243 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17244 and update call to aarch64_sub_sp.
17245 (aarch64_add_cfa_expression): New function.
17246 (aarch64_expand_prologue): Pass extra temporary registers to the
17247 functions above. Handle the case in which we need to emit new
17248 DW_CFA_expressions for registers that were originally saved
17249 relative to the stack pointer, but now have to be expressed
17250 relative to the frame pointer.
17251 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17253 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17254 IP0 and IP1 values for SVE frames.
17255 (aarch64_expand_vec_series): New function.
17256 (aarch64_expand_sve_widened_duplicate): Likewise.
17257 (aarch64_expand_sve_const_vector): Likewise.
17258 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17259 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17260 into the register, rather than emitting a SET directly.
17261 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17262 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17263 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17264 (offset_9bit_signed_scaled_p): New functions.
17265 (aarch64_replicate_bitmask_imm): New function.
17266 (aarch64_bitmask_imm): Use it.
17267 (aarch64_cannot_force_const_mem): Reject expressions involving
17268 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17269 (aarch64_classify_index): Handle SVE indices, by requiring
17270 a plain register index with a scale that matches the element size.
17271 (aarch64_classify_address): Handle SVE addresses. Assert that
17272 the mode of the address is VOIDmode or an integer mode.
17273 Update call to aarch64_classify_symbol.
17274 (aarch64_classify_symbolic_expression): Update call to
17275 aarch64_classify_symbol.
17276 (aarch64_const_vec_all_in_range_p): New function.
17277 (aarch64_print_vector_float_operand): Likewise.
17278 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17279 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17280 and the FP immediates 1.0 and 0.5.
17281 (aarch64_print_address_internal): Handle SVE addresses.
17282 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17283 (aarch64_regno_regclass): Handle predicate registers.
17284 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17286 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17287 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17288 (aarch64_convert_sve_vector_bits): New function.
17289 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17290 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17291 rather than an rtx.
17292 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17293 Handle SVE vector and predicate modes. Accept VL-based constants
17294 that need only one temporary register, and VL offsets that require
17295 no temporary registers.
17296 (aarch64_conditional_register_usage): Mark the predicate registers
17297 as fixed if SVE isn't available.
17298 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17299 Return true for SVE vector and predicate modes.
17300 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17301 rather than an unsigned int. Handle SVE modes.
17302 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17304 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17306 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17307 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17308 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17309 (aarch64_sve_float_mul_immediate_p): New functions.
17310 (aarch64_sve_valid_immediate): New function.
17311 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17312 Explicitly reject structure modes. Check for INDEX constants.
17313 Handle PTRUE and PFALSE constants.
17314 (aarch64_check_zero_based_sve_index_immediate): New function.
17315 (aarch64_simd_imm_zero_p): Delete.
17316 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17317 vector modes. Accept constants in the range of CNT[BHWD].
17318 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17319 ask for an Advanced SIMD mode.
17320 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17321 (aarch64_simd_vector_alignment): Handle SVE predicates.
17322 (aarch64_vectorize_preferred_vector_alignment): New function.
17323 (aarch64_simd_vector_alignment_reachable): Use it instead of
17325 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17326 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17328 (MAX_VECT_LEN): Delete.
17329 (expand_vec_perm_d): Add a vec_flags field.
17330 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17331 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17332 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17334 (aarch64_evpc_rev): Rename to...
17335 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17336 (aarch64_evpc_rev_global): New function.
17337 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17338 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17340 (aarch64_evpc_sve_tbl): New function.
17341 (aarch64_expand_vec_perm_const_1): Update after rename of
17342 aarch64_evpc_rev. Handle SVE permutes too, trying
17343 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17344 than aarch64_evpc_tbl.
17345 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17346 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17347 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17348 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17349 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17350 (aarch64_expand_sve_vcond): New functions.
17351 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17352 of aarch64_vector_mode_p.
17353 (aarch64_dwarf_poly_indeterminate_value): New function.
17354 (aarch64_compute_pressure_classes): Likewise.
17355 (aarch64_can_change_mode_class): Likewise.
17356 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17357 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17358 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17359 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17360 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17361 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17362 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17363 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17365 (Dn, Dl, Dr): Accept const as well as const_vector.
17366 (Dz): Likewise. Compare against CONST0_RTX.
17367 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17368 of "vector" where appropriate.
17369 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17370 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17371 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17372 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17373 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17374 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17375 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17376 (v_int_equiv): Extend to SVE modes.
17377 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17379 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17380 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17381 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17382 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17383 (SVE_COND_FP_CMP): New int iterators.
17384 (perm_hilo): Handle the new unpack unspecs.
17385 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17387 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17388 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17389 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17390 (aarch64_equality_operator, aarch64_constant_vector_operand)
17391 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17392 (aarch64_sve_nonimmediate_operand): Likewise.
17393 (aarch64_sve_general_operand): Likewise.
17394 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17395 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17396 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17397 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17398 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17399 (aarch64_sve_float_arith_immediate): Likewise.
17400 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17401 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17402 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17403 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17404 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17405 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17406 (aarch64_sve_float_arith_operand): Likewise.
17407 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17408 (aarch64_sve_float_mul_operand): Likewise.
17409 (aarch64_sve_vec_perm_operand): Likewise.
17410 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17411 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17412 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17413 as well as const_vector.
17414 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17415 in file. Use CONST0_RTX and CONSTM1_RTX.
17416 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17417 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17418 Use aarch64_simd_imm_zero.
17419 * config/aarch64/aarch64-sve.md: New file.
17420 * config/aarch64/aarch64.md: Include it.
17421 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17422 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17423 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17424 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17425 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17426 (sve): New attribute.
17427 (enabled): Disable instructions with the sve attribute unless
17429 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17430 aarch64_expand_mov_immediate.
17431 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17432 CNT[BHSD] immediates.
17433 (movti): Split CONST_POLY_INT moves into two halves.
17434 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17435 Split additions that need a temporary here if the destination
17436 is the stack pointer.
17437 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17438 (*add<mode>3_poly_1): New instruction.
17439 (set_clobber_cc): New expander.
17441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17443 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17444 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17445 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17446 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17447 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17448 Change innermode from fixed_mode_size to machine_mode.
17449 (simplify_subreg): Update call accordingly. Handle a constant-sized
17450 subreg of a variable-length CONST_VECTOR.
17452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17453 Alan Hayward <alan.hayward@arm.com>
17454 David Sherwood <david.sherwood@arm.com>
17456 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17457 (add_offset_to_base): New function, split out from...
17458 (create_mem_ref): ...here. When handling a scale other than 1,
17459 check first whether the address is valid without the offset.
17460 Add it into the base if so, leaving the index and scale as-is.
17462 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17465 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17466 fold_for_warn before checking if arg2 is INTEGER_CST.
17468 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17470 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17471 (store_multiple_operation): Delete.
17472 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17473 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17474 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17475 guarded by TARGET_STRING.
17476 (rs6000_output_load_multiple): Delete.
17477 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17478 OPTION_MASK_STRING / TARGET_STRING handling.
17479 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17480 (const rs6000_opt_masks) <"string">: Change mask to 0.
17481 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17482 (MASK_STRING): Delete.
17483 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17485 (load_multiple): Delete.
17492 (store_multiple): Delete.
17499 (movmemsi_8reg): Delete.
17500 (corresponding unnamed define_insn): Delete.
17501 (movmemsi_6reg): Delete.
17502 (corresponding unnamed define_insn): Delete.
17503 (movmemsi_4reg): Delete.
17504 (corresponding unnamed define_insn): Delete.
17505 (movmemsi_2reg): Delete.
17506 (corresponding unnamed define_insn): Delete.
17507 (movmemsi_1reg): Delete.
17508 (corresponding unnamed define_insn): Delete.
17509 * config/rs6000/rs6000.opt (mno-string): New.
17510 (mstring): Replace by deprecation warning stub.
17511 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17513 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17515 * regrename.c (regrename_do_replace): If replacing the same
17516 reg multiple times, try to reuse last created gen_raw_REG.
17519 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17520 main to workaround a bug in GDB.
17522 2018-01-12 Tom de Vries <tom@codesourcery.com>
17525 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17527 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17529 PR rtl-optimization/80481
17530 * ira-color.c (get_cap_member): New function.
17531 (allocnos_conflict_by_live_ranges_p): Use it.
17532 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17533 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17535 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17538 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17539 (*saddl_se_1): Ditto.
17540 (*ssubsi_1): Ditto.
17541 (*ssubl_se_1): Ditto.
17543 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17545 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17546 rather than wi::to_widest for DR_INITs.
17547 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17548 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17549 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17551 (vect_analyze_group_access_1): Note that here.
17553 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17555 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17556 polynomial type sizes.
17558 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17560 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17561 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17562 (gimple_add_tmp_var): Likewise.
17564 2018-01-12 Martin Liska <mliska@suse.cz>
17566 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17567 (gimple_alloc_sizes): Likewise.
17568 (dump_gimple_statistics): Use PRIu64 in printf format.
17569 * gimple.h: Change uint64_t to int.
17571 2018-01-12 Martin Liska <mliska@suse.cz>
17573 * tree-core.h: Use uint64_t instead of int.
17574 * tree.c (tree_node_counts): Likewise.
17575 (tree_node_sizes): Likewise.
17576 (dump_tree_statistics): Use PRIu64 in printf format.
17578 2018-01-12 Martin Liska <mliska@suse.cz>
17580 * Makefile.in: As qsort_chk is implemented in vec.c, add
17581 vec.o to linkage of gencfn-macros.
17582 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17583 passing the info to record_node_allocation_statistics.
17584 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17586 * ggc-common.c (struct ggc_usage): Add operator== and use
17587 it in operator< and compare function.
17588 * mem-stats.h (struct mem_usage): Likewise.
17589 * vec.c (struct vec_usage): Remove operator< and compare
17590 function. Can be simply inherited.
17592 2018-01-12 Martin Jambor <mjambor@suse.cz>
17595 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17596 * tree-ssa-math-opts.c: Include domwalk.h.
17597 (convert_mult_to_fma_1): New function.
17598 (fma_transformation_info): New type.
17599 (fma_deferring_state): Likewise.
17600 (cancel_fma_deferring): New function.
17601 (result_of_phi): Likewise.
17602 (last_fma_candidate_feeds_initial_phi): Likewise.
17603 (convert_mult_to_fma): Added deferring logic, split actual
17604 transformation to convert_mult_to_fma_1.
17605 (math_opts_dom_walker): New type.
17606 (math_opts_dom_walker::after_dom_children): New method, body moved
17607 here from pass_optimize_widening_mul::execute, added deferring logic
17609 (pass_optimize_widening_mul::execute): Moved most of code to
17610 math_opts_dom_walker::after_dom_children.
17611 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17612 * config/i386/i386.c (ix86_option_override_internal): Added
17613 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17615 2018-01-12 Richard Biener <rguenther@suse.de>
17618 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17619 inline instance vars.
17621 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17624 * config/rx/rx.c (rx_is_restricted_memory_address):
17625 Handle SUBREG case.
17627 2018-01-12 Richard Biener <rguenther@suse.de>
17629 PR tree-optimization/80846
17630 * target.def (split_reduction): New target hook.
17631 * targhooks.c (default_split_reduction): New function.
17632 * targhooks.h (default_split_reduction): Declare.
17633 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17634 target requests first reduce vectors by combining low and high
17636 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17637 (get_vectype_for_scalar_type_and_size): Export.
17638 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17639 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17640 * doc/tm.texi: Regenerate.
17641 * config/i386/i386.c (ix86_split_reduction): Implement
17642 TARGET_VECTORIZE_SPLIT_REDUCTION.
17644 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17647 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17648 in PIC mode except for TARGET_VXWORKS_RTP.
17649 * config/sparc/sparc.c: Include cfgrtl.h.
17650 (TARGET_INIT_PIC_REG): Define.
17651 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17652 (sparc_pic_register_p): New predicate.
17653 (sparc_legitimate_address_p): Use it.
17654 (sparc_legitimize_pic_address): Likewise.
17655 (sparc_delegitimize_address): Likewise.
17656 (sparc_mode_dependent_address_p): Likewise.
17657 (gen_load_pcrel_sym): Remove 4th parameter.
17658 (load_got_register): Adjust call to above. Remove obsolete stuff.
17659 (sparc_expand_prologue): Do not call load_got_register here.
17660 (sparc_flat_expand_prologue): Likewise.
17661 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17662 (sparc_use_pseudo_pic_reg): New function.
17663 (sparc_init_pic_reg): Likewise.
17664 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17665 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17667 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17669 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17670 Add item for branch_cost.
17672 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17674 PR rtl-optimization/83565
17675 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17676 not extend the result to a larger mode for rotate operations.
17677 (num_sign_bit_copies1): Likewise.
17679 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17682 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17684 Use values-Xc.o for -pedantic.
17685 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17687 2018-01-12 Martin Liska <mliska@suse.cz>
17690 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17692 (possible_polymorphic_call_targets): Use it.
17693 (ipa_devirt): Likewise.
17695 2018-01-12 Martin Liska <mliska@suse.cz>
17697 * profile-count.h (enum profile_quality): Use 0 as invalid
17698 enum value of profile_quality.
17700 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17702 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17703 -mext-string options.
17705 2018-01-12 Richard Biener <rguenther@suse.de>
17707 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17708 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17709 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17711 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17713 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17715 * configure.ac (--with-long-double-format): Add support for the
17716 configuration option to change the default long double format on
17718 * config.gcc (powerpc*-linux*-*): Likewise.
17719 * configure: Regenerate.
17720 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17721 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17722 used without modification.
17724 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17726 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17727 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17728 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17729 MISC_BUILTIN_SPEC_BARRIER.
17730 (rs6000_init_builtins): Likewise.
17731 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17733 (speculation_barrier): New define_insn.
17734 * doc/extend.texi: Document __builtin_speculation_barrier.
17736 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17739 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17740 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17741 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17743 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17744 integral modes instead of "ss" and "sd".
17745 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17746 vectors with 32-bit and 64-bit elements.
17747 (vecdupssescalarmodesuffix): New mode attribute.
17748 (vec_dup<mode>): Use it.
17750 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17753 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17754 frame if argument is passed on stack.
17756 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17759 * ree.c (combine_reaching_defs): Optimize also
17760 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17761 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17763 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17765 PR middle-end/83189
17766 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17768 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17770 PR middle-end/83718
17771 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17772 after they are computed.
17774 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17776 PR tree-optimization/83695
17777 * gimple-loop-linterchange.cc
17778 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17779 reset cached scev information after interchange.
17780 (pass_linterchange::execute): Remove call to scev_reset_htab.
17782 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17784 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17785 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17786 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17787 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17788 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17789 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17790 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17791 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17792 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17793 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17794 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17795 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17796 (V_lane_reg): Likewise.
17797 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17799 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17800 (vfmal_lane_low<mode>_intrinsic,
17801 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17802 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17803 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17804 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17805 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17806 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17808 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17810 * config/arm/arm-cpus.in (fp16fml): New feature.
17811 (ALL_SIMD): Add fp16fml.
17812 (armv8.2-a): Add fp16fml as an option.
17813 (armv8.3-a): Likewise.
17814 (armv8.4-a): Add fp16fml as part of fp16.
17815 * config/arm/arm.h (TARGET_FP16FML): Define.
17816 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17818 * config/arm/arm-modes.def (V2HF): Define.
17819 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17820 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17821 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17822 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17823 vfmsl_low, vfmsl_high): New set of builtins.
17824 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17825 (vfml_op): New code attribute.
17826 (VFMLHALVES): New int iterator.
17827 (VFML, VFMLSEL): New mode attributes.
17828 (V_reg): Define mapping for V2HF.
17829 (V_hi, V_lo): New mode attributes.
17830 (VF_constraint): Likewise.
17831 (vfml_half, vfml_half_selector): New int attributes.
17832 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17834 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17835 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17837 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17838 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17839 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17840 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17842 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17843 Document new effective target and option set.
17845 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17847 * config/arm/arm-cpus.in (armv8_4): New feature.
17848 (ARMv8_4a): New fgroup.
17849 (armv8.4-a): New arch.
17850 * config/arm/arm-tables.opt: Regenerate.
17851 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17852 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17853 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17854 Add matching rules for -march=armv8.4-a and extensions.
17855 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17857 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17860 * config/rx/rx.md (BW): New mode attribute.
17861 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17863 2018-01-11 Richard Biener <rguenther@suse.de>
17865 PR tree-optimization/83435
17866 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17867 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17868 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17870 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17871 Alan Hayward <alan.hayward@arm.com>
17872 David Sherwood <david.sherwood@arm.com>
17874 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17876 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17877 (aarch64_print_address_internal): Use it to check for a zero offset.
17879 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17880 Alan Hayward <alan.hayward@arm.com>
17881 David Sherwood <david.sherwood@arm.com>
17883 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17884 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17885 Return a poly_int64 rather than a HOST_WIDE_INT.
17886 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17887 rather than a HOST_WIDE_INT.
17888 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17889 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17890 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17891 final_offset from HOST_WIDE_INT to poly_int64.
17892 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17893 to_constant when getting the number of units in an Advanced SIMD
17895 (aarch64_builtin_vectorized_function): Check for a constant number
17897 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17899 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17900 attribute instead of GET_MODE_NUNITS.
17901 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17902 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17903 GET_MODE_SIZE for fixed-size registers.
17904 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17905 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17906 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17907 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17908 (aarch64_print_operand, aarch64_print_address_internal)
17909 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17910 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17911 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17912 Handle polynomial GET_MODE_SIZE.
17913 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17914 wider than SImode without modification.
17915 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17916 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17917 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17918 passing and returning SVE modes.
17919 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17920 rather than GEN_INT.
17921 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17922 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17923 (aarch64_allocate_and_probe_stack_space): Likewise.
17924 (aarch64_layout_frame): Cope with polynomial offsets.
17925 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17926 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17927 polynomial offsets.
17928 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17929 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17930 poly_int64 rather than a HOST_WIDE_INT.
17931 (aarch64_get_separate_components, aarch64_process_components)
17932 (aarch64_expand_prologue, aarch64_expand_epilogue)
17933 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17934 (aarch64_anchor_offset): New function, split out from...
17935 (aarch64_legitimize_address): ...here.
17936 (aarch64_builtin_vectorization_cost): Handle polynomial
17937 TYPE_VECTOR_SUBPARTS.
17938 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17940 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17941 number of elements from the PARALLEL rather than the mode.
17942 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17943 rather than GET_MODE_BITSIZE.
17944 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17945 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17946 (aarch64_expand_vec_perm_const_1): Handle polynomial
17947 d->perm.length () and d->perm elements.
17948 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17949 Apply to_constant to d->perm elements.
17950 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17951 polynomial CONST_VECTOR_NUNITS.
17952 (aarch64_move_pointer): Take amount as a poly_int64 rather
17954 (aarch64_progress_pointer): Avoid temporary variable.
17955 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17956 the mode attribute instead of GET_MODE.
17958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17959 Alan Hayward <alan.hayward@arm.com>
17960 David Sherwood <david.sherwood@arm.com>
17962 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17963 x exists before using it.
17964 (aarch64_add_constant_internal): Rename to...
17965 (aarch64_add_offset_1): ...this. Replace regnum with separate
17966 src and dest rtxes. Handle the case in which they're different,
17967 including when the offset is zero. Replace scratchreg with an rtx.
17968 Use 2 additions if there is no spare register into which we can
17969 move a 16-bit constant.
17970 (aarch64_add_constant): Delete.
17971 (aarch64_add_offset): Replace reg with separate src and dest
17972 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17973 Use aarch64_add_offset_1.
17974 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17975 an rtx rather than an int. Take the delta as a poly_int64
17976 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17977 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17978 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17979 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17980 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17981 and aarch64_add_sp.
17982 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17983 aarch64_add_constant.
17985 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17987 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17988 Use scalar_float_mode.
17990 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17992 * config/aarch64/aarch64-simd.md
17993 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17994 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17995 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17996 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17997 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17998 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17999 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18000 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18001 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18002 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18004 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18007 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18008 targ_options->x_arm_arch_string is non NULL.
18010 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18012 * config/aarch64/aarch64.h
18013 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18015 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18018 * expmed.c (emit_store_flag_force): Swap if const op0
18019 and change VOIDmode to mode of op0.
18021 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18023 PR rtl-optimization/83761
18024 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18025 than bytes to mode_for_size.
18027 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18029 PR middle-end/83189
18030 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18031 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18034 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18036 PR middle-end/83575
18037 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18038 when in layout mode.
18039 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18040 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18043 2018-01-10 Michael Collison <michael.collison@arm.com>
18045 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18046 * config/aarch64/aarch64-option-extension.def: Add
18047 AARCH64_OPT_EXTENSION of 'fp16fml'.
18048 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18049 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18050 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18051 * config/aarch64/constraints.md (Ui7): New constraint.
18052 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18053 (VFMLA_SEL_W): Ditto.
18056 (VFMLA16_LOW): New int iterator.
18057 (VFMLA16_HIGH): Ditto.
18058 (UNSPEC_FMLAL): New unspec.
18059 (UNSPEC_FMLSL): Ditto.
18060 (UNSPEC_FMLAL2): Ditto.
18061 (UNSPEC_FMLSL2): Ditto.
18062 (f16mac): New code attribute.
18063 * config/aarch64/aarch64-simd-builtins.def
18064 (aarch64_fmlal_lowv2sf): Ditto.
18065 (aarch64_fmlsl_lowv2sf): Ditto.
18066 (aarch64_fmlalq_lowv4sf): Ditto.
18067 (aarch64_fmlslq_lowv4sf): Ditto.
18068 (aarch64_fmlal_highv2sf): Ditto.
18069 (aarch64_fmlsl_highv2sf): Ditto.
18070 (aarch64_fmlalq_highv4sf): Ditto.
18071 (aarch64_fmlslq_highv4sf): Ditto.
18072 (aarch64_fmlal_lane_lowv2sf): Ditto.
18073 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18074 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18075 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18076 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18077 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18078 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18079 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18080 (aarch64_fmlal_lane_highv2sf): Ditto.
18081 (aarch64_fmlsl_lane_highv2sf): Ditto.
18082 (aarch64_fmlal_laneq_highv2sf): Ditto.
18083 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18084 (aarch64_fmlalq_lane_highv4sf): Ditto.
18085 (aarch64_fmlsl_lane_highv4sf): Ditto.
18086 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18087 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18088 * config/aarch64/aarch64-simd.md:
18089 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18090 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18091 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18092 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18093 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18094 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18095 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18096 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18097 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18098 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18099 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18100 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18101 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18102 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18103 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18104 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18105 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18106 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18107 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18108 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18109 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18110 (vfmlsl_low_u32): Ditto.
18111 (vfmlalq_low_u32): Ditto.
18112 (vfmlslq_low_u32): Ditto.
18113 (vfmlal_high_u32): Ditto.
18114 (vfmlsl_high_u32): Ditto.
18115 (vfmlalq_high_u32): Ditto.
18116 (vfmlslq_high_u32): Ditto.
18117 (vfmlal_lane_low_u32): Ditto.
18118 (vfmlsl_lane_low_u32): Ditto.
18119 (vfmlal_laneq_low_u32): Ditto.
18120 (vfmlsl_laneq_low_u32): Ditto.
18121 (vfmlalq_lane_low_u32): Ditto.
18122 (vfmlslq_lane_low_u32): Ditto.
18123 (vfmlalq_laneq_low_u32): Ditto.
18124 (vfmlslq_laneq_low_u32): Ditto.
18125 (vfmlal_lane_high_u32): Ditto.
18126 (vfmlsl_lane_high_u32): Ditto.
18127 (vfmlal_laneq_high_u32): Ditto.
18128 (vfmlsl_laneq_high_u32): Ditto.
18129 (vfmlalq_lane_high_u32): Ditto.
18130 (vfmlslq_lane_high_u32): Ditto.
18131 (vfmlalq_laneq_high_u32): Ditto.
18132 (vfmlslq_laneq_high_u32): Ditto.
18133 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18134 (AARCH64_FL_FOR_ARCH8_4): New.
18135 (AARCH64_ISA_F16FML): New ISA flag.
18136 (TARGET_F16FML): New feature flag for fp16fml.
18137 (doc/invoke.texi): Document new fp16fml option.
18139 2018-01-10 Michael Collison <michael.collison@arm.com>
18141 * config/aarch64/aarch64-builtins.c:
18142 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18143 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18144 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18145 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18146 (AARCH64_ISA_SHA3): New ISA flag.
18147 (TARGET_SHA3): New feature flag for sha3.
18148 * config/aarch64/iterators.md (sha512_op): New int attribute.
18149 (CRYPTO_SHA512): New int iterator.
18150 (UNSPEC_SHA512H): New unspec.
18151 (UNSPEC_SHA512H2): Ditto.
18152 (UNSPEC_SHA512SU0): Ditto.
18153 (UNSPEC_SHA512SU1): Ditto.
18154 * config/aarch64/aarch64-simd-builtins.def
18155 (aarch64_crypto_sha512hqv2di): New builtin.
18156 (aarch64_crypto_sha512h2qv2di): Ditto.
18157 (aarch64_crypto_sha512su0qv2di): Ditto.
18158 (aarch64_crypto_sha512su1qv2di): Ditto.
18159 (aarch64_eor3qv8hi): Ditto.
18160 (aarch64_rax1qv2di): Ditto.
18161 (aarch64_xarqv2di): Ditto.
18162 (aarch64_bcaxqv8hi): Ditto.
18163 * config/aarch64/aarch64-simd.md:
18164 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18165 (aarch64_crypto_sha512su0qv2di): Ditto.
18166 (aarch64_crypto_sha512su1qv2di): Ditto.
18167 (aarch64_eor3qv8hi): Ditto.
18168 (aarch64_rax1qv2di): Ditto.
18169 (aarch64_xarqv2di): Ditto.
18170 (aarch64_bcaxqv8hi): Ditto.
18171 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18172 (vsha512h2q_u64): Ditto.
18173 (vsha512su0q_u64): Ditto.
18174 (vsha512su1q_u64): Ditto.
18175 (veor3q_u16): Ditto.
18176 (vrax1q_u64): Ditto.
18177 (vxarq_u64): Ditto.
18178 (vbcaxq_u16): Ditto.
18179 * config/arm/types.md (crypto_sha512): New type attribute.
18180 (crypto_sha3): Ditto.
18181 (doc/invoke.texi): Document new sha3 option.
18183 2018-01-10 Michael Collison <michael.collison@arm.com>
18185 * config/aarch64/aarch64-builtins.c:
18186 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18187 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18188 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18189 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18190 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18191 (AARCH64_ISA_SM4): New ISA flag.
18192 (TARGET_SM4): New feature flag for sm4.
18193 * config/aarch64/aarch64-simd-builtins.def
18194 (aarch64_sm3ss1qv4si): Ditto.
18195 (aarch64_sm3tt1aq4si): Ditto.
18196 (aarch64_sm3tt1bq4si): Ditto.
18197 (aarch64_sm3tt2aq4si): Ditto.
18198 (aarch64_sm3tt2bq4si): Ditto.
18199 (aarch64_sm3partw1qv4si): Ditto.
18200 (aarch64_sm3partw2qv4si): Ditto.
18201 (aarch64_sm4eqv4si): Ditto.
18202 (aarch64_sm4ekeyqv4si): Ditto.
18203 * config/aarch64/aarch64-simd.md:
18204 (aarch64_sm3ss1qv4si): Ditto.
18205 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18206 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18207 (aarch64_sm4eqv4si): Ditto.
18208 (aarch64_sm4ekeyqv4si): Ditto.
18209 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18210 (sm3part_op): Ditto.
18211 (CRYPTO_SM3TT): Ditto.
18212 (CRYPTO_SM3PART): Ditto.
18213 (UNSPEC_SM3SS1): New unspec.
18214 (UNSPEC_SM3TT1A): Ditto.
18215 (UNSPEC_SM3TT1B): Ditto.
18216 (UNSPEC_SM3TT2A): Ditto.
18217 (UNSPEC_SM3TT2B): Ditto.
18218 (UNSPEC_SM3PARTW1): Ditto.
18219 (UNSPEC_SM3PARTW2): Ditto.
18220 (UNSPEC_SM4E): Ditto.
18221 (UNSPEC_SM4EKEY): Ditto.
18222 * config/aarch64/constraints.md (Ui2): New constraint.
18223 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18224 * config/arm/types.md (crypto_sm3): New type attribute.
18225 (crypto_sm4): Ditto.
18226 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18227 (vsm3tt1aq_u32): Ditto.
18228 (vsm3tt1bq_u32): Ditto.
18229 (vsm3tt2aq_u32): Ditto.
18230 (vsm3tt2bq_u32): Ditto.
18231 (vsm3partw1q_u32): Ditto.
18232 (vsm3partw2q_u32): Ditto.
18233 (vsm4eq_u32): Ditto.
18234 (vsm4ekeyq_u32): Ditto.
18235 (doc/invoke.texi): Document new sm4 option.
18237 2018-01-10 Michael Collison <michael.collison@arm.com>
18239 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18240 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18241 (AARCH64_FL_FOR_ARCH8_4): New.
18242 (AARCH64_FL_V8_4): New flag.
18243 (doc/invoke.texi): Document new armv8.4-a option.
18245 2018-01-10 Michael Collison <michael.collison@arm.com>
18247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18248 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18249 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18250 * config/aarch64/aarch64-option-extension.def: Add
18251 AARCH64_OPT_EXTENSION of 'sha2'.
18252 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18253 (crypto): Disable sha2 and aes if crypto disabled.
18254 (crypto): Enable aes and sha2 if enabled.
18255 (simd): Disable sha2 and aes if simd disabled.
18256 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18258 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18259 (TARGET_SHA2): New feature flag for sha2.
18260 (TARGET_AES): New feature flag for aes.
18261 * config/aarch64/aarch64-simd.md:
18262 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18263 conditional on TARGET_AES.
18264 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18265 (aarch64_crypto_sha1hsi): Make pattern conditional
18267 (aarch64_crypto_sha1hv4si): Ditto.
18268 (aarch64_be_crypto_sha1hv4si): Ditto.
18269 (aarch64_crypto_sha1su1v4si): Ditto.
18270 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18271 (aarch64_crypto_sha1su0v4si): Ditto.
18272 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18273 (aarch64_crypto_sha256su0v4si): Ditto.
18274 (aarch64_crypto_sha256su1v4si): Ditto.
18275 (doc/invoke.texi): Document new aes and sha2 options.
18277 2018-01-10 Martin Sebor <msebor@redhat.com>
18279 PR tree-optimization/83781
18280 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18283 2018-01-11 Martin Sebor <msebor@gmail.com>
18284 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18286 PR tree-optimization/83501
18287 PR tree-optimization/81703
18289 * tree-ssa-strlen.c (get_string_cst): Rename...
18290 (get_string_len): ...to this. Handle global constants.
18291 (handle_char_store): Adjust.
18293 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18294 Jim Wilson <jimw@sifive.com>
18296 * config/riscv/riscv-protos.h (riscv_output_return): New.
18297 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18298 (riscv_attribute_table, riscv_output_return),
18299 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18300 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18301 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18303 (riscv_expand_prologue): Add early return for naked function.
18304 (riscv_expand_epilogue): Likewise.
18305 (riscv_function_ok_for_sibcall): Return false for naked function.
18306 (riscv_set_current_function): New.
18307 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18308 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18309 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18310 * doc/extend.texi (RISC-V Function Attributes): New.
18312 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18314 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18315 check for 128-bit long double before checking TCmode.
18316 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18317 128-bit long doubles before checking TFmode or TCmode.
18318 (FLOAT128_IBM_P): Likewise.
18320 2018-01-10 Martin Sebor <msebor@redhat.com>
18322 PR tree-optimization/83671
18323 * builtins.c (c_strlen): Unconditionally return zero for the empty
18325 Use -Warray-bounds for warnings.
18326 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18327 for non-constant array indices with COMPONENT_REF, arrays of
18328 arrays, and pointers to arrays.
18329 (gimple_fold_builtin_strlen): Determine and set length range for
18330 non-constant character arrays.
18332 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18334 PR middle-end/81897
18335 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18338 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18340 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18342 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18345 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18346 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18347 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18348 indexed_or_indirect_operand predicate.
18349 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18350 (*vsx_le_perm_load_v8hi): Likewise.
18351 (*vsx_le_perm_load_v16qi): Likewise.
18352 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18353 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18354 (*vsx_le_perm_store_v8hi): Likewise.
18355 (*vsx_le_perm_store_v16qi): Likewise.
18356 (eight unnamed splitters): Likewise.
18358 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18360 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18361 * config/rs6000/emmintrin.h: Likewise.
18362 * config/rs6000/mmintrin.h: Likewise.
18363 * config/rs6000/xmmintrin.h: Likewise.
18365 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18368 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18370 * tree.c (tree_nop_conversion): Return true for location wrapper
18372 (maybe_wrap_with_location): New function.
18373 (selftest::check_strip_nops): New function.
18374 (selftest::test_location_wrappers): New function.
18375 (selftest::tree_c_tests): Call it.
18376 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18377 (maybe_wrap_with_location): New decl.
18378 (EXPR_LOCATION_WRAPPER_P): New macro.
18379 (location_wrapper_p): New inline function.
18380 (tree_strip_any_location_wrapper): New inline function.
18382 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18385 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18386 stack_realign_offset for the largest alignment of stack slot
18388 (ix86_find_max_used_stack_alignment): New function.
18389 (ix86_finalize_stack_frame_flags): Use it. Set
18390 max_used_stack_alignment if we don't realign stack.
18391 * config/i386/i386.h (machine_function): Add
18392 max_used_stack_alignment.
18394 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18396 * config/arm/arm.opt (-mbranch-cost): New option.
18397 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18400 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18403 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18404 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18406 2018-01-10 Richard Biener <rguenther@suse.de>
18409 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18410 early out so it also covers the case where we have a non-NULL
18413 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18415 PR tree-optimization/83753
18416 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18417 for non-strided grouped accesses if the number of elements is 1.
18419 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18422 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18423 * i386.h (TARGET_USE_GATHER): Define.
18424 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18426 2018-01-10 Martin Liska <mliska@suse.cz>
18429 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18430 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18432 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18433 CLEANUP_NO_PARTITIONING is not set.
18435 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18437 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18438 for vectors, as a partial revert of r254296.
18439 * rtl.h (const_vec_p): Delete.
18440 (const_vec_duplicate_p): Don't test for vector CONSTs.
18441 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18442 * expmed.c (make_tree): Likewise.
18445 * common.md (E, F): Use CONSTANT_P instead of checking for
18447 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18448 checking for CONST_VECTOR.
18450 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18452 PR middle-end/83575
18453 * predict.c (force_edge_cold): Handle in more sane way edges
18454 with no prediction.
18456 2018-01-09 Carl Love <cel@us.ibm.com>
18458 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18460 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18461 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18462 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18463 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18464 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18465 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18466 * config/rs6000/rs6000-protos.h: Add extern defition for
18467 rs6000_generate_float2_double_code.
18468 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18470 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18471 (float2_v2df): Add define_expand.
18473 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18476 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18477 op_mode in the force_to_mode call.
18479 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18481 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18482 instead of checking each element individually.
18483 (aarch64_evpc_uzp): Likewise.
18484 (aarch64_evpc_zip): Likewise.
18485 (aarch64_evpc_ext): Likewise.
18486 (aarch64_evpc_rev): Likewise.
18487 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18488 instead of checking each element individually. Return true without
18490 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18491 whether all selected elements come from the same input, instead of
18492 checking each element individually. Remove calls to gen_rtx_REG,
18493 start_sequence and end_sequence and instead assert that no rtl is
18496 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18498 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18499 order of HIGH and CONST checks.
18501 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18503 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18504 if the destination isn't an SSA_NAME.
18506 2018-01-09 Richard Biener <rguenther@suse.de>
18508 PR tree-optimization/83668
18509 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18511 (canonicalize_loop_form): ... here, renamed from ...
18512 (canonicalize_loop_closed_ssa_form): ... this and amended to
18513 swap successor edges for loop exit blocks to make us use
18514 the RPO order we need for initial schedule generation.
18516 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18518 PR tree-optimization/64811
18519 * match.pd: When optimizing comparisons with Inf, avoid
18520 introducing or losing exceptions from comparisons with NaN.
18522 2018-01-09 Martin Liska <mliska@suse.cz>
18525 * asan.c (shadow_mem_size): Add gcc_assert.
18527 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18529 Don't save registers in main().
18532 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18533 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18534 * config/avr/avr.c (avr_set_current_function): Don't error if
18535 naked, OS_task or OS_main are specified at the same time.
18536 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18538 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18540 * common/config/avr/avr-common.c (avr_option_optimization_table):
18541 Switch on -mmain-is-OS_task for optimizing compilations.
18543 2018-01-09 Richard Biener <rguenther@suse.de>
18545 PR tree-optimization/83572
18546 * graphite.c: Include cfganal.h.
18547 (graphite_transform_loops): Connect infinite loops to exit
18548 and remove fake edges at the end.
18550 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18552 * ipa-inline.c (edge_badness): Revert accidental checkin.
18554 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18557 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18558 symbols; not inline clones.
18560 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18563 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18564 hard registers. Formatting fixes.
18566 PR preprocessor/83722
18567 * gcc.c (try_generate_repro): Pass
18568 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18569 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18572 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18573 Kito Cheng <kito.cheng@gmail.com>
18575 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18576 (riscv_leaf_function_p): Delete.
18577 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18579 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18581 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18583 (do_ifelse): New function.
18584 (do_isel): New function.
18585 (do_sub3): New function.
18586 (do_add3): New function.
18587 (do_load_mask_compare): New function.
18588 (do_overlap_load_compare): New function.
18589 (expand_compare_loop): New function.
18590 (expand_block_compare): Call expand_compare_loop() when appropriate.
18591 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18592 option description.
18593 (-mblock-compare-inline-loop-limit): New option.
18595 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18598 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18599 Reverse order of second and third operands in first alternative.
18600 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18601 of first and second elements in UNSPEC_VPERMR vector.
18602 (altivec_expand_vec_perm_le): Likewise.
18604 2018-01-08 Jeff Law <law@redhat.com>
18606 PR rtl-optimizatin/81308
18607 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18608 (process_switch): If group_case_labels makes a change, then set
18610 (pass_convert_switch::execute): If a switch is converted, then
18611 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18613 PR rtl-optimization/81308
18614 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18617 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18619 PR target/83663 - Revert r255946
18620 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18621 generation for cases where splatting a value is not useful.
18622 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18623 across a vec_duplicate and a paradoxical subreg forming a vector
18624 mode to a vec_concat.
18626 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18628 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18629 -march=armv8.3-a variants.
18630 * config/arm/t-multilib: Likewise.
18631 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18633 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18635 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18637 (cceq_ior_compare_complement): Give it a name so I can use it, and
18638 change boolean_or_operator predicate to boolean_operator so it can
18639 be used to generate a crand.
18640 (eqne): New code iterator.
18641 (bd/bd_neg): New code_attrs.
18642 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18643 a single define_insn.
18644 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18645 decrement (bdnzt/bdnzf/bdzt/bdzf).
18646 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18647 with the new names of the branch decrement patterns, and added the
18648 names of the branch decrement conditional patterns.
18650 2018-01-08 Richard Biener <rguenther@suse.de>
18652 PR tree-optimization/83563
18653 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18656 2018-01-08 Richard Biener <rguenther@suse.de>
18658 PR middle-end/83713
18659 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18661 2018-01-08 Richard Biener <rguenther@suse.de>
18663 PR tree-optimization/83685
18664 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18665 references to abnormals.
18667 2018-01-08 Richard Biener <rguenther@suse.de>
18670 * dwarf2out.c (output_indirect_strings): Handle empty
18671 skeleton_debug_str_hash.
18672 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18674 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18676 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18677 (emit_store_direct): Likewise.
18678 (arc_trampoline_adjust_address): Likewise.
18679 (arc_asm_trampoline_template): New function.
18680 (arc_initialize_trampoline): Use asm_trampoline_template.
18681 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18682 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18683 * config/arc/arc.md (flush_icache): Delete pattern.
18685 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18687 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18688 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18691 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18694 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18695 by not USED_FOR_TARGET.
18696 (make_pass_resolve_sw_modes): Likewise.
18698 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18700 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18703 2018-01-08 Richard Biener <rguenther@suse.de>
18705 PR middle-end/83580
18706 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18708 2018-01-08 Richard Biener <rguenther@suse.de>
18710 PR middle-end/83517
18711 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18713 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18715 PR middle-end/81897
18716 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18717 basic blocks with a small number of successors.
18718 (convert_control_dep_chain_into_preds): Improve handling of
18720 (dump_predicates): Split apart into...
18721 (dump_pred_chain): ...here...
18722 (dump_pred_info): ...and here.
18723 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18724 (can_chain_union_be_invalidated_p): Improve check for invalidation
18726 (uninit_uses_cannot_happen): Avoid unnecessary if
18727 convert_control_dep_chain_into_preds yielded nothing.
18729 2018-01-06 Martin Sebor <msebor@redhat.com>
18731 PR tree-optimization/83640
18732 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18733 subtracting negative offset from size.
18734 (builtin_access::overlap): Adjust offset bounds of the access to fall
18735 within the size of the object if possible.
18737 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18739 PR rtl-optimization/83699
18740 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18741 extract_bit_field_as_subreg to cases in which the extracted
18742 value is also a vector.
18744 * lra-constraints.c (process_alt_operands): Test for the equivalence
18745 substitutions when detecting a possible reload cycle.
18747 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18750 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18751 by default if flag_selective_schedling{,2}. Formatting fixes.
18753 PR rtl-optimization/83682
18754 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18755 if it has non-VECTOR_MODE element mode.
18756 (vec_duplicate_p): Likewise.
18758 PR middle-end/83694
18759 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18760 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18762 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18765 * config/i386/i386-builtin.def
18766 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18767 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18768 Require also OPTION_MASK_ISA_AVX512F in addition to
18769 OPTION_MASK_ISA_GFNI.
18770 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18771 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18772 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18773 to OPTION_MASK_ISA_GFNI.
18774 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18775 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18776 OPTION_MASK_ISA_AVX512BW.
18777 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18778 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18779 addition to OPTION_MASK_ISA_GFNI.
18780 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18781 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18782 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18783 to OPTION_MASK_ISA_GFNI.
18784 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18785 a requirement for all ISAs rather than any of them with a few
18787 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18789 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18790 bitmasks to be enabled with 3 exceptions, instead of requiring any
18791 enabled ISA with lots of exceptions.
18792 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18793 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18794 Change avx512bw in isa attribute to avx512f.
18795 * config/i386/sgxintrin.h: Add license boilerplate.
18796 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18797 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18798 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18799 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18801 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18802 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18803 temporarily sse2 rather than sse if not enabled already.
18806 * config/i386/sse.md (VI248_VLBW): Rename to ...
18807 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18808 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18809 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18810 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18811 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18812 mode iterator instead of VI248_VLBW.
18814 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18816 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18817 (record_modified): Skip clobbers; add debug output.
18818 (param_change_prob): Use sreal frequencies.
18820 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18822 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18823 punt for user-aligned variables.
18825 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18827 * tree-chrec.c (chrec_contains_symbols): Return true for
18830 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18833 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18834 of (x|y) == x for BICS pattern.
18836 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18838 PR tree-optimization/83605
18839 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18840 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18843 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18845 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18846 * config/epiphany/rtems.h: New file.
18848 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18849 Uros Bizjak <ubizjak@gmail.com>
18852 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18853 QIreg_operand instead of register_operand predicate.
18854 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18855 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18856 comments instead of -fmitigate[-_]rop.
18858 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18861 * cgraphunit.c (symbol_table::compile): Switch to text_section
18862 before calling assembly_start debug hook.
18863 * run-rtl-passes.c (run_rtl_passes): Likewise.
18866 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18868 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18869 range_int_cst_p rather than !symbolic_range_p before calling
18870 extract_range_from_multiplicative_op_1.
18872 2018-01-04 Jeff Law <law@redhat.com>
18874 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18875 redundant test in assertion.
18877 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18879 * doc/rtl.texi: Document machine_mode wrapper classes.
18881 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18883 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18884 using tree_to_uhwi.
18886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18888 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18889 the VEC_PERM_EXPR fold to fail.
18891 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18894 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18895 to switched_sections.
18897 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18900 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18901 test for d.testing.
18903 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18906 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18907 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18909 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18912 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18913 is BLKmode and bitpos not zero or mode change is needed.
18915 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18918 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18921 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18924 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18925 instead of MULT rtx. Update all corresponding splitters.
18926 (*saddl_se): Ditto.
18927 (*ssub<modesuffix>): Ditto.
18928 (*ssubl_se): Ditto.
18929 (*cmp_sadd_di): Update split patterns.
18930 (*cmp_sadd_si): Ditto.
18931 (*cmp_sadd_sidi): Ditto.
18932 (*cmp_ssub_di): Ditto.
18933 (*cmp_ssub_si): Ditto.
18934 (*cmp_ssub_sidi): Ditto.
18935 * config/alpha/predicates.md (const23_operand): New predicate.
18936 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18937 Look for ASHIFT, not MULT inner operand.
18938 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18940 2018-01-04 Martin Liska <mliska@suse.cz>
18942 PR gcov-profile/83669
18943 * gcov.c (output_intermediate_file): Add version to intermediate
18945 * doc/gcov.texi: Document new field 'version' in intermediate
18946 file format. Fix location of '-k' option of gcov command.
18948 2018-01-04 Martin Liska <mliska@suse.cz>
18951 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18953 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18955 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18957 2018-01-03 Martin Sebor <msebor@redhat.com>
18959 PR tree-optimization/83655
18960 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18961 checking calls with invalid arguments.
18963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18965 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18966 (vectorizable_mask_load_store): Delete.
18967 (vectorizable_call): Return false for masked loads and stores.
18968 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18969 instead of gimple_assign_rhs1.
18970 (vectorizable_load): Handle IFN_MASK_LOAD.
18971 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18975 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18977 (vectorizable_mask_load_store): ...here.
18978 (vectorizable_load): ...and here.
18980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18982 * tree-vect-stmts.c (vect_build_all_ones_mask)
18983 (vect_build_zero_merge_argument): New functions, split out from...
18984 (vectorizable_load): ...here.
18986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18988 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18990 (vectorizable_mask_load_store): ...here.
18991 (vectorizable_store): ...and here.
18993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18995 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18997 (vectorizable_mask_load_store): ...here.
18999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19001 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19002 (vect_model_store_cost): Take a vec_load_store_type instead of a
19004 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19005 (vect_model_store_cost): Take a vec_load_store_type instead of a
19007 (vectorizable_mask_load_store): Update accordingly.
19008 (vectorizable_store): Likewise.
19009 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19013 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19014 IFN_MASK_LOAD calls here rather than...
19015 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19018 Alan Hayward <alan.hayward@arm.com>
19019 David Sherwood <david.sherwood@arm.com>
19021 * expmed.c (extract_bit_field_1): For vector extracts,
19022 fall back to extract_bit_field_as_subreg if vec_extract
19025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19026 Alan Hayward <alan.hayward@arm.com>
19027 David Sherwood <david.sherwood@arm.com>
19029 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19030 they are variable or constant sized.
19031 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19032 slots for constant-sized data.
19034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19035 Alan Hayward <alan.hayward@arm.com>
19036 David Sherwood <david.sherwood@arm.com>
19038 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19039 handling COND_EXPRs with boolean comparisons, try to find a better
19040 basis for the mask type than the boolean itself.
19042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19044 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19045 is calculated and how it can be overridden.
19046 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19047 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19049 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19053 Alan Hayward <alan.hayward@arm.com>
19054 David Sherwood <david.sherwood@arm.com>
19056 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19057 Remove the mode argument.
19058 (aarch64_simd_valid_immediate): Remove the mode and inverse
19060 * config/aarch64/iterators.md (bitsize): New iterator.
19061 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19062 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19063 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19064 aarch64_simd_valid_immediate.
19065 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19066 (aarch64_reg_or_bic_imm): Likewise.
19067 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19068 with an insn_type enum and msl with a modifier_type enum.
19069 Replace element_width with a scalar_mode. Change the shift
19070 to unsigned int. Add constructors for scalar_float_mode and
19071 scalar_int_mode elements.
19072 (aarch64_vect_float_const_representable_p): Delete.
19073 (aarch64_can_const_movi_rtx_p)
19074 (aarch64_simd_scalar_immediate_valid_for_move)
19075 (aarch64_simd_make_constant): Update call to
19076 aarch64_simd_valid_immediate.
19077 (aarch64_advsimd_valid_immediate_hs): New function.
19078 (aarch64_advsimd_valid_immediate): Likewise.
19079 (aarch64_simd_valid_immediate): Remove mode and inverse
19080 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19081 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19082 and aarch64_float_const_representable_p on the result.
19083 (aarch64_output_simd_mov_immediate): Remove mode argument.
19084 Update call to aarch64_simd_valid_immediate and use of
19085 simd_immediate_info.
19086 (aarch64_output_scalar_simd_mov_immediate): Update call
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_precision): Prefix with CONST_MODE_PRECISION.
19094 (mode_nunits): Likewise CONST_MODE_NUNITS.
19095 * machmode.def (ADJUST_NUNITS): Document.
19096 * genmodes.c (mode_data::need_nunits_adj): New field.
19097 (blank_mode): Update accordingly.
19098 (adj_nunits): New variable.
19099 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19101 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19102 listed in adj_nunits.
19103 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19104 listed in adj_nunits. Don't emit case statements for such modes.
19105 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19106 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19107 nothing if adj_nunits is nonnull.
19108 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19109 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19110 (emit_mode_fbit): Update use of print_maybe_const_decl.
19111 (emit_move_size): Likewise. Treat the array as non-const
19113 (emit_mode_adjustments): Handle adj_nunits.
19115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19117 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19118 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19119 (VECTOR_MODES): Use it.
19120 (make_vector_modes): Take the prefix as an argument.
19122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19123 Alan Hayward <alan.hayward@arm.com>
19124 David Sherwood <david.sherwood@arm.com>
19126 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19127 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19128 for MODE_VECTOR_BOOL.
19129 * machmode.def (VECTOR_BOOL_MODE): Document.
19130 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19131 (make_vector_bool_mode): New function.
19132 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19134 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19135 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19137 * stor-layout.c (int_mode_for_mode): Likewise.
19138 * tree.c (build_vector_type_for_mode): Likewise.
19139 * varasm.c (output_constant_pool_2): Likewise.
19140 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19141 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19142 for MODE_VECTOR_BOOL.
19143 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19144 of mode class checks.
19145 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19146 instead of a list of mode class checks.
19147 (expand_vector_scalar_condition): Likewise.
19148 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19151 Alan Hayward <alan.hayward@arm.com>
19152 David Sherwood <david.sherwood@arm.com>
19154 * machmode.h (mode_size): Change from unsigned short to
19156 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19157 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19158 or if measurement_type is not polynomial.
19159 (fixed_size_mode::includes_p): Check for constant-sized modes.
19160 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19161 return a poly_uint16 rather than an unsigned short.
19162 (emit_mode_size): Change the type of mode_size from unsigned short
19163 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19164 (emit_mode_adjustments): Cope with polynomial vector sizes.
19165 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19167 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19169 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19170 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19171 * caller-save.c (setup_save_areas): Likewise.
19172 (replace_reg_with_saved_mem): Likewise.
19173 * calls.c (emit_library_call_value_1): Likewise.
19174 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19175 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19176 (gen_lowpart_for_combine): Likewise.
19177 * convert.c (convert_to_integer_1): Likewise.
19178 * cse.c (equiv_constant, cse_insn): Likewise.
19179 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19180 (cselib_subst_to_values): Likewise.
19181 * dce.c (word_dce_process_block): Likewise.
19182 * df-problems.c (df_word_lr_mark_ref): Likewise.
19183 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19184 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19185 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19186 (rtl_for_decl_location): Likewise.
19187 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19188 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19189 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19190 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19191 (expand_expr_real_1): Likewise.
19192 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19193 (pad_below): Likewise.
19194 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19195 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19196 * ira.c (get_subreg_tracking_sizes): Likewise.
19197 * ira-build.c (ira_create_allocno_objects): Likewise.
19198 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19199 (ira_sort_regnos_for_alter_reg): Likewise.
19200 * ira-costs.c (record_operand_costs): Likewise.
19201 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19202 (resolve_simple_move): Likewise.
19203 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19204 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19205 (lra_constraints): Likewise.
19206 (CONST_POOL_OK_P): Reject variable-sized modes.
19207 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19208 (add_pseudo_to_slot, lra_spill): Likewise.
19209 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19210 * optabs-query.c (get_best_extraction_insn): Likewise.
19211 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19212 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19213 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19214 * recog.c (offsettable_address_addr_space_p): Likewise.
19215 * regcprop.c (maybe_mode_change): Likewise.
19216 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19217 * regrename.c (build_def_use): Likewise.
19218 * regstat.c (dump_reg_info): Likewise.
19219 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19220 (find_reloads, find_reloads_subreg_address): Likewise.
19221 * reload1.c (eliminate_regs_1): Likewise.
19222 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19223 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19224 (simplify_binary_operation_1, simplify_subreg): Likewise.
19225 * targhooks.c (default_function_arg_padding): Likewise.
19226 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19227 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19228 (verify_gimple_assign_ternary): Likewise.
19229 * tree-inline.c (estimate_move_cost): Likewise.
19230 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19231 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19232 (get_address_cost_ainc): Likewise.
19233 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19234 (vect_supportable_dr_alignment): Likewise.
19235 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19236 (vectorizable_reduction): Likewise.
19237 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19238 (vectorizable_operation, vectorizable_load): Likewise.
19239 * tree.c (build_same_sized_truth_vector_type): Likewise.
19240 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19241 * var-tracking.c (emit_note_insn_var_location): Likewise.
19242 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19243 (ADDR_VEC_ALIGN): Likewise.
19245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19246 Alan Hayward <alan.hayward@arm.com>
19247 David Sherwood <david.sherwood@arm.com>
19249 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19251 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19252 or if measurement_type is polynomial.
19253 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19254 * combine.c (make_extraction): Likewise.
19255 * dse.c (find_shift_sequence): Likewise.
19256 * dwarf2out.c (mem_loc_descriptor): Likewise.
19257 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19258 (extract_bit_field, extract_low_bits): Likewise.
19259 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19260 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19261 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19262 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19263 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19264 * reload.c (find_reloads): Likewise.
19265 * reload1.c (alter_reg): Likewise.
19266 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19267 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19268 * tree-if-conv.c (predicate_mem_writes): Likewise.
19269 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19270 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19271 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19272 * valtrack.c (dead_debug_insert_temp): Likewise.
19273 * varasm.c (mergeable_constant_section): Likewise.
19274 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19277 Alan Hayward <alan.hayward@arm.com>
19278 David Sherwood <david.sherwood@arm.com>
19280 * expr.c (expand_assignment): Cope with polynomial mode sizes
19281 when assigning to a CONCAT.
19283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19284 Alan Hayward <alan.hayward@arm.com>
19285 David Sherwood <david.sherwood@arm.com>
19287 * machmode.h (mode_precision): Change from unsigned short to
19289 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19291 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19292 or if measurement_type is not polynomial.
19293 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19294 in which the mode is already known to be a scalar_int_mode.
19295 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19296 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19298 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19299 for GET_MODE_PRECISION.
19300 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19301 for GET_MODE_PRECISION.
19302 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19304 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19305 (expand_field_assignment, make_extraction): Likewise.
19306 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19307 (get_last_value): Likewise.
19308 * convert.c (convert_to_integer_1): Likewise.
19309 * cse.c (cse_insn): Likewise.
19310 * expr.c (expand_expr_real_1): Likewise.
19311 * lra-constraints.c (simplify_operand_subreg): Likewise.
19312 * optabs-query.c (can_atomic_load_p): Likewise.
19313 * optabs.c (expand_atomic_load): Likewise.
19314 (expand_atomic_store): Likewise.
19315 * ree.c (combine_reaching_defs): Likewise.
19316 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19317 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19318 * tree.h (type_has_mode_precision_p): Likewise.
19319 * ubsan.c (instrument_si_overflow): Likewise.
19321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19322 Alan Hayward <alan.hayward@arm.com>
19323 David Sherwood <david.sherwood@arm.com>
19325 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19326 polynomial numbers of units.
19327 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19328 (valid_vector_subparts_p): New function.
19329 (build_vector_type): Remove temporary shim and take the number
19330 of units as a poly_uint64 rather than an int.
19331 (build_opaque_vector_type): Take the number of units as a
19332 poly_uint64 rather than an int.
19333 * tree.c (build_vector_from_ctor): Handle polynomial
19334 TYPE_VECTOR_SUBPARTS.
19335 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19336 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19337 (build_vector_from_val): If the number of units is variable,
19338 use build_vec_duplicate_cst for constant operands and
19339 VEC_DUPLICATE_EXPR otherwise.
19340 (make_vector_type): Remove temporary is_constant ().
19341 (build_vector_type, build_opaque_vector_type): Take the number of
19342 units as a poly_uint64 rather than an int.
19343 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19345 * cfgexpand.c (expand_debug_expr): Likewise.
19346 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19347 (store_constructor, expand_expr_real_1): Likewise.
19348 (const_scalar_mask_from_tree): Likewise.
19349 * fold-const-call.c (fold_const_reduction): Likewise.
19350 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19351 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19352 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19353 (fold_relational_const): Likewise.
19354 (native_interpret_vector): Likewise. Change the size from an
19355 int to an unsigned int.
19356 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19357 TYPE_VECTOR_SUBPARTS.
19358 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19359 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19360 duplicating a non-constant operand into a variable-length vector.
19361 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19362 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19363 * ipa-icf.c (sem_variable::equals): Likewise.
19364 * match.pd: Likewise.
19365 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19366 * print-tree.c (print_node): Likewise.
19367 * stor-layout.c (layout_type): Likewise.
19368 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19369 * tree-cfg.c (verify_gimple_comparison): Likewise.
19370 (verify_gimple_assign_binary): Likewise.
19371 (verify_gimple_assign_ternary): Likewise.
19372 (verify_gimple_assign_single): Likewise.
19373 * tree-pretty-print.c (dump_generic_node): Likewise.
19374 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19375 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19376 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19377 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19378 (vect_shift_permute_load_chain): Likewise.
19379 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19380 (expand_vector_condition, optimize_vector_constructor): Likewise.
19381 (lower_vec_perm, get_compute_type): Likewise.
19382 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19383 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19384 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19385 (vect_recog_mask_conversion_pattern): Likewise.
19386 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19387 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19388 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19389 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19390 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19391 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19392 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19393 (supportable_widening_operation): Likewise.
19394 (supportable_narrowing_operation): Likewise.
19395 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19397 * varasm.c (output_constant): Likewise.
19399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19400 Alan Hayward <alan.hayward@arm.com>
19401 David Sherwood <david.sherwood@arm.com>
19403 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19404 so that both the length == 3 and length != 3 cases set up their
19405 own permute vectors. Add comments explaining why we know the
19406 number of elements is constant.
19407 (vect_permute_load_chain): Likewise.
19409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19410 Alan Hayward <alan.hayward@arm.com>
19411 David Sherwood <david.sherwood@arm.com>
19413 * machmode.h (mode_nunits): Change from unsigned char to
19415 (ONLY_FIXED_SIZE_MODES): New macro.
19416 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19417 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19418 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19420 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19421 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19422 or if measurement_type is not polynomial.
19423 * genmodes.c (ZERO_COEFFS): New macro.
19424 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19426 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19427 Use ZERO_COEFFS when emitting initializers.
19428 * data-streamer.h (bp_pack_poly_value): New function.
19429 (bp_unpack_poly_value): Likewise.
19430 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19431 for GET_MODE_NUNITS.
19432 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19433 for GET_MODE_NUNITS.
19434 * tree.c (make_vector_type): Remove temporary shim and make
19435 the real function take the number of units as a poly_uint64
19436 rather than an int.
19437 (build_vector_type_for_mode): Handle polynomial nunits.
19438 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19439 * emit-rtl.c (const_vec_series_p_1): Likewise.
19440 (gen_rtx_CONST_VECTOR): Likewise.
19441 * fold-const.c (test_vec_duplicate_folding): Likewise.
19442 * genrecog.c (validate_pattern): Likewise.
19443 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19444 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19445 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19446 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19447 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19448 * rtlanal.c (subreg_get_info): Likewise.
19449 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19450 (vect_grouped_load_supported): Likewise.
19451 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19452 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19453 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19454 (simplify_const_unary_operation, simplify_binary_operation_1)
19455 (simplify_const_binary_operation, simplify_ternary_operation)
19456 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19457 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19458 instead of CONST_VECTOR_NUNITS.
19459 * varasm.c (output_constant_pool_2): Likewise.
19460 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19461 explicit-encoded elements in the XVEC for variable-length vectors.
19463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19465 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19468 Alan Hayward <alan.hayward@arm.com>
19469 David Sherwood <david.sherwood@arm.com>
19471 * coretypes.h (fixed_size_mode): Declare.
19472 (fixed_size_mode_pod): New typedef.
19473 * builtins.h (target_builtins::x_apply_args_mode)
19474 (target_builtins::x_apply_result_mode): Change type to
19475 fixed_size_mode_pod.
19476 * builtins.c (apply_args_size, apply_result_size, result_vector)
19477 (expand_builtin_apply_args_1, expand_builtin_apply)
19478 (expand_builtin_return): Update accordingly.
19480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19482 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19483 * cselib.c (cselib_hash_rtx): Likewise.
19484 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19485 CONST_VECTOR encoding.
19487 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19488 Jeff Law <law@redhat.com>
19491 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19492 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19493 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19494 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19497 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19498 explicitly probe *sp in a noreturn function if there were any callee
19499 register saves or frame pointer is needed.
19501 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19504 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19505 BLKmode for ternary, binary or unary expressions.
19508 * var-tracking.c (delete_vta_debug_insn): New inline function.
19509 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19510 insns from get_insns () to NULL instead of each bb separately.
19511 Use delete_vta_debug_insn. No longer static.
19512 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19513 delete_vta_debug_insns callers.
19514 * rtl.h (delete_vta_debug_insns): Declare.
19515 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19516 instead of variable_tracking_main.
19518 2018-01-03 Martin Sebor <msebor@redhat.com>
19520 PR tree-optimization/83603
19521 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19522 arguments past the endof the argument list in functions declared
19523 without a prototype.
19524 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19525 Avoid checking when arguments are null.
19527 2018-01-03 Martin Sebor <msebor@redhat.com>
19530 * doc/extend.texi (attribute const): Fix a typo.
19531 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19532 issuing -Wsuggest-attribute for void functions.
19534 2018-01-03 Martin Sebor <msebor@redhat.com>
19536 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19537 offset_int::from instead of wide_int::to_shwi.
19538 (maybe_diag_overlap): Remove assertion.
19539 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19540 * gimple-ssa-sprintf.c (format_directive): Same.
19541 (parse_directive): Same.
19542 (sprintf_dom_walker::compute_format_length): Same.
19543 (try_substitute_return_value): Same.
19545 2018-01-03 Jeff Law <law@redhat.com>
19547 PR middle-end/83654
19548 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19549 non-constant residual for zero at runtime and avoid probing in
19550 that case. Reorganize code for trailing problem to mirror handling
19553 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19555 PR tree-optimization/83501
19556 * tree-ssa-strlen.c (get_string_cst): New.
19557 (handle_char_store): Call get_string_cst.
19559 2018-01-03 Martin Liska <mliska@suse.cz>
19561 PR tree-optimization/83593
19562 * tree-ssa-strlen.c: Include tree-cfg.h.
19563 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19564 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19565 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19567 (strlen_dom_walker::before_dom_children): Call
19568 gimple_purge_dead_eh_edges. Dump tranformation with details
19570 (strlen_dom_walker::before_dom_children): Update call by adding
19571 new argument cleanup_eh.
19572 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19574 2018-01-03 Martin Liska <mliska@suse.cz>
19577 * cif-code.def (VARIADIC_THUNK): New enum value.
19578 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19581 2018-01-03 Jan Beulich <jbeulich@suse.com>
19583 * sse.md (mov<mode>_internal): Tighten condition for when to use
19584 vmovdqu<ssescalarsize> for TI and OI modes.
19586 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19588 Update copyright years.
19590 2018-01-03 Martin Liska <mliska@suse.cz>
19593 * ipa-visibility.c (function_and_variable_visibility): Skip
19594 functions with noipa attribure.
19596 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19598 * gcc.c (process_command): Update copyright notice dates.
19599 * gcov-dump.c (print_version): Ditto.
19600 * gcov.c (print_version): Ditto.
19601 * gcov-tool.c (print_version): Ditto.
19602 * gengtype.c (create_file): Ditto.
19603 * doc/cpp.texi: Bump @copying's copyright year.
19604 * doc/cppinternals.texi: Ditto.
19605 * doc/gcc.texi: Ditto.
19606 * doc/gccint.texi: Ditto.
19607 * doc/gcov.texi: Ditto.
19608 * doc/install.texi: Ditto.
19609 * doc/invoke.texi: Ditto.
19611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19613 * vector-builder.h (vector_builder::m_full_nelts): Change from
19614 unsigned int to poly_uint64.
19615 (vector_builder::full_nelts): Update prototype accordingly.
19616 (vector_builder::new_vector): Likewise.
19617 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19618 (vector_builder::operator ==): Likewise.
19619 (vector_builder::finalize): Likewise.
19620 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19621 Take the number of elements as a poly_uint64 rather than an
19623 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19624 from unsigned int to poly_uint64.
19625 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19626 (vec_perm_indices::new_vector): Likewise.
19627 (vec_perm_indices::length): Likewise.
19628 (vec_perm_indices::nelts_per_input): Likewise.
19629 (vec_perm_indices::input_nelts): Likewise.
19630 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19631 number of elements per input as a poly_uint64 rather than an
19632 unsigned int. Use the original encoding for variable-length
19633 vectors, rather than clamping each individual element.
19634 For the second and subsequent elements in each pattern,
19635 clamp the step and base before clamping their sum.
19636 (vec_perm_indices::series_p): Handle polynomial element counts.
19637 (vec_perm_indices::all_in_range_p): Likewise.
19638 (vec_perm_indices_to_tree): Likewise.
19639 (vec_perm_indices_to_rtx): Likewise.
19640 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19641 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19642 (tree_vector_builder::new_binary_operation): Handle polynomial
19643 element counts. Return false if we need to know the number
19644 of elements at compile time.
19645 * fold-const.c (fold_vec_perm): Punt if the number of elements
19646 isn't known at compile time.
19648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19650 * vec-perm-indices.h (vec_perm_builder): Change element type
19651 from HOST_WIDE_INT to poly_int64.
19652 (vec_perm_indices::element_type): Update accordingly.
19653 (vec_perm_indices::clamp): Handle polynomial element_types.
19654 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19655 (vec_perm_indices::all_in_range_p): Likewise.
19656 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19658 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19659 polynomial vec_perm_indices element types.
19660 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19661 * fold-const.c (fold_vec_perm): Likewise.
19662 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19663 * tree-vect-generic.c (lower_vec_perm): Likewise.
19664 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19665 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19666 element type to HOST_WIDE_INT.
19668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19669 Alan Hayward <alan.hayward@arm.com>
19670 David Sherwood <david.sherwood@arm.com>
19672 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19673 rather than an int. Use plus_constant.
19674 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19675 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19678 Alan Hayward <alan.hayward@arm.com>
19679 David Sherwood <david.sherwood@arm.com>
19681 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19682 a HOST_WIDE_INT to a poly_int64.
19684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19685 Alan Hayward <alan.hayward@arm.com>
19686 David Sherwood <david.sherwood@arm.com>
19688 * calls.c (load_register_parameters): Cope with polynomial
19689 mode sizes. Require a constant size for BLKmode parameters
19690 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19691 forces a parameter to be padded at the lsb end in order to
19692 fill a complete number of words, require the parameter size
19693 to be ordered wrt UNITS_PER_WORD.
19695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19696 Alan Hayward <alan.hayward@arm.com>
19697 David Sherwood <david.sherwood@arm.com>
19699 * reload1.c (spill_stack_slot_width): Change element type
19700 from unsigned int to poly_uint64_pod.
19701 (alter_reg): Treat mode sizes as polynomial.
19703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19704 Alan Hayward <alan.hayward@arm.com>
19705 David Sherwood <david.sherwood@arm.com>
19707 * reload.c (complex_word_subreg_p): New function.
19708 (reload_inner_reg_of_subreg, push_reload): Use it.
19710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19711 Alan Hayward <alan.hayward@arm.com>
19712 David Sherwood <david.sherwood@arm.com>
19714 * lra-constraints.c (process_alt_operands): Reject matched
19715 operands whose sizes aren't ordered.
19716 (match_reload): Refer to this check here.
19718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19719 Alan Hayward <alan.hayward@arm.com>
19720 David Sherwood <david.sherwood@arm.com>
19722 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19723 that the mode size is in the set {1, 2, 4, 8, 16}.
19725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19726 Alan Hayward <alan.hayward@arm.com>
19727 David Sherwood <david.sherwood@arm.com>
19729 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19730 Use plus_constant instead of gen_rtx_PLUS.
19732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19733 Alan Hayward <alan.hayward@arm.com>
19734 David Sherwood <david.sherwood@arm.com>
19736 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19737 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19738 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19739 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19740 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19741 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19742 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19743 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19744 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19745 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19747 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19748 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19749 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19750 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19751 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19752 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19753 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19754 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19755 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19756 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19758 * expr.c (emit_move_resolve_push): Treat the input and result
19759 of PUSH_ROUNDING as a poly_int64.
19760 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19761 (emit_push_insn): Likewise.
19762 * lra-eliminations.c (mark_not_eliminable): Likewise.
19763 * recog.c (push_operand): Likewise.
19764 * reload1.c (elimination_effects): Likewise.
19765 * rtlanal.c (nonzero_bits1): Likewise.
19766 * calls.c (store_one_arg): Likewise. Require the padding to be
19767 known at compile time.
19769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19770 Alan Hayward <alan.hayward@arm.com>
19771 David Sherwood <david.sherwood@arm.com>
19773 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19774 Use plus_constant instead of gen_rtx_PLUS.
19776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19777 Alan Hayward <alan.hayward@arm.com>
19778 David Sherwood <david.sherwood@arm.com>
19780 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19781 rather than an int.
19783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19784 Alan Hayward <alan.hayward@arm.com>
19785 David Sherwood <david.sherwood@arm.com>
19787 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19788 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19789 via stack temporaries. Treat the mode size as polynomial too.
19791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19792 Alan Hayward <alan.hayward@arm.com>
19793 David Sherwood <david.sherwood@arm.com>
19795 * expr.c (expand_expr_real_2): When handling conversions involving
19796 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19797 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19798 as a poly_uint64 too.
19800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19801 Alan Hayward <alan.hayward@arm.com>
19802 David Sherwood <david.sherwood@arm.com>
19804 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19807 Alan Hayward <alan.hayward@arm.com>
19808 David Sherwood <david.sherwood@arm.com>
19810 * combine.c (can_change_dest_mode): Handle polynomial
19811 REGMODE_NATURAL_SIZE.
19812 * expmed.c (store_bit_field_1): Likewise.
19813 * expr.c (store_constructor): Likewise.
19814 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19815 and polynomial REGMODE_NATURAL_SIZE.
19816 (gen_lowpart_common): Likewise.
19817 * reginfo.c (record_subregs_of_mode): Likewise.
19818 * rtlanal.c (read_modify_subreg_p): Likewise.
19820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19821 Alan Hayward <alan.hayward@arm.com>
19822 David Sherwood <david.sherwood@arm.com>
19824 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19825 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
19833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19834 Alan Hayward <alan.hayward@arm.com>
19835 David Sherwood <david.sherwood@arm.com>
19837 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19838 in a POINTER_PLUS_EXPR.
19840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19841 Alan Hayward <alan.hayward@arm.com>
19842 David Sherwood <david.sherwood@arm.com>
19844 * omp-simd-clone.c (simd_clone_subparts): New function.
19845 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19846 (ipa_simd_modify_function_body): Likewise.
19848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19849 Alan Hayward <alan.hayward@arm.com>
19850 David Sherwood <david.sherwood@arm.com>
19852 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19853 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19854 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19855 (expand_vector_condition, vector_element): Likewise.
19856 (subparts_gt): New function.
19857 (get_compute_type): Use subparts_gt.
19858 (count_type_subparts): Delete.
19859 (expand_vector_operations_1): Use subparts_gt instead of
19860 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
19867 (vect_compile_time_alias): ...this new function. Do the calculation
19868 on poly_ints rather than trees.
19869 (vect_prune_runtime_alias_test_list): Update call accordingly.
19871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19872 Alan Hayward <alan.hayward@arm.com>
19873 David Sherwood <david.sherwood@arm.com>
19875 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19877 (vect_schedule_slp_instance): Likewise.
19879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19880 Alan Hayward <alan.hayward@arm.com>
19881 David Sherwood <david.sherwood@arm.com>
19883 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19884 constant and extern definitions for variable-length vectors.
19885 (vect_get_constant_vectors): Note that the number of units
19886 is known to be constant.
19888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19889 Alan Hayward <alan.hayward@arm.com>
19890 David Sherwood <david.sherwood@arm.com>
19892 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19893 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
19901 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19908 vectors as polynomial. Use build_index_vector for
19909 IFN_GOMP_SIMD_LANE.
19911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19912 Alan Hayward <alan.hayward@arm.com>
19913 David Sherwood <david.sherwood@arm.com>
19915 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19916 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19917 for variable-length vectors.
19918 (vectorizable_mask_load_store): Treat the number of units as
19919 polynomial, asserting that it is constant if the condition has
19920 already been enforced.
19921 (vectorizable_store, vectorizable_load): Likewise.
19923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19924 Alan Hayward <alan.hayward@arm.com>
19925 David Sherwood <david.sherwood@arm.com>
19927 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19928 of units as polynomial. Punt if we can't tell at compile time
19929 which vector contains the final result.
19931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19932 Alan Hayward <alan.hayward@arm.com>
19933 David Sherwood <david.sherwood@arm.com>
19935 * tree-vect-loop.c (vectorizable_induction): Treat the number
19936 of units as polynomial. Punt on SLP inductions. Use an integer
19937 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19938 cast of such a series for variable-length floating-point
19941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19942 Alan Hayward <alan.hayward@arm.com>
19943 David Sherwood <david.sherwood@arm.com>
19945 * tree.h (build_index_vector): Declare.
19946 * tree.c (build_index_vector): New function.
19947 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19948 of units as polynomial, forcibly converting it to a constant if
19949 vectorizable_reduction has already enforced the condition.
19950 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19951 to create a {1,2,3,...} vector.
19952 (vectorizable_reduction): Treat the number of units as polynomial.
19953 Choose vectype_in based on the largest scalar element size rather
19954 than the smallest number of units. Enforce the restrictions
19957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19958 Alan Hayward <alan.hayward@arm.com>
19959 David Sherwood <david.sherwood@arm.com>
19961 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19962 number of units as polynomial.
19964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19965 Alan Hayward <alan.hayward@arm.com>
19966 David Sherwood <david.sherwood@arm.com>
19968 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19969 * target.def (autovectorize_vector_sizes): Return the vector sizes
19970 by pointer, using vector_sizes rather than a bitmask.
19971 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19972 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19973 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19975 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19976 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19977 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19978 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19979 * omp-general.c (omp_max_vf): Likewise.
19980 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19981 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19982 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19983 * tree-vect-slp.c (vect_slp_bb): Likewise.
19984 * doc/tm.texi: Regenerate.
19985 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19987 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19988 the vector size as a poly_uint64 rather than an unsigned int.
19989 (current_vector_size): Change from an unsigned int to a poly_uint64.
19990 (get_vectype_for_scalar_type): Update accordingly.
19991 * tree.h (build_truth_vector_type): Take the size and number of
19992 units as a poly_uint64 rather than an unsigned int.
19993 (build_vector_type): Add a temporary overload that takes
19994 the number of units as a poly_uint64 rather than an unsigned int.
19995 * tree.c (make_vector_type): Likewise.
19996 (build_truth_vector_type): Take the number of units as a poly_uint64
19997 rather than an unsigned int.
19999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20000 Alan Hayward <alan.hayward@arm.com>
20001 David Sherwood <david.sherwood@arm.com>
20003 * target.def (get_mask_mode): Take the number of units and length
20004 as poly_uint64s rather than unsigned ints.
20005 * targhooks.h (default_get_mask_mode): Update accordingly.
20006 * targhooks.c (default_get_mask_mode): Likewise.
20007 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20008 * doc/tm.texi: Regenerate.
20010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20011 Alan Hayward <alan.hayward@arm.com>
20012 David Sherwood <david.sherwood@arm.com>
20014 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20015 * omp-general.c (omp_max_vf): Likewise.
20016 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20017 (expand_omp_simd): Handle polynomial safelen.
20018 * omp-low.c (omplow_simd_context): Add a default constructor.
20019 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20020 (lower_rec_simd_input_clauses): Update accordingly.
20021 (lower_rec_input_clauses): Likewise.
20023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20024 Alan Hayward <alan.hayward@arm.com>
20025 David Sherwood <david.sherwood@arm.com>
20027 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20028 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20029 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20030 (vect_analyze_slp_cost): Likewise.
20031 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20032 (vect_model_load_cost): Likewise.
20034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20035 Alan Hayward <alan.hayward@arm.com>
20036 David Sherwood <david.sherwood@arm.com>
20038 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20039 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20040 from an unsigned int * to a poly_uint64_pod *.
20041 (calculate_unrolling_factor): New function.
20042 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20045 Alan Hayward <alan.hayward@arm.com>
20046 David Sherwood <david.sherwood@arm.com>
20048 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20049 from an unsigned int to a poly_uint64.
20050 (_loop_vec_info::slp_unrolling_factor): Likewise.
20051 (_loop_vec_info::vectorization_factor): Change from an int
20053 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20054 (vect_get_num_vectors): New function.
20055 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20056 (vect_get_num_copies): Use vect_get_num_vectors.
20057 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20058 to an unsigned int *.
20059 (vect_analyze_data_refs): Change min_vf from an int * to a
20061 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20062 than an unsigned HOST_WIDE_INT.
20063 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20064 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20065 to an unsigned int *.
20066 (vect_analyze_data_ref_dependences): Likewise.
20067 (vect_compute_data_ref_alignment): Handle polynomial vf.
20068 (vect_enhance_data_refs_alignment): Likewise.
20069 (vect_prune_runtime_alias_test_list): Likewise.
20070 (vect_shift_permute_load_chain): Likewise.
20071 (vect_supportable_dr_alignment): Likewise.
20072 (dependence_distance_ge_vf): Take the vectorization factor as a
20073 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20074 (vect_analyze_data_refs): Change min_vf from an int * to a
20076 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20077 vfm1 as a poly_uint64 rather than an int. Make the same change
20078 for the returned bound_scalar.
20079 (vect_gen_vector_loop_niters): Handle polynomial vf.
20080 (vect_do_peeling): Likewise. Update call to
20081 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20082 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20084 * tree-vect-loop.c (vect_determine_vectorization_factor)
20085 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20086 (vect_get_known_peeling_cost): Likewise.
20087 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20088 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20089 (vect_transform_loop): Likewise. Use the lowest possible VF when
20090 updating the upper bounds of the loop.
20091 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20092 rather than an int.
20093 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20094 polynomial unroll factors.
20095 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20096 (vect_make_slp_decision): Likewise.
20097 (vect_supported_load_permutation_p): Likewise, and polynomial
20099 (vect_analyze_slp_cost): Handle polynomial vf.
20100 (vect_slp_analyze_node_operations): Likewise.
20101 (vect_slp_analyze_bb_1): Likewise.
20102 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20103 than an unsigned HOST_WIDE_INT.
20104 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20105 (vectorizable_load): Handle polynomial vf.
20106 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20108 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20111 Alan Hayward <alan.hayward@arm.com>
20112 David Sherwood <david.sherwood@arm.com>
20114 * match.pd: Handle bit operations involving three constants
20115 and try to fold one pair.
20117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20119 * tree-vect-loop-manip.c: Include gimple-fold.h.
20120 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20121 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20122 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20123 Add a path that uses a step of VF instead of 1, but disable it
20125 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20126 and niters_no_overflow parameters. Update calls to
20127 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20128 Create a new SSA name if the latter choses to use a ste other
20129 than zero, and return it via niters_vector_mult_vf_var.
20130 * tree-vect-loop.c (vect_transform_loop): Update calls to
20131 vect_do_peeling, vect_gen_vector_loop_niters and
20132 slpeel_make_loop_iterate_ntimes.
20133 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20134 (vect_gen_vector_loop_niters): Update declarations after above changes.
20136 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20138 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20139 128-bit round to integer instructions.
20140 (ceil<mode>2): Likewise.
20141 (btrunc<mode>2): Likewise.
20142 (round<mode>2): Likewise.
20144 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20146 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20147 unaligned VSX load/store on P8/P9.
20148 (expand_block_clear): Allow the use of unaligned VSX
20149 load/store on P8/P9.
20151 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20153 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20155 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20156 swap associated with both a load and a store.
20158 2018-01-02 Andrew Waterman <andrew@sifive.com>
20160 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20161 * config/riscv/riscv.md (clear_cache): Use it.
20163 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20165 * web.c: Remove out-of-date comment.
20167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20169 * expr.c (fixup_args_size_notes): Check that any existing
20170 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20171 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20172 (emit_single_push_insn): ...here.
20174 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20176 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20177 (const_vector_encoded_nelts): New function.
20178 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20179 (const_vector_int_elt, const_vector_elt): Declare.
20180 * emit-rtl.c (const_vector_int_elt_1): New function.
20181 (const_vector_elt): Likewise.
20182 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20183 of CONST_VECTOR_ELT.
20185 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20187 * expr.c: Include rtx-vector-builder.h.
20188 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20189 directly on the tree encoding.
20190 (const_vector_from_tree): Likewise.
20191 * optabs.c: Include rtx-vector-builder.h.
20192 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20193 sequence of "u" values.
20194 * vec-perm-indices.c: Include rtx-vector-builder.h.
20195 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20196 directly on the vec_perm_indices encoding.
20198 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20200 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20201 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20202 * rtx-vector-builder.h: New file.
20203 * rtx-vector-builder.c: Likewise.
20204 * rtl.h (rtx_def::u2): Add a const_vector field.
20205 (CONST_VECTOR_NPATTERNS): New macro.
20206 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20207 (CONST_VECTOR_DUPLICATE_P): Likewise.
20208 (CONST_VECTOR_STEPPED_P): Likewise.
20209 (CONST_VECTOR_ENCODED_ELT): Likewise.
20210 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20211 (unwrap_const_vec_duplicate): Likewise.
20212 (const_vec_series_p): Check for a non-duplicated vector encoding.
20213 Say that the function only returns true for integer vectors.
20214 * emit-rtl.c: Include rtx-vector-builder.h.
20215 (gen_const_vec_duplicate_1): Delete.
20216 (gen_const_vector): Call gen_const_vec_duplicate instead of
20217 gen_const_vec_duplicate_1.
20218 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20219 (gen_const_vec_duplicate): Use rtx_vector_builder.
20220 (gen_const_vec_series): Likewise.
20221 (gen_rtx_CONST_VECTOR): Likewise.
20222 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20223 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20224 Build a new vector rather than modifying a CONST_VECTOR in-place.
20225 (handle_special_swappables): Update call accordingly.
20226 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20227 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20228 Build a new vector rather than modifying a CONST_VECTOR in-place.
20229 (handle_special_swappables): Update call accordingly.
20231 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20233 * simplify-rtx.c (simplify_const_binary_operation): Use
20234 CONST_VECTOR_ELT instead of XVECEXP.
20236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20238 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20239 the selector elements to be different from the data elements
20240 if the selector is a VECTOR_CST.
20241 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20242 ssizetype for the selector.
20244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20246 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20247 before testing each element individually.
20248 * tree-vect-generic.c (lower_vec_perm): Likewise.
20250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20252 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20253 * selftest-run-tests.c (selftest::run_tests): Call it.
20254 * vector-builder.h (vector_builder::operator ==): New function.
20255 (vector_builder::operator !=): Likewise.
20256 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20257 (vec_perm_indices::all_from_input_p): New function.
20258 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20259 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20260 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20261 instead of reading the VECTOR_CST directly. Detect whether both
20262 vector inputs are the same before constructing the vec_perm_indices,
20263 and update the number of inputs argument accordingly. Use the
20264 utility functions added above. Only construct sel2 if we need to.
20266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20268 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20269 the broadcast of the low byte.
20270 (expand_mult_highpart): Use an explicit encoding for the permutes.
20271 * optabs-query.c (can_mult_highpart_p): Likewise.
20272 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20273 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20274 (vectorizable_bswap): Likewise.
20275 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20276 explicit encoding for the power-of-2 permutes.
20277 (vect_permute_store_chain): Likewise.
20278 (vect_grouped_load_supported): Likewise.
20279 (vect_permute_load_chain): Likewise.
20281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20283 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20284 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20285 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20286 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20287 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20288 (vect_gen_perm_mask_any): Likewise.
20290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20292 * int-vector-builder.h: New file.
20293 * vec-perm-indices.h: Include int-vector-builder.h.
20294 (vec_perm_indices): Redefine as an int_vector_builder.
20295 (auto_vec_perm_indices): Delete.
20296 (vec_perm_builder): Redefine as a stand-alone class.
20297 (vec_perm_indices::vec_perm_indices): New function.
20298 (vec_perm_indices::clamp): Likewise.
20299 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20300 (vec_perm_indices::new_vector): New function.
20301 (vec_perm_indices::new_expanded_vector): Update for new
20302 vec_perm_indices class.
20303 (vec_perm_indices::rotate_inputs): New function.
20304 (vec_perm_indices::all_in_range_p): Operate directly on the
20305 encoded form, without computing elided elements.
20306 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20307 encoding. Update for new vec_perm_indices class.
20308 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20309 the given vec_perm_builder.
20310 (expand_vec_perm_var): Update vec_perm_builder constructor.
20311 (expand_mult_highpart): Use vec_perm_builder instead of
20312 auto_vec_perm_indices.
20313 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20314 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20315 or double series encoding as appropriate.
20316 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20317 vec_perm_indices instead of auto_vec_perm_indices.
20318 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20319 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20320 (vect_permute_store_chain): Likewise.
20321 (vect_grouped_load_supported): Likewise.
20322 (vect_permute_load_chain): Likewise.
20323 (vect_shift_permute_load_chain): Likewise.
20324 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20325 (vect_transform_slp_perm_load): Likewise.
20326 (vect_schedule_slp_instance): Likewise.
20327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20328 (vectorizable_mask_load_store): Likewise.
20329 (vectorizable_bswap): Likewise.
20330 (vectorizable_store): Likewise.
20331 (vectorizable_load): Likewise.
20332 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20333 vec_perm_indices instead of auto_vec_perm_indices. Use
20334 tree_to_vec_perm_builder to read the vector from a tree.
20335 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20336 vec_perm_builder instead of a vec_perm_indices.
20337 (have_whole_vector_shift): Use vec_perm_builder and
20338 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20339 truncation to calc_vec_perm_mask_for_shift.
20340 (vect_create_epilog_for_reduction): Likewise.
20341 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20342 from auto_vec_perm_indices to vec_perm_indices.
20343 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20344 instead of changing individual elements.
20345 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20346 the vector in d.perm.
20347 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20348 from auto_vec_perm_indices to vec_perm_indices.
20349 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20350 instead of changing individual elements.
20351 (arm_vectorize_vec_perm_const): Use new_vector to install
20352 the vector in d.perm.
20353 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20354 Update vec_perm_builder constructor.
20355 (rs6000_expand_interleave): Likewise.
20356 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20357 (rs6000_expand_interleave): Likewise.
20359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20361 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20362 to qimode could truncate the indices.
20363 * optabs.c (expand_vec_perm_var): Likewise.
20365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20367 * Makefile.in (OBJS): Add vec-perm-indices.o.
20368 * vec-perm-indices.h: New file.
20369 * vec-perm-indices.c: Likewise.
20370 * target.h (vec_perm_indices): Replace with a forward class
20372 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20373 * optabs.h: Include vec-perm-indices.h.
20374 (expand_vec_perm): Delete.
20375 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20376 (expand_vec_perm_const): Declare.
20377 * target.def (vec_perm_const_ok): Replace with...
20378 (vec_perm_const): ...this new hook.
20379 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20380 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20381 * doc/tm.texi: Regenerate.
20382 * optabs.def (vec_perm_const): Delete.
20383 * doc/md.texi (vec_perm_const): Likewise.
20384 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20385 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20386 expand_vec_perm for constant permutation vectors. Assert that
20387 the mode of variable permutation vectors is the integer equivalent
20388 of the mode that is being permuted.
20389 * optabs-query.h (selector_fits_mode_p): Declare.
20390 * optabs-query.c: Include vec-perm-indices.h.
20391 (selector_fits_mode_p): New function.
20392 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20393 is defined, instead of checking whether the vec_perm_const_optab
20394 exists. Use targetm.vectorize.vec_perm_const instead of
20395 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20396 fit in the vector mode before using a variable permute.
20397 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20398 vec_perm_indices instead of an rtx.
20399 (expand_vec_perm): Replace with...
20400 (expand_vec_perm_const): ...this new function. Take the selector
20401 as a vec_perm_indices rather than an rtx. Also take the mode of
20402 the selector. Update call to shift_amt_for_vec_perm_mask.
20403 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20404 Use vec_perm_indices::new_expanded_vector to expand the original
20405 selector into bytes. Check whether the indices fit in the vector
20406 mode before using a variable permute.
20407 (expand_vec_perm_var): Make global.
20408 (expand_mult_highpart): Use expand_vec_perm_const.
20409 * fold-const.c: Includes vec-perm-indices.h.
20410 * tree-ssa-forwprop.c: Likewise.
20411 * tree-vect-data-refs.c: Likewise.
20412 * tree-vect-generic.c: Likewise.
20413 * tree-vect-loop.c: Likewise.
20414 * tree-vect-slp.c: Likewise.
20415 * tree-vect-stmts.c: Likewise.
20416 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20418 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20419 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20420 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20421 (aarch64_vectorize_vec_perm_const): ...this new function.
20422 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20423 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20424 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20425 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20426 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20427 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20428 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20430 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20431 check for NEON modes.
20432 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20433 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20434 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20435 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20437 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20438 the old VEC_PERM_CONST conditions.
20439 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20440 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20441 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20442 (ia64_vectorize_vec_perm_const_ok): Merge into...
20443 (ia64_vectorize_vec_perm_const): ...this new function.
20444 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20445 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20446 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20447 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20448 * config/mips/mips.c (mips_expand_vec_perm_const)
20449 (mips_vectorize_vec_perm_const_ok): Merge into...
20450 (mips_vectorize_vec_perm_const): ...this new function.
20451 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20452 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20453 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20454 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20455 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20456 (rs6000_expand_vec_perm_const): Delete.
20457 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20459 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20460 (altivec_expand_vec_perm_const_le): Take each operand individually.
20461 Operate on constant selectors rather than rtxes.
20462 (altivec_expand_vec_perm_const): Likewise. Update call to
20463 altivec_expand_vec_perm_const_le.
20464 (rs6000_expand_vec_perm_const): Delete.
20465 (rs6000_vectorize_vec_perm_const_ok): Delete.
20466 (rs6000_vectorize_vec_perm_const): New function.
20467 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20468 an element count and rtx array.
20469 (rs6000_expand_extract_even): Update call accordingly.
20470 (rs6000_expand_interleave): Likewise.
20471 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20472 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20473 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20474 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20475 (rs6000_expand_vec_perm_const): Delete.
20476 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20477 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20478 (altivec_expand_vec_perm_const_le): Take each operand individually.
20479 Operate on constant selectors rather than rtxes.
20480 (altivec_expand_vec_perm_const): Likewise. Update call to
20481 altivec_expand_vec_perm_const_le.
20482 (rs6000_expand_vec_perm_const): Delete.
20483 (rs6000_vectorize_vec_perm_const_ok): Delete.
20484 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20485 reference to the SPE evmerge intructions.
20486 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20487 an element count and rtx array.
20488 (rs6000_expand_extract_even): Update call accordingly.
20489 (rs6000_expand_interleave): Likewise.
20490 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20491 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20493 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20495 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20497 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20498 vector mode and that that mode matches the mode of the data
20500 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20501 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20502 directly using expand_vec_perm_1 when forcing selectors into
20504 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20508 * optabs-query.h (can_vec_perm_p): Delete.
20509 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20510 * optabs-query.c (can_vec_perm_p): Split into...
20511 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20512 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20513 particular selector is valid.
20514 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20515 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20516 (vect_grouped_load_supported): Likewise.
20517 (vect_shift_permute_load_chain): Likewise.
20518 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20519 (vect_transform_slp_perm_load): Likewise.
20520 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20521 (vectorizable_bswap): Likewise.
20522 (vect_gen_perm_mask_checked): Likewise.
20523 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20524 implementations of variable permutation vectors into account
20525 when deciding which selector to use.
20526 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20527 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20528 with a false third argument.
20529 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20530 to test whether the constant selector is valid and can_vec_perm_var_p
20531 to test whether a variable selector is valid.
20533 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20535 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20536 * optabs-query.c (can_vec_perm_p): Likewise.
20537 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20538 instead of vec_perm_indices.
20539 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20540 (vect_gen_perm_mask_checked): Likewise,
20541 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20542 (vect_gen_perm_mask_checked): Likewise,
20544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20546 * optabs-query.h (qimode_for_vec_perm): Declare.
20547 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20548 (qimode_for_vec_perm): ...this new function.
20549 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20551 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20553 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20554 does not have a conditional at the top.
20556 2018-01-02 Richard Biener <rguenther@suse.de>
20558 * ipa-inline.c (big_speedup_p): Fix expression.
20560 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20563 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20566 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20570 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20571 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20572 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20573 cond_taken_branch_cost 3->4.
20575 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20577 PR tree-optimization/83581
20578 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20579 TODO_cleanup_cfg if any changes have been made.
20581 PR middle-end/83608
20582 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20583 convert_modes if target mode has the right side, but different mode
20586 PR middle-end/83609
20587 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20588 last argument when extracting from CONCAT. If either from_real or
20589 from_imag is NULL, use expansion through memory. If result is not
20590 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20591 the parts directly to inner mode, if even that fails, use expansion
20594 PR middle-end/83623
20595 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20596 check for bswap in mode rather than HImode and use that in expand_unop
20599 Copyright (C) 2018 Free Software Foundation, Inc.
20601 Copying and distribution of this file, with or without modification,
20602 are permitted in any medium without royalty provided the copyright
20603 notice and this notice are preserved.