1 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3 * pretty-print.c (selftest::test_pp_format): Move save and restore
4 of quotes to class auto_fix_quotes, and add an instance.
5 * selftest.c: Include "intl.h".
6 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
8 * selftest.h (selftest::auto_fix_quotes): New class.
10 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
12 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13 (aarch64_sve_prepare_conditional_op): Remove.
14 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
15 Allow aarch64_simd_reg_or_zero as select operand; remove
16 the aarch64_sve_prepare_conditional_op call.
17 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
18 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
19 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
20 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
21 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
22 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
23 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
24 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
25 and a splitters to match all of the *_any patterns.
26 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
28 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
29 (SVE_COND_FP_BINARY_REV): Remove.
30 (sve_int_op_rev, sve_fp_op_rev): New.
31 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
32 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
33 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
34 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
35 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
36 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
37 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
38 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
39 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
41 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
42 Remove match_dup 1 from the inner unspec.
43 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
45 * config/aarch64/aarch64.md (movprfx): New attr.
46 (length): Default movprfx to 8.
47 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
48 (*madd<SVE_I>, *msub<SVE_I): Likewise.
49 (*<su>mul<SVE_I>3_highpart): Likewise.
50 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
51 (*v<ASHIFT><SVE_I>3): Likewise.
52 (*<su><MAXMIN><SVE_I>3): Likewise.
53 (*<su><MAXMIN><SVE_F>3): Likewise.
54 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
55 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
56 (*div<SVE_F>4): Likewise.
58 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
60 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
63 2018-07-02 Richard Biener <rguenther@suse.de>
65 PR tree-optimization/86363
66 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
67 memset argument refers to a non-variable address.
69 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
71 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
72 {PLUS,MINUS}_EXPR code to...
73 (adjust_symbolic_bound): ...here,
74 (combine_bound): ...here,
75 (set_value_range_with_overflow): ...and here.
77 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
79 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
81 (extract_range_from_abs_expr): ...here.
83 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
85 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
86 -fno-omit-frame-pointer when not optimizing.
88 2018-07-02 Martin Liska <mliska@suse.cz>
91 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
92 (propagate_nothrow): Likewise.
94 2018-07-02 Martin Liska <mliska@suse.cz>
97 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
99 2018-07-02 David Malcolm <dmalcolm@redhat.com>
101 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
102 function in r262149, changing "loc" param from source_location to
103 const dump_location_t &.
104 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
105 declaration, as above.
107 2018-07-01 Paul Koning <ni1d@arrl.net>
109 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
110 -munit-asm, -mgnu-asm, -mdec-asm.
111 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
112 (pdp11_output_labelref): New.
113 (pdp11_output_def): New.
114 (pdp11_output_addr_vec_elt): New.
115 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
116 %# and %@ format codes.
117 (pdp11_option_override): New.
118 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
119 (pdp11_output_ident): New.
120 (pdp11_asm_named_section): New.
121 (pdp11_asm_init_sections): New.
122 (pdp11_file_start): New.
123 (pdp11_file_end): New.
124 (output_ascii): Use .ascii/.asciz for -mdec-asm.
125 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
126 %o, like %c but octal.
127 (pdp11_option_override): New.
128 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
130 (DATA_SECTION_ASM_OP): Ditto.
131 (READONLY_DATA_SECTION_ASM_OP): New.
132 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
133 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
134 (ASM_OUTPUT_LABELREF): Ditto.
135 (ASM_OUTPUT_DEF): Ditto.
136 (ASM_OUTPUT_EXTERNAL): New.
137 (ASM_OUTPUT_SOURCE_FILENAME): New.
138 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
139 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
140 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
141 %# and %@ format codes.
142 * config/pdp11/pdp11.opt (mgnu-asm): New.
143 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
144 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
145 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
147 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
149 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
150 dereferencing path[] beyond its length.
152 (debug_all_paths): New.
153 (rewire_first_differing_edge): New.
154 (adjust_paths_after_duplication): New.
155 (duplicate_thread_path): Call adjust_paths_after_duplication.
157 (thread_through_all_blocks): Add new argument to
158 duplicate_thread_path.
160 2018-06-30 Jim Wilson <jimw@sifive.com>
162 * config/riscv/predicates.md (p2m1_shift_operand): New.
163 (high_mask_shift_operand): New.
164 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
165 pattern using p2m1_shift_operand.
166 (lshsi3_zero_extend_3+2): New combiner pattern using
167 high_mask_shift_operand.
169 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
171 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
173 (vect_recog_rotate_pattern): ...here.
174 (vect_convert_input): Try to insert casts of invariants in the
176 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
177 preheader to be empty.
179 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
181 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
182 vector type. If given, install it in the new statement's
184 (vect_element_precision): New function.
185 (vect_unpromoted_value): New struct.
186 (vect_unpromoted_value::vect_unpromoted_value): New function.
187 (vect_unpromoted_value::set_op): Likewise.
188 (vect_look_through_possible_promotion): Likewise.
189 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
190 (vect_widened_op_tree, vect_convert_input): Likewise.
191 (vect_convert_inputs, vect_convert_output): Likewise.
192 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
193 to handle the optional cast of the multiplication result and
194 vect_widened_op_tree to detect the widened multiplication itself.
195 Do not require the input and output of promotion casts to have
196 the same sign, but base the signedness of the operation on the
197 input rather than the result. If the pattern includes two
198 promotions, check that those promotions have the same sign.
199 Do not restrict the MULT_EXPR handling to a double-width result;
200 handle quadruple-width results and wider. Use vect_convert_inputs
201 to convert the inputs to the common type.
202 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
203 to handle the optional cast of the ABS result. Also allow a sign
204 change or a sign extension between the ABS and MINUS.
205 Use vect_widened_op_tree to detect the widened subtraction and use
206 vect_convert_inputs to convert the inputs to the common type.
207 (vect_handle_widen_op_by_const): Delete.
208 (vect_recog_widen_op_pattern): New function.
209 (vect_recog_widen_mult_pattern): Use it.
210 (vect_recog_widen_shift_pattern): Likewise.
211 (vect_recog_widen_sum_pattern): Use
212 vect_look_through_possible_promotion to handle the promoted
215 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
217 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
218 the containing gimple_seq *.
219 * gimple-iterator.h (gsi_for_stmt): Declare it.
220 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
221 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
222 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
223 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
224 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
226 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
228 (vect_mark_pattern_stmts): ...here. Handle cases in which the
229 statement being replaced is part of an existing pattern
230 definition sequence, inserting the new pattern statements before
232 (vect_pattern_recog_1): Don't return a bool. If the statement
233 is already part of a pattern, instead apply pattern matching
234 to the pattern definition statements. Don't clear the
235 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
236 (vect_pattern_recog): Don't break after the first match;
237 continue processing the pattern definition statements instead.
238 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
240 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
242 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
243 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
244 (vect_recog_widen_sum_pattern): Use it.
246 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
248 * tree-vect-loop.c (vectorizable_reduction): Assert that the
249 phi is not a pattern statement and has not been replaced by
251 * tree-vect-patterns.c (type_conversion_p): Don't check
252 STMT_VINFO_IN_PATTERN_P.
253 (vect_recog_vector_vector_shift_pattern): Likewise.
254 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
255 the pattern statement rather than the original statement; check
256 directly for a WIDEN_MULT_EXPR here.
257 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
258 vect_is_simple_use to return the pattern statement rather
259 than the original statement; use is_pattern_stmt_p to check
260 for such a pattern statement.
261 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
262 to return the pattern statement rather than the original statement;
263 don't do the same transformation here.
264 (vect_is_simple_use): If the defining statement has been replaced
265 by a pattern statement, return the pattern statement instead.
266 Remove the corresponding (local) transformation from the vectype
269 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
271 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
272 end and default to null.
273 * tree-vect-loop.c (vect_create_epilog_for_reduction)
274 (vectorizable_reduction): Update calls accordingly, dropping the
275 gimple ** argument if the passed-back statement isn't needed.
276 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
277 (vect_recog_rotate_pattern): Likewise.
278 (vect_recog_mask_conversion_pattern): Likewise.
279 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
280 (vect_mask_constant_operand_p): Likewise.
281 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
282 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
283 (get_group_load_store_type, get_load_store_type): Likewise.
284 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
285 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
286 (vectorizable_conversion, vectorizable_assignment): Likewise.
287 (vectorizable_shift, vectorizable_operation): Likewise.
288 (vectorizable_store, vect_is_simple_cond): Likewise.
289 (vectorizable_condition, vectorizable_comparison): Likewise.
290 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
291 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
292 and move it to the end. Cope with null def_stmt_outs.
294 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
296 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
298 2018-06-29 Jeff Law <law@redhat.com>
300 * config/v850/v850.c (v850_legitimate_address_p): Handle large
301 displacements for TARGET_V850E2V3 and newer.
302 (TARGET_LRA_P): Remove. Defaults to LRA now.
303 * config/v850/v850.md (sign23byte_load): Remove.
304 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
305 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
307 2018-06-29 Martin Liska <mliska@suse.cz>
310 * coverage.c (coverage_init): Mangle full path name.
311 * doc/invoke.texi: Document the change.
312 * gcov-io.c (mangle_path): New.
313 * gcov-io.h (mangle_path): Likewise.
314 * gcov.c (mangle_name): Use mangle_path for path mangling.
316 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
318 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
319 if starting source register is not even.
321 2018-06-29 Martin Liska <mliska@suse.cz>
323 PR tree-optimization/86263
324 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
325 Make edge redirection.
327 2018-06-29 David Malcolm <dmalcolm@redhat.com>
329 * dumpfile.c (dump_loc): Add indentation based on scope depth.
330 (dump_scope_depth): New variable.
331 (get_dump_scope_depth): New function.
332 (dump_begin_scope): New function.
333 (dump_end_scope): New function.
334 * dumpfile.h (get_dump_scope_depth): New declaration.
335 (dump_begin_scope): New declaration.
336 (dump_end_scope): New declaration.
337 (class auto_dump_scope): New class.
338 (AUTO_DUMP_SCOPE): New macro.
339 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
342 2018-06-29 Richard Biener <rguenther@suse.de>
344 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
345 compute_all_dependences succeeds.
346 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
347 exceed --param loop-max-datarefs-for-datadeps.
349 2018-06-29 Jakub Jelinek <jakub@redhat.com>
351 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
353 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
356 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
357 alternative 0 in preferred_for_speed attribute.
359 2018-06-28 Paul Koning <ni1d@arrl.net>
361 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
362 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
363 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
364 * config/pdp11/pdp11.md: Correct "length" attribute calculation
365 for shift insn patterns.
367 2018-06-28 David Malcolm <dmalcolm@redhat.com>
369 * cgraph.c (cgraph_node::get_body): Replace assignments to
370 "dump_file" with calls to set_dump_file.
371 * dumpfile.c (alt_dump_file): Make static, and group with...
372 (alt_flags): ...this definition.
373 (dumps_are_enabled): New variable.
374 (refresh_dumps_are_enabled): New function.
375 (set_dump_file): New function.
376 (set_alt_dump_file): New function.
377 (gcc::dump_manager::dump_start): Replace assignments to
378 "dump_file" and "alt_dump_file" with calls to set_dump_file and
380 (gcc::dump_manager::dump_finish): Likewise.
381 * dumpfile.h (alt_dump_file): Delete decl.
382 (dumps_are_enabled): New variable decl.
383 (set_dump_file): New function decl.
384 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
386 * tree-nested.c (lower_nested_functions): Replace assignments to
387 "dump_file" with calls to set_dump_file.
389 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
391 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
392 goto_locus of each outgoing edge of each basic block.
394 2018-06-28 Richard Biener <rguenther@suse.de>
396 * dwarf2out.c (decl_scope_table): Remove.
397 (push_decl_scope): Likewise.
398 (pop_decl_scope): Likewise.
399 (gen_type_die_for_member): Do not call push/pop_decl_scope.
400 (gen_struct_or_union_type_die): Likewise.
401 (gen_tagged_type_die): Likewise.
402 (dwarf2out_init): Do not initialize decl_scope_table.
403 (dwarf2out_c_finalize): Do not free it.
405 2018-06-28 Richard Biener <rguenther@suse.de>
407 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
408 deciding whether to not re-use a DIE.
410 2018-06-28 Richard Biener <rguenther@suse.de>
412 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
413 DW_AT_abstract_origin attribute.
415 2018-06-28 Martin Liska <mliska@suse.cz>
417 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
418 Use newly introduced constants.
419 * tree-switch-conversion.h (struct jump_table_cluster):
420 Define max_ratio_for_size and max_ratio_for_speed.
422 2018-06-28 Martin Liska <mliska@suse.cz>
424 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
425 Add new checking assert to catch invalid state.
426 (jump_table_cluster::can_be_handled): Handle single case
428 (jump_table_cluster::is_beneficial): Bail out for such case.
429 (bit_test_cluster::find_bit_tests):
430 Add new checking assert to catch invalid state.
431 (bit_test_cluster::can_be_handled): Handle single case
433 (bit_test_cluster::is_beneficial): Bail out for such case.
434 (switch_decision_tree::analyze_switch_statement):
437 2018-06-28 Martin Liska <mliska@suse.cz>
439 * common.opt: Introduce -completion option.
440 * gcc.c (driver_handle_option): Handle it.
441 (driver::main): Print completions if completion
443 * opt-suggestions.c (option_proposer::get_completions):
445 (option_proposer::suggest_completion): Likewise.
446 (option_proposer::find_param_completions): Likewise.
447 (verify_autocompletions): Likewise.
448 (test_completion_valid_options): Likewise.
449 (test_completion_valid_params): Likewise.
450 (in_completion_p): Likewise.
451 (empty_completion_p): Likewise.
452 (test_completion_partial_match): Likewise.
453 (test_completion_garbage): Likewise.
454 (opt_proposer_c_tests): Likewise.
455 * opt-suggestions.h: Declare new functions.
456 * opts.c (common_handle_option): Handle OPT__completion_.
457 * selftest-run-tests.c (selftest::run_tests): Add
458 opt_proposer_c_tests.
459 * selftest.c (assert_str_startswith): New.
460 * selftest.h (assert_str_startswith): Likewise.
461 (opt_proposer_c_tests): New.
462 (ASSERT_STR_STARTSWITH): Likewise.
464 2018-06-28 Martin Liska <mliska@suse.cz>
466 * Makefile.in: Add opt-suggestions.o.
467 * gcc-main.c: Include opt-suggestions.h.
468 * gcc.c (driver::driver): Likewise.
469 (driver::~driver): Remove m_option_suggestions.
470 (driver::build_option_suggestions): Moved to option_proposer.
471 (driver::suggest_option): Likewise.
472 (driver::handle_unrecognized_options): Use option_proposer.
473 * gcc.h (class driver): Add new memver m_option_proposer.
474 * opt-suggestions.c: New file.
475 * opt-suggestions.h: New file.
477 2018-06-28 Martin Liska <mliska@suse.cz>
479 * vec.h (class auto_string_vec): New (moved from auto_argvec).
480 (auto_string_vec::~auto_string_vec): Likewise.
482 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
484 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
485 prevent_decl_creation_for_types fields up and add reset_location field.
486 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
487 statement if id->reset_location is true.
488 (copy_edges_for_bb): Do not set goto_locus on the new edges if
489 id->reset_location is true.
490 (copy_phis_for_bb): Force input_location on the arguments if
491 id->reset_location is true.
492 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
493 is set on the function to be inlined.
495 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
497 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
499 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
501 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
503 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
504 hard registers for the clobbered pseudo.
506 2018-06-27 Paul Koning <ni1d@arrl.net>
508 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
509 mutually exclusive options.
510 * config/pdp11/constraints.md (h): New constraint.
511 (O): Update definition to match shift code generation.
513 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
515 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
517 (output_jump): Change arguments.
518 (pdp11_fixed_cc_regs): New function.
519 (pdp11_cc_mode): Ditto.
520 (pdp11_expand_shift): Ditto.
521 (pdp11_assemble_shift): Ditto.
522 (pdp11_small_shift): Ditto.
523 (pdp11_branch_cost): Remove.
524 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
526 (pdp11_register_move_cost): Update for CC registers.
527 (pdp11_rtx_costs): Add case for LSHIFTRT.
528 (pdp11_output_jump): Add CCNZ mode conditional branches.
529 (notice_update_cc_on_set): Remove.
530 (pdp11_cc_mode): New function.
531 (simple_memory_operand): Correct pre/post decrement case.
532 (no_side_effect_operand): New function.
533 (pdp11_regno_reg_class): Add CC_REGS class.
534 (pdp11_fixed_cc_regs): New function.
535 (pdp11_small_shift): New function.
536 (pdp11_expand_shift): New function to expand shift insns.
537 (pdp11_assemble_shift): New function to output shifts.
538 (pdp11_branch_cost): Remove.
539 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
540 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
542 (PTRDIFF_TYPE): Ditto.
543 (ADJUST_INSN_LENGTH): New macro.
544 (FIXED_REGISTERS): Add CC registers.
545 (CALL_USED_REGISTERS): Ditto.
547 (REG_CLASS_NAMES): Ditto.
548 (REG_CLASS_CONTENTS): Ditto.
549 (SELECT_CC_MODE): Use new function.
550 (TARGET_FLAGS_REGNUM): New macro.
551 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
552 (cc0_reg_rtx): Remove.
553 (CC_STATUS_MDEP): Remove.
554 (CC_STATUS_MDEFP_INIT): Remove.
556 (NOTICE_UPDATE_CC): Remove.
557 (REGISTER_NAMES): Add CC registers.
558 (BRANCH_COST): Change to constant 1.
559 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
561 * config/pdp11/pdp11.opt (mbcopy): Remove.
562 (mbcopy-builtin): Remove.
563 (mbranch-cheap): Remove.
564 (mbranch-expensive): Remove.
565 * config/pdp11/predicates.md (expand_shift_operand): Update to
566 match shift code generation.
567 (ccnz_operator): New predicate.
568 * doc/invoke.texi (PDP-11 Options): Remove deleted options
569 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
570 Remove non-existent option -mabshi, -mno-abshi. Document mutually
572 * doc/md.texi (PDP-11): Document new D and h constraints. Update
573 description of O constraint.
575 2018-06-27 Jeff Law <law@redhat.com>
576 Austin Law <austinklaw@gmail.com>
578 * config/v850/v850.md (addsi3_set_flags): New pattern.
579 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
580 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
581 (zero_extendhisi2_v850_set_flags): Likewise.
582 (zero_extendqisi2_v850_set_flags): Likewise.
583 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
584 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
585 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
587 * config/v850/v850-protos.h (notice_update_cc): Remove.
588 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
589 (v850_print_operand): Handle 'D' and "d".
590 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
591 Add handling of arithmetic/logical operations compared against zero.
592 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
593 Do not look at v850_compare_op, instead get mode from last argument.
594 (v850_gen_compare): Remove
595 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
596 after reload for prologue insns.
597 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
599 (construct_save_jarl): Likewise.
600 (TARGET_FLAGS_REGNUM): Define.
601 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
602 (NOTICE_UPDATE_CC): Remove.
603 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
604 than cc0. Conditionalize on reload_completed.
605 (cmpsi_insn, setfcc_insn): Likewise.
606 (tst1 splitter): Turn into define_and_split which sets the flags
608 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
609 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
610 (cstoresf4, cstoredf4): Clobber the flags.
611 (cmpsi, cmpsf, cmpdf): Remove expanders.
612 (setf_insn): Remove pattern.
613 (addsi3): Turn into define_and_split which clobbers the flags after
614 reload and a suitable pattern (addsi3_clobber_flags) for use after
616 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
617 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
618 (ashrsi3, ashrsi3_v850e2): Likewise.
619 (bins): Clobber the flags.
620 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
621 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
622 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
623 (call_value_internal_short, call_value_internal_long): Likewise.
624 (callt_save_interrupt, callt_return_interrupt): Likewise.
625 (save_interrupt, return_interrupt): Likewise.
626 (callt_save_all_interrupt, save_all_interrupt): Likewise.
627 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
628 (restore_all_interrupt, _restore_all_interrupt): Likewise.
629 (All FP comparisons): Only allow after reload has completed.
631 (divh, divhu): Tweak output template.
632 (branch_z_normal, branch_z_invert): Remove
633 (branch_nz_normal, branch_nz_invert): Likewise.
634 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
636 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
637 * config/v850/v850.c (notice_update_cc): Remove.
638 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
639 (CC_NO_CARRY): Likewise.
640 (NOTICE_UPDATE_CC): Define to nothing.
641 * config/v850/v850.md: Remove block comment on cc0 handling
642 Remove "cc" attribute from all patterns. Remove cc_status handling
643 from all patterns. Minor formatting fixes.
645 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
647 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
648 (cortex-a76.cortex-a55): Likewise.
649 * config/aarch64/aarch64-tune.md: Regenerate.
650 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
651 cortex-a76.cortex-a55.
653 2018-06-27 Jeff Law <law@redhat.com>
655 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
656 (MULTILIB_DIRNAMES): Similarly.
658 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
660 * gimple.h (gimple_return_retbnd): Delete.
661 (gimple_return_set_retbnd): Likewise.
662 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
663 gimple_return_set_retbnd.
664 * gimple-pretty-print.c (dump_gimple_return): Remove call to
665 gimple_return_retbnd and adjust.
666 * tree-inline.h (struct copy_body_data): Remove retbnd field.
667 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
668 Explicitly return NULL in a couple more cases. Move assertion
669 on debug statements and remove unreachable code.
670 (reset_debug_binding): Do not test id->retbnd.
671 (expand_call_inline): Do not set it.
673 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
675 * configure.ac: Add --disable-gcov option.
676 * configure: Regenerate.
677 * Makefile.in: Honour @enable_gcov@.
678 * doc/install.texi: Document --disable-gcov.
680 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
682 * config/arm/arm-cpus.in (cortex-a76): New entry.
683 (cortex-a76.cortex-a55): Likewise.
684 * config/arm/arm-tables.opt: Regenerate.
685 * config/arm/arm-tune.md: Likewise.
686 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
687 * doc/invoke.texi (ARM Options): Document cortex-a76 and
688 cortex-a76.cortex-a55.
690 2018-06-27 Tamar Christina <tamar.christina@arm.com>
693 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
695 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
697 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
699 (EPILOGUE_USES): Likewise.
701 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
703 * tree-inline.c (remap_location): New function extracted from...
704 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
705 (copy_phis_for_bb): ...here. Call remap_location.
706 (copy_cfg_body): Adjust call to copy_edges_for_bb.
708 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
710 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
711 unaligned vsx for 16B memset.
713 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
716 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
717 ieee128_float_type_node to long_double_type_node unless
718 TARGET_LONG_DOUBLE_128 is set.
720 2018-06-26 David Malcolm <dmalcolm@redhat.com>
722 * cfgloop.c (get_loop_location): Convert return type from
723 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
724 by implicit construction from rtx_insn *, and using
725 dump_user_location_t::from_function_decl for the fallback case.
726 * cfgloop.h (get_loop_location): Convert return type from
727 location_t to dump_user_location_t.
728 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
729 dump_printf_loc to pass in a dump_location_t rather than a
730 location_t, via the gimple stmt.
731 * coverage.c (get_coverage_counts): Update calls to
732 dump_printf_loc to pass in dump_location_t rather than a
734 * doc/optinfo.texi (Dump types): Convert example of
735 dump_printf_loc from taking "locus" to taking "insn". Update
736 description of the "_loc" calls to cover dump_location_t.
737 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
739 (dump_user_location_t::dump_user_location_t): New constructors,
740 from gimple *stmt and rtx_insn *.
741 (dump_user_location_t::from_function_decl): New function.
742 (dump_loc): Make static.
743 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
744 const dump_location_t &.
745 (dump_generic_expr_loc): Delete.
746 (dump_printf_loc): Convert param "loc" from location_t to
747 const dump_location_t &.
748 (selftest::test_impl_location): New function.
749 (selftest::dumpfile_c_tests): New function.
750 * dumpfile.h: Include "profile-count.h".
751 (class dump_user_location_t): New class.
752 (struct dump_impl_location_t): New struct.
753 (class dump_location_t): New class.
754 (dump_printf_loc): Convert 2nd param from source_location to
755 const dump_location_t &.
756 (dump_generic_expr_loc): Delete.
757 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
758 const dump_location_t &.
759 * gimple-fold.c (fold_gimple_assign): Update call to
760 dump_printf_loc to pass in a dump_location_t rather than a
761 location_t, via the gimple stmt.
762 (gimple_fold_call): Likewise.
763 * gimple-loop-interchange.cc
764 (loop_cand::analyze_iloop_reduction_var): Update for change to
765 check_reduction_path.
766 (tree_loop_interchange::interchange): Update for change to
768 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
769 change in return-type of find_loop_location.
770 (graphite_regenerate_ast_isl): Likewise.
771 * graphite-optimize-isl.c (optimize_isl): Likewise.
772 * graphite.c (graphite_transform_loops): Likewise.
773 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
774 pass in a dump_location_t rather than a location_t, via the
776 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
777 * ipa.c (walk_polymorphic_call_targets): Likewise.
778 * loop-unroll.c (report_unroll): Convert "locus" param from
779 location_t to dump_location_t.
780 (decide_unrolling): Update for change to get_loop_location's
782 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
783 location_t to dump_user_location_t.
784 (grid_find_single_omp_among_assignments_1): Updates calls to
785 dump_printf_loc to pass in a dump_location_t rather than a
786 location_t, via the gimple stmt.
787 (grid_parallel_clauses_gridifiable): Convert "tloc" from
788 location_t to dump_location_t. Updates calls to dump_printf_loc
789 to pass in a dump_location_t rather than a location_t, via the
791 (grid_inner_loop_gridifiable_p): Likewise.
792 (grid_dist_follows_simple_pattern): Likewise.
793 (grid_gfor_follows_tiling_pattern): Likewise.
794 (grid_target_follows_gridifiable_pattern): Likewise.
795 (grid_attempt_target_gridification): Convert initialization
796 of local "grid" from memset to zero-initialization; FIXME: does
797 this require C++11? Update call to dump_printf_loc to pass in a
798 optinfo_location rather than a location_t, via the gimple stmt.
799 * profile.c (read_profile_edge_counts): Updates call to
800 dump_printf_loc to pass in a dump_location_t rather than a
802 (compute_branch_probabilities): Likewise.
803 * selftest-run-tests.c (selftest::run_tests): Call
805 * selftest.h (dumpfile_c_tests): New decl.
806 * tree-loop-distribution.c (pass_loop_distribution::execute):
807 Update for change in return type of find_loop_location.
808 * tree-parloops.c (parallelize_loops): Likewise.
809 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
810 "locus" from location_t to dump_user_location_t.
811 (canonicalize_loop_induction_variables): Likewise.
812 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
813 for change in return type of find_loop_location.
814 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
815 to dump_printf_loc to pass in a dump_location_t rather than a
816 location_t, via the stmt.
817 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
819 * tree-vect-loop-manip.c (find_loop_location): Convert return
820 type from source_location to dump_user_location_t.
821 (vect_do_peeling): Update for above change.
822 (vect_loop_versioning): Update for change in type of
824 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
825 from location_t to dump_user_location_t.
826 (vect_estimate_min_profitable_iters): Update for change in type
828 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
829 location_t to dump_location_t.
830 (vect_slp_bb): Update for change in type of vect_location.
831 * tree-vectorizer.c (vect_location): Convert from source_location
832 to dump_user_location_t.
833 (try_vectorize_loop_1): Update for change in vect_location's type.
834 (vectorize_loops): Likewise.
835 (increase_alignment): Likewise.
836 * tree-vectorizer.h (vect_location): Convert from source_location
837 to dump_user_location_t.
838 (find_loop_location): Convert return type from source_location to
839 dump_user_location_t.
840 (check_reduction_path): Convert 1st param from location_t to
841 dump_user_location_t.
842 * value-prof.c (check_counter): Update call to dump_printf_loc to
843 pass in a dump_user_location_t rather than a location_t; update
844 call to error_at for change in type of "locus".
845 (check_ic_target): Update call to dump_printf_loc to
846 pass in a dump_user_location_t rather than a location_t, via the
849 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
851 * config/s390/s390.h (enum processor_flags): Do not use
852 default tune parameter when -march was specified.
854 2018-06-26 Jakub Jelinek <jakub@redhat.com>
857 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
858 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
861 2018-06-26 Richard Biener <rguenther@suse.de>
863 PR tree-optimization/86287
865 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
866 (vect_analyze_loop): Initialize n_stmts.
868 2018-06-26 Richard Biener <rguenther@suse.de>
871 * fold-const.c (fold_convertible_p): Pointer extension
874 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
877 * dwarf2out.c (loc_list_has_views): Adjust comments.
878 (dw_loc_list): Split single cross-partition range with
881 2018-06-25 Jeff Law <law@redhat.com>
883 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
884 on -mbig-switch by default.
886 * config/v850/predicates.md (const_float_1_operand): Fix match_code
888 (const_float_0_operand): Remove unused predicate.
889 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
890 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
891 (recipsf2): New expander. Original pattern now called
893 (recipdf2, recipdf2_insn): Similarly.
894 (rsqrtsf2, rsqrtsf2_insn): Similarly
895 (rsqrtdf2, rsqrtdf2_insn): Similarly
897 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
899 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
900 Simplify logic for FreeBSD (twice).
902 2018-06-25 Martin Sebor <msebor@redhat.com>
904 PR tree-optimization/86204
905 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
906 a strnlen result if it's less than the length of the string.
908 2018-06-25 Martin Sebor <msebor@redhat.com>
910 PR tree-optimization/85700
911 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
912 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
913 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
915 2018-06-25 Martin Sebor <msebor@redhat.com>
917 * doc/extend.texi (Zero-length arrays): Update and clarify.
919 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
921 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
922 added IEEE/IBM long double multilib support on PowerPC little
923 endian Linux systems.
924 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
925 (MULTILIB_DEFAULTS): Likewise.
926 * config/rs6000/rs6000.c (rs6000_option_override_internal):
928 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
929 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
930 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
932 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
935 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
936 (REORDER_45): Likewise.
938 2018-06-25 Jeff Law <law@redhat.com>
940 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
941 dividend to 32 bits. Adjust length.
942 (udivmodhi4): Cleanup output template. Fix length.
944 2018-06-25 Carl Love <cel@us.ibm.com>
946 * config/rs6000/vsx.md: Change word selector to prefered location.
948 2018-06-25 Richard Biener <rguenther@suse.de>
950 PR tree-optimization/86304
951 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
952 epilogue-if-converted loops as well.
954 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
956 * lto-section-out.c (lto_begin_section): Do not print section
957 name for noaddr and unnumbered dumps.
959 2018-06-25 Richard Biener <rguenther@suse.de>
961 * tree-vectorizer.h (struct vec_info_shared): New structure
962 with parts split out from struct vec_info and loop_nest from
963 struct _loop_vec_info.
964 (struct vec_info): Adjust accordingly.
965 (struct _loop_vec_info): Likewise.
966 (LOOP_VINFO_LOOP_NEST): Adjust.
967 (LOOP_VINFO_DATAREFS): Likewise.
968 (LOOP_VINFO_DDRS): Likewise.
969 (struct _bb_vec_info): Likewise.
970 (BB_VINFO_DATAREFS): Likewise.
971 (BB_VINFO_DDRS): Likewise.
972 (struct _stmt_vec_info): Add dr_aux member.
973 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
974 (DR_MISALIGNMENT_UNINITIALIZED): New.
975 (set_dr_misalignment): Adjust.
976 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
977 (vect_analyze_loop): Adjust prototype.
978 (vect_analyze_loop_form): Likewise.
979 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
980 Compute dependences lazily.
981 (vect_record_base_alignments): Use shared datarefs/ddrs.
982 (vect_verify_datarefs_alignment): Likewise.
983 (vect_analyze_data_refs_alignment): Likewise.
984 (vect_analyze_data_ref_accesses): Likewise.
985 (vect_analyze_data_refs): Likewise.
986 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
987 constructor parameter for shared part.
988 (vect_analyze_loop_form): Pass in shared part and adjust.
989 (vect_analyze_loop_2): Pass in storage for the number of
990 stmts. Move loop nest finding to the caller. Compute
992 (vect_analyze_loop): Pass in shared part.
993 (vect_transform_loop): Verify shared datarefs are unchanged.
994 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
995 constructor parameter for shared part.
996 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
997 (vect_slp_bb): Verify shared datarefs are unchanged before
999 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1001 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1002 DR_MISALIGNMENT_UNINITIALIZED.
1003 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1004 parameter for shared part.
1005 (vec_info::~vec_info): Adjust.
1006 (vec_info_shared::vec_info_shared): New.
1007 (vec_info_shared::~vec_info_shared): Likewise.
1008 (vec_info_shared::save_datarefs): Likewise.
1009 (vec_info_shared::check_datarefs): Likewise.
1010 (try_vectorize_loop_1): Construct shared part live for analyses
1011 of a single loop for multiple vector sizes.
1012 * tree-parloops.c (gather_scalar_reductions): Adjust.
1014 2018-06-25 Richard Biener <rguenther@suse.de>
1016 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1017 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1018 (vect_analyze_data_refs): Remove similar code from here and
1019 simplify accordingly.
1021 2018-06-25 Richard Biener <rguenther@suse.de>
1023 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1024 for reverse storage order accesses rather than asserting
1025 they cannot happen here.
1027 2018-06-25 Tom de Vries <tdevries@suse.de>
1030 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1031 Use data16 instead of .byte for insn prefix.
1033 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1036 * parser.c (make_char_string_pack): Pass this literal chars
1037 through cpp_interpret_string.
1038 (cp_parser_userdef_numeric_literal): Check the result of
1039 make_char_string_pack.
1041 2018-06-24 Maya Rashish <coypu@sdf.org>
1043 * ginclude/stddef.h: Simplify conditions around avoiding
1044 re-definition of __size_t.
1046 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1048 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1049 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1051 2018-06-22 Maya Rashish <coypu@sdf.org>
1053 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1054 made to OpenBSD and NetBSD through the years.
1056 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1058 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1059 behavior of vec_pack (vector double, vector double) to match
1060 behavior of vec_float2 (vector double, vector double).
1062 2018-06-22 Olivier Hainque <hainque@adacore.com>
1064 * gimplify.c (gimplify_function_tree): Prevent creation
1065 of a trampoline for the address of the current function
1066 passed to entry/exit instrumentation hooks.
1068 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1071 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1074 2018-06-22 Martin Liska <mliska@suse.cz>
1076 PR tree-optimization/86263
1077 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1078 Bail out if is_enabled is false.
1079 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1081 (jump_table_cluster::is_enabled): New function.
1083 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1085 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1086 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1087 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1088 (lto_input_ts_binfo_tree_pointers): Likewise.
1089 * tree-streamer-out.c (streamer_write_tree_bitfields,
1090 write_ts_binfo_tree_pointers): Likewise.
1091 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1093 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1095 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1097 2018-06-22 Martin Liska <mliska@suse.cz>
1099 * symbol-summary.h (get): Make it pure and inline move
1100 functionality from ::get function.
1101 (get): Remove and inline into ::get and ::get_create.
1102 (get_create): Move code from ::get function.
1104 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1107 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1108 -x assembler-with-cpp.
1110 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1112 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1113 _FILE_OFFSET_BITS=64 for C++.
1115 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1117 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1118 conversion insn that shows up when pr85657-3.c is compiled using
1119 IEEE 128-bit long double.
1120 (neg<mode>2_internal): Use the correct mode to check whether the
1121 mode is IBM extended.
1122 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1123 multiply and divide external functions from being created more
1126 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1128 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1130 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1131 the edge can be forwarded.
1132 (cfg_layout_merge_blocks): Likewise.
1134 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1136 * except.c (finish_eh_generation): Commit edge insertions only after
1137 the EH edges have been redirected from post-landing to landing pads.
1139 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1141 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1142 create_tmp_var_for to create the FRAME decl.
1143 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1145 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1147 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1148 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1149 debug statement when resetting its value.
1150 (expand_call_inline): Copy the locus of the call onto the assignment
1151 of the return value, if any. Use local variable in more cases.
1153 2018-06-21 Martin Liska <mliska@suse.cz>
1155 * ipa-pure-const.c (propagate_nothrow): Use
1156 funct_state_summaries->get.
1157 (dump_malloc_lattice): Likewise.
1158 (propagate_malloc): Likewise.
1160 2018-06-21 Richard Biener <rguenther@suse.de>
1162 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1163 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1164 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1166 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1167 BLOCK_ABSTRACT_ORIGIN unconditionally.
1169 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1171 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1173 * ipa-reference.c (ipa_reference_c_finalize): Delete
1174 ipa_ref_opt_sum_summaries and set it to NULL.
1176 2018-06-21 Tom de Vries <tdevries@suse.de>
1178 PR tree-optimization/85859
1179 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1180 test with comment from bb_no_side_effects_p.
1182 2018-06-21 Richard Biener <rguenther@suse.de>
1184 PR tree-optimization/86232
1185 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1186 max for constant niter.
1188 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1190 * config/aarch64/aarch64-simd.md
1191 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1193 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1195 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1196 Make opernads of the unspec commutative.
1198 2018-06-21 Richard Biener <rguenther@suse.de>
1200 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1201 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1202 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1203 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1205 (vect_record_base_alignments): Do not record base alignment
1206 for gathers or scatters.
1207 (vect_compute_data_ref_alignment): Drop return value that is always
1208 true. Bail out early for gathers or scatters.
1209 (vect_enhance_data_refs_alignment): Bail out early for gathers
1211 (vect_find_same_alignment_drs): Likewise.
1212 (vect_analyze_data_refs_alignment): Remove dead code.
1213 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1214 (vect_analyze_data_refs): For possible gathers or scatters do
1215 not create an alternate DR, just check their possible validity
1216 and mark them. Adjust DECL_NONALIASED handling to not rely
1218 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1219 update inits of gathers or scatters.
1220 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1221 Also copy gather/scatter flag to pattern vinfo.
1223 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1225 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1226 behavior of vec_packsu (vector unsigned long long, vector unsigned
1227 long long) to match behavior of vec_packs with same signature.
1229 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1230 Thomas Schwinge <thomas@codesourcery.com>
1231 Cesar Philippidis <cesar@codesourcery.com>
1233 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1234 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1235 (gimplify_adjust_omp_clauses): Likewise.
1236 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1237 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1238 (gimplify_omp_target_update): Update handling of acc update and
1240 * omp-low.c (install_var_field): Remove unused parameter
1241 base_pointers_restrict.
1242 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1243 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1245 (omp_target_base_pointers_restrict_p): Delete.
1246 (scan_omp_target): Update call to scan_sharing_clauses.
1247 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1249 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1250 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1251 (convert_local_omp_clauses): Likewise.
1252 * tree-pretty-print.c (dump_omp_clause): Likewise.
1253 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1255 (omp_clause_code_name): Likewise.
1257 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1260 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1263 PR tree-optimization/86231
1264 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1265 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1267 2018-06-20 Tom de Vries <tdevries@suse.de>
1269 PR tree-optimization/86097
1270 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1271 iv type if signedness of iv type is not the same as that of *nit.
1273 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1275 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1276 EDGE_EH edges, verify they are all EDGE_EH.
1278 2018-06-20 Maya Rashish <coypu@sdf.org>
1280 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1282 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1284 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1285 * config/aarch64/aarch64.c (xgene1_tunings): Add
1286 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1287 (aarch64_mode_valid_for_sched_fusion_p):
1288 Allow 16-byte modes.
1289 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1290 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1292 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1294 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1295 * config/aarch64/iterators.md (VQ2): New mode iterator.
1297 2018-06-20 Martin Liska <mliska@suse.cz>
1299 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1300 Change default ratio from 10 to 8.
1302 2018-06-20 Martin Liska <mliska@suse.cz>
1304 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1306 (bit_test_cluster::find_bit_tests): Likewise.
1307 (switch_decision_tree::analyze_switch_statement): Find clusters.
1308 * tree-switch-conversion.h (struct jump_table_cluster): Document
1311 2018-06-20 Martin Liska <mliska@suse.cz>
1313 * tree-switch-conversion.c (switch_conversion::collect):
1314 Record m_uniq property.
1315 (switch_conversion::expand): Bail out for special conditions.
1316 (group_cluster::~group_cluster): New.
1317 (group_cluster::group_cluster): Likewise.
1318 (group_cluster::dump): Likewise.
1319 (jump_table_cluster::emit): New.
1320 (switch_decision_tree::fix_phi_operands_for_edges): New.
1321 (struct case_node): Remove struct.
1322 (jump_table_cluster::can_be_handled): New.
1323 (case_values_threshold): Moved to header.
1324 (reset_out_edges_aux): Likewise.
1325 (jump_table_cluster::is_beneficial): New.
1326 (bit_test_cluster::can_be_handled): Likewise.
1327 (add_case_node): Remove.
1328 (bit_test_cluster::is_beneficial): New.
1329 (case_bit_test::cmp): New.
1330 (bit_test_cluster::emit): New.
1331 (expand_switch_as_decision_tree_p): Remove.
1332 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1333 (fix_phi_operands_for_edge): Likewise.
1334 (switch_decision_tree::analyze_switch_statement): New.
1335 (compute_cases_per_edge): Move ...
1336 (switch_decision_tree::compute_cases_per_edge): ... here.
1337 (try_switch_expansion): Likewise.
1338 (switch_decision_tree::try_switch_expansion): Likewise.
1339 (record_phi_operand_mapping): Likewise.
1340 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1341 (emit_case_decision_tree): Likewise.
1342 (switch_decision_tree::emit): Likewise.
1343 (balance_case_nodes): Likewise.
1344 (switch_decision_tree::balance_case_nodes): Likewise.
1345 (dump_case_nodes): Likewise.
1346 (switch_decision_tree::dump_case_nodes): Likewise.
1347 (emit_jump): Likewise.
1348 (switch_decision_tree::emit_jump): Likewise.
1349 (emit_cmp_and_jump_insns): Likewise.
1350 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1351 (emit_case_nodes): Likewise.
1352 (switch_decision_tree::emit_case_nodes): Likewise.
1353 (conditional_probability): Remove.
1354 * tree-switch-conversion.h (enum cluster_type): New.
1356 (struct cluster): Likewise.
1357 (cluster::cluster): Likewise.
1358 (struct simple_cluster): Likewise.
1359 (simple_cluster::simple_cluster): Likewise.
1360 (struct group_cluster): Likewise.
1361 (struct jump_table_cluster): Likewise.
1362 (struct bit_test_cluster): Likewise.
1363 (struct min_cluster_item): Likewise.
1364 (struct case_tree_node): Likewise.
1365 (case_tree_node::case_tree_node): Likewise.
1366 (jump_table_cluster::case_values_threshold): Likewise.
1367 (struct case_bit_test): Likewise.
1368 (struct switch_decision_tree): Likewise.
1369 (struct switch_conversion): Likewise.
1370 (switch_decision_tree::reset_out_edges_aux): Likewise.
1372 2018-06-20 Martin Liska <mliska@suse.cz>
1374 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1375 (hoist_edge_and_branch_if_true): Likewise.
1376 (expand_switch_using_bit_tests_p): Likewise.
1377 (struct case_bit_test): Likewise.
1378 (case_bit_test_cmp): Likewise.
1379 (emit_case_bit_tests): Likewise.
1380 (switch_conversion::switch_conversion): New class.
1381 (struct switch_conv_info): Remove old struct.
1382 (collect_switch_conv_info): More to ...
1383 (switch_conversion::collect): ... this.
1384 (check_range): Likewise.
1385 (switch_conversion::check_range): Likewise.
1386 (check_all_empty_except_final): Likewise.
1387 (switch_conversion::check_all_empty_except_final): Likewise.
1388 (check_final_bb): Likewise.
1389 (switch_conversion::check_final_bb): Likewise.
1390 (create_temp_arrays): Likewise.
1391 (switch_conversion::create_temp_arrays): Likewise.
1392 (free_temp_arrays): Likewise.
1393 (gather_default_values): Likewise.
1394 (switch_conversion::gather_default_values): Likewise.
1395 (build_constructors): Likewise.
1396 (switch_conversion::build_constructors): Likewise.
1397 (constructor_contains_same_values_p): Likewise.
1398 (switch_conversion::contains_same_values_p): Likewise.
1399 (array_value_type): Likewise.
1400 (switch_conversion::array_value_type): Likewise.
1401 (build_one_array): Likewise.
1402 (switch_conversion::build_one_array): Likewise.
1403 (build_arrays): Likewise.
1404 (switch_conversion::build_arrays): Likewise.
1405 (gen_def_assigns): Likewise.
1406 (switch_conversion::gen_def_assigns): Likewise.
1407 (prune_bbs): Likewise.
1408 (switch_conversion::prune_bbs): Likewise.
1409 (fix_phi_nodes): Likewise.
1410 (switch_conversion::fix_phi_nodes): Likewise.
1411 (gen_inbound_check): Likewise.
1412 (switch_conversion::gen_inbound_check): Likewise.
1413 (process_switch): Use the newly created class.
1414 (switch_conversion::expand): New.
1415 (switch_conversion::~switch_conversion): New.
1416 * tree-switch-conversion.h: New file.
1418 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1420 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1421 tree-vect-patterns.c.
1422 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1423 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1424 (vect_recog_sad_pattern): Likewise.
1425 (vect_recog_widen_sum_pattern): Likewise.
1426 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1427 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1428 (vect_recog_rotate_pattern): Likewise.
1429 (vect_recog_mult_pattern): Likewise.
1430 (vect_recog_vector_vector_shift_pattern): Likewise.
1431 (vect_recog_divmod_pattern): Likewise.
1432 (vect_recog_mixed_size_cond_pattern): Likewise.
1433 (vect_recog_bool_pattern): Likewise.
1434 (vect_recog_mask_conversion_pattern): Likewise.
1435 (vect_try_gather_scatter_pattern): Likewise.
1436 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1437 (vect_recog_over_widening_pattern): Likewise.
1438 (vect_recog_gather_scatter_pattern): Likewise.
1439 (vect_recog_func_ptr): Move from tree-vectorizer.h
1440 (vect_vect_recog_func_ptrs): Move further down the file.
1441 (vect_recog_func): Likewise. Remove the third argument.
1442 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1443 (vect_pattern_recog_1): Expect the pattern function to do any
1444 necessary target tests. Also expect it to provide a vector type.
1445 Remove the type_in handling.
1447 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1449 * tree-vect-patterns.c (vect_pattern_detected): New function.
1450 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1451 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1452 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1453 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1454 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1455 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1456 (vect_recog_mask_conversion_pattern)
1457 (vect_try_gather_scatter_pattern): Likewise.
1459 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1461 * tree-vect-patterns.c (vect_get_internal_def): New function.
1462 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1463 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1464 (search_type_for_mask_1): Use it.
1466 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1468 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1469 redundant WIDEN_SUM_EXPR handling.
1470 (vect_recog_sad_pattern): Likewise.
1472 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1474 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1475 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1476 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1477 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1478 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1480 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1482 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1483 use the stmt_vec_info of the original bb statement for the
1484 new zero assignment, even if the call is part of a pattern.
1486 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1488 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1489 that the sequence is attached to the original statement rather
1490 than the pattern statement.
1491 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1492 PATTERN_DEF_SEQ from the original statement rather than
1493 the main pattern statement.
1494 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1495 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1496 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1498 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1500 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1501 definition statements before the early exit for statements that aren't
1503 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1505 (vect_transform_loop): ...here. Process pattern definition
1506 statements without first checking whether the main pattern
1507 statement is live or relevant.
1509 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1511 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1512 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1514 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1516 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1517 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1518 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1519 (expand_block_compare): Change select_block_compare_mode call.
1520 (expand_strncmp_align_check): Use new functions, fix comment.
1521 (emit_final_str_compare_gpr): New function.
1522 (expand_strn_compare): Refactor and clean up code.
1523 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1525 2018-06-19 Tony Reix <tony.reix@atos.com>
1526 Damien Bergamini <damien.bergamini@atos.com>
1527 David Edelsohn <dje.gcc@gmail.com>
1529 * collect2.c (static_obj): New variable.
1530 (static_libs): New variable.
1531 (is_in_list): Uncomment declaration.
1532 (main): Track AIX libraries linked statically.
1533 (is_in_list): Uncomment definition.
1534 (scan_prog_file): Don't add AIX shared libraries initializer
1535 to constructor list if linking statically.
1537 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1539 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1541 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1543 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1545 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1548 2018-06-19 Martin Liska <mliska@suse.cz>
1550 * config/i386/i386.c (ix86_can_inline_p): Do not use
1551 ipa_fn_summaries::get_create.
1552 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1554 (devirtualization_time_bonus): Likewise.
1555 (ipcp_propagate_stage): Likewise.
1556 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1557 (edge_set_predicate): Likewise.
1558 (evaluate_conditions_for_known_args): Likewise.
1559 (evaluate_properties_for_edge): Likewise.
1560 (ipa_call_summary::reset): Tranform to ...
1561 (ipa_call_summary::~ipa_call_summary): ... this.
1562 (ipa_fn_summary::reset): Transform to ...
1563 (ipa_fn_summary::~ipa_fn_summary): ... this.
1564 (ipa_fn_summary_t::remove): Rename to ...
1565 (ipa_fn_summary_t::remove_callees): ... this.
1566 (ipa_fn_summary_t::duplicate): Use placement new
1567 instead of memory copy.
1568 (ipa_call_summary_t::duplicate): Likewise.
1569 (ipa_call_summary_t::remove): Remove.
1570 (dump_ipa_call_summary): Change get_create to get.
1571 (ipa_dump_fn_summary): Dump only when summary exists.
1572 (analyze_function_body): Use symbol_summary::get instead
1574 (compute_fn_summary): Likewise.
1575 (estimate_edge_devirt_benefit): Likewise.
1576 (estimate_edge_size_and_time): Likewise.
1577 (inline_update_callee_summaries): Likewise.
1578 (remap_edge_change_prob): Likewise.
1579 (remap_edge_summaries): Likewise.
1580 (ipa_merge_fn_summary_after_inlining): Likewise.
1581 (write_ipa_call_summary): Likewise.
1582 (ipa_fn_summary_write): Likewise.
1583 (ipa_free_fn_summary): Likewise.
1584 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1585 (struct ipa_call_summary): Likewise.
1586 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1588 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1589 (estimate_size_after_inlining): Likewise.
1590 (estimate_growth): Likewise.
1591 (growth_likely_positive): Likewise.
1592 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1593 (inline_call): Likewise.
1594 * ipa-inline.c (caller_growth_limits): Likewise.
1595 (can_inline_edge_p): Likewise.
1596 (can_inline_edge_by_limits_p): Likewise.
1597 (compute_uninlined_call_time): Likewise.
1598 (compute_inlined_call_time): Likewise.
1599 (want_inline_small_function_p): Likewise.
1600 (edge_badness): Likewise.
1601 (update_caller_keys): Likewise.
1602 (update_callee_keys): Likewise.
1603 (inline_small_functions): Likewise.
1604 (inline_to_all_callers_1): Likewise.
1605 (dump_overall_stats): Likewise.
1606 (early_inline_small_functions): Likewise.
1607 (early_inliner): Likewise.
1608 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1609 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1610 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1611 * ipa-split.c (execute_split_functions): Likewise.
1612 * symbol-summary.h: Likewise.
1613 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1615 2018-06-19 Richard Biener <rguenther@suse.de>
1617 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1618 (vectorize_loops): ... here. Fix dbgcnt handling.
1619 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1621 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1624 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1625 ieee128 argument takes up only one (vector) register, not two (floating
1628 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1630 * gimplify.c (gimplify_init_constructor): Really never clear for an
1631 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1633 2018-06-19 Richard Biener <rguenther@suse.de>
1635 PR tree-optimization/86179
1636 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1637 after failed recognition.
1639 2018-06-18 Martin Sebor <msebor@redhat.com>
1642 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1643 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1644 Handle integer subtraction.
1645 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1646 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1648 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1650 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1651 param from rtx to rtx_insn *.
1652 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1654 (frv_ifcvt_modify_insn): Likwise.
1655 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1656 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1657 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1658 as_a <rtx_insn *> cast to local "unprotected_region" once
1659 it's been established that it's not NULL or pc_rtx.
1660 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1661 param "sethi" from rtx to rtx_insn *.
1662 (nds32_group_float_insns): Likewise for param "insn".
1663 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1665 (vax_output_int_subtract): Likewise.
1666 * config/vax/vax.c (vax_output_int_add): Likewise for param
1668 (vax_output_int_subtract): Likewise.
1669 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1670 (emit_pattern_after): Likewise for param "after".
1671 (emit_insn_after): Likewise.
1672 (emit_jump_insn_after): Likewise.
1673 (emit_call_insn_after): Likewise.
1674 (emit_debug_insn_after): Likewise.
1675 (emit_pattern_before): Likewise for param "before".
1676 (emit_insn_before): Likewise.
1677 (emit_jump_insn_before): Likewise.
1678 * final.c (get_insn_template): Likewise for param "insn", removing
1680 * output.h (get_insn_template): Likewise for 2nd param.
1681 * rtl.h (emit_insn_before): Likewise.
1682 (emit_jump_insn_before): Likewise.
1683 (emit_debug_insn_before_noloc): Likewise.
1684 (emit_insn_after): Likewise.
1685 (emit_jump_insn_after): Likewise.
1686 (emit_call_insn_after): Likewise.
1687 (emit_debug_insn_after): Likewise.
1688 (set_insn_deleted): Likewise for param.
1690 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1693 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1694 floating point modes, so that IFmode is numerically greater than
1695 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1696 to declare the ordering. This prevents IFmode from being
1697 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1698 machine. Include rs6000-modes.h to share the fractional values
1699 between genmodes* and the rest of the compiler.
1703 * config/rs6000/rs6000-modes.h: New file.
1704 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1705 meaning of rs6000_long_double_size so that 126..128 selects an
1706 appropriate 128-bit floating point type.
1707 (rs6000_option_override_internal): Likewise.
1708 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1709 (TARGET_LONG_DOUBLE_128): Change the meaning of
1710 rs6000_long_double_size so that 126..128 selects an appropriate
1711 128-bit floating point type.
1712 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1713 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1714 source and destination to match the standard usage.
1715 (truncifkf2): Likewise.
1716 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1717 ISA 2.07 to use an explicit clobber, instead of passing in a
1719 (copysign<mode>3_soft): Likewise.
1721 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1723 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1724 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1725 (vect_slp_analyze_instance_dependence): Likewise.
1726 (vect_enhance_data_refs_alignment): Likewise.
1727 (vect_analyze_data_refs_alignment): Likewise.
1728 (vect_slp_analyze_and_verify_instance_alignment
1729 (vect_analyze_data_ref_accesses): Likewise.
1730 (vect_prune_runtime_alias_test_list): Likewise.
1731 (vect_analyze_data_refs): Likewise.
1732 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1733 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1734 (vect_analyze_scalar_cycles_1): Likewise.
1735 (vect_get_loop_niters): Likewise.
1736 (vect_analyze_loop_form_1): Likewise.
1737 (vect_update_vf_for_slp): Likewise.
1738 (vect_analyze_loop_operations): Likewise.
1739 (vect_analyze_loop): Likewise.
1740 (vectorizable_induction): Likewise.
1741 (vect_transform_loop): Likewise.
1742 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1743 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1744 (vect_make_slp_decision): Likewise.
1745 (vect_detect_hybrid_slp): Likewise.
1746 (vect_slp_analyze_operations): Likewise.
1747 (vect_slp_bb): Likewise.
1748 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1749 (vectorizable_bswap): Likewise.
1750 (vectorizable_call): Likewise.
1751 (vectorizable_simd_clone_call): Likewise.
1752 (vectorizable_conversion): Likewise.
1753 (vectorizable_assignment): Likewise.
1754 (vectorizable_shift): Likewise.
1755 (vectorizable_operation): Likewise.
1756 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1758 2018-06-18 Martin Sebor <msebor@redhat.com>
1760 PR tree-optimization/81384
1761 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1762 * builtins.c (expand_builtin_strnlen): New function.
1763 (expand_builtin): Call it.
1764 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1765 * builtins.def (BUILT_IN_STRNLEN): New.
1766 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1767 Warn for bounds in excess of maximum object size.
1768 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1769 single-value ranges. Handle strnlen.
1770 (handle_builtin_strlen): Handle strnlen.
1771 (strlen_check_and_optimize_stmt): Same.
1772 * doc/extend.texi (Other Builtins): Document strnlen.
1774 2018-06-18 Maya Rashish <coypu@sdf.org>
1776 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1777 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1778 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1780 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1782 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1784 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1786 * tree.c (escaped_string::escape): Replace cast to char * by
1787 const_cast<char *> (unescaped).
1789 2018-06-18 Nick Clifton <nickc@redhat.com>
1792 * tree.c (escaped_string): New class. Converts an unescaped
1793 string into its escaped equivalent.
1794 (warn_deprecated_use): Use the new class to convert the
1795 deprecation message, if present.
1796 (test_escaped_strings): New self test.
1797 (test_c_tests): Add test_escaped_strings.
1798 * doc/extend.texi (deprecated): Add a note that the
1799 deprecation message is affected by the -fmessage-length
1800 option, and that control characters will be escaped.
1801 (#pragma GCC error): Document this pragma.
1802 (#pragma GCC warning): Likewise.
1803 * doc/invoke.texi (-fmessage-length): Document this option's
1804 effect on the #warning and #error preprocessor directives and
1805 the deprecated attribute.
1807 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1809 * tree.c (decl_value_expr_lookup): Revert latest change.
1810 (decl_value_expr_insert): Likewise.
1812 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1814 * gimplify.c (nonlocal_vlas): Delete.
1815 (nonlocal_vla_vars): Likewise.
1816 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1818 (gimplify_body): Do not create and destroy nonlocal_vlas.
1819 * tree-nested.c: Include diagnostic.h.
1820 (use_pointer_in_frame): Tweak.
1821 (lookup_field_for_decl): Add assertion and declare the transformation.
1822 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1823 internal error when the reference is in a wrong context. Do not
1824 create a debug decl by default.
1825 (note_nonlocal_block_vlas): Delete.
1826 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1827 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1828 create a debug decl by default.
1829 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1830 call is in a wrong context.
1831 (fixup_vla_decls): New function.
1832 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1833 debug variables were created.
1834 * tree.c (decl_value_expr_lookup): Add checking assertion.
1835 (decl_value_expr_insert): Likewise.
1837 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1840 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1841 * tree-scalar-evolution.c (interpret_expr): Likewise.
1842 (expression_expensive_p): Likewise.
1843 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1844 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1845 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1846 (ssa_defined_by_minus_one_stmt_p): New.
1848 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1851 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1852 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1853 * dojump.c (do_jump): Likewise.
1854 * expr.c (expand_expr_real_2): Check operand type's sign.
1855 * fold-const.c (const_unop): Handle ABSU_EXPR.
1856 (fold_abs_const): Likewise.
1857 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1858 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1859 (strip_sign_op_1): Likesise.
1860 * match.pd: Add new pattern to generate ABSU_EXPR.
1861 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1862 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1863 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1864 * tree-inline.c (estimate_operator_cost): Likewise.
1865 * tree-pretty-print.c (dump_generic_node): Likewise.
1866 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1867 * tree.def (ABSU_EXPR): New.
1869 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1872 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1873 documented as preserved for backward compatibility only.
1874 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1876 PR rtl-optimization/86108
1877 * bb-reorder.c (create_forwarder_block): Renamed to ...
1878 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1879 jump from new landing pad to the second part.
1880 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1883 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1886 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1887 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1888 Only call store_expr for halves if the mode is the same.
1891 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1892 Fix up comment formatting.
1894 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1896 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1897 (typed_splay_tree::closure,
1898 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1899 (typed_splay_tree::typed_splay_tree,
1900 typed_splay_tree::operator =): Declared private.
1901 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1902 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1903 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1904 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1905 typed_splay_tree::splay_tree_splay,
1906 typed_splay_tree::splay_tree_foreach_helper,
1907 typed_splay_tree::splay_tree_insert,
1908 typed_splay_tree::splay_tree_remove,
1909 typed_splay_tree::splay_tree_lookup,
1910 typed_splay_tree::splay_tree_predecessor,
1911 typed_splay_tree::splay_tree_successor,
1912 typed_splay_tree::splay_tree_min,
1913 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1914 (typed_splay_tree::root, typed_splay_tree::comp,
1915 typed_splay_tree::delete_key,
1916 typed_splay_tree::delete_value): New data members.
1917 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1918 typed_splay_tree::remove.
1920 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1922 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1923 -mginv and -mno-ginv to the assembler.
1924 * config/mips/mips.opt (-mcrc): New option.
1926 * doc/invoke.text (-mcrc): Document.
1929 2018-06-15 Nick Clifton <nickc@redhat.com>
1932 * tree.c (escaped_string): New class. Converts an unescaped
1933 string into its escaped equivalent.
1934 (warn_deprecated_use): Use the new class to convert the
1935 deprecation message, if present.
1936 (test_escaped_strings): New self test.
1937 (test_c_tests): Add test_escaped_strings.
1938 * doc/extend.texi (deprecated): Add a note that the
1939 deprecation message is affected by the -fmessage-length
1940 option, and that control characters will be escaped.
1941 (#pragma GCC error): Document this pragma.
1942 (#pragma GCC warning): Likewise.
1943 * doc/invoke.texi (-fmessage-length): Document this option's
1944 effect on the #warning and #error preprocessor directives and
1945 the deprecated attribute.
1947 2018-06-15 Richard Biener <rguenther@suse.de>
1949 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1950 here, also noting vector size used.
1951 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1952 size used in MSG_OPTIMIZED_LOCATIONS dump.
1953 (pass_slp_vectorize::execute): Adjust.
1955 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1958 * config/arc/arc.c (arc_return_address_register): Fix
1961 2018-06-15 Richard Biener <rguenther@suse.de>
1964 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1965 leave useless conversion stripping to force_gimple_operand_gsi.
1966 (gimplify_build2): Likewise.
1967 (gimplify_build1): Likewise.
1969 2018-06-15 Richard Biener <rguenther@suse.de>
1972 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1973 before adjusting their block.
1975 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1977 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1978 multilibs for *-*-rtems*.
1979 * config/riscv/t-rtems: New file.
1981 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1984 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1985 unsigned_type_for returns NULL.
1988 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1989 subregs of multi-word pseudos unless the float mode has word size.
1991 2018-06-14 Richard Biener <rguenther@suse.de>
1994 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1995 duplicate and harmful type_hash_canon.
1996 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1998 2018-06-14 Richard Biener <rguenther@suse.de>
2001 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2004 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2006 * config/rtems.h (STDINT_LONG32): Define.
2008 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2009 Prachi Godbole <prachi.godbole@imgtec.com>
2011 * config/mips/mips-cpus.def: Define P6600.
2012 * config/mips/mips-tables.opt: Regenerate.
2013 * config/mips/mips.c (mips_ucbranch_type): New enum.
2014 (mips_rtx_cost_data): Add support for P6600.
2015 (mips_issue_rate): Likewise.
2016 (mips_multipass_dfa_lookahead): Likewise.
2017 (mips_avoid_hazard): Likewise.
2018 (mips_reorg_process_insns): Likewise.
2019 (mips_classify_branch_p6600): New function.
2020 * config/mips/mips.h (TUNE_P6600): New define.
2021 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2022 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2023 * config/mips/mips.md: Include p6600.md.
2024 (processor): Add p6600.
2025 * config/mips/p6600.md: New file.
2026 * doc/invoke.texi: Add p6600 to supported architectures.
2028 2018-06-13 Martin Sebor <msebor@redhat.com>
2030 PR tree-optimization/86114
2031 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2033 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2035 2018-06-13 Richard Biener <rguenther@suse.de>
2037 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2038 Properly set vector type of the intermediate stmt.
2039 * tree-vect-stmts.c (vectorizable_operation): The destination
2040 var always has vectype_out type.
2042 2018-06-13 Jeff Law <law@redhat.com>
2044 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2045 integer 0 for argument to print_rtl_with_bb.
2046 (rl78_reorg): Likewise.
2048 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2050 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2051 from rtx to rtx_insn *.
2052 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2054 (add_sched_insns_for_speculation): Likewise for local "target",
2055 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2056 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2057 from rtx_insn ** to rtx_code_label **.
2058 (reorg_emit_nops): Likewise.
2059 (c6x_reorg): Likewise for local "call_labels".
2060 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2062 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2063 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2064 the loops over LABEL_REFS.
2065 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2067 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2068 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2069 (split_branches): Strengthen local "olabel" from rtx to
2070 rtx_insn *, adding a safe_as_a cast.
2071 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2073 (add_insn_after): Likewise for first two params.
2074 (add_insn_before): Likewise.
2075 (remove_insn): Likewise for param.
2076 (emit_pattern_before_noloc): Likewise for second and third params.
2077 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2078 (emit_call_insn_before_noloc): Likewise.
2079 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2081 (emit_barrier_before): Likewise.
2082 (emit_label_before): Strengthen "label" param from "rtx" to
2083 "rtx_code_label *". Strengthen "before" param from "rtx" to
2085 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2087 (emit_pattern_after_noloc): Likewise.
2088 (emit_insn_after_noloc): Likewise.
2089 (emit_jump_insn_after_noloc): Likewise.
2090 (emit_call_insn_after_noloc): Likewise.
2091 (emit_debug_insn_after_noloc): Likewise.
2092 (emit_barrier_after): Likewise.
2093 (emit_label_after): Likewise for both params.
2094 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2095 "loc" param from "int" to "location_t".
2096 (emit_insn_after_setloc): Likewise.
2097 (emit_jump_insn_after_setloc): Likewise.
2098 (emit_call_insn_after_setloc): Likewise.
2099 (emit_debug_insn_after_setloc): Likewise.
2100 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2101 "loc" param from "int" to "location_t".
2102 (emit_pattern_before): Convert NULL_RTX to NULL.
2103 (emit_insn_before_setloc): Convert "loc" param from "int" to
2105 (emit_jump_insn_before_setloc): Likewise.
2106 (emit_call_insn_before_setloc): Likewise.
2107 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2108 rtx_insn *. Convert "loc" param from "int" to "location_t".
2109 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2110 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2111 Convert 3rd param from "int" to "location_t".
2112 (emit_barrier_before, emit_barrier_after, next_real_insn):
2113 Strengthen param from rtx to rtx_insn *.
2114 (emit_label_before): Strengthen 1st param from "rtx" to
2115 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2117 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2118 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2119 Strengthen 2nd param from "rtx" to "rtx_insn *".
2120 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2121 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2122 Likewise. Convert 3rd param from "int" to "location_t".
2123 (emit_label_after): Strengthen 1st param from "rtx" to
2125 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2127 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2128 from "rtx" to "rtx_insn *".
2130 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2132 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2133 bodies streamed in with -Q.
2134 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2135 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2136 * gimple-streamer-out.c: Include gimple-pretty-print.h
2137 (output_bb): Dump stmts streamed.
2138 * lto-section-out.c: Include print-tree.h
2139 (lto_begin_section): Dump sections created.
2140 (lto_output_decl_index): Dump decl encoded.
2141 * lto-streamer-out.c: Include print-tree.h
2142 (create_output_block): Dump output block created.
2143 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2144 (output_function): Dump function output.
2145 (output_constructor): Dump constructor streamed.
2146 (write_global_stream): Output indexes encoded.
2147 (produce_asm_for_decls): Dump streams encoded.
2148 * lto-streamer.c (streamer_dump_file): New global var.
2149 * lto-streamer.h (streamer_dump_file): Declare.
2150 * passes.c (ipa_write_summaries): Initialize streamer dump.
2151 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2154 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2157 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2158 offsets for register save directives. Emit a second batch of save
2159 directives, if need be, when the function accesses prior frames.
2161 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2163 * config/arc/fpu.md (fmasf4): Force operand to register.
2164 (fnmasf4): Likewise.
2166 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2168 * config/arc/arc-protos.h (arc_pad_return): Remove.
2169 * config/arc/arc.c (machine_function): Remove force_short_suffix
2171 (arc_print_operand): Adjust printing of '&'.
2172 (arc_verify_short): Remove conditional printing of short suffix.
2173 (arc_final_prescan_insn): Remove reference to size_reason.
2174 (pad_return): New function.
2175 (arc_reorg): Call pad_return.
2176 (arc_pad_return): Remove.
2177 (arc_init_machine_status): Remove reference to force_short_suffix.
2178 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2179 (attr length): When attribute iscompact is true force to 2
2180 regardless; in the case of maybe check if we want to force the
2181 instruction to have 4 bytes length.
2182 (nopv): Change it to generate 4 byte long nop as well.
2183 (blockage): New pattern.
2184 (simple_return): Remove call to arc_pad_return.
2185 (p_return_i): Likewise.
2187 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2189 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2191 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2193 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2196 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2198 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2199 for ARC700 and ARCv2.
2201 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2204 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2205 operands[2] instead of operands[1].
2208 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2210 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2211 case, check whether the outer register overlaps an unallocatable
2212 register, not just whether it fits the required class.
2214 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2216 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2218 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2219 (memrefs_conflict_p): Likewise.
2220 (init_alias_analysis): Likewise.
2221 * cfgexpand.c (expand_debug_expr): Likewise.
2222 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2223 * cse.c (fold_rtx): Likewise.
2224 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2225 * expr.c (emit_block_move_hints): Likewise.
2226 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2227 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2228 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2229 (emit_group_store): Likewise.
2230 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2231 to read the PRE/POST_MODIFY increment.
2232 * calls.c (store_one_arg): Use strip_offset.
2233 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2235 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2237 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2238 (simplify_binary_operation_1): Extend CONST_INT handling to
2240 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2241 than a HOST_WIDE_INT.
2242 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2244 (adjust_mems, add_stores): Update accodingly.
2245 (vt_canonicalize_addr): Track polynomial offsets.
2246 (emit_note_insn_var_location): Likewise.
2247 (vt_add_function_parameter): Likewise.
2248 (vt_initialize): Likewise.
2250 2018-06-12 Jeff Law <law@redhat.com>
2252 * config.gcc (alpha*-*-freebsd*): Remove.
2253 * config/alpha/freebsd.h: Remove.
2255 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2258 * spellcheck-tree.c (levenshtein_distance): Rename to...
2259 (get_edit_distance): ...this, and update for underlying renaming.
2260 * spellcheck-tree.h (levenshtein_distance): Rename to...
2261 (get_edit_distance): ...this.
2262 * spellcheck.c (levenshtein_distance): Rename to...
2263 (get_edit_distance): ...this. Convert from Levenshtein distance
2264 to Damerau-Levenshtein distance by supporting transpositions of
2265 adjacent characters. Rename "v1" to "v_next" and "v0" to
2267 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2268 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2269 update for underlying renaming.
2270 (selftest::levenshtein_distance_unit_test): Rename to...
2271 (selftest::test_get_edit_distance_unit): ...this, and update for
2272 underlying renaming.
2273 (selftest::test_find_closest_string): Add example from PR 69968
2274 where transposition helps
2275 (selftest::test_metric_conditions): Update for renaming.
2276 (selftest::test_metric_conditions): Likewise.
2277 (selftest::spellcheck_c_tests): Likewise.
2278 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2279 (get_edit_distance): ...this.
2280 (best_match::consider): Update for renaming.
2282 2018-06-12 Martin Sebor <msebor@redhat.com>
2284 PR tree-optimization/85259
2285 * builtins.c (compute_objsize): Handle constant offsets.
2286 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2287 true iff a warning has been issued.
2288 * gimple.h (gimple_nonartificial_location): New function.
2289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2290 gimple_nonartificial_location and handle -Wno-system-headers.
2291 (handle_builtin_stxncpy): Same.
2293 2018-06-12 Martin Sebor <msebor@redhat.com>
2296 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2298 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2300 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2301 BUILTIN_VEC_XST entries for pointer to double and long long.
2303 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2306 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2308 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2311 2018-06-12 Martin Liska <mliska@suse.cz>
2313 * doc/options.texi: Document IntegerRange.
2315 2018-06-12 Martin Liska <mliska@suse.cz>
2317 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2318 * opt-functions.awk: Handle Deprecated flag.
2319 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2321 (read_cmdline_option): Report warning for a deprecated option.
2322 * opts.h (struct cl_option): Add new field cl_deprecated.
2323 (CL_ERR_DEPRECATED): New.
2325 2018-06-12 Martin Liska <mliska@suse.cz>
2327 * doc/options.texi: Document Deprecated option flag.
2329 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2331 * config/arc/arc-arch.h (arc_extras): New enum.
2332 (arc_cpu_t):Add field extra.
2333 (arc_cpu_types): Consider the extras.
2334 * config/arc/arc-cpus.def: Add extras info.
2335 * config/arc/arc-opts.h (processor_type): Consider extra field.
2336 * config/arc/arc.c (arc_override_options): Handle extra field.
2338 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2340 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2341 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2343 * config/arc/arc.c (arc_init): Update pic warning.
2344 (irq_range): Update irq range parsing warnings.
2345 (arc_override_options): Update various warning messages.
2346 (arc_handle_aux_attribute): Likewise.
2348 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2350 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2352 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2354 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2355 compiler messages directives.
2357 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2359 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2360 * config/mips/mips-tables.opt: Regenerate.
2361 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2363 * doc/invoke.texi: Document -march=i6500.
2365 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2367 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2368 (i6400_gpmul): Add cpu_unit.
2369 (i6400_gpdiv): Likewise.
2370 (i6400_msa_add_d): Update reservations.
2371 (i6400_msa_int_add) Likewise.
2372 (i6400_msa_short_logic3) Likewise.
2373 (i6400_msa_short_logic2) Likewise.
2374 (i6400_msa_short_logic) Likewise.
2375 (i6400_msa_move) Likewise.
2376 (i6400_msa_cmp) Likewise.
2377 (i6400_msa_short_float2) Likewise.
2378 (i6400_msa_div_d) Likewise.
2379 (i6400_msa_long_logic1) Likewise.
2380 (i6400_msa_long_logic2) Likewise.
2381 (i6400_msa_mult) Likewise.
2382 (i6400_msa_long_float2) Likewise.
2383 (i6400_msa_long_float4) Likewise.
2384 (i6400_msa_long_float5) Likewise.
2385 (i6400_msa_long_float8) Likewise.
2386 (i6400_fpu_fadd): Include frint type.
2387 (i6400_fpu_store): New define_insn_reservation.
2388 (i6400_fpu_load): Likewise.
2389 (i6400_fpu_move): Likewise.
2390 (i6400_fpu_fcmp): Likewise.
2391 (i6400_fpu_fmadd): Likewise.
2392 (i6400_int_mult): Include imul3nc type and update reservation.
2393 (i6400_int_div): Include idiv3 type and update reservation.
2394 (i6400_int_load): Update to check type not move_type.
2395 (i6400_int_store): Likewise.
2396 (i6400_int_prefetch): Set zero latency.
2398 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2400 * gcc.c: Document new %@{...} sequence.
2401 (LINK_COMMAND_SPEC): Use it for the -L switches.
2402 (cpp_unique_options): Use it for the -I switches.
2403 (at_file_argbuf): New global variable.
2404 (in_at_file): Likewise.
2405 (alloc_args): Create at_file_argbuf.
2406 (clear_args): Truncate at_file_argbuf.
2407 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2408 (open_at_file): New function.
2409 (close_at_file): Likewise.
2410 (create_at_file): Delete.
2411 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2414 (validate_switches_from_spec): Deal with %@{...} sequence.
2415 (validate_switches): Likewise.
2416 (driver::finalize): Call clear_args.
2418 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2420 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2422 2018-06-11 Martin Sebor <msebor@redhat.com>
2424 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2425 (Wno-class-memaccess): Add @opindex.
2426 (Wno-templates, Wno-multiple-inheritance): Same.
2427 (Wno-virtual-inheritance, Wno-namespaces): Same.
2428 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2429 (Wno-format-overflow, Wno-format-truncation): Same.
2430 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2431 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2432 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2433 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2434 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2435 (Wno-misspelled-isr): Same.
2437 2018-06-11 Martin Sebor <msebor@redhat.com>
2439 * PR tree-optimization/86083
2440 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2442 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2444 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2446 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2449 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2450 on the correct operand.
2451 (*movdi_internal64): Ditto.
2453 2018-06-11 Martin Liska <mliska@suse.cz>
2455 PR tree-optimization/86089
2456 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2458 2018-06-11 Julia Koval <julia.koval@intel.com>
2460 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2461 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2462 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2464 2018-06-11 Olivier Hainque <hainque@adacore.com>
2466 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2467 for Ada with strict dwarf2.
2469 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2472 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2475 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2477 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2479 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2481 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2484 2018-06-08 Martin Liska <mliska@suse.cz>
2486 * tree-cfg.h (debug_function): Fix argument type to match
2489 2018-06-08 Martin Liska <mliska@suse.cz>
2491 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2492 Remove usage of MPX-related (and removed) fields.
2493 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2495 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2497 * cfg.c (debug): Use TDF_NONE rather than 0.
2498 * cfghooks.c (debug): Likewise.
2499 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2500 (struct dump_option_value_info): Convert to...
2501 (struct kv_pair): ...this template type.
2502 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2504 (optinfo_verbosity_options): Likewise.
2505 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2507 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2508 than int for "optgroup_flags" param.
2509 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2511 (dump_dec): Likewise.
2512 (dump_finish): Use TDF_NONE rather than 0.
2513 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2514 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2515 than 0. Update for change to option_ptr.
2516 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2517 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2518 0. Update for changes to optinfo_verbosity_options and
2520 (opt_info_switch_p): Convert optgroup_flags from int to
2522 (dump_basic_block): Use dump_flags_t rather than int
2523 for "dump_kind" param.
2524 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2525 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2526 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2527 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2528 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2529 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2530 TDF_NONE): Convert from macros to...
2531 (enum dump_flag): ...this new enum.
2532 (dump_flags_t): Update to use enum.
2533 (operator|, operator&, operator~, operator|=, operator&=):
2534 Implement for dump_flags_t.
2535 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2536 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2537 Convert from macros to...
2538 (enum optgroup_flag): ...this new enum.
2539 (optgroup_flags_t): New typedef.
2540 (operator|, operator|=): Implement for optgroup_flags_t.
2541 (struct dump_file_info): Convert field "alt_flags" to
2542 dump_flags_t. Convert field "optgroup_flags" to
2544 (dump_basic_block): Use dump_flags_t rather than int for param.
2545 (dump_generic_expr_loc): Likewise.
2546 (dump_dec): Likewise.
2547 (dump_register): Convert param "optgroup_flags" to
2549 (opt_info_enable_passes): Likewise.
2550 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2552 * gimple-pretty-print.c (debug): Likewise.
2553 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2554 (merged_store_group::apply_stores): Likewise.
2555 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2556 * gimple.c (verify_gimple_pp): Likewise.
2557 * graphite-poly.c (print_pbb_body): Likewise.
2558 * passes.c (pass_manager::register_one_dump_file): Convert
2559 local "optgroup_flags" to optgroup_flags_t.
2560 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2562 (debug_body): Likewise.
2563 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2564 to optgroup_flags_t.
2565 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2567 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2568 (convert_mult_to_fma): Likewise.
2569 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2570 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2571 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2572 "dump_kind" to dump_flags_t.
2574 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2576 * config/rs6000/rs6000.c (min, max): Delete.
2578 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2580 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2581 -mabi=spe and -mabi=no-spe.
2583 2018-06-08 Martin Liska <mliska@suse.cz>
2585 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2586 where we expect an existing summary.
2588 2018-06-08 Martin Liska <mliska@suse.cz>
2590 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2591 * ipa-inline.h (estimate_edge_growth): Likewise.
2593 2018-06-08 Martin Liska <mliska@suse.cz>
2595 * cgraph.c (function_version_hasher::hash): Use
2596 cgraph_node::get_uid ().
2597 (function_version_hasher::equal):
2598 * cgraph.h (cgraph_node::get_uid): New method.
2599 * ipa-inline.c (update_caller_keys): Use
2600 cgraph_node::get_uid ().
2601 (update_callee_keys): Likewise.
2602 * ipa-utils.c (searchc): Likewise.
2603 (ipa_reduced_postorder): Likewise.
2604 * lto-cgraph.c (input_node): Likewise.
2605 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2606 * symbol-summary.h (symtab_insertion): Likewise.
2607 (symtab_removal): Likewise.
2608 (symtab_duplication): Likewise.
2609 * tree-pretty-print.c (dump_function_header): Likewise.
2610 * tree-sra.c (convert_callers_for_node): Likewise.
2612 2018-06-08 Martin Liska <mliska@suse.cz>
2614 * cgraph.c (symbol_table::create_edge): Always assign a new
2616 (symbol_table::free_edge): Do not recycle numbers.
2617 * cgraph.h (cgraph_edge::get): New method.
2618 * symbol-summary.h (symtab_removal): Use it.
2619 (symtab_duplication): Likewise.
2620 (call_summary::hashable_uid): Remove.
2622 2018-06-08 Martin Liska <mliska@suse.cz>
2624 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2625 (initialize_growth_caches): Remove.
2626 (free_growth_caches): Likewise.
2627 (do_estimate_edge_time): Use edge_growth_cache.
2628 (do_estimate_edge_size): Likewise.
2629 (do_estimate_edge_hints): Likewise.
2630 * ipa-inline.c (reset_edge_caches): Likewise.
2631 (recursive_inlining): Likewise.
2632 (inline_small_functions): Likewise.
2633 * ipa-inline.h (initialize_growth_caches): Remove.
2634 (estimate_edge_size): Likewise.
2635 (estimate_edge_time): Likewise.
2636 (estimate_edge_hints): Likewise.
2637 (reset_edge_growth_cache): Likewise.
2638 * symbol-summary.h (call_summary::remove): New method.
2640 2018-06-08 Martin Liska <mliska@suse.cz>
2642 * ipa-cp.c (class edge_clone_summary): New summary.
2643 (grow_edge_clone_vectors): Remove.
2644 (ipcp_edge_duplication_hook): Remove.
2645 (class edge_clone_summary_t): New call_summary class.
2646 (ipcp_edge_removal_hook): Remove.
2647 (edge_clone_summary_t::duplicate): New function.
2648 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2649 (create_specialized_node): Likewise.
2650 (ipcp_driver): Initialize edge_clone_summaries and do not
2653 2018-06-08 Martin Liska <mliska@suse.cz>
2655 * symbol-summary.h (get): New function.
2656 (call_summary::m_initialize_when_cloning): New class member.
2658 2018-06-08 Martin Liska <mliska@suse.cz>
2660 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2661 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2662 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2663 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2664 (remove_cgraph_node_from_order): Use the removed_nodes set.
2665 (do_per_function_toporder): Likwise.
2666 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2667 instead of summary_uid.
2668 (symtab_removal): Likewise.
2669 (symtab_duplication): Likewise.
2671 2018-06-08 Martin Liska <mliska@suse.cz>
2673 * ipa-cp.c (ipcp_store_bits_results): Use
2674 ipcp_transformation_sum.
2675 (ipcp_store_vr_results): Likewise.
2676 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2678 (ipcp_transformation_initialize): ... this.
2679 (ipa_set_node_agg_value_chain):
2680 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2681 (write_ipcp_transformation_info): Likewise.
2682 (read_ipcp_transformation_info): Likewise.
2683 (ipcp_update_bits): Likewise.
2684 (ipcp_update_vr): Likewise.
2685 (ipcp_transform_function): Likewise.
2686 * ipa-prop.h: Rename ipcp_transformation_summary to
2687 ipcp_transformation.
2688 (class ipcp_transformation_t): New function summary.
2689 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2690 (ipa_get_agg_replacements_for_node): Likewise.
2692 2018-06-08 Martin Liska <mliska@suse.cz>
2694 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2696 (class funct_state_summary_t): New function_summary class.
2697 (has_function_state): Remove.
2698 (get_function_state): Likewise.
2699 (set_function_state): Likewise.
2700 (add_new_function): Likewise.
2701 (funct_state_summary_t::insert): New function.
2702 (duplicate_node_data): Remove.
2703 (remove_node_data): Remove.
2704 (funct_state_summary_t::duplicate): New function.
2705 (register_hooks): Create new funct_state_summaries.
2706 (pure_const_generate_summary): Use it.
2707 (pure_const_write_summary): Likewise.
2708 (pure_const_read_summary): Likewise.
2709 (propagate_pure_const): Likewise.
2710 (propagate_nothrow): Likewise.
2711 (dump_malloc_lattice): Likewise.
2712 (propagate_malloc): Likewise.
2713 (execute): Do not register hooks, just remove summary
2715 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2718 2018-06-08 Martin Liska <mliska@suse.cz>
2720 * ipa-reference.c (remove_node_data): Remove.
2721 (duplicate_node_data): Likewise.
2722 (class ipa_ref_var_info_summary_t): New class.
2723 (class ipa_ref_opt_summary_t): Likewise.
2724 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2725 (get_reference_optimization_summary): Use
2726 ipa_ref_opt_sum_summaries.
2727 (set_reference_vars_info): Remove.
2728 (set_reference_optimization_summary): Likewise.
2729 (ipa_init): Create summaries.
2730 (init_function_info): Use function summary.
2731 (ipa_ref_opt_summary_t::duplicate): New function.
2732 (ipa_ref_opt_summary_t::remove): New function.
2733 (get_read_write_all_from_node): Fix GNU coding style.
2734 (propagate): Use function summary.
2735 (write_node_summary_p): Fix GNU coding style.
2736 (stream_out_bitmap): Likewise.
2737 (ipa_reference_read_optimization_summary): Use function summary.
2738 (ipa_reference_c_finalize): Do not release hooks.
2740 2018-06-08 Martin Liska <mliska@suse.cz>
2742 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2743 (analyze_function_body): Extract multiple calls of get_create.
2744 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2745 * ipa-inline.c (recursive_inlining): Use ::get method.
2746 * ipa-inline.h (estimate_edge_growth): Likewise.
2748 2018-06-08 Martin Liska <mliska@suse.cz>
2750 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2752 (hsa_function_summary::hsa_function_summary): Use the new enum
2754 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2755 * hsa-gen.c (hsa_get_host_function): Likewise.
2756 (get_brig_function_name): Likewise.
2757 * ipa-hsa.c (process_hsa_functions): Likewise.
2758 (ipa_hsa_write_summary): Likewise.
2759 * symbol-summary.h (symtab_duplication): Use ::get function/
2760 (get): New function.
2762 2018-06-08 Martin Liska <mliska@suse.cz>
2764 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2766 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2767 (hsa_register_kernel): Likewise.
2768 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2769 * hsa-gen.c (hsa_get_host_function): Likewise.
2770 (get_brig_function_name): Likewise.
2771 (generate_hsa): Likewise.
2772 (pass_gen_hsail::execute): Likewise.
2773 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2774 (devirtualization_time_bonus): Likewise.
2775 (ipcp_propagate_stage): Likewise.
2776 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2777 (edge_set_predicate): Likewise.
2778 (evaluate_conditions_for_known_args): Likewise.
2779 (evaluate_properties_for_edge): Likewise.
2780 (ipa_fn_summary::reset): Likewise.
2781 (ipa_fn_summary_t::duplicate): Likewise.
2782 (dump_ipa_call_summary): Likewise.
2783 (ipa_dump_fn_summary): Likewise.
2784 (analyze_function_body): Likewise.
2785 (compute_fn_summary): Likewise.
2786 (estimate_edge_devirt_benefit): Likewise.
2787 (estimate_edge_size_and_time): Likewise.
2788 (estimate_calls_size_and_time): Likewise.
2789 (estimate_node_size_and_time): Likewise.
2790 (inline_update_callee_summaries): Likewise.
2791 (remap_edge_change_prob): Likewise.
2792 (remap_edge_summaries): Likewise.
2793 (ipa_merge_fn_summary_after_inlining): Likewise.
2794 (ipa_update_overall_fn_summary): Likewise.
2795 (read_ipa_call_summary): Likewise.
2796 (inline_read_section): Likewise.
2797 (write_ipa_call_summary): Likewise.
2798 (ipa_fn_summary_write): Likewise.
2799 (ipa_free_fn_summary): Likewise.
2800 * ipa-hsa.c (process_hsa_functions): Likewise.
2801 (ipa_hsa_write_summary): Likewise.
2802 (ipa_hsa_read_section): Likewise.
2803 * ipa-icf.c (sem_function::merge): Likewise.
2804 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2805 (do_estimate_edge_time): Likewise.
2806 (estimate_size_after_inlining): Likewise.
2807 (estimate_growth): Likewise.
2808 (growth_likely_positive): Likewise.
2809 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2810 (inline_call): Likewise.
2811 * ipa-inline.c (caller_growth_limits): Likewise.
2812 (can_inline_edge_p): Likewise.
2813 (can_inline_edge_by_limits_p): Likewise.
2814 (compute_uninlined_call_time): Likewise.
2815 (compute_inlined_call_time): Likewise.
2816 (want_inline_small_function_p): Likewise.
2817 (edge_badness): Likewise.
2818 (update_caller_keys): Likewise.
2819 (update_callee_keys): Likewise.
2820 (recursive_inlining): Likewise.
2821 (inline_small_functions): Likewise.
2822 (inline_to_all_callers_1): Likewise.
2823 (dump_overall_stats): Likewise.
2824 (early_inline_small_functions): Likewise.
2825 (early_inliner): Likewise.
2826 * ipa-inline.h (estimate_edge_growth): Likewise.
2827 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2828 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2829 * ipa-prop.h (IPA_NODE_REF): Likewise.
2830 (IPA_EDGE_REF): Likewise.
2831 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2832 (propagate_malloc): Likewise.
2833 * ipa-split.c (execute_split_functions): Likewise.
2834 * symbol-summary.h: Rename get to get_create.
2836 (get_create): Likewise.
2837 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2839 2018-06-08 Martin Liska <mliska@suse.cz>
2841 * symbol-summary.h (release): Move definition out of class
2843 (symtab_removal): Likewise.
2844 (symtab_duplication): Likewise.
2846 2018-06-08 Martin Liska <mliska@suse.cz>
2848 * symbol-summary.h (function_summary): Move constructor
2849 implementation out of class declaration.
2850 (release): Likewise.
2851 (symtab_insertion): Likewise.
2852 (symtab_removal): Likewise.
2853 (symtab_duplication): Likewise.
2856 2018-06-08 Martin Liska <mliska@suse.cz>
2858 * Makefile.in: Remove support for MPX (macros, related functions,
2859 fields in cgraph_node, ...).
2860 * builtin-types.def (BT_BND): Likewise.
2861 (BT_FN_BND_CONST_PTR): Likewise.
2862 (BT_FN_CONST_PTR_BND): Likewise.
2863 (BT_FN_VOID_PTR_BND): Likewise.
2864 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2865 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2866 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2867 (expand_builtin_mempcpy_with_bounds): Likewise.
2868 (expand_builtin_memset_with_bounds): Likewise.
2869 (expand_builtin_memset_args): Likewise.
2870 (std_expand_builtin_va_start): Likewise.
2871 (expand_builtin): Likewise.
2872 (expand_builtin_with_bounds): Likewise.
2873 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2874 (DEF_LIB_BUILTIN_CHKP): Likewise.
2875 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2876 (DEF_CHKP_BUILTIN): Likewise.
2877 (BUILT_IN_MEMCPY): Likewise.
2878 (BUILT_IN_MEMMOVE): Likewise.
2879 (BUILT_IN_MEMPCPY): Likewise.
2880 (BUILT_IN_MEMSET): Likewise.
2881 (BUILT_IN_STPCPY): Likewise.
2882 (BUILT_IN_STRCAT): Likewise.
2883 (BUILT_IN_STRCHR): Likewise.
2884 (BUILT_IN_STRCPY): Likewise.
2885 (BUILT_IN_STRLEN): Likewise.
2886 (BUILT_IN_MEMCPY_CHK): Likewise.
2887 (BUILT_IN_MEMMOVE_CHK): Likewise.
2888 (BUILT_IN_MEMPCPY_CHK): Likewise.
2889 (BUILT_IN_MEMSET_CHK): Likewise.
2890 (BUILT_IN_STPCPY_CHK): Likewise.
2891 (BUILT_IN_STRCAT_CHK): Likewise.
2892 (BUILT_IN_STRCPY_CHK): Likewise.
2893 * calls.c (store_bounds): Likewise.
2894 (emit_call_1): Likewise.
2895 (special_function_p): Likewise.
2896 (maybe_warn_nonstring_arg): Likewise.
2897 (initialize_argument_information): Likewise.
2898 (finalize_must_preallocate): Likewise.
2899 (compute_argument_addresses): Likewise.
2900 (expand_call): Likewise.
2901 * cfgexpand.c (expand_call_stmt): Likewise.
2902 (expand_return): Likewise.
2903 (expand_gimple_stmt_1): Likewise.
2904 (pass_expand::execute): Likewise.
2905 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2906 (cgraph_node::remove): Likewise.
2907 (cgraph_node::dump): Likewise.
2908 (cgraph_node::verify_node): Likewise.
2909 * cgraph.h (chkp_function_instrumented_p): Likewise.
2910 (symtab_node::get_alias_target): Likewise.
2911 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2912 (cgraph_local_p): Likewise.
2913 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2914 (cgraph_edge::rebuild_references): Likewise.
2915 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2916 (walk_polymorphic_call_targets): Likewise.
2917 (cgraph_node::expand_thunk): Likewise.
2918 (symbol_table::output_weakrefs): Likewise.
2919 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2920 (ix86_handle_option): Likewise.
2921 * config/i386/constraints.md: Likewise.
2922 * config/i386/i386-builtin-types.def (BND): Likewise.
2926 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2927 (BDESC_FIRST): Likewise.
2929 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2930 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2931 * config/i386/i386.c (enum reg_class): Likewise.
2932 (ix86_target_string): Likewise.
2933 (ix86_option_override_internal): Likewise.
2934 (ix86_conditional_register_usage): Likewise.
2935 (ix86_valid_target_attribute_inner_p): Likewise.
2936 (ix86_set_indirect_branch_type): Likewise.
2937 (ix86_set_current_function): Likewise.
2938 (ix86_function_arg_regno_p): Likewise.
2939 (init_cumulative_args): Likewise.
2940 (ix86_function_arg_advance): Likewise.
2941 (ix86_function_arg): Likewise.
2942 (ix86_pass_by_reference): Likewise.
2943 (ix86_function_value_regno_p): Likewise.
2944 (ix86_function_value_1): Likewise.
2945 (ix86_function_value_bounds): Likewise.
2946 (ix86_return_in_memory): Likewise.
2947 (ix86_setup_incoming_vararg_bounds): Likewise.
2948 (ix86_va_start): Likewise.
2949 (indirect_thunk_need_prefix): Likewise.
2950 (print_reg): Likewise.
2951 (ix86_print_operand): Likewise.
2952 (ix86_expand_call): Likewise.
2953 (ix86_output_function_return): Likewise.
2954 (reg_encoded_number): Likewise.
2955 (BDESC_VERIFYS): Likewise.
2956 (ix86_init_mpx_builtins): Likewise.
2957 (ix86_init_builtins): Likewise.
2958 (ix86_emit_cmove): Likewise.
2959 (ix86_emit_move_max): Likewise.
2960 (ix86_expand_builtin): Likewise.
2961 (ix86_builtin_mpx_function): Likewise.
2962 (ix86_get_arg_address_for_bt): Likewise.
2963 (ix86_load_bounds): Likewise.
2964 (ix86_store_bounds): Likewise.
2965 (ix86_load_returned_bounds): Likewise.
2966 (ix86_store_returned_bounds): Likewise.
2967 (ix86_class_likely_spilled_p): Likewise.
2968 (ix86_hard_regno_mode_ok): Likewise.
2969 (x86_order_regs_for_local_alloc): Likewise.
2970 (ix86_mitigate_rop): Likewise.
2971 (ix86_bnd_prefixed_insn_p): Likewise.
2972 (ix86_mpx_bound_mode): Likewise.
2973 (ix86_make_bounds_constant): Likewise.
2974 (ix86_initialize_bounds): Likewise.
2975 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2976 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2977 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2978 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2979 (TARGET_CHKP_BOUND_MODE): Likewise.
2980 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2981 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2982 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2983 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2984 * config/i386/i386.h (TARGET_MPX): Likewise.
2985 (TARGET_MPX_P): Likewise.
2986 (VALID_BND_REG_MODE): Likewise.
2987 (FIRST_BND_REG): Likewise.
2988 (LAST_BND_REG): Likewise.
2989 (enum reg_class): Likewise.
2990 (BND_REG_P): Likewise.
2991 (BND_REGNO_P): Likewise.
2992 (BNDmode): Likewise.
2993 (ADJUST_INSN_LENGTH): Likewise.
2994 * config/i386/i386.md: Likewise.
2995 * config/i386/i386.opt: Likewise.
2996 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2997 (defined): Likewise.
2998 (LINK_MPX): Likewise.
2999 (MPX_SPEC): Likewise.
3000 (LIBMPX_SPEC): Likewise.
3001 (LIBMPXWRAPPERS_SPEC): Likewise.
3002 (CHKP_SPEC): Likewise.
3003 * config/i386/predicates.md: Likewise.
3004 * dbxout.c (dbxout_type): Likewise.
3005 * doc/extend.texi: Likewise.
3006 * doc/invoke.texi: Likewise.
3007 * doc/md.texi: Likewise.
3008 * doc/tm.texi: Likewise.
3009 * doc/tm.texi.in: Likewise.
3010 * dwarf2out.c (is_base_type): Likewise.
3011 (gen_formal_types_die): Likewise.
3012 (gen_subprogram_die): Likewise.
3013 (gen_type_die_with_usage): Likewise.
3014 (gen_decl_die): Likewise.
3015 (dwarf2out_late_global_decl): Likewise.
3016 * expr.c (expand_assignment): Likewise.
3017 (emit_storent_insn): Likewise.
3018 (store_expr_with_bounds): Likewise.
3019 (store_expr): Likewise.
3020 (expand_expr_real_1): Likewise.
3021 * expr.h (store_expr_with_bounds): Likewise.
3022 * function.c (use_register_for_decl): Likewise.
3023 (struct bounds_parm_data): Likewise.
3024 (assign_parms_augmented_arg_list): Likewise.
3025 (assign_parm_find_entry_rtl): Likewise.
3026 (assign_parm_is_stack_parm): Likewise.
3027 (assign_parm_load_bounds): Likewise.
3028 (assign_bounds): Likewise.
3029 (assign_parms): Likewise.
3030 (expand_function_start): Likewise.
3031 * gcc.c (CHKP_SPEC): Likewise.
3032 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3033 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3034 (wrestrict_dom_walker::check_call): Likewise.
3035 * gimple.c (gimple_build_call_from_tree): Likewise.
3036 * gimple.h (enum gf_mask): Likewise.
3037 (gimple_call_with_bounds_p): Likewise.
3038 (gimple_call_set_with_bounds): Likewise.
3039 * gimplify.c (gimplify_init_constructor): Likewise.
3040 * ipa-cp.c (initialize_node_lattices): Likewise.
3041 (propagate_constants_across_call): Likewise.
3042 (find_more_scalar_values_for_callers_subset): Likewise.
3043 * ipa-hsa.c (process_hsa_functions): Likewise.
3044 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3045 * ipa-icf.c (sem_function::merge): Likewise.
3046 * ipa-inline.c (early_inliner): Likewise.
3047 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3048 (warn_function_cold): Likewise.
3049 (propagate_pure_const): Likewise.
3050 * ipa-ref.h (enum GTY): Likewise.
3051 * ipa-split.c (find_retbnd): Likewise.
3052 (consider_split): Likewise.
3053 (split_function): Likewise.
3054 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3055 * ipa.c (walk_polymorphic_call_targets): Likewise.
3056 (symbol_table::remove_unreachable_nodes): Likewise.
3057 (process_references): Likewise.
3058 (cgraph_build_static_cdtor_1): Likewise.
3059 * lto-cgraph.c (lto_output_node): Likewise.
3060 (output_refs): Likewise.
3061 (compute_ltrans_boundary): Likewise.
3062 (input_overwrite_node): Likewise.
3063 (input_node): Likewise.
3064 (input_cgraph_1): Likewise.
3065 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3066 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3067 (class pass_chkp_instrumentation_passes): Likewise.
3068 (make_pass_chkp_instrumentation_passes): Likewise.
3069 * passes.def: Likewise.
3070 * rtl.h (struct GTY): Likewise.
3071 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3072 * stor-layout.c (layout_type): Likewise.
3073 * symtab.c: Likewise.
3074 * target.def: Likewise.
3075 * targhooks.c (default_chkp_bound_type): Likewise.
3076 (default_chkp_bound_mode): Likewise.
3077 (default_builtin_chkp_function): Likewise.
3078 (default_chkp_function_value_bounds): Likewise.
3079 (default_chkp_make_bounds_constant): Likewise.
3080 (default_chkp_initialize_bounds): Likewise.
3081 * targhooks.h (default_chkp_bound_type): Likewise.
3082 (default_chkp_bound_mode): Likewise.
3083 (default_builtin_chkp_function): Likewise.
3084 (default_chkp_function_value_bounds): Likewise.
3085 (default_chkp_make_bounds_constant): Likewise.
3086 (default_chkp_initialize_bounds): Likewise.
3087 * toplev.c (compile_file): Likewise.
3088 (process_options): Likewise.
3089 * tree-core.h (DEF_BUILTIN): Likewise.
3090 (DEF_BUILTIN_CHKP): Likewise.
3091 * tree-inline.c (declare_return_variable): Likewise.
3092 (remap_gimple_stmt): Likewise.
3093 (copy_bb): Likewise.
3094 (initialize_inlined_parameters): Likewise.
3095 (expand_call_inline): Likewise.
3096 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3097 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3098 (make_pass_ipa_chkp_produce_thunks): Likewise.
3099 (make_pass_chkp): Likewise.
3100 (make_pass_chkp_opt): Likewise.
3101 (make_pass_chkp_instrumentation_passes): Likewise.
3102 * tree-pretty-print.c (dump_generic_node): Likewise.
3103 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3104 * tree-ssa-dce.c (propagate_necessity): Likewise.
3105 (eliminate_unnecessary_stmts): Likewise.
3106 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3107 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3108 * tree-ssa-sccvn.h: Likewise.
3109 * tree-ssa-strlen.c (get_string_length): Likewise.
3110 (valid_builtin_call): Likewise.
3111 (adjust_last_stmt): Likewise.
3112 (handle_builtin_strchr): Likewise.
3113 (handle_builtin_strcpy): Likewise.
3114 (handle_builtin_stxncpy): Likewise.
3115 (handle_builtin_memcpy): Likewise.
3116 (handle_builtin_strcat): Likewise.
3117 (strlen_check_and_optimize_stmt): Likewise.
3118 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3119 * tree-streamer-in.c: Likewise.
3120 * tree-streamer.c (record_common_node): Likewise.
3121 * tree.c (tree_code_size): Likewise.
3122 (wide_int_to_tree_1): Likewise.
3123 (type_contains_placeholder_1): Likewise.
3124 (build_common_tree_nodes): Likewise.
3125 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3126 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3127 (POINTER_BOUNDS_P): Likewise.
3128 (BOUNDED_TYPE_P): Likewise.
3129 (BOUNDED_P): Likewise.
3130 (CALL_WITH_BOUNDS_P): Likewise.
3131 (pointer_bounds_type_node): Likewise.
3132 * value-prof.c (gimple_ic): Likewise.
3133 * var-tracking.c (vt_add_function_parameters): Likewise.
3134 * varasm.c (make_decl_rtl): Likewise.
3135 (assemble_start_function): Likewise.
3136 (output_constant): Likewise.
3137 (maybe_assemble_visibility): Likewise.
3138 * varpool.c (ctor_for_folding): Likewise.
3139 * chkp-builtins.def: Remove.
3140 * ipa-chkp.c: Remove.
3141 * ipa-chkp.h: Remove.
3142 * rtl-chkp.c: Remove.
3143 * rtl-chkp.h: Remove.
3144 * tree-chkp-opt.c: Remove.
3145 * tree-chkp.c: Remove.
3146 * tree-chkp.h: Remove.
3148 2018-06-07 Carl Love <cel@us.ibm.com>
3150 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3151 vextract_fp_from_shortl): Add BE support.
3153 2018-06-07 Paul Koning <ni1d@arrl.net>
3155 * compare-elim.c (try_merge_compare): Don't merge compare if
3156 address contains a side effect.
3157 (try_eliminate_compare): Likewise.
3159 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3161 * config.gcc: Support "tremont".
3162 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3163 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3165 * config/i386/i386.c (m_TREMONT): Define.
3166 (processor_target_table): Add "tremont".
3167 (PTA_TREMONT): Define.
3168 (ix86_lea_outperforms): Add TARGET_TREMONT.
3169 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3170 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3171 and M_INTEL_GOLDMONT_PLUS.
3172 (fold_builtin_cpu): Add "tremont".
3173 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3174 (ix86_option_override_internal): Add "tremont".
3175 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3176 (processor_type): Add PROCESSOR_TREMONT.
3177 * config/i386/x86-tune.def: Add m_TREMONT.
3178 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3180 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3182 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3183 symbol defined for msp430i* devices to be lower case.
3185 gcc/testsuite/gcc.target/msp430/
3186 * msp430i-device-symbol.c: New test.
3187 * msp430f-device-symbol.c: New test.
3188 * msp430.h: New test header file.
3190 2018-06-07 Richard Biener <rguenther@suse.de>
3192 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3193 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3194 Properly wrap signed arithmetic if overflow wraps.
3196 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3198 PR tree-optimization/69615
3199 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3200 of a cast from a same precision integral SSA_NAME in a bb dominated
3201 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3202 cast to utype if rhs2 has already a compatible type.
3204 2018-06-07 Richard Biener <rguenther@suse.de>
3206 PR tree-optimization/85935
3207 * graphite-scop-detection.c (find_params_in_bb): Analyze
3208 condition operands with respect to the correct loop. Assert
3209 the analysis doesn't fail.
3211 2018-06-04 Carl Love <cel@us.ibm.com>
3213 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3214 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3215 as it is slightly cheaper.
3216 (first_match_or_eos_index_<mode>):
3217 Calculate index using natural element order.
3218 (first_match_index_<mode>):
3219 Calculate index using natural element order.
3220 (first_match_or_eos_index_<mode>):
3221 Calculate index using natural order.
3222 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3223 for BE and LE modes.
3224 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3225 P9V_BUILTIN_VCLZLSBB_V16QI.
3226 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3229 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3231 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3232 indentation and line wrap for many prototypes. Add missing
3233 @smallexample directives around block of prototypes for vec_xl and
3236 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3238 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3239 track if we pass or return IEEE 128-bit floating point.
3240 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3241 C++ mangling that is compatible with GCC 8.1.
3242 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3243 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3244 floating point types.
3245 (rs6000_function_arg_advance_1): Likewise.
3246 (rs6000_mangle_type): Optionally generate mangled names that match
3247 what GCC 8.1 generated for IEEE 128-bit floating point types.
3248 (rs6000_globalize_decl_name): If we have an external function that
3249 passes or returns IEEE 128-bit types, generate a weak reference
3250 from the mangled name used in GCC 8.1 to the current mangled
3252 (rs6000_init_builtins): Make __ibm128 use the long double type if
3253 long double is IBM extended double. Make __float128 use the long
3254 double type if long double is IEEE 128-bit.
3257 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3258 macro for __ibm128 built-in functions.
3259 (PACK_IF): Add __ibm128 pack/unpack functions.
3260 (UNPACK_IF): Likewise.
3261 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3262 enable long double built-in functions if long double is IEEE
3263 128-bit floating point.
3264 (rs6000_invalid_builtin): Update long double built-in function
3266 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3267 functions, adjust the built-in function to use the long double
3268 built-in function if __ibm128 and long double are the same type.
3269 * doc/extend.texi (PowerPC builtins): Update documention for
3270 __builtin_{,un}pack_longdouble. Add documentation for
3271 __builtin_{,un}pack_ibm128.
3273 2018-06-06 Jim Wilson <jimw@sifive.com>
3275 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3276 (struct machine_function): New field interrupt_mode.
3277 (riscv_handle_type_attribute): New function. Add forward declaration.
3278 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3279 (riscv_expand_epilogue): Check interrupt_mode field.
3280 (riscv_set_current_function): Check interrupt attribute args and
3281 set interrupt_mode field.
3282 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3283 (riscv_sret, riscv_uret): New.
3284 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3285 new arguments to interrupt attribute.
3287 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3290 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3291 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3293 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3295 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3296 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3297 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3298 several redundant entries.
3300 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3302 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3303 type from "rtx" to "rtx_insn *".
3304 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3305 for local "call_insn", removing cast.
3306 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3308 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3310 PR tree-optimization/86066
3311 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3312 for BIT_INSERT_EXPR stores.
3314 2018-06-06 Richard Biener <rguenther@suse.de>
3316 PR tree-optimization/86062
3317 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3318 component refs ontop
3319 of to be offsetted base.
3321 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3323 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3324 to be static and remove check on interrupt attribute name.
3326 gcc/testsuite/gcc.target/msp430/
3327 * function-attributes-4.c: New test.
3328 * static-interrupts.c: New test.
3330 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3332 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3333 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3335 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3338 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3340 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3341 Remove second argument, change how error is called.
3342 (aarch64_layout_arg): Remove second argument from
3343 aarch64_err_no_fpadvsimd call.
3344 (aarch64_init_cumulative_args): Ditto.
3345 (aarch64_gimplify_va_arg_expr): Ditto.
3346 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3348 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3350 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3351 (*simple_return_indirect_internal<mode>): Rename from
3352 simple_return_indirect_internal. Use W mode iterator.
3353 (rstorssp): New expander.
3354 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3355 (clrssbsy): New expander.
3356 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3358 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3360 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3362 (cmse_check_pointed_object): Likewise.
3364 2018-06-05 Martin Liska <mliska@suse.cz>
3366 PR gcov-profile/47618
3367 * doc/invoke.texi: Document how -fprofile-dir format
3370 2018-06-05 Richard Biener <rguenther@suse.de>
3372 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3373 removal pretend DOM info isn't available so we do not update
3374 it and only remove edges, not dominated blocks. Actually free
3375 DOM info in case we removed something. Remove unreachable blocks.
3376 (mfb_keep_latches): Work with either DOM info or marked backedges.
3377 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3378 first. Mark backedges if DOM info isn't available.
3379 (Re-)compute DOM info after cleanup_control_flow_pre.
3381 2018-06-05 Richard Biener <rguenther@suse.de>
3383 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3384 (locus_discrim_hasher::hash): Adjust.
3385 (locus_discrim_hasher::equal): Likewise.
3386 (next_discriminator_for_locus): Work on line directly.
3387 (same_line_p): Pass in expanded locus1 as well.
3388 (assign_discriminators): Avoid redundant location expansions.
3390 2018-06-05 Richard Biener <rguenther@suse.de>
3392 PR tree-optimization/86046
3393 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3394 if required after clearing TREE_ADDRESSABLE.
3396 2018-06-05 Richard Biener <rguenther@suse.de>
3398 PR tree-optimization/86047
3399 * tree-ssa-loop.c (for_each_index): Glob handling of all
3400 decls and constants and really handle all of them.
3402 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3405 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3406 qualifier_void_pointer and qualifier_const_void_pointer.
3407 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3408 (arm_init_builtins): Handle the above.
3409 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3410 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3411 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3414 2018-06-05 Martin Liska <mliska@suse.cz>
3416 * auto-profile.c (read_autofdo_file): Do not use
3417 gcov_ctr_summary struct.
3418 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3419 * coverage.c (struct counts_entry): Likewise.
3420 (read_counts_file): Read just single summary entry.
3421 (get_coverage_counts): Use gcov_summary struct.
3422 * coverage.h (get_coverage_counts): Likewise.
3423 * gcov-dump.c (dump_working_sets): Likewise.
3424 (tag_summary): Dump just single summary.
3425 * gcov-io.c (gcov_write_summary): Write just histogram
3427 (gcov_read_summary): Read just single summary.
3428 (compute_working_sets): Use gcov_summary struct.
3429 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3430 of GCOV_COUNTERS_SUMMABLE.
3431 (GCOV_COUNTERS_SUMMABLE): Remove.
3432 (GCOV_FIRST_VALUE_COUNTER): Replace with
3433 GCOV_COUNTER_V_INTERVAL.
3434 (struct gcov_ctr_summary): Remove.
3435 (struct gcov_summary): Directly use fields of former
3437 (compute_working_sets): Use gcov_summary struct.
3438 * gcov.c (read_count_file): Do not use ctrs fields.
3439 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3441 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3443 * profile.c: Likewise.
3444 * profile.h: Likewise.
3446 2018-06-05 Martin Liska <mliska@suse.cz>
3448 PR gcov-profile/84846
3449 * gcov.c (output_lines): Print working directory only
3450 in intermediate format.
3452 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3454 * config/s390/s390-builtin-types.def: Add void function type.
3455 * config/s390/s390-builtins.def: Use the function type for the
3458 2018-06-04 Jim Wilson <jimw@sifive.com>
3460 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3462 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3463 and maybe_eh_return. Change regno to unsigned int. Use new args to
3464 handle EH_RETURN_DATA_REGNO registers properly.
3465 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3466 (riscv_expand_epilogue): Update comment. Change argument name and
3467 type. Update code to use new name and type. Pass new args to
3468 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3470 * config/riscv/riscv.md (NORMAL_RETURN): New.
3471 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3472 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3473 (eh_return): Call gen_eh_return_internal and emit barrier.
3474 (eh_return_internal): Call riscv_expand_epilogue.
3476 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3478 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3479 bit_insertion field and declare can_be_merged_into method.
3480 (merged_store_group::can_be_merged_into): New method.
3481 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3482 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3483 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3485 2018-06-04 Richard Biener <rguenther@suse.de>
3487 PR tree-optimization/85955
3488 * builtins.c (fold_builtin_sincos): Convert pointers to
3489 destination to appropriate type before dereferencing.
3491 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3493 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3495 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3497 * expr.c (expand_expr_real_1): Force the operand into memory if
3498 its TYPE_MODE is BLKmode and if there is no integer mode for
3499 the number of bits being extracted.
3501 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3505 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3506 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3508 2018-06-04 Richard Biener <rguenther@suse.de>
3510 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3511 (cleanup_tree_cfg_noloop): ... single caller. Do
3512 start_recording_case_labels later.
3514 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3516 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3517 to _IMMINTRIN_H_INCLUDED.
3518 * config/i386/pconfigintrin.h: Ditto.
3519 * config/i386/waitpkgintrin.h: Ditto.
3520 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3521 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3522 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3523 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3524 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3525 waitpkgintrin.h and cldemoteintrin.h.
3527 2018-06-04 Richard Biener <rguenther@suse.de>
3529 PR tree-optimization/86038
3530 * tracer.c (find_best_successor): Check probability for
3531 being initialized, bail out if not.
3533 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3536 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3537 of bits to ignore when comparing architectures.
3539 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3541 PR tree-optimization/69615
3542 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3543 maximum or minimum of the type, try to merge it also as if
3544 range1 is + [-, x - 1] or + [x + 1, -].
3547 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3549 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3551 PR tree-optimization/86034
3552 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3553 the unsigned bitfield type in a bit insertion sequence if it does not
3554 have a larger precision than the bitfield size.
3555 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3557 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3559 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3561 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3563 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3564 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3565 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3566 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3568 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3570 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3571 Disable -fdelete-null-pointer-checks for ELF toolchain.
3573 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3574 Kito Cheng <kito.cheng@gmail.com>
3576 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3577 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3578 (nds32*-*-*): Add glibc and uclibc conditions.
3579 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3580 (TARGET_EXCEPT_UNWIND_INFO): Define.
3581 * config/nds32/elf.h: New file.
3582 * config/nds32/linux.h: New file.
3583 * config/nds32/nds32-elf.opt: New file.
3584 * config/nds32/nds32-linux.opt: New file.
3585 * config/nds32/nds32-fp-as-gp.c
3586 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3587 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3589 (nds32_asm_file_end): Ditto.
3590 (nds32_print_operand): Ditto.
3591 (nds32_insert_attributes): Ditto.
3592 (nds32_init_libfuncs): New function.
3593 (TARGET_HAVE_TLS): Define.
3594 (TARGET_INIT_LIBFUNCS): Define.
3595 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3597 (TARGET_ELF): Apply different mcmodel setting.
3598 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3599 been migrated into elf.h and linux.h files.
3600 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3601 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3602 (mcmodel): The content has been migrated into nds32-elf.opt and
3603 nds32-linux.opt files.
3604 * config/nds32/t-elf: New file.
3605 * config/nds32/t-linux: New file.
3607 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3608 Shiva Chen <shiva0217@gmail.com>
3610 * config/nds32/constants.md (unspec_volatile_element): Add
3611 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3612 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3614 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3615 (make_pass_nds32_fp_as_gp): Declare.
3616 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3618 (nds32_asm_function_end_prologue): Remove unused asm output.
3619 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3620 (nds32_asm_file_start): Output necessary fp_as_gp information.
3621 (nds32_option_override): Adjust register usage.
3622 (nds32_expand_prologue): Consider fp_as_gp situation.
3623 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3624 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3627 (simple_return): Ditto.
3628 (omit_fp_begin): Output special directive for fp_as_gp.
3629 (omit_fp_end): Output special directive for fp_as_gp.
3630 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3631 mforbid-fp-as-gp): New options.
3633 2018-06-01 Mark Wielaard <mark@klomp.org>
3635 * dwarf2out.c (dwarf2out_finish): Remove generation of
3636 DW_AT_loclists_base.
3638 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3640 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3641 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3642 (struct merged_store_group): Add bit_insertion field.
3643 (dump_char_array): Use standard hexadecimal format.
3644 (merged_store_group::merged_store_group): Set bit_insertion to false.
3645 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3646 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3647 also print the mask in the dump file.
3648 (pass_store_merging::gate): Minor tweak.
3649 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3650 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3651 stores with INTEGER_CST stores.
3652 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3653 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3654 and use it throughout. Generate bit insertion sequences if need be.
3655 (pass_store_merging::process_store): Remove redundant condition.
3656 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3658 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3660 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3661 the 128-bit floating point types. Fix function comment.
3663 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3665 * config/aarch64/aarch64-simd.md
3666 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3668 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3671 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3673 PR tree-optimization/85989
3674 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3676 (backprop::intersect_uses): Check it when deciding whether this
3677 is a backedge reference.
3678 (backprop::process_block): Add each phi to m_visited_phis
3679 after visiting it, then clear it at the end.
3681 2018-06-01 Richard Biener <rguenther@suse.de>
3683 * tree-vectorizer.h (vect_dr_stmt): New function.
3684 (vect_get_load_cost): Adjust.
3685 (vect_get_store_cost): Likewise.
3686 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3687 Use vect_dr_stmt instead of DR_SMTT.
3688 (vect_record_base_alignments): Likewise.
3689 (vect_calculate_target_alignment): Likewise.
3690 (vect_compute_data_ref_alignment): Likewise and make static.
3691 (vect_update_misalignment_for_peel): Likewise.
3692 (vect_verify_datarefs_alignment): Likewise.
3693 (vector_alignment_reachable_p): Likewise.
3694 (vect_get_data_access_cost): Likewise. Pass down
3695 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3696 (vect_get_peeling_costs_all_drs): Likewise.
3697 (vect_peeling_hash_get_lowest_cost): Likewise.
3698 (vect_enhance_data_refs_alignment): Likewise.
3699 (vect_find_same_alignment_drs): Likewise.
3700 (vect_analyze_data_refs_alignment): Likewise.
3701 (vect_analyze_group_access_1): Likewise.
3702 (vect_analyze_group_access): Likewise.
3703 (vect_analyze_data_ref_access): Likewise.
3704 (vect_analyze_data_ref_accesses): Likewise.
3705 (vect_vfa_segment_size): Likewise.
3706 (vect_small_gap_p): Likewise.
3707 (vectorizable_with_step_bound_p): Likewise.
3708 (vect_prune_runtime_alias_test_list): Likewise.
3709 (vect_analyze_data_refs): Likewise.
3710 (vect_supportable_dr_alignment): Likewise.
3711 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3712 (vect_gen_prolog_loop_niters): Likewise.
3713 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3714 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3716 (vect_recog_mask_conversion_pattern): Likewise.
3717 (vect_try_gather_scatter_pattern): Likewise.
3718 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3719 to vect_get_store_cost.
3720 (vect_get_store_cost): Get stmt_info instead of DR.
3721 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3722 (vect_get_load_cost): Get stmt_info instead of DR.
3724 2018-06-01 Richard Biener <rguenther@suse.de>
3727 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3728 vars wrapped in MEM_REFs.
3730 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3732 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3733 Fix subreg tests so that we only return a choice between
3734 GENERAL_REGS and FP_REGS if the original classes included both.
3736 2018-06-01 Richard Biener <rguenther@suse.de>
3739 * tree-ssa-structalias.c (get_function_part_constraint):
3740 Handle NULL fi->decl.
3741 (find_func_aliases_for_call): Properly handle indirect
3742 fi from direct call.
3743 (find_func_clobbers): Likewise.
3744 (ipa_pta_execute): Likewise.
3745 (create_variable_info_for): For functions that are ifunc_resolver
3746 resolve to a varinfo that contains the result of the resolver call.
3747 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3750 2018-05-31 Michael Collison <michael.collison@arm.com>
3752 * config/aarch64/aarch64.md:
3753 (*fix_to_zero_extenddfdi2): New pattern.
3754 * gcc.target/aarch64/fix_extend1.c: New testcase.
3756 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3760 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3761 and BUILT_IN_STRNCMP_EQ.
3762 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3763 BUILT_IN_STRNCMP_EQ.
3764 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3765 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3766 (gimple_fold_builtin): Likewise.
3767 * tree-ssa-strlen.c (compute_string_length): New function.
3768 (determine_min_obsize): New function.
3769 (handle_builtin_string_cmp): New function to handle calls to
3770 string compare functions.
3771 (strlen_optimize_stmt): Add handling to builtin string compare
3773 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3774 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3775 * tree.c (build_common_builtin_nodes): Add new defines of
3776 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3778 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3781 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3782 functions with naked attribute.
3784 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3786 * config/i386/sse.md (avx_vec_concat<mode>):
3787 Substitute concat_tg_mode mode attribute with xtg_mode.
3788 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3789 (concat_tg_mode): Remove mode attribute.
3791 2018-05-31 Martin Sebor <msebor@redhat.com>
3794 * calls.c (alloc_max_size): Correct a logic error/typo.
3795 Treat excessive arguments as infinite. Warn for invalid arguments.
3796 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3798 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3801 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3802 and movx for Haswell.
3804 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3805 Cesar Philippidis <cesar@codesourcery.com>
3808 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3809 when emitting error on private/firstprivate reductions.
3810 * omp-low.c (lower_omp_target): Avoid reference-type processing
3811 on pointers for firstprivate clause.
3813 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3815 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3818 * config/aarch64/aarch64-simd.md
3819 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3820 (aarch64_ld1_x3_<mode>): Likewise
3821 (aarch64_st1x2<VALLDIF:mode>): Likewise
3822 (aarch64_st1_x2_<mode>): Likewise
3823 (aarch64_st1x3<VALLDIF:mode>): Likewise
3824 (aarch64_st1_x3_<mode>): Likewise
3825 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3826 (vld1_s8_x3): Likewise.
3827 (vld1_u16_x3): Likewise.
3828 (vld1_s16_x3): Likewise.
3829 (vld1_u32_x3): Likewise.
3830 (vld1_s32_x3): Likewise.
3831 (vld1_u64_x3): Likewise.
3832 (vld1_s64_x3): Likewise.
3833 (vld1_f16_x3): Likewise.
3834 (vld1_f32_x3): Likewise.
3835 (vld1_f64_x3): Likewise.
3836 (vld1_p8_x3): Likewise.
3837 (vld1_p16_x3): Likewise.
3838 (vld1_p64_x3): Likewise.
3839 (vld1q_u8_x3): Likewise.
3840 (vld1q_s8_x3): Likewise.
3841 (vld1q_u16_x3): Likewise.
3842 (vld1q_s16_x3): Likewise.
3843 (vld1q_u32_x3): Likewise.
3844 (vld1q_s32_x3): Likewise.
3845 (vld1q_u64_x3): Likewise.
3846 (vld1q_s64_x3): Likewise.
3847 (vld1q_f16_x3): Likewise.
3848 (vld1q_f32_x3): Likewise.
3849 (vld1q_f64_x3): Likewise.
3850 (vld1q_p8_x3): Likewise.
3851 (vld1q_p16_x3): Likewise.
3852 (vld1q_p64_x3): Likewise.
3853 (vst1_s64_x2): Likewise.
3854 (vst1_u64_x2): Likewise.
3855 (vst1_f64_x2): Likewise.
3856 (vst1_s8_x2): Likewise.
3857 (vst1_p8_x2): Likewise.
3858 (vst1_s16_x2): Likewise.
3859 (vst1_p16_x2): Likewise.
3860 (vst1_s32_x2): Likewise.
3861 (vst1_u8_x2): Likewise.
3862 (vst1_u16_x2): Likewise.
3863 (vst1_u32_x2): Likewise.
3864 (vst1_f16_x2): Likewise.
3865 (vst1_f32_x2): Likewise.
3866 (vst1_p64_x2): Likewise.
3867 (vst1q_s8_x2): Likewise.
3868 (vst1q_p8_x2): Likewise.
3869 (vst1q_s16_x2): Likewise.
3870 (vst1q_p16_x2): Likewise.
3871 (vst1q_s32_x2): Likewise.
3872 (vst1q_s64_x2): Likewise.
3873 (vst1q_u8_x2): Likewise.
3874 (vst1q_u16_x2): Likewise.
3875 (vst1q_u32_x2): Likewise.
3876 (vst1q_u64_x2): Likewise.
3877 (vst1q_f16_x2): Likewise.
3878 (vst1q_f32_x2): Likewise.
3879 (vst1q_f64_x2): Likewise.
3880 (vst1q_p64_x2): Likewise.
3881 (vst1_s64_x3): Likewise.
3882 (vst1_u64_x3): Likewise.
3883 (vst1_f64_x3): Likewise.
3884 (vst1_s8_x3): Likewise.
3885 (vst1_p8_x3): Likewise.
3886 (vst1_s16_x3): Likewise.
3887 (vst1_p16_x3): Likewise.
3888 (vst1_s32_x3): Likewise.
3889 (vst1_u8_x3): Likewise.
3890 (vst1_u16_x3): Likewise.
3891 (vst1_u32_x3): Likewise.
3892 (vst1_f16_x3): Likewise.
3893 (vst1_f32_x3): Likewise.
3894 (vst1_p64_x3): Likewise.
3895 (vst1q_s8_x3): Likewise.
3896 (vst1q_p8_x3): Likewise.
3897 (vst1q_s16_x3): Likewise.
3898 (vst1q_p16_x3): Likewise.
3899 (vst1q_s32_x3): Likewise.
3900 (vst1q_s64_x3): Likewise.
3901 (vst1q_u8_x3): Likewise.
3902 (vst1q_u16_x3): Likewise.
3903 (vst1q_u32_x3): Likewise.
3904 (vst1q_u64_x3): Likewise.
3905 (vst1q_f16_x3): Likewise.
3906 (vst1q_f32_x3): Likewise.
3907 (vst1q_f64_x3): Likewise.
3908 (vst1q_p64_x3): Likewise.
3910 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3912 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3913 user_label_prefix to name.
3915 * tree-core.h: Update comment about the format of NAME string
3916 passed to handler in attribute_spec.
3918 * config/msp430/msp430.md: Remove erroneous subreg expression from
3919 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3920 zero_extend{q,h}isi2.
3922 2018-05-30 Borislav Petkov <bp@suse.de>
3924 * doc/extend.texi: Document some architecture specific
3925 constraints and sort entries.
3927 2018-05-30 Martin Sebor <msebor@redhat.com>
3930 * builtins.c (expand_builtin_stpcpy_1): New function.
3931 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3932 only if the former succeeds.
3934 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3936 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3939 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3941 * doc/invoke.texi (-flinker-output): Document
3943 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3945 * passes.c (ipa_write_summaries): Only modify statements if body
3947 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3948 incrementally linking.
3949 (ipa_passes): Likewise.
3950 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3951 pass down resolution info.
3952 * common.opt (flag_incremental_link): Update info.
3953 * gcc.c (plugin specs): Turn flinker-output=* to
3954 -plugin-opt=-linker-output-known
3955 * toplev.c (compile_file): Also cut compilation when doing incremental
3957 * flag-types. (enum lto_partition_model): Add
3958 LTO_LINKER_OUTPUT_NOLTOREL.
3959 (invoke.texi): Add -flinker-output docs.
3960 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3961 link same way as WPA; do not stream in dead initializers.
3963 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3964 darf2out_register_external_decl): Support incremental link.
3966 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3968 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3970 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3972 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3973 it down to simple_object_copy_lto_debug_sections.
3974 (run_gcc): Determine incremental LTO link time and configure
3975 lto1 into non-wpa mode, disable renaming of debug sections.
3977 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3979 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3980 descriptions of various incorrectly documented functions.
3982 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3985 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3986 address check not strict.
3988 2018-05-30 Richard Biener <rguenther@suse.de>
3990 PR tree-optimization/85964
3991 * tracer.c (better_p): Drop initialized count check, we only
3992 call the function with initialized counts now.
3993 (find_best_successor): Do find a best edge if one
3994 has uninitialized count.
3995 (find_best_predecessor): Likewise. Do BB frequency check only
3996 if count is initialized.
3998 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4000 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4001 (aarch64_ldrstr_offset_compare): New.
4002 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4003 load/store orderings.
4004 (aarch64_gen_adjusted_ldpstp): Likewise.
4006 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4008 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4009 Check for subset of GENERAL_REGS and FP_REGS.
4010 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4013 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4015 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4016 and wi::to_poly_offset. Add the current offset and then check
4017 whether the sum fits, rather than using an unchecked addition of
4018 a checked term. Check for a shwi rather than a uhwi.
4019 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4020 (store_constructor): Use poly_int_tree_p.
4021 (expand_expr_real_1): Likewise.
4022 * function.c (assign_temp): Likewise.
4023 * fold-const.c (const_binop): Use poly_int_tree_p and
4025 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4027 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4028 to_poly_offset for MEM offsets.
4029 * ipa-icf.c (sem_variable::equals): Likewise.
4030 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4031 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4032 wi::to_poly_offset for BIT_FIELD_REF offsets.
4033 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4035 * var-tracking.c (emit_note_insn_var_location): Use
4036 tree_to_poly_uint64.
4038 2018-05-29 Jim Wilson <jimw@sifive.com>
4040 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4042 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4045 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4046 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4048 (sse4_1_round<mode>2): Use nonimmediate_operand
4049 for operand 1 predicate.
4051 2018-05-29 Martin Sebor <msebor@redhat.com>
4052 Richard Biener <rguenther@suse.de>
4055 * calls.c (get_size_range): Call determine_value_range instead
4056 of get_value_range..
4057 * tree-vrp.h (determine_value_range): Declared new function.
4058 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4060 2018-05-29 Richard Biener <rguenther@suse.de>
4062 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4063 sure to use non-pattern stmts for get_earlier_stmt arguments.
4064 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4065 called on pattern stmts.
4066 (get_later_stmt): Likewise.
4068 2018-05-29 Martin Liska <mliska@suse.cz>
4070 PR gcov-profile/85759
4071 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4074 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4076 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4078 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4082 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4083 VEC_PACK_FLOAT_EXPR): New tree codes.
4084 * tree-pretty-print.c (op_code_prio): Handle
4085 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4086 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4087 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4088 * tree-inline.c (estimate_operator_cost): Likewise.
4089 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4090 * fold-const.c (const_binop): Likewise.
4091 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4092 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4093 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4094 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4095 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4096 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4097 * expr.c (expand_expr_real_2): Likewise.
4098 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4099 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4100 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4102 * optabs.c (expand_widen_pattern_expr): For
4103 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4104 sign from result type rather than operand's type.
4105 (expand_binop_directly): For vec_packu_float_optab and
4106 vec_packs_float_optab allow result type to be different from operand's
4108 * optabs-tree.c (optab_for_tree_code): Handle
4109 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4110 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4111 * tree-vect-generic.c (expand_vector_operations_1): Handle
4112 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4113 VEC_PACK_FLOAT_EXPR.
4114 * tree-vect-stmts.c (supportable_widening_operation): Handle
4116 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4117 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4118 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4119 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4120 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4122 (vec_pack<floatprefix>_float_<mode>): New expander.
4123 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4125 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4126 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4127 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4128 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4129 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4131 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4132 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4133 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4134 VEC_PACK_FLOAT_EXPR): Document.
4136 2018-05-29 Richard Biener <rguenther@suse.de>
4138 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4140 (stmt_vec_info_vec): Make pointer.
4141 (init_stmt_vec_info_vec): Remove.
4142 (free_stmt_vec_info_vec): Likewise.
4143 (set_stmt_vec_info_vec): New function.
4144 (free_stmt_vec_infos): Likewise.
4145 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4146 (set_vinfo_for_stmt): Likewise.
4147 (get_earlier_stmt): Likewise.
4148 (get_later_stmt): Likewise.
4149 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4150 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4151 (vec_info::~vec_info): Free stmt_vec_infos.
4152 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4153 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4154 (pass_slp_vectorize::execute): Likewise.
4155 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4156 (free_stmt_vec_info_vec): Likewise.
4157 (set_stmt_vec_info_vec): New function.
4158 (free_stmt_vec_infos): Likewise.
4159 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4160 the global stmt_vec_info_vec.
4161 * tree-parloops.c (gather_scalar_reductions): Use
4162 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4165 2018-05-29 Richard Biener <rguenther@suse.de>
4167 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4169 2018-05-29 Martin Liska <mliska@suse.cz>
4170 David Malcolm <dmalcolm@redhat.com>
4172 * vec.c (test_reverse): New.
4173 (vec_c_tests): Add new test.
4174 * vec.h (vl_ptr>::reverse): New function.
4176 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4178 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4180 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4183 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4185 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4187 2018-05-28 Richard Biener <rguenther@suse.de>
4189 PR tree-optimization/85933
4190 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4191 look at stmts marked as vectorizable.
4193 2018-05-28 Richard Biener <rguenther@suse.de>
4195 PR tree-optimization/85934
4196 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4197 vector boolean check before scalar optimization.
4199 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4201 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4204 2018-05-28 Mark Wielaard <mark@klomp.org>
4206 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4207 if it is an expression containing a minus sign.
4209 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4211 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4213 2018-05-27 Paul Koning <ni1d@arrl.net>
4215 * config/pdp11/pdp11.md (truncsihi2): Remove.
4217 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4218 Chung-Ju Wu <jasonwucj@gmail.com>
4220 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4222 (unaligned_store_dw): Ditto.
4223 * config/nds32/nds32-memory-manipulation.c
4224 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4225 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4226 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4227 (emit_setmem_word_loop): Rename to ...
4228 (emit_setmem_doubleword_loop): ... this.
4229 (nds32_gen_dup_4_byte_to_word_value): New function.
4230 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4231 (nds32_expand_setmem_loop): Refine implementation.
4232 (nds32_expand_setmem_loop_v3m): Ditto.
4233 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4236 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4238 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4240 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4242 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4243 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4244 machine->attr_no_prologue_p.
4245 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4246 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4247 (nds32_expand_epilogue): Consider attr_naked_p.
4248 (nds32_expand_epilogue_v3pop): Likewise.
4249 (nds32_can_use_return_insn): Likewise.
4250 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4251 attr_no_prologue_p fields.
4252 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4254 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4257 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4259 * config/i386/sse.md
4260 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4262 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4264 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4266 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4268 (*<floatsuffix>floatv2div2sf2): Rename to ...
4269 (*float<floatunssuffix>v2div2sf2): ... this.
4270 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4271 (float<floatunssuffix>v2div2sf2_mask): ... this.
4272 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4273 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4274 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4276 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4278 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4280 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4282 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4284 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4286 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4287 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4288 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4289 gen_ufix_truncv8dfv8si2.
4290 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4291 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4292 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4293 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4294 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4295 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4296 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4297 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4299 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4303 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4305 2018-05-25 Jim Wilson <jimw@sifive.com>
4307 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4308 * config/riscv/riscv.c (struct machine_function): Add
4309 interrupt_handler_p and attribute_checked_p fields.
4310 (riscv_attribute_table): Add interrupt.
4311 (riscv_interrupt_type_p): New.
4312 (riscv_save_reg_p): Save extra regs for interrupt handler.
4313 (riscv_use_save_libcall): Return false for interrupt handler.
4314 (riscv_first_stack_step): Add forward declaration.
4315 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4316 for interrupt handler with large frame. Use it for saved reg list.
4317 (riscv_expand_prologue): Move flag_stack_usage_info support to
4318 eliminate duplication.
4319 (riscv_expand_epilogue): Generate mret for interrupt handler.
4320 (riscv_epilogue_uses): New.
4321 (riscv_can_use_return_insn): Return false for interrupt handler.
4322 (riscv_function_ok_for_sibcall): Likewise.
4323 (riscv_set_current_function): Add interrupt handler support.
4324 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4325 * config/riscv/riscv.md (UNSPECV_MRET): New.
4327 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4329 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4331 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4333 PR tree-optimization/85712
4334 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4335 this candidate has already been replaced in-situ by a copy.
4337 2018-05-25 Jason Merrill <jason@redhat.com>
4339 PR c++/80485 - inline function non-zero address.
4340 * symtab.c (nonzero_address): Check DECL_COMDAT.
4342 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4345 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4346 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4347 extension of SImode operation. Use const123_operand predicate.
4348 (*saddsi_1): Remove.
4349 (*saddl_se_1): Ditto.
4351 (*ssubl_se_1): Ditto.
4352 * config/alpha/predicates.md (const123_operand): New predicate.
4353 * config/alpha/constraints.md (P): Use IN_RANGE.
4355 2018-05-25 Richard Biener <rguenther@suse.de>
4357 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4359 (ref_maybe_used_by_stmt_p): Likewise.
4360 (stmt_may_clobber_ref_p): Likewise.
4361 (stmt_may_clobber_ref_p_1): Likewise.
4362 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4364 (ref_maybe_used_by_stmt_p): Likewise.
4365 (stmt_may_clobber_ref_p): Likewise.
4366 (stmt_may_clobber_ref_p_1): Likewise.
4367 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4368 the alias oracle to disambiguate DRs with stmts DR analysis
4370 (vect_analyze_data_refs): Do not give up on not analyzable
4371 DRs for BB vectorization. Remove code truncating the dataref
4374 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4377 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4378 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4379 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4381 2018-05-25 Richard Biener <rguenther@suse.de>
4383 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4384 function, combining stmt data ref gathering and fatal analysis
4386 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4387 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4388 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4389 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4390 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4392 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4394 PR tree-optimization/85720
4395 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4396 SCC if all partitions are builtins.
4397 (version_loop_by_alias_check): New parameter. Generate cancelable
4398 runtime alias check if all partitions are builtins.
4399 (distribute_loop): Update call to above function.
4401 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4403 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4404 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4405 (parm_default_def_partition_arg): Ditto.
4406 (set_parm_default_def_partition): Ditto.
4407 (get_parm_default_def_partitions): Ditto and make it static.
4408 (get_undefined_value_partitions): Ditto and make it static.
4409 (remove_ssa_form): Refactor call to init_var_map here.
4410 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4411 computation for loop region.
4412 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4413 (register_default_def): Delete.
4414 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4415 (parm_default_def_partition_arg): Ditto.
4416 (set_parm_default_def_partition): Ditto.
4417 (get_parm_default_def_partitions): Ditto and make it static.
4418 (get_undefined_value_partitions): Ditto and make it static.
4419 (coalesce_with_default, coalesce_with_default): Update comment.
4420 (create_coalesce_list_for_region): New func factored out from
4421 create_outofssa_var_map.
4422 (populate_coalesce_list_for_outofssa): New func factored out from
4423 create_outofssa_var_map and coalesce_ssa_name.
4424 (create_outofssa_var_map): Delete.
4425 (coalesce_ssa_name): Refactor to support live range computation.
4426 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4427 (get_parm_default_def_partitions): Delete.
4428 (get_undefined_value_partitions): Ditto.
4429 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4430 computation for loop region.
4431 (new_tree_live_info, loe_visit_block): Ditto.
4432 (live_worklist, set_var_live_on_entry): Ditto.
4433 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4434 * tree-ssa-live.h (struct _var_map): New fields.
4435 (init_var_map): Change decl.
4436 (region_contains_p): New.
4438 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4440 * tree-ssa-live.h (live_merge_and_clear): Delete.
4442 2018-05-25 Richard Biener <rguenther@suse.de>
4445 * tree-dump.c (dequeue_and_dump): Remove access to removed
4446 operand 2 of a SWITCH_EXPR.
4448 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4450 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4451 multiplication and division.
4452 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4453 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4454 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4455 (cond_udiv_optab, cond_umod_optab): New optabs.
4456 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4457 (IFN_COND_RDIV): New internal functions.
4458 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4459 TRUNC_MOD_EXPR and RDIV_EXPR.
4460 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4461 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4463 (SVE_INT_BINARY): Include mult.
4464 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4465 (optab, sve_int_op): Handle mult.
4466 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4468 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4469 for SVE_INT_BINARY_SD.
4471 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4473 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4474 (optab, sve_int_op): Handle div and udiv.
4475 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4476 for SVE_INT_BINARY_SD.
4477 (*<optab><mode>3): New insn for the same.
4479 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4481 * tree-vect-patterns.c: Include predict.h.
4482 (vect_recog_divmod_pattern): Restrict check for division support
4483 to when optimizing for size.
4485 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4487 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4488 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4489 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4490 (gimple_match_op::set_op): Likewise.
4491 (gimple_resimplify4): Declare.
4492 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4493 (expr::gen_transform): Likewise.
4494 (decision_tree::gen): Generate a simplification routine for 4 operands.
4495 * gimple-match-head.c (gimple_simplify): Add an overload for
4496 4 operands. In the top-level function, handle up to 4 call
4497 arguments and call gimple_resimplify4.
4498 (gimple_resimplify4): New function.
4499 (build_call_internal): Pass a fourth operand.
4500 (maybe_push_to_seq): Likewise.
4501 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4502 Fold VEC_COND_EXPRs of an operation and a default value into
4503 an IFN_COND_* function if possible.
4504 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4506 (SVE_COND_FP_BINARY): Include them.
4507 (optab, sve_fp_op): Handle them.
4508 (SVE_INT_BINARY_REV): New code iterator.
4509 (SVE_COND_FP_BINARY_REV): New int iterator.
4510 (commutative): New int attribute.
4511 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4513 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4515 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4516 (*cond_<optab><mode>): New patterns for reversed operands.
4518 2018-05-25 Richard Biener <rguenther@suse.de>
4520 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4521 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4522 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4523 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4524 * tree-vect-data-refs.c (everywhere): Adjust users.
4525 * tree-vect-loop.c (everywhere): Likewise.
4526 * tree-vect-slp.c (everywhere): Likewise.
4527 * tree-vect-stmts.c (everywhere): Likewise.
4528 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4530 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4532 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4534 (gcc_cv_as_section_exclude): ... this.
4535 Try Solaris as #exclude syntax.
4536 * configure: Regenerate.
4537 * config.in: Regenerate.
4538 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4540 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4541 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4543 * varasm.c (default_elf_asm_named_section): Don't check if
4544 HAVE_GAS_SECTION_EXCLUDE is defined.
4546 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4548 * doc/md.texi: Update the documentation of the cond_* optabs
4549 to mention the new final operand. Fix GET_MODE_NUNITS call.
4550 Describe the scalar case too.
4551 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4552 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4554 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4555 (get_conditional_internal_fn): Update comment.
4556 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4557 accumulator value as a final argument to conditional functions.
4558 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4559 a define_expand and add an "else" operand. Assert for now that
4560 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4561 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4562 (*cond_<optab><mode>): New patterns.
4563 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4564 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4565 (UNSPEC_COND_EOR): Delete.
4566 (optab): Remove associated mappings.
4567 (SVE_INT_BINARY): New code iterator.
4568 (sve_int_op): Remove int attribute and add "minus" to the code
4570 (SVE_COND_INT_OP): Delete.
4571 (SVE_COND_FP_OP): Rename to...
4572 (SVE_COND_FP_BINARY): ...this.
4574 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4576 * optabs.c (can_reuse_operands_p): New function.
4577 (maybe_legitimize_operands): Try to reuse the results for
4580 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4582 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4583 Add {q} suffix to insn mnemonic.
4585 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4587 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4588 (msp430_warn_func_return): New.
4590 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4592 * fold-const.c (tree_nonzero_bits): New function.
4593 * fold-const.h (tree_nonzero_bits): Likewise.
4594 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4595 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4597 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4601 * varasm.c (assemble_alias): Check ifunc_resolver only on
4604 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4607 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4608 when memory input operand is handled.
4610 2018-05-24 Luis Machado <luis.machado@linaro.org>
4612 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4614 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4616 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4618 * match.pd: Delay FMA folds until after vectorization.
4620 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4623 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4624 address check not strict.
4626 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4628 * gimple-match.h (gimple_match_op): New class.
4629 (mprts_hook): Replace parameters with a gimple_match_op *.
4630 (maybe_build_generic_op): Likewise.
4631 (gimple_simplified_result_is_gimple_val): Replace parameters with
4632 a const gimple_match_op *.
4633 (gimple_simplify): Replace code_helper * and tree * parameters with
4634 a gimple_match_op * parameter.
4635 (gimple_resimplify1): Replace code_helper *, tree and tree *
4636 parameters with a gimple_match_op * parameter.
4637 (gimple_resimplify2): Likewise.
4638 (gimple_resimplify3): Likewise.
4639 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4640 parameters with a gimple_match_op * parameter.
4641 * gimple-match-head.c (gimple_simplify): Change prototypes of
4642 auto-generated functions to take a gimple_match_op * instead of
4643 separate code_helper * and tree * parameters. Make the same
4644 change in the top-level overload and update calls to the
4645 gimple_resimplify routines. Update calls to the auto-generated
4646 functions and to maybe_push_res_to_seq in the publicly-facing
4647 operation-specific gimple_simplify overloads.
4648 (gimple_match_op::MAX_NUM_OPS): Define.
4649 (gimple_resimplify1): Replace rcode and ops with a single res_op
4650 parameter. Update call to gimple_simplify.
4651 (gimple_resimplify2): Likewise.
4652 (gimple_resimplify3): Likewise.
4653 (mprts_hook): Replace parameters with a gimple_match_op *.
4654 (maybe_build_generic_op): Likewise.
4655 (build_call_internal): Replace type, nargs and ops with
4656 a gimple_match_op *.
4657 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4658 with a single gimple_match_op *. Update calls to mprts_hook,
4659 build_call_internal and gimple_simplified_result_is_gimple_val.
4660 Factor out code that is common to the tree_code and combined_fn cases.
4661 * genmatch.c (expr::gen_transform): Replace tem_code and
4662 tem_ops with a gimple_match_op called tem_op. Update calls
4663 to the gimple_resimplify functions and maybe_push_res_to_seq.
4664 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4665 res_ops. Update call to the gimple_resimplify functions.
4666 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4667 (decision_tree::gen): Make the functions take a gimple_match_op *
4668 called res_op instead of separate res_code and res_ops parameters.
4669 Update call accordingly.
4670 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4671 and ops with a single res_op parameter. Update calls to
4672 maybe_build_generic_op and maybe_push_res_to_seq.
4673 (fold_stmt_1): Update calls to gimple_simplify and
4674 replace_stmt_with_simplification.
4675 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4676 and gimple_simplified_result_is_gimple_val.
4677 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4679 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4680 with a gimple_match_op *.
4681 (vn_nary_build_or_lookup): Likewise. Update call to
4682 vn_nary_build_or_lookup_1.
4683 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4684 gimple_match_op *. Update calls to the gimple_resimplify routines
4685 and to gimple_simplified_result_is_gimple_val.
4686 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4687 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4688 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4689 (visit_nary_op): Likewise.
4690 (visit_reference_op_load): Likewise.
4692 2018-05-23 Luis Machado <luis.machado@linaro.org>
4694 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4695 modifier for printing the step amount.
4697 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4700 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4703 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4705 * doc/sourcebuild.texi (Endianness): New subsubsection.
4707 2018-05-23 Luis Machado <luis.machado@linaro.org>
4709 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4710 <prefetch_dynamic_strides>: New const bool field.
4711 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4712 prefetch_dynamic_strides.
4713 (exynosm1_prefetch_tune): Likewise.
4714 (thunderxt88_prefetch_tune): Likewise.
4715 (thunderx_prefetch_tune): Likewise.
4716 (thunderx2t99_prefetch_tune): Likewise.
4717 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4719 (aarch64_override_options_internal): Update to set
4720 PARAM_PREFETCH_DYNAMIC_STRIDES.
4721 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4722 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4723 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4724 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4725 prefetch-dynamic-strides setting.
4727 2018-05-23 Luis Machado <luis.machado@linaro.org>
4729 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4730 <minimum_stride>: New const int field.
4731 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4732 minimum_stride field defaulting to -1.
4733 (exynosm1_prefetch_tune): Likewise.
4734 (thunderxt88_prefetch_tune): Likewise.
4735 (thunderx_prefetch_tune): Likewise.
4736 (thunderx2t99_prefetch_tune): Likewise.
4737 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4738 <default_opt_level>: Set to 3.
4739 (aarch64_override_options_internal): Update to set
4740 PARAM_PREFETCH_MINIMUM_STRIDE.
4741 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4742 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4743 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4744 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4745 stride is constant and is below the minimum stride threshold.
4747 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4749 * config/arm/arm-cpus.in (mode26): Delete.
4750 (armv4): Delete mode26 reference.
4751 * config/arm/arm.c (arm_configure_build_target): Delete use of
4754 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4756 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4758 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4759 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4761 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4762 (floatunsdidf2): Ditto.
4764 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4766 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4767 (fixuns_trunc<mode>si2_avx512f): Ditto.
4768 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4769 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4770 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4772 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4774 PR rtl-optimization/79985
4775 * df-scan.c (df_insn_refs_collect): Remove special case for
4776 global registers and asm statements.
4778 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4780 * extend.texi (Global Register Variables): Rewrite the bullet list.
4781 Note that the register is available for allocation. Note that access
4782 via inline asm must use constraints. Add note about async-signal
4783 handlers. Remove paragraph about automagic register selection.
4785 2018-05-23 Richard Biener <rguenther@suse.de>
4787 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4788 of fixed offset from memset VN.
4790 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4792 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4794 (alloc_cand_and_find_basis): Initialize first_interp field.
4795 (slsr_process_mul): Modify first_interp field.
4796 (slsr_process_add): Likewise.
4797 (slsr_process_cast): Modify first_interp field for each new
4799 (slsr_process_copy): Likewise.
4800 (dump_candidate): Dump first_interp field.
4801 (replace_mult_candidate): Process all interpretations, not just
4803 (replace_rhs_if_not_dup): Likewise.
4804 (replace_one_candidate): Likewise.
4806 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4808 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4810 (aarch64_needs_frame_chain): New function.
4811 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4813 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4816 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4817 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4818 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4819 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4820 as true for strict-align.
4821 (aarch64_can_inline_p): Perform checks even when callee has no
4822 attributes to check for strict alignment.
4823 * doc/extend.texi (AArch64 Function Attributes): Document
4825 * doc/invoke.texi: (AArch64 Options): Likewise.
4827 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4829 PR tree-optimization/85853
4830 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4831 the handling of the root of the node to...
4832 (vect_slp_analyze_node_operations_1): ...this new function,
4833 and run the whole thing with the child nodes' def types
4834 set according to their SLP node's def type.
4836 2018-05-23 Richard Biener <rguenther@suse.de>
4839 * tree-data-ref.c (create_runtime_alias_checks): Defer
4840 and ignore overflow warnings.
4842 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4844 PR tree-optimization/85822
4845 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4848 2018-05-23 Richard Biener <rguenther@suse.de>
4850 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4851 memset constants via native_interpret_expr.
4853 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4856 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4858 (cgraph_node::create_alias): Likewise.
4859 (cgraph_node::get_availability): Check ifunc_resolver instead
4860 of looking up ifunc attribute.
4861 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4862 * varasm.c (do_assemble_alias): Likewise.
4863 (assemble_alias): Likewise.
4864 (default_binds_local_p_3): Likewise.
4865 * cgraph.h (cgraph_node): Add ifunc_resolver.
4866 (cgraph_node::only_called_directly_or_aliased_p): Return false
4868 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4870 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4871 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4872 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4873 instead of looking up ifunc attribute.
4875 2018-05-22 Luis Machado <luis.machado@linaro.org>
4877 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4879 2018-05-22 Martin Sebor <msebor@redhat.com>
4882 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4883 only when expasion succeeds.
4884 (expand_builtin_strcmp): Same.
4885 (expand_builtin_strncmp): Same.
4887 2018-05-22 Martin Sebor <msebor@redhat.com>
4889 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4891 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4892 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4894 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4895 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4897 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4898 Add check for legitimate_address.
4899 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4900 (aarch64_swap_ldrstr_operands): New.
4901 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4904 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4905 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4907 * config/aarch64/aarch64.md: New patterns to generate stp
4909 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4910 single words and double words.
4911 (load_pair_sw, load_pair_dw): Likewise.
4912 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4914 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4916 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4917 for different mode ldpstp and add peephole for merged zero stores.
4919 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4921 (aarch64_gen_store_pair): Rename calls to match new patterns.
4922 (aarch64_gen_load_pair): Rename calls to match new patterns.
4923 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4924 (load_pair<DREG:mode><DREG2:mode>): ... This.
4925 (store_pair<mode>): Rename to...
4926 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4927 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4929 (V_INT_EQUIV): Handle SImode.
4930 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4933 2018-05-22 Martin Sebor <msebor@redhat.com>
4936 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4937 or ajust the presumed bound on an operation to avoid unnecessary
4940 2018-05-22 Martin Sebor <msebor@redhat.com>
4942 PR tree-optimization/85826
4943 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4944 assuming that a DECL necesarily has a constant size.
4946 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4949 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4951 2018-05-22 Richard Biener <rguenther@suse.de>
4953 PR tree-optimization/85834
4954 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4955 non-constant and non-zero memset arguments.
4957 2018-05-22 Martin Liska <mliska@suse.cz>
4960 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4962 2018-05-22 Richard Biener <rguenther@suse.de>
4964 PR tree-optimization/85863
4965 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4966 comparisons when vectype is specified.
4967 (vectorizable_condition): Do not specify vectype for
4968 vect_is_simple_cond when SLP vectorizing.
4970 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4973 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4974 define __ibm128 as long double.
4975 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4976 as a distinct type when IEEE 128-bit support is enabled.
4977 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4978 128-bit types to use the correct functions.
4979 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4980 convert between 128-bit floating point types that have different
4981 modes but the same representation, instead of using gen_lowpart to
4983 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4985 (IFKF_reg): New attributes to give the register constraints for
4987 (extend<mode>tf2_internal): New insns to mark an explicit
4988 conversion between 128-bit floating point types that have a
4989 different mode but share the same representation.
4991 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4993 PR tree-optimization/85814
4994 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4995 a null return from get_strinfo when unsharing the next
4996 strinfo in the chain.
4998 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5001 * varasm.c (weak_finish): Clean up weak_decls.
5003 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5005 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5006 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5007 UNSPEC_UADALP values.
5008 * config/aarch64/iterators.md (ABAL): New int iterator.
5011 (sur): Add mappings for the above.
5012 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5014 (aarch64_<sur>abal<mode>_4): Likewise.
5015 (aarch64_<sur>adalp<mode>_3): Likewise.
5016 (<sur>sadv16qi): New define_expand.
5018 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5020 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5021 (*movdf_internal): Ditto.
5022 (*rcpsf2_sse): Ditto.
5023 (*rsqrtsf2_sse): Ditto.
5024 (*sqrt<mode>2_sse): Ditto.
5026 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5028 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5030 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5031 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5032 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5033 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5035 * config/aarch64/arm_neon.h: Likewise.
5036 * config/aarch64/iterators.md (VQ_I): New.
5038 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5040 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5042 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5045 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5047 * config/nds32/constraints.md (S): New constraint.
5048 * config/nds32/nds32.md (call_internal): Use constraint S.
5049 (call_value_internal): Likewise.
5050 (sibcall_internal): Likewise.
5051 (sibcall_value_internal): Likewise.
5053 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5054 Chung-Ju Wu <jasonwucj@gmail.com>
5056 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5059 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5060 Chung-Ju Wu <jasonwucj@gmail.com>
5062 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5063 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5064 (nds32_rtx_costs_impl): Simplify.
5065 (nds32_address_cost_impl): Simplify.
5066 (nds32_init_rtx_costs): New function.
5067 (nds32_rtx_costs_speed_prefer): Likewise.
5068 (nds32_rtx_costs_size_prefer): Likewise.
5069 (nds32_address_cost_speed_prefer): Likewise.
5070 (nds32_address_cost_speed_fwprop): Likewise.
5071 (nds32_address_cost_size_prefer): Likewise.
5072 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5073 * config/nds32/nds32.c (nds32_option_override): Use
5074 nds32_init_rtx_costs function.
5076 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5078 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5079 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5080 (TARGET_PIPELINE_N8): Likewise.
5081 (TARGET_PIPELINE_N10): Likewise.
5082 (TARGET_PIPELINE_N13): Likewise.
5083 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5085 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5087 * config/nds32/nds32-fpu.md: Update copyright year.
5089 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5091 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5093 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5095 * config/nds32/nds32.c
5096 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5097 * config/nds32/nds32.opt (minline-asm-r15): New option.
5099 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5101 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5103 * config/nds32/nds32.md (abssi2): New pattern.
5105 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5107 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5108 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5109 Merge insn pattern from sse_cvtsi2ss<round_name> and
5110 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5111 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5112 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5113 using SWI48 mode iterator.
5114 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5115 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5116 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5117 pattern from sse_cvttss2si<round_saeonly_name>
5118 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5119 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5120 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5121 using SWI48 mode iterator.
5122 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5123 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5124 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5125 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5126 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5127 using SWI48 mode iterator.
5128 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5129 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5130 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5131 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5132 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5133 SWI48 mode iterator.
5134 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5135 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5136 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5137 pattern from sse_cvttsd2si<round_saeonly_name>
5138 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5140 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5142 * config/nds32/nds32-md-auxiliary.c
5143 (nds32_valid_smw_lwm_base_p): Refine.
5144 (nds32_output_smw_single_word): Refine.
5145 (nds32_output_smw_double_word): New.
5146 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5148 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5151 (nds32_output_stack_pop): Refine.
5152 (nds32_expand_unaligned_load): Refine.
5153 (nds32_expand_unaligned_store): Refine.
5155 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5156 Chung-Ju Wu <jasonwucj@gmail.com>
5158 * config/nds32/constants.md: Add TP_REGNUM constant.
5159 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5160 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5162 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5163 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5164 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5165 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5166 and PIC code generation.
5167 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5169 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5171 * config/nds32/nds32.md: Support TLS and PIC.
5172 * config/nds32/nds32.c: Support TLS and PIC.
5173 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5174 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5177 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5179 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5180 mode with E_ prefix.
5182 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5183 Chung-Ju Wu <jasonwucj@gmail.com>
5185 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5186 * config/nds32/nds32-md-auxiliary.c
5187 (symbolic_reference_mentioned_p): New.
5188 (nds32_legitimize_ict_address): New.
5189 (nds32_expand_ict_move): New.
5190 (nds32_indirect_call_referenced_p): New.
5191 (nds32_symbol_binds_local_p): Delete.
5192 (nds32_long_call_p): Modify.
5193 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5194 * config/nds32/nds32-protos.h
5195 (symbolic_reference_mentioned_p): Declare.
5196 (nds32_legitimize_ict_address): Declare.
5197 (nds32_expand_ict_move): Declare.
5198 (nds32_indirect_call_referenced_p): Declare.
5199 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5200 (nds32_relax_group): Use nds32_ict_const_p as condition.
5201 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5202 (nds32_asm_file_start): Output ict_model directive in asm code.
5203 (nds32_legitimate_address_p): Consider indirect call.
5204 (nds32_print_operand): Consider indirect call.
5205 (nds32_print_operand_address): Consider indirect call.
5206 (nds32_insert_attributes): Handle "indirect_call" attribute.
5207 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5208 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5209 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5210 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5211 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5212 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5213 (TARGET_ICT_MODEL_SMALL): Define.
5214 (TARGET_ICT_MODEL_LARGE): Define.
5215 * config/nds32/nds32.md (movsi): Consider ict model.
5216 (call, call_value): Consider ict model.
5217 (sibcall, sibcall_value): Consider ict model.
5218 * config/nds32/nds32.opt (mict-model): New option.
5219 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5222 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5223 Monk Chiang <sh.chiang04@gmail.com>
5224 Jim Wilson <jimw@sifive.com>
5226 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5227 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5228 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5229 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5230 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5231 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5232 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5233 compute save_libcall_adjustment properly.
5234 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5235 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5236 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5237 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5238 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5239 (ABI_SPEC): Handle mabi=ilp32e.
5240 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5241 (RVE): Add RVE mask.
5242 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5243 <-march>: Add rv32e as an example.
5245 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5248 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5249 (intra_create_variable_infos): Handle C++ constructors.
5251 2018-05-18 Martin Liska <mliska@suse.cz>
5253 * passes.def: Remove a redundant pass.
5255 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5258 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5260 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5262 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5264 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5266 (armv2, armv2a, armv3, armv3m): Delete architectures.
5267 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5268 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5269 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5271 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5272 (*mulsidi3adddi): Likewise.
5273 (mulsidi3): Likewise.
5274 (*mulsidi3_nov6): Likewise.
5275 (umulsidi3): Likewise.
5276 (umulsidi3_nov6): Likewise.
5277 (umaddsidi4): Likewise.
5278 (*umulsidi3adddi): Likewise.
5279 (smulsi3_highpart): Likewise.
5280 (*smulsi3_highpart_nov6): Likewise.
5281 (umulsi3_highpart): Likewise.
5282 (*umulsi3_highpart_nov6): Likewise.
5283 * config/arm/arm.h (arm_arch3m): Delete.
5284 * config/arm/arm.c (arm_arch3m): Delete.
5285 (arm_option_override_internal): Update armv3-related comment.
5286 (arm_configure_build_target): Delete use of isa_bit_mode32.
5287 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5288 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5289 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5291 (mulusa3): Likewise.
5292 * config/arm/arm-protos.h (arm_arch3m): Delete.
5293 * config/arm/arm-tables.opt: Regenerate.
5294 * config/arm/arm-tune.md: Likewise.
5295 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5296 deleted architectures.
5298 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5300 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5301 (armv5t, armv5te): New features.
5302 (ARMv5, ARMv5e): Delete fgroups.
5303 (ARMv5t, ARMv5te): Adjust for above changes.
5305 (armv5, armv5e): Delete arches.
5306 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5308 (*call_reg_arm): Likewise.
5309 (*call_value_reg_armv5): Likewise.
5310 (*call_value_reg_arm): Likewise.
5311 (*call_symbol): Likewise.
5312 (*call_value_symbol): Likewise.
5313 (*sibcall_insn): Likewise.
5314 (*sibcall_value_insn): Likewise.
5316 (prefetch): Likewise.
5317 (define_split and define_peephole2 dependent on arm_arch5):
5319 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5321 (TARGET_ARM_QBIT): Likewise.
5322 (TARGET_DSP_MULTIPLY): Likewise.
5323 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5324 (arm_arch5, arm_arch5e): Delete.
5325 (arm_arch5t, arm_arch5te): Declare.
5326 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5327 (arm_arch5t): Declare.
5328 (arm_option_reconfigure_globals): Update for the above.
5329 (arm_options_perform_arch_sanity_checks): Update comment, replace
5330 use of arm_arch5 with arm_arch5t.
5331 (use_return_insn): Likewise.
5332 (arm_emit_call_insn): Likewise.
5333 (output_return_instruction): Likewise.
5334 (arm_final_prescan_insn): Likewise.
5335 (arm_coproc_builtin_available): Likewise.
5336 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5337 arm_arch5e with arm_arch5t and arm_arch5te.
5338 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5339 (arm_arch5t, arm_arch5te): Declare.
5340 * config/arm/arm-tables.opt: Regenerate.
5341 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5342 * config/arm/t-multilib: Likewise.
5343 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5344 instead of arm_arch5.
5345 (*call_reg_thumb1): Likewise.
5346 (*call_value_reg_thumb1_v5): Likewise.
5347 (*call_value_reg_thumb1): Likewise.
5348 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5350 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5352 2018-05-18 Martin Liska <mliska@suse.cz>
5354 PR gcov-profile/84846
5355 * doc/gcov.texi: Document -t option of gcov tool.
5357 2018-05-18 Martin Liska <mliska@suse.cz>
5359 PR gcov-profile/84846
5360 * gcov.c (print_usage): Add new -t option.
5361 (process_args): Handle the option.
5362 (generate_results): Use stdout as output when requested by
5365 2018-05-18 Martin Liska <mliska@suse.cz>
5367 PR gcov-profile/84846
5368 * coverage.c (coverage_init): Write PWD to .gcno file.
5369 * doc/gcov.texi: Document how working directory is printed.
5370 * gcov-dump.c (dump_gcov_file): Print PWD.
5371 * gcov.c (output_intermediate_file): Likewise.
5372 (read_graph_file): Read PWD string.
5373 (output_lines): Print PWD.
5375 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5378 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5379 for retval and return false if all args to phi are zero.
5381 2018-05-18 Richard Biener <rguenther@suse.de>
5383 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5385 (evrp_dom_walker::before_dom_children): Call it.
5387 2018-05-18 Richard Biener <rguenther@suse.de>
5389 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5390 results when processing array refs with variable index.
5392 2018-05-18 Toon Moene <toon@moene.org>
5394 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5395 directly after that of -floop-interchange. Indicate that both
5396 options are enabled by default when specifying -O3.
5398 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5400 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5401 iterator. Delete separate integer-mode vec_set<mode> expander.
5402 (aarch64_simd_vec_setv2di): Delete.
5403 (vec_setv2di): Delete.
5404 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5405 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5406 the "w, r" alternative.
5408 2018-05-18 Martin Liska <mliska@suse.cz>
5410 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5411 * tree-pass.h (make_pass_lower_switch_O0): New function.
5412 * tree-switch-conversion.c (node_has_low_bound): Remove.
5413 (node_has_high_bound): Likewise.
5414 (node_is_bounded): Likewise.
5415 (class pass_lower_switch): Make it a template type and create
5417 (pass_lower_switch::execute): Add template argument.
5418 (make_pass_lower_switch): New function.
5419 (make_pass_lower_switch_O0): New function.
5420 (do_jump_if_equal): Remove.
5421 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5422 all the hard work to tree optimization passes.
5424 2018-05-18 Martin Liska <mliska@suse.cz>
5426 * dbgcnt.c (limit_low): Renamed from limit.
5427 (limit_high): New variable.
5428 (dbg_cnt_is_enabled): Check for upper limit.
5429 (dbg_cnt): Adjust dumping.
5430 (dbg_cnt_set_limit_by_index): Add new argument for high
5432 (dbg_cnt_set_limit_by_name): Likewise.
5433 (dbg_cnt_process_single_pair): Parse new format.
5434 (dbg_cnt_process_opt): Use strtok.
5435 (dbg_cnt_list_all_counters): Remove 'value' and add
5437 * doc/invoke.texi: Document changes.
5439 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5441 * doc/sourcebuild.texi (scalar_all_fma): Document.
5442 * tree.def (FMA_EXPR): Delete.
5443 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5444 * internal-fn.c (ternary_direct): New macro.
5445 (expand_ternary_optab_fn): Likewise.
5446 (direct_ternary_optab_supported_p): Likewise.
5447 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5448 * builtins.c (fold_builtin_fma): Delete.
5449 (fold_builtin_3): Don't call it.
5450 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5451 * expr.c (expand_expr_real_2): Likewise.
5452 * fold-const.c (operand_equal_p): Likewise.
5453 (fold_ternary_loc): Likewise.
5454 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5455 * gimple.c (DEFTREECODE): Likewise.
5456 * gimplify.c (gimplify_expr): Likewise.
5457 * optabs-tree.c (optab_for_tree_code): Likewise.
5458 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5459 * tree-eh.c (operation_could_trap_p): Likewise.
5460 (stmt_could_throw_1_p): Likewise.
5461 * tree-inline.c (estimate_operator_cost): Likewise.
5462 * tree-pretty-print.c (dump_generic_node): Likewise.
5463 (op_code_prio): Likewise.
5464 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5465 * tree-ssa-operands.c (get_expr_operands): Likewise.
5466 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5467 * fold-const-call.h (fold_fma): Delete.
5468 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5469 CFN_FNMA and CFN_FNMS.
5471 * genmatch.c (combined_fn): New enum.
5472 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5473 (commutative_op): New function.
5474 (commutate): Use it. Handle more than 2 operands.
5475 (dt_operand::gen_gimple_expr): Use commutative_op.
5476 (parser::parse_expr): Allow :c to be used with non-binary
5477 operators if the commutative operand is known.
5478 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5479 CFN_FMS, CFN_FNMA and CFN_FNMS.
5480 (backprop::process_assign_use): Remove FMA_EXPR handling.
5481 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5482 (gen_hsa_fma): New function.
5483 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5484 IFN_FNMA and IFN_FNMS.
5485 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5486 * gimple-fold.h (follow_all_ssa_edges): Declare.
5487 * gimple-fold.c (follow_all_ssa_edges): New function.
5488 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5489 gimple_build interface and use follow_all_ssa_edges to fold the result.
5490 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5491 instead of checking for optabs directly.
5492 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5493 rather than FMA_EXPRs.
5494 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5495 call to IFN_FMA instead of an FMA_EXPR.
5497 2018-05-17 Jim Wilson <jimw@sifive.com>
5499 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5500 sign extended promoted subreg, and the range does not have the sign bit
5501 set, then do a sign extend.
5503 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5504 test, check for sign extended subreg and/or constant operands, and
5505 do a sign extend in that case.
5507 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5509 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5510 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5512 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5513 Change logics_shift_reg to logics_shift_imm.
5514 (thunderx2t99_fp_loadpair_basic): Delete.
5515 (thunderx2t99_fp_storepair_basic): Delete.
5516 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5517 (thunderx2t99_asimd_polynomial): Delete.
5518 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5519 and neon_fp_mul_d_scalar_q.
5520 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5521 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5522 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5523 (thunderx2t99_asimd_lut): Add missing tbl types.
5524 (thunderx2t99_asimd_ext): Delete.
5525 (thunderx2t99_asimd_load1_1_mult): Delete.
5526 (thunderx2t99_asimd_load1_2_mult): Delete.
5527 (thunderx2t99_asimd_load1_ldp): New.
5528 (thunderx2t99_asimd_load1): New.
5529 (thunderx2t99_asimd_load2): Add missing *load2* types.
5530 (thunderx2t99_asimd_load3): New.
5531 (thunderx2t99_asimd_load4): New.
5532 (thunderx2t99_asimd_store1_1_mult): Delete.
5533 (thunderx2t99_asimd_store1_2_mult): Delete.
5534 (thunderx2t99_asimd_store2_mult): Delete.
5535 (thunderx2t99_asimd_store2_onelane): Delete.
5536 (thunderx2t99_asimd_store_stp): New.
5537 (thunderx2t99_asimd_store1): New.
5538 (thunderx2t99_asimd_store2): New.
5539 (thunderx2t99_asimd_store3): New.
5540 (thunderx2t99_asimd_store4): New.
5542 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5544 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5545 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5547 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5548 Segher Boessenkool <segher@kernel.crashing.org>
5551 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5554 2018-05-17 Richard Biener <rguenther@suse.de>
5556 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5557 for pruning loop and prune defs feeding only already visited PHIs.
5559 2018-05-17 Richard Biener <rguenther@suse.de>
5561 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5563 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5564 Richard Biener <rguenther@suse.de>
5566 PR tree-optimization/85793
5567 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5568 for VMAT_ELEMENTWISE.
5570 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5572 * internal-fn.h (lookup_internal_fn): Declare
5573 * internal-fn.c (lookup_internal_fn): New function.
5574 * gimple.c (gimple_build_call_from_tree): Handle calls to
5576 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5577 internal function names.
5578 * tree-pretty-print.c (dump_generic_node): Likewise.
5579 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5581 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5583 * gimple-fold.h (gimple_build): Make the function forms take
5584 combined_fn rather than built_in_function.
5585 (gimple_simplify): Likewise.
5586 * gimple-match-head.c (gimple_simplify): Likewise.
5587 * gimple-fold.c (gimple_build): Likewise.
5588 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5589 rather than gimple_build_call_internal.
5590 (get_initial_defs_for_reduction): Likewise.
5591 (vect_create_epilog_for_reduction): Likewise.
5592 (vectorizable_live_operation): Likewise.
5594 2018-05-17 Martin Liska <mliska@suse.cz>
5596 * gimple-ssa-sprintf.c (format_directive): Do not use
5597 space in between 'G_' and '('.
5599 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5602 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5603 even if the mask is not all ones.
5606 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5608 (ix86_gimple_fold_builtin): Likewise.
5611 * config/i386/i386.c: Include tree-vector-builder.h.
5612 (ix86_vector_shift_count): New function.
5613 (ix86_fold_builtin): Fold shift builtins by scalar count.
5614 (ix86_gimple_fold_builtin): Likewise.
5616 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5617 _mm512_setzero): New intrinsics.
5619 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5620 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5622 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5623 code generation for cases where splatting a value is not useful.
5624 * simplify-rtx.c (simplify_ternary_operation): Simplify
5625 vec_merge across a vec_duplicate and a paradoxical subreg forming
5626 a vector mode to a vec_concat.
5628 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5630 * config.gcc: Support "goldmont-plus".
5631 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5633 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5634 PROCESSOR_GOLDMONT_PLUS.
5635 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5636 (processor_target_table): Add "goldmont-plus".
5637 (PTA_GOLDMONT_PLUS): Define.
5638 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5639 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5640 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5641 (fold_builtin_cpu): Add "goldmont-plus".
5642 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5643 (ix86_option_override_internal): Add "goldmont-plus".
5644 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5645 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5646 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5647 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5649 2018-05-17 Richard Biener <rguenther@suse.de>
5651 PR tree-optimization/85757
5652 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5653 remove defs that only feed that PHI from further processing.
5655 2018-05-16 Jim Wilson <jimw@sifive.com>
5657 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5659 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5661 2018-05-16 Mark Wielaard <mark@klomp.org>
5663 * dwarf2out.c (count_index_strings): New function.
5664 (output_indirect_strings): Call count_index_strings and generate
5665 header for dwarf_version >= 5.
5667 2018-05-16 Mark Wielaard <mark@klomp.org>
5669 * dwarf2out.c (dwarf_FORM): New function.
5670 (set_indirect_string): Use dwarf_FORM.
5671 (reset_indirect_string): Likewise.
5672 (size_of_die): Likewise.
5673 (value_format): Likewise.
5674 (output_die): Likewise.
5675 (add_skeleton_AT_string): Likewise.
5676 (output_macinfo_op): Likewise.
5677 (index_string): Likewise.
5678 (output_index_string_offset): Likewise.
5679 (output_index_string): Likewise.
5680 (count_index_strings): Likewise.
5682 2018-05-16 Carl Love <cel@us.ibm.com>
5684 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5685 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5687 2018-05-16 Martin Jambor <mjambor@suse.cz>
5689 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5690 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5692 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5694 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5695 (fnma<mode>4): Likewise.
5696 (fms<mode>4): Likewise.
5697 (fnms<mode>4): Likewise.
5698 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5699 (aarch64_fnma<mode>4): Likewise.
5700 (aarch64_fms<mode>4): Likewise.
5701 (aarch64_fnms<mode>4): Likewise.
5702 (aarch64_fnmadd<mode>4): Likewise.
5704 2018-05-16 Jason Merrill <jason@redhat.com>
5706 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5708 2018-05-16 Richard Biener <rguenther@suse.de>
5710 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5711 (dump_stmt_cost): Declare.
5712 (add_stmt_cost): Dump cost we add.
5713 (add_stmt_costs): New function.
5714 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5716 (vect_analyze_stmt): Adjust prototype.
5717 (vectorizable_condition): Likewise.
5718 (vectorizable_live_operation): Likewise.
5719 (vectorizable_reduction): Likewise.
5720 (vectorizable_induction): Likewise.
5721 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5722 cost vector to pass to vectorizable_ and record afterwards.
5723 (vect_model_reduction_cost): Take cost vector argument and adjust.
5724 (vect_model_induction_cost): Likewise.
5725 (vectorizable_reduction): Likewise.
5726 (vectorizable_induction): Likewise.
5727 (vectorizable_live_operation): Likewise.
5728 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5729 SLP_TREE_NUMBER_OF_VEC_STMTS.
5730 (vect_analyze_slp_cost_1): Remove.
5731 (vect_analyze_slp_cost): Likewise.
5732 (vect_slp_analyze_node_operations): Take visited args and
5733 a target cost vector. Avoid processing already visited stmt sets.
5734 (vect_slp_analyze_operations): Use a local cost vector to gather
5735 costs and register those of non-discarded instances.
5736 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5737 (vect_schedule_slp_instance): Remove copying of
5738 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5740 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5741 adding cost. Record cost entry location.
5742 (vect_prologue_cost_for_slp_op): Function to compute cost of
5743 a constant or invariant generated for SLP vect in the prologue,
5744 split out from vect_analyze_slp_cost_1.
5745 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5746 (vect_model_promotion_demotion_cost): Likewise.
5747 (vect_model_store_cost): Likewise, make static.
5748 (vect_model_load_cost): Likewise.
5749 (vectorizable_bswap): Add cost vector arg and adjust.
5750 (vectorizable_call): Likewise.
5751 (vectorizable_simd_clone_call): Likewise.
5752 (vectorizable_conversion): Likewise.
5753 (vectorizable_assignment): Likewise.
5754 (vectorizable_shift): Likewise.
5755 (vectorizable_operation): Likewise.
5756 (vectorizable_store): Likewise.
5757 (vectorizable_load): Likewise.
5758 (vectorizable_condition): Likewise.
5759 (vectorizable_comparison): Likewise.
5760 (can_vectorize_live_stmts): Likewise.
5761 (vect_analyze_stmt): Likewise.
5762 (vect_transform_stmt): Adjust calls to vectorizable_*.
5763 * tree-vectorizer.c: Include gimple-pretty-print.h.
5764 (dump_stmt_cost): New function.
5766 2018-05-16 Richard Biener <rguenther@suse.de>
5768 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5769 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5770 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5771 (check_name): New callback.
5772 (dse_classify_store): Track cycles via a visited bitmap of PHI
5773 defs and simplify handling of in-loop and across loop dead stores
5774 and properly fail for loop-variant refs. Handle byte-tracking with
5775 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5778 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5780 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5781 (vect_get_mask_type_for_stmt): Likewise.
5782 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5784 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5785 to determine the statement's vector type and the vector type that
5786 should be used for calculating nunits. Deal with cases in which
5787 the type has to be deferred.
5788 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5789 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5790 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5791 (vect_determine_vf_for_stmt): New functions, split out from...
5792 (vect_determine_vectorization_factor): ...here.
5793 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5794 (vect_get_mask_type_for_stmt): New functions, split out from
5795 vect_determine_vectorization_factor.
5797 2018-05-16 Richard Biener <rguenther@suse.de>
5799 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5800 verify the [VEC_]COND_EXPR embedded comparison.
5802 2018-05-15 Martin Sebor <msebor@redhat.com>
5804 PR tree-optimization/85753
5805 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5806 RECORD_TYPE in addition to ARRAY_TYPE.
5808 2018-05-15 Martin Sebor <msebor@redhat.com>
5811 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5813 2018-05-15 Richard Biener <rguenther@suse.de>
5815 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5816 add by_clobber_p one. Change algorithm to collect all defs
5817 representing uses we need to walk and try reducing them to
5818 a single one before failing.
5819 (dse_dom_walker::dse_optimize_stmt): Adjust.
5821 2018-05-13 Mark Wielaard <mark@klomp.org>
5823 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5824 (size_of_loc_descr): Likewise.
5825 (output_loc_operands): Likewise.
5826 (output_loc_operands_raw): Likewise.
5827 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5828 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5829 (hash_loc_operands): Likewise.
5830 (compare_loc_operands): Likewise.
5832 2018-05-14 Mark Wielaard <mark@klomp.org>
5834 * dwarf2out.c (count_index_addrs): New function.
5835 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5837 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5839 PR tree-optimization/83648
5840 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5841 return value as malloc candidate.
5843 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5846 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5847 param as true in call to suggest_attribute.
5849 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5851 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5852 -mreadonly-in-sdata.
5854 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5856 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5858 (aarch64_crypto_aesd_fused): Likewise.
5860 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5862 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5863 (movsi_aarch64): Likewise.
5864 (load_pairsi): Likewise.
5865 (load_pairdi): Likewise.
5866 (store_pairsi): Likewise.
5867 (store_pairdi): Likewise.
5868 (load_pairsf): Likewise.
5869 (load_pairdf): Likewise.
5870 (store_pairsf): Likewise.
5871 (store_pairdf): Likewise.
5872 (zero_extend): Likewise.
5873 (trunc): Swap alternatives.
5874 (fcvt_target): Add '?' to prefer w over r.
5876 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5879 * config/i386/i386.md: Disallow non-commutative arithmetics in
5880 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5881 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5882 in the peephole2 before it.
5884 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5886 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5887 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5888 (ix86_handle_option): Handle -mcldemote.
5889 * config.gcc: New header.
5890 * config/i386/cldemoteintrin.h: New file.
5891 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5892 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5894 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5895 OPTION_MASK_ISA_CLDEMOTE.
5896 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5897 (ix86_valid_target_attribute_inner_p): Ditto.
5898 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5899 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5900 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5901 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5902 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5904 * config/i386/i386.opt: Add -mcldemote.
5905 * config/i386/x86intrin.h: New header.
5906 * doc/invoke.texi: Add -mcldemote.
5908 2018-05-14 Richard Biener <rguenther@suse.de>
5910 * doc/match-and-simplify.texi: Adjust :s documentation.
5912 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5914 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5915 intended memcpy size.
5916 (REORDER_45): Likewise.
5918 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5920 * sort.cc: New file.
5921 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5922 * vec.c (qsort_chk): Use gcc_qsort.
5923 * Makefile.in (OBJS-libcommon): Add sort.o.
5924 (build/sort.o): New target. Use it...
5925 (BUILD_RTL): ... here, and...
5926 (build/gencfn-macros): ... here, and...
5927 (build/genmatch): ... here.
5929 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5930 Chung-Ju Wu <jasonwucj@gmail.com>
5932 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5933 * config/nds32/nds32-graywolf.md: New file.
5934 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5935 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5937 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5938 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5939 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5940 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5941 * config/nds32/pipelines.md: Include n15 settings.
5943 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5944 Chung-Ju Wu <jasonwucj@gmail.com>
5946 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5947 * config/nds32/nds32-n13.md: New file.
5948 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5949 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5951 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5952 * config/nds32/nds32.md (pipeline_model): Add n13.
5953 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5954 * config/nds32/pipelines.md: Include n13 settings.
5956 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5957 Chung-Ju Wu <jasonwucj@gmail.com>
5959 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5960 * config/nds32/nds32-n10.md: New file.
5961 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5962 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5964 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5965 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5966 * config/nds32/nds32.md (pipeline_model): Add n10.
5967 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5968 * config/nds32/pipelines.md: Include n10 settings.
5970 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5971 Kito Cheng <kito.cheng@gmail.com>
5972 Chung-Ju Wu <jasonwucj@gmail.com>
5974 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5975 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5976 Add enum values for DSP extension instructions.
5977 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5979 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5980 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5982 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5983 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5984 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5985 * config/nds32/nds32-intrinsic.md: Likewise.
5986 * config/nds32/nds32_intrinsic.h: Likewise.
5987 * config/nds32/nds32-md-auxiliary.c: Likewise.
5988 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5989 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5990 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5991 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5992 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5993 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5994 TYPE_DMAC in switch statement.
5995 * config/nds32/nds32.c: New checking and implementation for DSP
5996 extension instructions.
5997 * config/nds32/nds32.h: Likewise.
5998 * config/nds32/nds32.md: Likewise.
5999 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6000 * config/nds32/predicates.md: Implement new predicates for DSP
6003 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6005 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6006 Reformat alternatives and attributes so it is easier to identify
6007 which constraints/attributes go with which instruction.
6008 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6009 (mov<mode>_softfloat32, FMOVE64): Likewise.
6010 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6011 (mov<mode>_softfloat64, FMOVE64): Likewise.
6013 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6015 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6017 (Basic PowerPC Built-in Functions): The new name of the
6018 subsection previously known as "PowerPC Built-in Functions".
6019 (Basic PowerPC Built-in Functions Available on all Configurations):
6021 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6022 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6023 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6024 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6026 2018-05-11 Martin Jambor <mjambor@suse.cz>
6029 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6032 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6035 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6037 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6039 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6040 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6041 (ix86_handle_option): Handle -mwaitpkg.
6042 * config.gcc: New header.
6043 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6044 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6045 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6047 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6048 OPTION_MASK_ISA_WAITPKG.
6049 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6050 (ix86_option_override_internal): Add PTA_WAITPKG.
6051 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6052 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6053 IX86_BUILTIN_TPAUSE.
6054 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6055 __builtin_ia32_umwait and __builtin_ia32_tpause.
6056 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6057 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6058 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6059 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6060 UNSPECV_TPAUSE): New.
6061 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6062 * config/i386/i386.opt: Add -mwaitpkg.
6063 * config/i386/waitpkgintrin.h: New file.
6064 * config/i386/x86intrin.h: New header.
6065 * doc/invoke.texi: Add -mwaitpkg.
6067 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6070 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6072 (cortex-m0): Use armv6s-m isa.
6073 (cortex-m0plus): Likewise.
6074 (cortex-m1): Likewise.
6075 (cortex-m0.small-multiply): Likewise.
6076 (cortex-m0plus.small-multiply): Likewise.
6077 (cortex-m1.small-multiply): Likewise.
6079 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6080 Jakub Jelinek <jakub@redhat.com>
6082 PR tree-optimization/85692
6083 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6084 source permute as well.
6086 2018-05-11 Martin Liska <mliska@suse.cz>
6089 * doc/extend.texi: Document LLVM style format for no_sanitize
6092 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6094 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6095 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6096 (mode_supports_vsx_dform_quad): Likewise.
6097 (mode_supports_vmx_dform): Move these functions to be next to the
6098 other mode_supports functions.
6099 (mode_supports_dq_form): Likewise.
6100 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6101 mode_supports_dq_form.
6102 (reg_offset_addressing_ok_p): Likewise.
6103 (offsettable_ok_by_alignment): Likewise.
6104 (rs6000_legitimate_offset_address_p): Likewise.
6105 (legitimate_lo_sum_address_p): Likewise.
6106 (rs6000_legitimize_address): Likewise.
6107 (rs6000_legitimize_reload_address): Likewise.
6108 (rs6000_secondary_reload_inner): Likewise.
6109 (rs6000_preferred_reload_class): Likewise.
6110 (rs6000_output_move_128bit): Likewise.
6112 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6114 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6115 Generate SImode target register for null target.
6116 <case IX86_BUILTIN_XGETBV>: Ditto.
6117 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6118 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6120 2018-05-10 Carl Love <cel@us.ibm.com>
6122 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6123 dcbtt and dcbtstt if operands[2] is 0.
6125 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6128 * config/i386/sse.md (usadv64qi): New expander.
6130 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6132 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6133 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6134 -maltivec=be support.
6135 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6136 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6137 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6138 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6139 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6140 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6141 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6142 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6143 altivec_vsumsws): Adjust.
6144 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6145 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6146 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6148 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6149 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6150 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6151 (altivec_lve<VI_char>x): Delete expand.
6152 (*altivec_lve<VI_char>x_internal): Rename to...
6153 (altivec_lve<VI_char>x): ... this.
6154 (altivec_lvxl_<mode>): Delete expand.
6155 (*altivec_lvxl_<mode>_internal): Rename to ...
6156 (altivec_lvxl_<mode>): ... this.
6157 (altivec_stvxl_<mode>): Delete expand.
6158 (*altivec_stvxl_<mode>_internal): Rename to ...
6159 (altivec_stvxl_<mode>): ... this.
6160 (altivec_stve<VI_char>x): Delete expand.
6161 (*altivec_stve<VI_char>x_internal): Rename to ...
6162 (altivec_stve<VI_char>x): ... this.
6163 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6164 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6165 reduc_plus_scal_<mode>): Adjust.
6166 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6168 (rs6000_cpu_cpp_builtins): Adjust.
6169 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6170 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6171 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6172 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6173 -maltivec=be support.
6174 (rs6000_split_vec_extract_var): Adjust.
6175 (rs6000_split_v4si_init): Adjust.
6176 (swap_selector_for_mode): Delete.
6177 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6178 altivec_expand_stvex_be): Delete.
6179 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6180 -maltivec=be support.
6181 (rs6000_gimple_fold_builtin): Ditto.
6182 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6184 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6185 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6186 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6187 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6188 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6189 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6190 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6191 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6192 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6193 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6194 anonymous split): Adjust.
6195 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6196 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6198 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6200 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6201 when --with-gxx-include-dir is also specified.
6202 * configure: Regenerate.
6204 2018-05-09 Jim Wilson <jimw@sifive.com>
6207 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6208 * config/riscv/t-withmultilib: New.
6209 * config/riscv/withmultilib.h: New.
6210 * doc/install.texi: Document RISC-V --with-multilib-list support.
6212 2018-05-09 Richard Biener <rguenther@suse.de>
6214 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6216 (vect_bb_vectorization_profitable_p): Adjust. Compute
6217 actual scalar cost using the cost vector and the add_stmt_cost
6220 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6222 PR rtl-optimization/85645
6223 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6224 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6226 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6228 PR rtl-optimization/85645
6229 * shrink-wrap.c (spread_components): Return a boolean saying if
6230 anything was changed.
6231 (try_shrink_wrapping_separate): Iterate spread_components until
6232 nothing changes anymore.
6234 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6236 PR rtl-optimization/85645
6237 * regrename.c (build_def_use): Also kill the chains that include the
6238 destination of a REG_CFA_REGISTER note.
6240 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6242 PR rtl-optimization/85645
6243 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6244 insn that has a REG_CFA_REGISTER note.
6246 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6248 * cfgexpand.c (expand_clobber): New function.
6249 (expand_gimple_stmt_1): Use it.
6250 * tree-vect-stmts.c (vect_clobber_variable): New function,
6252 (vectorizable_simd_clone_call): ...here.
6253 (vectorizable_store): Emit a clobber either side of an
6254 IFN_STORE_LANES sequence.
6255 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6257 2018-05-09 Tom de Vries <tom@codesourcery.com>
6260 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6261 (define_insn "trap_if_false"): Add exit after trap.
6263 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6265 PR rtl-optimization/85638
6266 * bb-reorder.c: Include common/common-target.h.
6267 (create_forwarder_block): New function extracted from...
6268 (fix_up_crossing_landing_pad): ...here. Rename into...
6269 (dw2_fix_up_crossing_landing_pad): ...this.
6270 (sjlj_fix_up_crossing_landing_pad): New function.
6271 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6272 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6273 from both partitions and exit the loop after one iteration.
6275 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6278 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6280 (Basic PowerPC Built-in Functions): The new name of the
6281 subsection previously known as "PowerPC Built-in Functions".
6282 (Basic PowerPC Built-in Functions Available on all Configurations):
6284 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6286 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6288 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6290 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6293 2018-05-08 Jim Wilson <jimw@sifive.com>
6295 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6296 (LD_EMUL_SUFFIX): New.
6297 (LINK_SPEC): Use it.
6299 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6301 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6303 (Basic PowerPC Built-in Functions): The new name of the
6304 subsection previously known as "PowerPC Built-in Functions".
6305 (Basic PowerPC Built-in Functions Available on all Configurations):
6307 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6309 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6311 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6313 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6316 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6319 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6320 after cmpelim optimization.
6322 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6324 * config.gcc: Support "goldmont".
6325 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6326 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6328 * config/i386/i386.c (m_GOLDMONT): Define.
6329 (processor_target_table): Add "goldmont".
6330 (PTA_GOLDMONT): Define.
6331 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6332 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6333 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6334 (fold_builtin_cpu): Add "goldmont".
6335 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6336 (ix86_option_override_internal): Add "goldmont".
6337 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6338 (processor_type): Add PROCESSOR_GOLDMONT.
6339 * config/i386/i386.md: Add CPU "glm".
6340 * config/i386/glm.md: New file.
6341 * config/i386/x86-tune.def: Add m_GOLDMONT.
6342 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6344 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6347 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6349 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6350 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6351 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6354 * config/i386/i386.c (ix86_fold_builtin): Handle
6355 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6358 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6359 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6361 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6364 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6365 (check_arch): Likewise.
6366 (check_fpu): Return the result rather than printing it.
6367 (end arch): Fix operator precedence.
6368 (end cpu): Likewise.
6369 (END): Print the result from check_fpu.
6371 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6372 Alan Hayward <alan.hayward@arm.com>
6373 David Sherwood <david.sherwood@arm.com>
6375 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6376 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6377 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6378 (*fcmuo<mode>_and): New patterns.
6380 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6382 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6383 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6384 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6385 (cmp_op, sve_imm_con): New code attributes.
6386 (SVE_COND_INT_CMP, imm_con): Delete.
6387 (cmp_op): Remove above unspecs from int attribute.
6388 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6390 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6391 comparison-specific unspecs.
6392 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6393 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6394 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6395 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6396 (*vec_fcm<cmp_op><mode>): Rename to...
6397 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6398 (*vec_fcmuo<mode>): Rename to...
6399 (*fcmuo<mode>): ...this and adjust likewise.
6400 (*pred_fcm<cmp_op><mode>): New pattern.
6401 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6402 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6404 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6406 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6407 (aarch64_emit_sve_predicated_cond): New function.
6408 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6409 (aarch64_emit_unspec_cond_or): Replace with...
6410 (aarch64_emit_sve_or_conds): ...this new function. Use
6411 aarch64_emit_sve_ptrue_op for the individual comparisons and
6412 aarch64_emit_binop to OR them together.
6413 (aarch64_emit_inverted_unspec_cond): Replace with...
6414 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6415 aarch64_emit_sve_ptrue_op for the comparison and
6416 aarch64_emit_unop to invert the result.
6417 (aarch64_expand_sve_vec_cmp_float): Update after the above
6418 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6420 2018-05-07 Nathan Sidwell <nathan@acm.org>
6422 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6423 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6424 (Backwards Compatibility): Likewise.
6426 2018-05-07 Luis Machado <luis.machado@linaro.org>
6430 2018-05-07 Luis Machado <luis.machado@linaro.org>
6432 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6433 <prefetch_dynamic_strides>: New const bool field.
6434 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6435 prefetch_dynamic_strides.
6436 (exynosm1_prefetch_tune): Likewise.
6437 (thunderxt88_prefetch_tune): Likewise.
6438 (thunderx_prefetch_tune): Likewise.
6439 (thunderx2t99_prefetch_tune): Likewise.
6440 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6442 (aarch64_override_options_internal): Update to set
6443 PARAM_PREFETCH_DYNAMIC_STRIDES.
6444 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6445 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6446 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6447 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6448 prefetch-dynamic-strides setting.
6450 2018-05-07 Luis Machado <luis.machado@linaro.org>
6452 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6453 <minimum_stride>: New const int field.
6454 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6455 minimum_stride field.
6456 (exynosm1_prefetch_tune): Likewise.
6457 (thunderxt88_prefetch_tune): Likewise.
6458 (thunderx_prefetch_tune): Likewise.
6459 (thunderx2t99_prefetch_tune): Likewise.
6460 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6461 (aarch64_override_options_internal): Update to set
6462 PARAM_PREFETCH_MINIMUM_STRIDE.
6463 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6464 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6465 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6466 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6467 stride is constant and is below the minimum stride threshold.
6469 2018-05-07 Luis Machado <luis.machado@linaro.org>
6471 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6474 2018-05-07 Luis Machado <luis.machado@linaro.org>
6476 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6477 <prefetch_dynamic_strides>: New const bool field.
6478 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6479 prefetch_dynamic_strides.
6480 (exynosm1_prefetch_tune): Likewise.
6481 (thunderxt88_prefetch_tune): Likewise.
6482 (thunderx_prefetch_tune): Likewise.
6483 (thunderx2t99_prefetch_tune): Likewise.
6484 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6486 (aarch64_override_options_internal): Update to set
6487 PARAM_PREFETCH_DYNAMIC_STRIDES.
6488 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6489 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6490 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6491 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6492 prefetch-dynamic-strides setting.
6494 2018-05-07 Luis Machado <luis.machado@linaro.org>
6496 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6497 <minimum_stride>: New const int field.
6498 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6499 minimum_stride field.
6500 (exynosm1_prefetch_tune): Likewise.
6501 (thunderxt88_prefetch_tune): Likewise.
6502 (thunderx_prefetch_tune): Likewise.
6503 (thunderx2t99_prefetch_tune): Likewise.
6504 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6505 (aarch64_override_options_internal): Update to set
6506 PARAM_PREFETCH_MINIMUM_STRIDE.
6507 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6508 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6509 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6510 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6511 stride is constant and is below the minimum stride threshold.
6513 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6516 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6517 the type is addressable. Don't force op into register if it has
6520 2018-05-05 Roland McGrath <mcgrathr@google.com>
6523 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6524 any section for which we don't know a specific type it should have,
6525 regardless of name. Previously this was done only for the exact
6526 names ".init_array", ".fini_array", and ".preinit_array".
6527 (default_elf_asm_named_section): Add comment about
6528 relationship with default_section_type_flags and SECTION_NOTYPE.
6529 (get_section): Don't consider it a type conflict if one side has
6530 SECTION_NOTYPE and the other doesn't, as long as neither has the
6531 SECTION_BSS et al used in the default_section_type_flags logic.
6533 2018-05-05 Tom de Vries <tom@codesourcery.com>
6536 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6537 (workaround_barsyncs): New function.
6538 (nvptx_reorg): Use workaround_barsyncs.
6539 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6540 (define_expand "nvptx_membar_cta"): New define_expand.
6541 (define_insn "*nvptx_membar_cta"): New insn.
6543 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6545 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6546 To improve optimization opportunities.
6547 * builtin-types.def: The new needed builtin types for the above.
6549 2018-05-04 Richard Biener <rguenther@suse.de>
6551 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6552 * gimple-ssa-store-merging.c
6553 (imm_store_chain_info::output_merged_store): Remove redundant create,
6554 release split_store vector contents on failure.
6555 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6556 scalar stmt vector on cache hit.
6558 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6560 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6562 * config.gcc (powerpc-xilinx-eabi*): Remove.
6563 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6565 (fusion_addis_mem_combo_load): Ditto.
6566 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6568 (rs6000_cpu_cpp_builtins): Ditto.
6569 * config/rs6000/rs6000-linux.c
6570 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6571 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6572 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6574 (rs6000_setup_reg_addr_masks): Ditto.
6575 (rs6000_init_hard_regno_mode_ok): Ditto.
6576 (rs6000_option_override_internal): Ditto.
6577 (legitimate_lo_sum_address_p): Ditto.
6578 (rs6000_legitimize_address): Ditto.
6579 (rs6000_legitimize_reload_address): Ditto.
6580 (rs6000_legitimate_address_p): Ditto.
6581 (abi_v4_pass_in_fpr): Ditto.
6582 (setup_incoming_varargs): Ditto.
6583 (rs6000_gimplify_va_arg): Ditto.
6584 (rs6000_split_multireg_move): Ditto.
6585 (rs6000_savres_strategy): Ditto.
6586 (rs6000_emit_prologue_components): Ditto.
6587 (rs6000_emit_epilogue_components): Ditto.
6588 (rs6000_emit_prologue): Ditto.
6589 (rs6000_emit_epilogue): Ditto.
6590 (rs6000_elf_file_end): Ditto.
6591 (rs6000_function_value): Ditto.
6592 (rs6000_libcall_value): Ditto.
6593 * config/rs6000/rs6000.h: Ditto.
6594 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6595 (TARGET_MINMAX): ... this. New.
6596 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6597 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6598 (*movsi_internal1_single): Delete.
6599 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6600 mfpu=, mxilinx-fpu): Delete.
6601 * config/rs6000/singlefp.h: Delete.
6602 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6603 * config/rs6000/t-rs6000: Ditto.
6604 * config/rs6000/t-xilinx: Delete.
6605 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6606 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6607 (VStype_simple): Delete.
6608 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6609 * config/rs6000/xfpu.h: Delete.
6610 * config/rs6000/xfpu.md: Delete.
6611 * config/rs6000/xilinx.h: Delete.
6612 * config/rs6000/xilinx.opt: Delete.
6613 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6614 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6616 2018-05-04 Tom de Vries <tom@codesourcery.com>
6619 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6622 2018-05-04 Richard Biener <rguenther@suse.de>
6625 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6626 (expand_complex_div_wide): Likewise.
6627 (expand_complex_operations_1): Likewise.
6628 (expand_complex_libcall): Preserve EH info of the original stmt.
6629 (tree_lower_complex): Handle removed blocks.
6630 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6631 on complex multiplication and division libcall builtins.
6633 2018-05-04 Richard Biener <rguenther@suse.de>
6636 * fold-const.c (negate_expr_p): Restrict negation of operand
6637 zero of a division to when we know that can happen without
6639 (fold_negate_expr_1): Likewise.
6641 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6644 * real.h (real_nextafter): Declare.
6645 * real.c (real_nextafter): New function.
6646 * fold-const-call.c (fold_const_nextafter): New function.
6647 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6648 CASE_CFN_NEXTTOWARD.
6649 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6650 even when arg1_mode is different from arg0_mode.
6652 2018-05-03 Nathan Sidwell <nathan@acm.org>
6654 * doc/extend.texi (Deprecated Features): Remove
6656 (Backwards Compatibility): Likewise.
6657 * doc/invoke.texi (C++ Language Options): Likewise.
6658 (C++ Dialect Options): Likewise.
6660 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6663 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6664 _mm512_mask_mullox_epi64): New intrinsics.
6666 2018-05-03 Tom de Vries <tom@codesourcery.com>
6669 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6670 dump files): Add offload-tree.
6672 2018-05-03 Richard Biener <rguenther@suse.de>
6674 PR tree-optimization/85615
6675 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6676 to loops not nested in BBs loop father to avoid creating multi-entry
6679 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6681 PR tree-optimization/70291
6682 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6683 arguments. Change return type to tree. Emit libcall as a new
6684 statement rather than replacing existing one when inplace_p is true.
6685 (expand_complex_multiplication_components): New function.
6686 (expand_complex_multiplication): Expand floating-point complex
6687 multiplication using the above.
6688 (expand_complex_division): Rename inner_type parameter to type.
6689 Update expand_complex_libcall call-site.
6690 (expand_complex_operations_1): Update expand_complex_multiplication
6691 and expand_complex_division call-sites.
6693 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6696 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6697 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6698 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6699 the highest significant bit of the shift count mask is clear. In
6700 check whether and[sq]i3 is needed verify that all significant bits
6701 of the shift count other than the highest are set.
6703 2018-05-02 Tom de Vries <tom@codesourcery.com>
6706 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6707 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6708 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6709 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6710 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6711 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6712 __builtin_goacc_parlevel_size.
6714 2018-05-02 Richard Biener <rguenther@suse.de>
6716 PR tree-optimization/85597
6717 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6718 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6721 2018-05-02 Tom de Vries <tom@codesourcery.com>
6724 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6725 dump files): Add ltrans-tree.
6727 2018-05-02 Tom de Vries <tom@codesourcery.com>
6730 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6731 dump files): Add wpa-ipa.
6733 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6735 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6736 powerpc*-*-linux*paired* target.
6737 * config/rs6000/750cl.h: Delete.
6738 * config/rs6000/paired.h: Delete.
6739 * config/rs6000/paired.md: Delete.
6740 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6742 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6743 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6744 comment. Remove paired float support.
6745 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6746 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6748 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6749 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6751 * config/rs6000/rs6000.c: Remove paired float support.
6752 (paired_expand_vector_init, paired_expand_vector_move,
6753 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6754 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6755 paired_expand_builtin, paired_expand_predicate_builtin,
6756 paired_init_builtins): Delete.
6757 * config/rs6000/rs6000.h: Remove paired float support.
6758 * config/rs6000/rs6000.md: Remove paired float support.
6759 (move_from_CR_ov_bit): Delete.
6760 * config/rs6000/rs6000.opt (mpaired): Delete.
6761 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6762 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6764 2018-05-02 Richard Biener <rguenther@suse.de>
6767 * gimplify.c (gimplify_save_expr): When in SSA form allow
6768 SAVE_EXPRs to compute to SSA vars.
6770 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6773 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6774 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6775 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6776 clobber operands[2], instead use a new pseudo. Formatting fixes.
6778 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6780 PR tree-optimization/85586
6781 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6782 exit early for statements in the same group if the accesses are
6785 2018-05-02 Tom de Vries <tom@codesourcery.com>
6788 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6791 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6793 PR tree-optimization/85143
6794 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6796 2018-05-01 Tom de Vries <tom@codesourcery.com>
6799 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6800 not found" error message.
6802 2018-05-01 Tom de Vries <tom@codesourcery.com>
6805 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6806 * vec.c (test_ordered_remove_if): New function.
6807 (vec_c_tests): Call test_ordered_remove_if.
6808 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6809 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6810 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6811 VEC_ORDERED_REMOVE_IF.
6813 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6815 PR tree-optimization/82665
6816 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6817 pointer subtraction where arguments come from a memchr call.
6819 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6821 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6822 --push-state --as-needed and --pop-state instead of --as-needed and
6823 --no-as-needed if ld supports it.
6824 * configure: Regenerated.
6827 * doc/install.texi2html: Replace _002d with - and _002a with * in
6828 generated html files using sed.
6830 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6833 * gcc-rich-location.c (blank_line_before_p): New function.
6834 (use_new_line): New function.
6835 (gcc_rich_location::add_fixit_insert_formatted): New function.
6836 * gcc-rich-location.h
6837 (gcc_rich_location::add_fixit_insert_formatted): New function.
6839 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6841 * selftest.c (assert_streq): Rename "expected" and "actual" to
6842 "val1" and "val2". Extend NULL-handling to cover both inputs
6843 symmetrically, while still requiring both to be non-NULL for a pass.
6844 * selftest.h (assert_streq): Rename "expected" and "actual" to
6846 (ASSERT_EQ): Likewise.
6847 (ASSERT_EQ_AT): Likewise.
6848 (ASSERT_KNOWN_EQ): Likewise.
6849 (ASSERT_KNOWN_EQ_AT): Likewise.
6850 (ASSERT_NE): Likewise.
6851 (ASSERT_MAYBE_NE): Likewise.
6852 (ASSERT_MAYBE_NE_AT): Likewise.
6853 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6854 the assertion to pass.
6855 (ASSERT_STREQ_AT): Likewise.
6857 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6859 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6860 interaction with -pie.
6862 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6864 * selftest.h: Fix alphabetization of per-source-file selftest
6867 2018-04-30 Jason Merrill <jason@redhat.com>
6869 PR c++/61982 - dead stores to destroyed objects.
6870 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6873 2018-04-30 Jason Merrill <jason@redhat.com>
6875 * tree.c (build_clobber): New.
6876 * tree.h: Declare it.
6877 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6879 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6881 * diagnostic-show-locus.c (layout::layout): Update for
6882 location_get_source_line returning a char_span.
6883 (struct char_span): Move to input.h.
6884 (struct correction): Update for fields in char_span becoming
6886 (struct source_line): Update for location_get_source_line
6887 returning a char_span.
6888 (layout::print_line): Likewise.
6889 * edit-context.c (edited_file::print_content): Likewise.
6890 (edited_file::print_diff_hunk): Likewise.
6891 (edited_file::print_run_of_changed_lines): Likewise.
6892 (edited_file::get_num_lines): Likewise.
6893 (edited_line::edited_line): Likewise.
6894 * final.c (asm_show_source): Likewise.
6895 * input.c (location_get_source_line): Convert return type
6896 from const char * to char_span, losing the final "line_len"
6898 (dump_location_info): Update for the above.
6899 (get_substring_ranges_for_loc): Likewise. Use a char_span
6900 when handling the literal within the line.
6901 (test_reading_source_line): Update for location_get_source_line
6902 returning a char_span.
6903 * input.h (class char_span): Move here from
6904 diagnostic-show-locus.c, converting from a struct to a class.
6905 Make data members private.
6906 (char_span::operator bool): New.
6907 (char_span::length): New.
6908 (char_span::get_buffer): New.
6909 (char_span::operator[]): New.
6910 (char_span::subspan): Make const.
6911 (char_span::xstrdup): New.
6912 (location_get_source_line): Convert return type from const char *
6913 to char_span, losing the final "line_size" param.
6915 2018-04-30 Jan Hubicka <jh@suse.cz>
6917 * lto-wrapper.c (ltrans_priorities): New static var.
6918 (cmp_priority): New.
6919 (run_gcc): Read priorities and if doing parallel build order
6920 the Makefile by them.
6922 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6924 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6926 2018-04-30 Richard Biener <rguenther@suse.de>
6928 * tree-cfg.c (verify_address): Remove base argument, add
6929 flag whether to check TREE_ADDRESSABLE and do that.
6930 (verify_expr): Remove.
6931 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6932 (verify_gimple_assign_single): Likewise.
6933 (verify_gimple_switch): Likewise.
6934 (verify_expr_location_1): Dereference tp once. Add (disabled)
6935 piece from verify_expr.
6936 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6938 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6940 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6942 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6944 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6945 (small_data_pattern): Likewise.
6946 (arc_rewrite_small_data): Likewise.
6947 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6948 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6949 (get_symbol_alignment): New function.
6950 (legitimate_small_data_address_p): Likewise.
6951 (legitimate_scaled_address): Update, call
6952 legitimate_small_data_address_p.
6953 (output_sdata): New static variable.
6954 (arc_print_operand): Update how we handle small data operands.
6955 (arc_print_operand_address): Likewise.
6956 (arc_legitimate_address_p): Update, use
6957 legitimate_small_data_address_p.
6958 (arc_rewrite_small_data_p): Remove.
6959 (arc_rewrite_small_data_1): Likewise.
6960 (arc_rewrite_small_data): Likewise.
6961 (small_data_pattern): Likewise.
6962 (compact_sda_memory_operand): Update to use
6963 legitimate_small_data_address_p and get_symbol_alignment.
6964 (prepare_move_operands): Don't rewite sdata pattern.
6965 (prepare_extend_operands): Remove.
6966 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6968 (zero_extendqisi2): Likewise.
6969 (zero_extendhisi2): Likewise.
6970 (extendqihi2): Likewise.
6971 (extendqisi2): Likewise.
6972 (extendhisi2): Likewise.
6976 * config/arc/constraints.md (Usd): Change it to memory constraint.
6978 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6980 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6981 as source of std instructions.
6982 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6983 allow 6-bit constants as source for store instructions.
6984 (movdi_insn): Update instruction pattern to allow 6-bit constants
6985 as source for store instructions.
6987 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6989 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6991 2018-04-30 Nathan Sidwell <nathan@acm.org>
6992 Sandra Loosemore <sandra@codesourcery.com>
6994 * dumpfile.c (dump_open): Allow '-' for stdout.
6995 * doc/invoke.texi (Developer Options): Document dump filename
6996 determination early. Document stdin/stdout selection.
6998 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7000 Microblaze Target: PIC data text relative
7002 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7003 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7005 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7006 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7007 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7008 New addressing mode for data-text relative position indepenedent code.
7009 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7010 'ADDRESS_SYMBOLIC_TXT_REL'.
7011 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7012 (microblaze_legitimate_pic_operand): Exclude function calls from
7013 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7014 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7016 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7017 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7018 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7019 for 'address + offset'.
7020 (microblaze_expand_prologue): Add new function prologue call for
7022 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7023 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7024 table in case of TARGET_PIC_DATA_TEXT_REL.
7025 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7026 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7027 Add new macros 'UNSPEC_TEXT',
7028 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7029 + exclude function calls from 'UNSPEC_PLT' in case of data text
7031 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7032 new target hook for generating address diff vector tables in case of
7034 * doc/tm.texi : Regenerate.
7035 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7036 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7037 of addr diff vector generation.
7038 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7039 target hook definition.
7040 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7041 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7042 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7043 option for data text relative.
7045 2018-04-30 Richard Biener <rguenther@suse.de>
7047 * tree-chrec.h (evolution_function_is_constant_p): Remove
7049 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7051 2018-04-30 Richard Biener <rguenther@suse.de>
7054 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7056 2018-04-30 Richard Biener <rguenther@suse.de>
7058 PR tree-optimization/28364
7059 PR tree-optimization/85275
7060 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7061 copying first exit test.
7063 2018-04-28 Mark Wielaard <mark@klomp.org>
7065 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7067 (dwarf_AT): Handle DW_AT_addr_base.
7068 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7070 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7073 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7074 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7075 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7076 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7078 2018-04-28 Richard Biener <rguenther@suse.de>
7080 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7081 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7082 to reflect use. Only add interesting stmts.
7084 2018-04-27 Martin Jambor <mjambor@suse.cz>
7087 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7088 the jump function allows for passing through aggregate values.
7090 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7092 * input.h (in_system_header_at): Convert from macro to inline
7094 (from_macro_expansion_at): Likewise.
7095 (from_macro_definition_at): Likewise.
7097 2018-04-27 Jeff Law <law@redhat.com>
7099 * config.gcc: Mark tile* targets as deprecated/obsolete.
7101 2018-04-27 Richard Biener <rguenther@suse.de>
7103 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7106 2018-04-27 Richard Biener <rguenther@suse.de>
7108 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7110 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7112 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7113 with Yd constraint. Set "preferred_for_speed" attribute from
7114 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7116 (*movdi_internal): Ditto.
7117 (movti_interunit splitters): Remove
7118 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7119 (movdi_interunit splitters): Ditto.
7120 * config/i386/constraints.md (Ye): Remove.
7121 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7123 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7126 * config/aarch64/constraints.md (Usg): Limit to 31.
7129 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7131 PR tree-optimization/85529
7132 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7133 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7134 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7135 zero extension or masking of the MSB bit.
7136 (optimize_range_tests): Add FIRST_BB argument, pass it through
7137 to optimize_range_tests_var_bound.
7138 (maybe_optimize_range_tests, reassociate_bb): Adjust
7139 optimize_range_tests callers.
7141 2018-04-26 Richard Biener <rguenther@suse.de>
7142 Jakub Jelinek <jakub@redhat.com>
7144 * cgraph.h (symbol_table): Just declare debug method here.
7145 * symtab.c (symbol_table::debug): Define.
7147 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7149 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7151 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7153 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7154 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7155 (*movdi_internal): Substitute Yi and Yj constraint with x
7156 and Ym and Yn constraint with y constraint. Update "isa"
7157 attribute and set "preferred_for_speed" attribute from
7158 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7159 (*movsi_internal): Ditto.
7160 (*movdf_internal): Ditto.
7161 (*movsf_internal): Ditto.
7162 (*zero_extendsidi2): Ditto.
7163 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7164 (sse2_loadld): Ditto.
7165 (*vec_extract<ssevecmodelower>_0): Ditto.
7166 (*vec_extractv4si_0_zext_sse4): Ditto.
7167 (vec_concatv2di): Ditto.
7168 (*vec_dup<mode>): Ditto.
7169 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7170 * config/i386/constraints.md (Yi): Remove.
7175 2018-04-26 Nathan Sidwell <nathan@acm.org>
7177 * dumpfile.c (dump_open): New.
7178 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7179 (dump_finish): Detect stdio/stderr by value not name.
7181 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7183 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7185 2018-04-26 Tom de Vries <tom@codesourcery.com>
7188 * config/nvptx/nvptx.c (verify_neutering_jumps)
7189 (verify_neutering_labels): New function
7190 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7192 2018-04-26 Tom de Vries <tom@codesourcery.com>
7195 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7196 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7199 2018-04-26 Richard Biener <rguenther@suse.de>
7200 Tom de Vries <tom@codesourcery.com>
7203 * lto-streamer-out.c (output_function): Fixup loops if required to match
7204 discovery done in the reader.
7206 2018-04-26 Richard Biener <rguenther@suse.de>
7208 PR tree-optimization/85116
7209 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7210 have a loop exit from the single latch predecessor. Remove
7211 case of header with just condition.
7212 (ch_base::copy_headers): Exclude infinite loops from any
7214 (pass_ch::execute): Record exits.
7216 2018-04-26 Richard Biener <rguenther@suse.de>
7218 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7219 prologue cost vector and pass it to vect_get_load_cost.
7220 (vect_get_peeling_costs_all_drs): Likewise.
7221 (vect_peeling_hash_get_lowest_cost): Likewise.
7222 (vect_enhance_data_refs_alignment): Likewise.
7224 2018-04-26 Richard Biener <rguenther@suse.de>
7227 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7228 checking of integer<->pointer conversions.
7229 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7230 sign-/zero-extending pointer types.
7231 (expand_omp_for_static_chunk): Likewise.
7233 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7234 Jean Lee <xiaoyur347@gmail.com>
7236 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7237 (TARGET_ASAN_SHADOW_OFFSET): Define.
7238 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7239 true for -fsanitize=address.
7241 2018-04-25 Mark Wielaard <mark@klomp.org>
7243 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7246 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7248 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7249 than "alu", remove explicit "memory" and "imm_disp" attributes.
7250 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7253 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7254 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7255 gen_lowpart_no_emit.
7257 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7260 * config/i386/i386.c (ix86_expand_builtin): Change memory
7261 operand to XI, extend p0 to Pmode.
7262 * config/i386/i386.md: Change unspec volatile and operand
7263 1 mode to XI, change operand 0 mode to P.
7265 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7267 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7268 GET_MODE_MASK before any checking.
7269 (nds32_can_use_bset_p): Likewise.
7270 (nds32_can_use_btgl_p): Likewise.
7272 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7274 * config/nds32/nds32-doubleword.md: New define_split pattern for
7275 illegal register number.
7277 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7279 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7281 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7283 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7285 2018-04-25 Richard Biener <rguenther@suse.de>
7287 * lto-streamer.h (LTO_major_version): Bump to 8.
7289 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7291 * BASE-VER: Set to 9.0.0.
7293 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7295 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7296 in __abskf2 and __powikf2.
7298 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7301 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7302 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7303 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7304 Use the above on operand 2. Reindent.
7305 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7307 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7310 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7312 * config/i386/i386.opt (mcet): Removed.
7313 * doc/install.texi: Remove -mcet documentation.
7314 * doc/invoke.texi: Likewise.
7316 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7319 * doc/install.texi: Remove -mcet from bootstrap-cet.
7321 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7324 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7325 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7329 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7330 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7331 containing a CONST_VECTOR.
7333 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7335 * doc/install.texi: Update newlib dependency for nvptx.
7337 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7340 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7341 instead of INTVAL when shifting x left.
7343 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7345 PR tree-optimization/85478
7346 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7347 vect_grouped_store_supported for single element vectors.
7349 2018-04-24 Richard Biener <rguenther@suse.de>
7352 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7353 load cost increase to the case of non-constant step.
7355 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7358 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7359 destination if any_malformed_asm.
7361 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7364 * expr.c (store_field): In the bitfield case, if the value comes from
7365 a function call and is returned in registers by means of a PARALLEL,
7366 do not change the mode of the temporary unless BLKmode and VOIDmode.
7368 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7370 PR rtl-optimization/85423
7371 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7372 dependencies to debug insns when the previous insn is non-debug.
7374 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7376 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7377 enums into a single definition.
7378 (fls): Fix predicates and printing.
7381 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7383 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7384 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7385 and short u6 immediate.
7386 (check_if_valid_sleep_operand): Remove.
7387 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7389 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7391 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7392 flag_always_save_lp condition.
7393 * config/nds32/nds32.opt (malways-save-lp): New option.
7395 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7397 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7398 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7399 * config/nds32/nds32.h
7400 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7401 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7403 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7405 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7406 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7408 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7409 Chung-Ju Wu <jasonwucj@gmail.com>
7411 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7413 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7414 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7416 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7418 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7420 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7422 * config/nds32/nds32-protos.h (nds32_data_alignment,
7423 nds32_local_alignment): Declare.
7424 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7425 nds32_local_alignment): New functions.
7426 (TARGET_CONSTANT_ALIGNMENT): Define.
7427 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7429 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7431 * config/nds32/nds32.c
7432 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7433 (TARGET_MODES_TIEABLE_P): Likewise.
7435 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7437 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7440 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7441 Chung-Ju Wu <jasonwucj@gmail.com>
7443 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7444 for unaligned access.
7445 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7447 * config/nds32/nds32-intrinsic.md: Likewise.
7448 * config/nds32/nds32_intrinsic.h: Likewise.
7449 * config/nds32/nds32.h (nds32_builtins): Likewise.
7450 * config/nds32/nds32.opt (munaligned-access): New option.
7451 * config/nds32/nds32.c (nds32_asm_file_start): Display
7452 flag_unaligned_access status.
7454 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7456 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7457 -mno-relax is present.
7458 * config/riscv/linux.h (LINK_SPEC): Ditto.
7460 2018-04-20 Martin Sebor <msebor@redhat.com>
7463 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7465 (gimple_fold_builtin_stxcpy_chk): Same.
7466 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7468 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7471 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7472 __powikf2 when long double is IEEE 128-bit.
7474 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7476 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7477 step to make sure stack always aligned.
7479 2018-04-20 Carl Love <cel@us.ibm.com>
7482 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7483 size check for arg0.
7485 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7486 Tom de Vries <tom@codesourcery.com>
7489 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7490 Emit insns for calls too.
7491 (nvptx_find_par): Always look for worker-level predecessor insn.
7492 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7494 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7495 (nvptx_process_pars): Propagate frames for calls.
7497 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7500 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7502 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7503 (ix86_handle_option): Don't handle OPT_mibt.
7504 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7506 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7508 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7509 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7510 (ix86_target_macros): Define __CET__ with flag_cf_protection
7511 for -fcf-protection.
7512 * config/i386/i386.c (isa2_opts): Remove -mibt.
7513 * config/i386/i386.h (TARGET_IBT): Removed.
7514 (TARGET_IBT_P): Likewise.
7515 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7516 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7517 * config/i386/i386.opt (mcet): Update help message.
7520 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7521 -mcet as an alias for -mshstk.
7523 2018-04-20 Richard Biener <rguenther@suse.de>
7526 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7527 complexity by forcing a single use of the multiply operand.
7529 2018-04-20 Martin Jambor <mjambor@suse.cz>
7532 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7533 recursion dependency to only apply to non-clones.
7535 2018-04-20 Martin Jambor <mjambor@suse.cz>
7538 * ipa-cp.c (create_specialized_node): Check that clones of
7539 self-recursive edges exist during IPA-CP.
7541 2018-04-19 Toon Moene <toon@moene.org>
7543 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7546 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7548 PR tree-optimization/85467
7549 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7550 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7551 VECTOR_CST element to type.
7553 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7556 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7557 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7558 (builtin_longjmp): Likewise.
7559 (save_stack_nonlocal): New pattern.
7560 (restore_stack_nonlocal): Likewise.
7562 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7565 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7566 Replace ASM_OUTPUT_LABEL with fprintf.
7568 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7571 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7572 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7573 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7574 define __IBT__ and __SHSTK__ for -fcf-protection.
7575 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7577 (ix86_trampoline_init): Likewise.
7578 (x86_output_mi_thunk): Likewise.
7579 (ix86_notrack_prefixed_insn_p): Likewise.
7580 (ix86_option_override_internal): Don't disallow -fcf-protection.
7581 * config/i386/i386.md (rdssp<mode>): Also enable for
7583 (incssp<mode>): Likewise.
7584 (nop_endbr): Likewise.
7585 * config/i386/i386.opt (mcet): Change help message to built-in
7589 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7590 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7591 enable CET built-in functions.
7593 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7595 * common/config/i386/i386-common.c
7596 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7597 OPTION_MASK_ISA_MOVDIRI_UNSET,
7598 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7599 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7600 * config.gcc (movdirintrin.h): New header.
7601 * config/i386/cpuid.h (bit_MOVDIRI,
7602 bit_MOVDIR64B): New bits.
7603 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7605 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7606 (VOID, PVOID, PCVOID)): New function types.
7607 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7608 __builtin_ia32_directstoreu_u64,
7609 __builtin_ia32_movdir64b): New builtins.
7610 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7611 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7613 (ix86_valid_target_attribute_inner_p): Ditto.
7614 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7615 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7616 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7617 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7618 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7619 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7620 (movdiri<mode>, movdir64b_<mode>): New.
7621 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7622 * config/i386/immintrin.h: Include movdirintrin.h.
7623 * config/i386/movdirintrin.h: New file.
7624 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7626 2018-04-19 Richard Biener <rguenther@suse.de>
7629 * cfg.c (clear_bb_flags): When loop state says we have
7630 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7632 2018-04-19 Richard Biener <rguenther@suse.de>
7634 PR tree-optimization/84737
7635 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7636 copying restrict info.
7637 (vect_setup_realignment): Use it.
7638 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7639 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7640 the first DR to all generated stores.
7641 (vectorizable_load): Likewise for loads.
7643 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7645 PR tree-optimization/85446
7646 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7647 the integral and pointer types to have the same precision.
7649 * doc/install.texi: Document --disable-cet being the default and
7652 2018-04-18 Martin Liska <mliska@suse.cz>
7654 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7657 2018-04-18 Martin Liska <mliska@suse.cz>
7660 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7663 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7664 arguments if they are comparable.
7666 2018-04-18 Martin Liska <mliska@suse.cz>
7669 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7672 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7675 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7678 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7679 ENDBR after calling __morestack.
7681 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7684 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7685 by using gcc_base_ver to generate a gcc_driver_version, and use
7686 it when generating GCC_DRIVER_NAME.
7687 * configure: Regenerate.
7689 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7692 * config.gcc: Obsolete powerpc*-*-*spe*.
7694 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7697 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7698 (stabstr_D): Change type of unum from unsigned int to
7699 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7702 2018-04-17 Jim Wilson <jimw@sifive.com>
7705 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7706 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7707 Set arg_pointer_offset after using pretend_args_size.
7709 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7711 PR rtl-optimization/85431
7712 * dse.c (record_store): Ignore zero width stores.
7715 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7716 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7717 __builtin_stack_restore rather than after it.
7718 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7719 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7720 argument instead of virtual_dynamic_stack_rtx.
7722 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7724 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7726 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7727 Add note to error message to explain internal mapping of overloaded
7728 built-in function name to non-overloaded built-in function name.
7729 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7732 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7735 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7736 where the inputs overlap with the output.
7738 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7740 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7741 (=v, v) alternative and explicit "memory" attribute.
7742 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7743 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7745 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7746 "sselog1" type instead of "sselog".
7747 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7748 "sselog". Remove explicit "memory" attribute.
7749 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7750 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7752 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7753 "sselog1" type instead of "sselog". Remove explicit "memory"
7755 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7756 use "sselog1" type instead of "sselog". Remove explicit "memory"
7758 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7759 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7761 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7762 "sselog1" type instead of "sselog". Remove explicit "memory"
7764 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7765 use "sselog1" type instead of "sselog". Remove explicit "memory"
7769 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7772 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7775 2018-04-17 Martin Jambor <mjambor@suse.cz>
7778 * ipa-cp.c (create_specialized_node): Call
7779 expand_all_artificial_thunks if necessary.
7781 2018-04-17 Martin Liska <mliska@suse.cz>
7784 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7785 in message, remote space in between '_G' and '('.
7787 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7790 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7791 avx512f_vmcmp<mode>3<round_saeonly_name>,
7792 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7793 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7794 avx512f_rndscale<mode><round_saeonly_name>,
7795 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7796 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7797 Use %<iptr>2 instead of %2 for -masm=intel.
7798 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7799 avx512f_vcvttss2usi<round_saeonly_name>,
7800 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7802 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7803 avx512f_vcvttsd2usi<round_saeonly_name>,
7804 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7805 Use %q1 instead of %1 for -masm=intel.
7806 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7807 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7808 of %3 for -masm=intel.
7809 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7811 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7813 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7815 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7816 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7818 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7820 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7821 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7822 %g1 and one with %0 and %1.
7823 (avx512er_vmrcp28<mode><round_saeonly_name>,
7824 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7826 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7827 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7828 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7829 of %0 and %{%4%} for -masm=intel.
7830 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7831 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7832 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7833 order of %0 and %{%5%}%{z%} for -masm=intel.
7835 2018-04-17 Jan Hubicka <jh@suse.cz>
7838 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7840 2018-04-17 Martin Liska <mliska@suse.cz>
7843 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7844 for target_clone error message. Make default implementation
7845 clone to be a local declaration.
7846 (separate_attrs): Add new argument and check for an empty
7848 (expand_target_clones): Handle it.
7849 (ipa_target_clone): Make redirection just for target_clones
7852 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7853 Tom de Vries <tom@codesourcery.com>
7856 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7857 tiled basic blocks without omp continue statements.
7859 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7862 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7863 vec_extract expression as having side effects to make sure it gets
7866 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7869 * config/i386/i386.c (get_builtin_code_for_version): Check
7872 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7875 * gcc/config.gcc: Support "skylake".
7876 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7878 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7879 (processor_target_table): Add "skylake".
7880 (ix86_option_override_internal): Add "skylake".
7881 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7882 PROCESSOR_CANNONLAKE.
7883 (get_builtin_code_for_version): Fix priority for
7884 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7885 PROCESSOR_SKYLAKE-AVX512.
7886 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7887 (processor_type): Add PROCESSOR_SKYLAKE.
7889 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7890 Jason Merrill <jason@redhat.com>
7893 * convert.c (convert_to_integer_1): Use direct recursion for
7894 enumeral types and types with a precision less than the number
7895 of bits in their mode.
7897 2018-04-16 Julia Koval <julia.koval@intel.com>
7900 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7901 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7903 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7906 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7907 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7908 and -mno-direct-move.
7910 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7913 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7914 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7915 Ensure negative shifts result in {0}.
7917 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7919 PR rtl-optimization/79916
7920 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7921 regs (if any) to define how to gnerate SD moves when LRA is in
7924 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7926 PR rtl-optimization/85393
7927 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7928 * except.c (expand_dw2_landing_pad_for_region): Make static.
7929 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7930 a label and unconditional jump to old_bb, rather than
7931 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7934 PR rtl-optimization/85376
7935 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7936 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7937 instead of a specific value.
7939 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7940 Bin Cheng <bin.cheng@arm.com>
7942 PR tree-optimization/82965
7943 PR tree-optimization/83991
7944 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7945 by_profile_only parameter.
7946 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7947 information if the loop was predicted to iterate too many times.
7948 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7950 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7953 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7956 2018-04-13 Martin Liska <mliska@suse.cz>
7957 Jakub Jelinek <jakub@redhat.com>
7960 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7961 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7962 * builtins.c (expand_builtin_memory_copy_args): Use
7963 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7964 handle dest_addr == pc_rtx.
7966 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7969 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7970 asked to not generate direct moves.
7971 (fix_trunc<mode>si2_stfiwx): Similar.
7972 (fix_trunc<mode>si2_internal): Similar.
7974 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7977 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7978 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7979 lookup if dest in some wider mode is known to be const0_rtx and
7980 if so, record permanent equivalence for it to be ZERO_EXTEND of
7981 the narrower mode destination.
7983 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7985 * lto-streamer-out.c (output_function): Revert 259346.
7986 * omp-expand.c (expand_oacc_for): Likewise.
7988 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7990 PR rtl-optimization/85354
7991 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7992 * sel-sched.c (sel_global_init): ... here.
7994 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7997 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7998 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7999 mode for PE-COFF targets.
8000 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8001 (i386_pe_asm_lto_end): Likewise.
8002 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8003 (TARGET_ASM_LTO_END): Likewise.
8004 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8005 (i386_pe_asm_lto_start): New function.
8006 (i386_pe_asm_lto_end): Likewise.
8008 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8009 Richard Biener <rguenther@suse.de>
8012 * lto-streamer-out.c (output_function): Fix CFG loop state before
8014 * omp-expand.c (expand_oacc_for): Handle calls to internal
8015 functions like regular functions.
8017 2018-04-12 Richard Biener <rguenther@suse.de>
8020 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8021 for the early LTO debug to properly generate references to it
8022 during DIE emission. Do not re-use that for the skeleton for
8024 (dwarf2out_early_finish): Likewise.
8026 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8029 * config/i386/sse.md
8030 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8031 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8032 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8033 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8034 and output is a reg, avoid creating invalid lowpart subreg, but
8035 instead split into a 512-bit move. Don't split if not AVX512VL,
8036 input is xmm16+ reg and output is a mem.
8037 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8038 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8039 xmm16+ reg and output is a mem.
8041 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8043 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8044 also for flag_dwarf2_cfi_asm.
8046 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8048 PR rtl-optimization/85342
8049 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8050 a bool scalar var inside of the loop instead. Don't try to update
8051 recog_data.operand after failed apply_change_group.
8053 2018-04-12 Tom de Vries <tom@codesourcery.com>
8056 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8057 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8058 array with flexible array member as array without given dimension.
8059 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8060 argument for undefined param to true.
8062 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8065 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8066 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8067 from PowerPC section.
8068 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8069 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8070 help text that is too long.
8071 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8072 help text that is too long.
8073 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8074 help text that is too long.
8076 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8078 * config/alpha/alpha.md (stack_probe_internal): Rename
8079 from "probe_stack". Update all callers.
8081 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8083 PR rtl-optimization/84566
8084 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8085 sched_macro_fuse_insns.
8087 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8090 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8091 (compute_block_dependences): ... from here.
8093 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8095 PR tree-optimization/85331
8096 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8097 from int to HOST_WIDE_INT.
8099 2018-04-11 Martin Jambor <mjambor@suse.cz>
8102 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8103 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8104 not the same as the source val.
8105 (cgraph_edge_brings_value_p): New parameter.
8106 (gather_edges_for_value): Pass destination value to
8107 cgraph_edge_brings_value_p.
8108 (perhaps_add_new_callers): Likewise.
8109 (get_info_about_necessary_edges): Likewise and exclude values brought
8110 only by self-recursive edges.
8111 (create_specialized_node): Redirect only clones of self-calling edges.
8112 (+self_recursive_pass_through_p): New function.
8113 (find_more_scalar_values_for_callers_subset): Use it.
8114 (find_aggregate_values_for_callers_subset): Likewise.
8115 (known_aggs_to_agg_replacement_list): Removed.
8116 (decide_whether_version_node): Re-calculate known constants for all
8117 remaining context clones.
8119 2018-04-11 Richard Biener <rguenther@suse.de>
8122 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8123 from early DWARF output.
8124 (dwarf2out_early_finish): Output line info unconditionally into
8125 early DWARF and add reference to it.
8127 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8130 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8131 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8132 other than V2DFmode using iptr mode attribute.
8133 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8135 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8137 PR rtl-optimization/84659
8138 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8140 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8143 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8145 (output_loc_list): Pass address of a dummy size variable even in the
8146 locview handling loop.
8147 (index_location_lists): Add comment on why skip_loc_list_entry can't
8150 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8153 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8156 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8159 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8160 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8161 and -mstring-compare-inline-limit.
8163 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8166 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8167 for stack clash protection in a register whenever we need it to be in
8170 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8172 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8173 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8175 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8178 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8180 (mlong-double-): Ditto.
8181 * config/rs6000/sysv4.opt (msdata=): Ditto.
8182 (mtls-size=): Ditto.
8184 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8186 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8187 erroneous entries for
8188 "vector int vec_ldl (int, long int *)", and
8189 "vector unsigned int vec_ldl (int, unsigned long int *)".
8190 Add comments and entries for
8191 "vector bool char vec_ldl (int, bool char *)",
8192 "vector bool short vec_ldl (int, bool short *)",
8193 "vector bool int vec_ldl (int, bool int *)",
8194 "vector bool long long vec_ldl (int, bool long long *)",
8195 "vector pixel vec_ldl (int, pixel *)",
8196 "vector long long vec_ldl (int, long long *)",
8197 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8198 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8199 type tree bool_long_long_type_node and correct definition of
8200 bool_V2DI_type_node to make reference to this new type tree.
8201 (rs6000_mangle_type): Replace erroneous reference to
8202 bool_long_type_node with bool_long_long_type_node.
8203 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8204 comments to emphasize sign distinctions for char and int types and
8205 replace RS6000_BTI_bool_long constant with
8206 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8207 use of RS6000_BTI_pixel.
8208 (bool_long_type_node): Remove this macro definition.
8209 (bool_long_long_type_node): New macro definition
8211 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8213 PR rtl-optimization/85300
8214 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8215 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8216 simplify_unary_operation fails.
8218 2018-04-10 Martin Liska <mliska@suse.cz>
8220 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8221 cgraph_edge and ipa_ref.
8223 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8227 * config/i386/sse.md
8228 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8229 computation of the VEC_MERGE selector from mask.
8230 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8231 Fix decoding of the VEC_MERGE selector into mask.
8233 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8235 PR tree-optimization/85286
8236 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8238 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8240 * final.c (final_1): Set insn_last_address as well as
8241 insn_current_address.
8243 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8246 * explow.c (emit_stack_probe): Call validize_mem on memory location
8247 before passing it to gen_probe_stack. Create address operand and
8248 legitimize it for the probe_stack_address case.
8250 2018-04-09 Jan Hubicka <jh@suse.cz>
8253 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8254 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8255 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8257 (free_lang_data): Rebuild type inheritance graph.
8259 2018-04-09 Martin Sebor <msebor@redhat.com>
8261 * invoke.texi (-finline-small-functions): Mention other optimization
8263 (-findirect-inlining, -fpartial-inlining): Same.
8264 (-finline-functions-called-once): Same.
8265 (-freorder-blocks-and-partition): Same.
8267 2018-04-09 Jan Hubicka <jh@suse.cz>
8270 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8271 jumps; choose last target that matches the criteria (i.e.
8272 no partition changes for non-crossing jumps).
8273 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8274 support for redirecting crossing jumps to non-crossing.
8276 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8278 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8279 also for naked functions.
8281 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8283 * config/arc/arc.md (add_shift): New pattern.
8284 (add_shift2): Likewise.
8285 (sub_shift): Likewise.
8286 (sub_shift_cmp0_noout): Likewise.
8287 (compare_si_ashiftsi): Likewise.
8288 (xbfu_cmp0_noout): New combine pattern.
8289 (xbfu_cmp0"): Likewise.
8290 (movsi_set_cc_insn): Place the predicable variant first.
8291 (commutative_binary_cmp0_noout): Remove clobber.
8292 (commutative_binary_cmp0): New pattern.
8293 (noncommutative_binary_cmp0): Likewise.
8294 (noncommutative_binary_cmp0_noout): Likewise.
8295 (noncommutative_binary_comparison_result_used): Removed.
8296 (rsub_cmp0): New pattern.
8297 (rsub_cmp0_noout): Likewise.
8298 (extzvsi): Changed, keep only meaningful variants.
8299 (SQH, SEZ): New iterators.
8300 (SQH_postfix): New mode attribute.
8301 (SEZ_prefix): New code attribute.
8302 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8303 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8304 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8306 (noncommutative_operator): Check the availability of barrel
8309 2018-04-09 Richard Biener <rguenther@suse.de>
8311 PR tree-optimization/85284
8312 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8313 Only use the niter constraining form of simple_iv when the exit
8316 2018-04-09 Tom de Vries <tom@codesourcery.com>
8319 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8320 (define_expand "*memory_barrier"): New define_expand.
8321 (define_insn "memory_barrier"): New insn.
8323 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8325 PR rtl-optimization/80463
8326 PR rtl-optimization/83972
8327 PR rtl-optimization/83480
8329 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8330 correct producer for the insn.
8331 (tidy_control_flow): Fixup seqnos in case of debug insns.
8333 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8335 PR rtl-optimization/83913
8337 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8338 different sched-times when merging exprs.
8340 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8342 PR rtl-optimization/83962
8344 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8345 tidy_fallthru_edge and tidy_control_flow.
8347 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8349 PR rtl-optimization/83530
8351 * sel-sched.c (force_next_insn): New global variable.
8352 (remove_insn_for_debug): When force_next_insn is true, also leave only
8353 next insn in the ready list.
8354 (sel_sched_region): When the region wasn't scheduled, make another pass
8355 over it with force_next_insn set to 1.
8357 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8359 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8361 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8362 for interrupt control.
8363 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8364 functions for interrupt control.
8365 * config/nds32/nds32-intrinsic.md: Likewise.
8366 * config/nds32/nds32_intrinsic.h: Likewise.
8367 * config/nds32/nds32.h (nds32_builtins): Likewise.
8369 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8371 * config/nds32/nds32.c (nds32_init_machine_status,
8372 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8373 strict_aligned_p field.
8374 (nds32_expand_to_rtl_hook): New function.
8375 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8376 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8378 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8379 Chung-Ju Wu <jasonwucj@gmail.com>
8381 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8382 * config/nds32/nds32-n7.md: New file.
8383 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8384 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8386 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8387 * config/nds32/nds32.md (pipeline_model): Add n7.
8388 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8389 * config/nds32/pipelines.md: Include n7 settings.
8391 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8392 Chung-Ju Wu <jasonwucj@gmail.com>
8394 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8395 * config/nds32/nds32-e8.md: New file.
8396 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8397 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8399 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8400 * config/nds32/nds32.md (pipeline_model): Add e8.
8401 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8402 * config/nds32/pipelines.md: Include e8 settings.
8404 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8405 Chung-Ju Wu <jasonwucj@gmail.com>
8407 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8408 * config/nds32/nds32-n8.md: New file.
8409 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8410 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8412 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8413 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8414 * config/nds32/nds32.md (pipeline_model): Add n8.
8415 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8416 * config/nds32/pipelines.md: Include n8 settings.
8418 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8419 Chung-Ju Wu <jasonwucj@gmail.com>
8421 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8422 * config/nds32/nds32-n9-2r1w.md: New file.
8423 * config/nds32/nds32-n9-3r2w.md: New file.
8424 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8425 nds32_register_ports): New or modify for cpu n9.
8426 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8428 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8429 * config/nds32/nds32-utils.c: New file.
8430 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8431 TARGET_MUL_SLOW): Define.
8432 * config/nds32/nds32.md (pipeline_model): New attribute.
8433 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8434 New options that support cpu n9.
8435 * config/nds32/pipelines.md: Include n9 settings.
8436 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8438 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8440 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8441 information if necessary.
8442 (output_cond_branch_compare_zero): Likewise.
8443 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8444 (nds32_target_alignment): Refine for alignment.
8445 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8446 (FUNCTION_BOUNDARY): Modify.
8447 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8449 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8451 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8453 * config/nds32/constants.md (unspec_volatile_element): Add values for
8454 TLB operation and data prefetch.
8455 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8456 functions for TLB operation and data prefetch.
8457 * config/nds32/nds32-intrinsic.md: Likewise.
8458 * config/nds32/nds32_intrinsic.h: Likewise.
8459 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8460 (nds32_print_operand): Likewise.
8461 * config/nds32/nds32.h (nds32_builtins): Likewise.
8463 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8464 Andrew Pinski <pinsika@gcc.gnu.org>
8467 * match.pd: Use constant_boolean_node of correct type instead of
8468 boolean_true_node or boolean_false_node for simplifying
8469 pointer comparisons to zero.
8471 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8473 PR tree-optimization/80021
8474 * tree.c (verify_type_variant): Make error call in verify_variant_match
8475 translatable and remove final full stop.
8477 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8479 * config/nds32/constants.md (unspec_volatile_element): Add
8480 UNSPEC_VOLATILE_EH_RETURN.
8481 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8482 nds32_output_stack_pop): Support dwarf exception handling process.
8483 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8484 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8485 exception handling process.
8486 (nds32_compute_stack_frame): Likewise.
8487 (nds32_return_addr_rtx): Likewise.
8488 (nds32_initial_elimination_offset): Likewise.
8489 (nds32_expand_prologue): Likewise.
8490 (nds32_expand_epilogue): Likewise.
8491 (nds32_dynamic_chain_address): New function.
8492 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8494 (DYNAMIC_CHAIN_ADDRESS): Define.
8495 (EH_RETURN_DATA_REGNO): Define.
8496 (EH_RETURN_STACKADJ_RTX): Define.
8497 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8498 patterns for dwarf exception handling.
8500 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8502 * config/nds32/nds32.h: Clean up obsolete macros.
8504 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8506 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8507 Add enum values for particular instructions.
8508 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8509 particular intrinsic functions.
8510 * config/nds32/nds32-intrinsic.md: Likewise.
8511 * config/nds32/nds32_intrinsic.h: Likewise.
8512 * config/nds32/nds32.h (nds32_builtins): Likewise.
8513 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8514 (btst, ave): New patterns for particular instructions.
8516 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8518 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8519 Add enum values for atomic load/store and memory sync.
8520 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8522 * config/nds32/nds32-intrinsic.md: Likewise.
8523 * config/nds32/nds32_intrinsic.h: Likewise.
8524 * config/nds32/nds32.h (nds32_builtins): Likewise.
8526 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8528 PR tree-optimization/85257
8529 * fold-const.c (native_encode_vector): If not all elts could fit
8530 and off is -1, return 0 rather than offset.
8531 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8532 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8533 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8534 adjust buffer in native_interpret_expr call.
8536 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8538 * config/nds32/constants.md (unspec_volatile_element): Add cache
8539 control enum values.
8540 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8541 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8542 * config/nds32/nds32.c (nds32_cctl_names): New.
8543 (nds32_print_operand): Handle cache control register names.
8544 * config/nds32/nds32.h (nds32_builtins): New enum values.
8545 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8547 * config/nds32/nds32.md (type): Add mmu.
8548 * config/nds32/pipelines.md (simple_insn): Add mmu.
8550 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8552 * config/nds32/nds32.md (type): Remove call.
8553 * config/nds32/pipelines.md (simple_insn): Likewise.
8555 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8557 * config/nds32/constants.md (unspec_volatile_element): Add
8558 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8559 UNSPEC_VOLATILE_FMFCFG.
8560 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8561 description for fmfcfg and fmfcsr.
8562 (bdesc_1arg): Add fmtcsr.
8563 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8564 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8565 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8566 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8567 unspec_fmfcfg): New patterns.
8568 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8569 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8570 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8571 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8572 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8573 __nds32__fmfcfg): Define.
8575 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8577 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8578 intrinsic register names.
8579 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8580 intrinsic register enum values and macros.
8582 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8584 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8585 for load/store addressing form.
8586 (nds32_print_operand_address): Likewise.
8588 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8591 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8592 based on LABEL_REF. Remove useless assertion.
8593 (pic_address_needs_scratch): Fix formatting.
8594 (sparc_legitimize_pic_address): Minor tweaks.
8595 (sparc_delegitimize_address): Adjust assertion accordingly.
8596 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8597 into symbolic_operand.
8598 (movsi_high_pic_label_ref): Likewise.
8599 (movsi_lo_sum_pic_label_ref): Likewise.
8600 (movdi_pic_label_ref): Likewise.
8601 (movdi_high_pic_label_ref): Likewise.
8602 (movdi_lo_sum_pic_label_ref): Likewise.
8604 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8606 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8607 custom LIB_SPEC setup.
8609 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8610 Kito Cheng <kito.cheng@gmail.com>
8612 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8613 * config/riscv/freebsd.h: New.
8615 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8617 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8618 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8621 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8622 Kito Cheng <kito.cheng@gmail.com>
8624 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8625 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8626 * config/nds32/nds32-protos.h (nds32_output_call,
8627 nds32_output_return): Declare.
8628 * config/nds32/nds32.md: Refine all the call and return patterns.
8630 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8633 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8634 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8636 PR rtl-optimization/84872
8637 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8638 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8641 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8643 * expr.c (copy_blkmode_to_reg): Revert 254862.
8644 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8646 2018-04-06 Richard Biener <rguenther@suse.de>
8649 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8650 after seeing a component reference with an adjacent field. Treat
8651 refs to arrays at struct end of external decls similar to
8652 refs to unconstrained commons.
8654 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8657 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8658 look through SAVE_EXPRs with non-side-effects argument. Adjust
8660 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8663 2018-04-06 Richard Biener <rguenther@suse.de>
8666 * alias.c (find_base_term): New wrapper around find_base_term
8667 unwinding CSELIB_VAL_PTR changes.
8668 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8671 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8673 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8675 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8676 constant definitions.
8677 ("nop"): lr 0,0 -> nopr r0
8678 ("nop_lr0", "nop_lr1"): New insn definitions.
8680 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8682 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8683 NDS32_V3PUSH_AVAILABLE_P macro.
8685 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8686 Chung-Ju Wu <jasonwucj@gmail.com>
8688 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8689 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8690 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8691 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8692 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8693 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8694 * config/nds32/constraints.md: New constraints and checking for hard
8695 float configuration.
8696 * config/nds32/iterators.md: New mode iterator and attribute for hard
8697 float configuration.
8698 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8700 * config/nds32/nds32-fpu.md: New file.
8701 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8702 deal with hard float code generation.
8703 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8705 (abi_type, float_reg_number): New enum type.
8706 * config/nds32/nds32-predicates.c: New predicates for hard float.
8707 * config/nds32/nds32-protos.h: Declare functions for hard float.
8708 * config/nds32/nds32.c: Implementation for hard float configuration.
8709 * config/nds32/nds32.h: Definitions for hard float configuration.
8710 * config/nds32/nds32.md: Include hard float machine description and
8711 modify patterns for hard float configuration.
8712 * config/nds32/nds32.opt: New options for hard float configuration.
8713 * config/nds32/predicates.md: New predicates for hard float
8716 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8718 * common/config/nds32/nds32-common.c
8719 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8721 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8724 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8725 CONSTRUCTOR_ELT (ctor, ...)->value.
8727 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8730 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8732 2018-04-05 Tom de Vries <tom@codesourcery.com>
8735 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8738 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8739 Kito Cheng <kito.cheng@gmail.com>
8741 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8742 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8743 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8744 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8746 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8747 Kito Cheng <kito.cheng@gmail.com>
8749 * config/nds32/constraints.md (Ufe): New memory constraint.
8750 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8751 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8752 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8754 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8755 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8757 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8759 * config/nds32/nds32.md: Use optimize_size in the condition for
8760 alu-shift instructions.
8762 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8764 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8766 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8768 * config/nds32/nds32.md (negsi2): Refine pattern.
8770 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8771 Chung-Ju Wu <jasonwucj@gmail.com>
8773 * config/nds32/iterators.md (shift_rotate): New code iterator.
8774 (shift): New code attribute.
8775 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8776 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8777 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8778 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8779 bit-wise operations.
8780 (andsi3, *andsi3): Ditto.
8781 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8782 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8783 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8784 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8785 nds32_ior_operand, nds32_xor_operand): New predicates.
8787 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8789 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8790 (addsi3, subsi3): ... this.
8792 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8794 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8796 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8798 * config/nds32/nds32.md: Adjust indention.
8800 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8802 * config/nds32/nds32.md (feature): New attribute.
8804 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8806 * config/nds32/nds32.md (subtype): New attribute.
8808 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8811 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8812 expansion to perform a bitwise AND of the argument followed by a
8813 boolean negation of the result.
8815 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8817 PR rtl-optimization/84878
8818 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8819 the basic block. Assert the use reference is not artificial and that
8820 it has an associated insn.
8822 2018-04-04 Michael Matz <matz@suse.de>
8824 * builtins.c (compute_objsize): Pass correct operand
8825 to array_at_struct_end_p.
8827 2018-04-04 Richard Biener <rguenther@suse.de>
8830 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8831 from contexts for DINFO_LEVEL_TERSE and below.
8833 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8835 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8836 resiter_operand condition.
8837 * config/nds32/nds32.md (*move<mode>): Ditto.
8839 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8840 Monk Chiang <sh.chiang04@gmail.com>
8842 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8844 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8846 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8848 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8849 Kito Cheng <kito.cheng@gmail.com>
8851 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8852 nds32_cond_code_str, output_cond_branch,
8853 output_cond_branch_compare_zero, nds32_expand_cbranch,
8854 nds32_expand_cstore, nds32_expand_movcc,
8855 nds32_output_cbranchsi4_equality_zero,
8856 nds32_output_cbranchsi4_equality_reg,
8857 nds32_output_cbranchsi4_equality_reg_or_const_int,
8858 nds32_output_cbranchsi4_greater_less_zero: New functions.
8859 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8860 nds32_expand_cstore, nds32_expand_movcc,
8861 nds32_output_cbranchsi4_equality_zero,
8862 nds32_output_cbranchsi4_equality_reg,
8863 nds32_output_cbranchsi4_equality_reg_or_const_int,
8864 nds32_output_cbranchsi4_greater_less_zero): Declare.
8865 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8866 nds32_rimm11s_operand): New predicates.
8867 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8868 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8871 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8873 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8874 * config/nds32/nds32.md: Ditto.
8875 * config/nds32/pipelines.md: Ditto.
8877 2018-04-04 Richard Biener <rguenther@suse.de>
8879 PR tree-optimization/85168
8880 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8881 propagating abnormals.
8883 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8885 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8887 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8888 Kito Cheng <kito.cheng@gmail.com>
8890 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8891 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8892 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8893 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8894 * config/nds32/nds32.md (sibcall_internal): New.
8895 (sibcall_register): Remove.
8896 (sibcall_immediate): Remove.
8897 (sibcall_value_internal): New.
8898 (sibcall_value_register): Remove.
8899 (sibcall_value_immediate): Remove.
8900 * config/nds32/predicates.md (nds32_general_register_operand): New.
8901 (nds32_call_address_operand): New.
8903 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8905 PR rtl-optimization/85167
8906 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8907 bb_defs if *split_p, instead preinitialize it to NULL.
8909 PR tree-optimization/85156
8910 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8911 evaluating the argument multiple times.
8913 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8915 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8917 (_mm_cvtpd_ps): Likewise.
8918 (_mm_cvttpd_epi32): Likewise.
8919 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8920 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8921 vector, pixel, and bool following altivec.h include.
8923 2018-04-03 Martin Sebor <msebor@redhat.com>
8925 * doc/extend.texi (Common Function Attributes): Clarify.
8926 (const attribute): Likewise.
8927 (pure attribute): Likewise.
8929 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8932 * config/i386/i386.c (ix86_expand_vector_set): Use
8933 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8935 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8937 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8938 instructions when changing rounding bits to preserve precision bits
8939 in the x87 control word.
8941 2018-04-03 Martin Liska <mliska@suse.cz>
8943 PR tree-optimization/82491
8944 * rtl.h (strip_offset_and_add): Replace += suboffset with
8945 poly_uint64 () + suboffset.
8947 2018-03-29 Martin Liska <mliska@suse.cz>
8948 Martin Jambor <mjambor@suse.cz>
8951 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8952 param_type is not an integral or pointer type.
8954 2018-04-03 Richard Biener <rguenther@suse.de>
8956 * sese.h (recompute_all_dominators): Remove.
8958 2018-04-02 Martin Sebor <msebor@redhat.com>
8960 * doc/invoke.texi (-Wrestrict): Fix typos.
8962 2018-04-02 Jim Wilson <jimw@sifive.com>
8964 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8965 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8966 (<optab>di3, <optab>si3_extend): Likewise.
8967 (<optab>si3_mask, <optab>si3_mask_1): New.
8968 (<optab>di3_mask, <optab>di3_mask_1): New.
8969 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8970 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8971 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8973 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8975 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8978 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8980 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8981 (nds32_canonicalize_comparison): New function.
8983 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8984 Kito Cheng <kito.cheng@gmail.com>
8985 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8987 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8988 * config/nds32/constants.md (unspec_volatile_element): Add
8989 UNSPEC_VOLATILE_RELAX_GROUP.
8990 * config/nds32/nds32-relax-opt.c: New file.
8991 * config/nds32/nds32-predicates.c
8992 (nds32_symbol_load_store_p): New function.
8993 * config/nds32/nds32-protos.h
8994 (nds32_symbol_load_store_p): Declare function.
8995 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8996 * config/nds32/nds32.c
8997 (nds32_register_pass): New function to register pass.
8998 (nds32_register_passes): New function to register passes.
8999 * config/nds32/nds32.md (relax_group): New pattern.
9000 * config/nds32/nds32.opt (mrelax-hint): New option.
9001 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9003 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9005 * config/nds32/t-nds32: Modify files dependency.
9007 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9009 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9010 (PROFILE_HOOK): Define its implementation.
9012 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9014 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9015 type and 32-bit size.
9017 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9020 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9021 (V_128_256): New mode iterator.
9022 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9023 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9024 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9026 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9027 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9029 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9032 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9033 NaN inputs correctly.
9035 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9038 * config/rs6000/vsx.md (??r): New mode attribute.
9039 (*vsx_mov<mode>_64bit): Use it.
9040 (*vsx_mov<mode>_32bit): Likewise.
9042 2018-03-30 Martin Sebor <msebor@redhat.com>
9044 PR tree-optimization/84818
9045 * builtins.c (check_access): Use warning_n.
9047 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9050 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9052 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9055 2018-03-30 Julia Koval <julia.koval@intel.com>
9058 * x86-tune.def (movx, partial_reg_dependency): Enable for
9061 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9064 * lra-constraints.c (process_alt_operands): Move setting
9065 this_alternative_matches below.
9067 2018-03-29 Martin Liska <mliska@suse.cz>
9070 * doc/invoke.texi: Document how LTO works with debug info.
9071 Describe auto-load support of binutils. Mention 'x86-64'
9072 as valid option value of -march option.
9074 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9076 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9079 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9080 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9081 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9084 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9087 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9088 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9089 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9090 for RS6000_BTM_POWERPC64.
9091 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9092 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9093 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9098 2018-03-28 Carl Love <cel@us.ibm.com>
9101 2017-09-27 Carl Love <cel@us.ibm.com>
9103 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9104 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9105 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9108 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9110 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9111 instead of __vector bool.
9112 (_mm_max_pu8): Likewise.
9113 (_mm_min_pi16): Likewise.
9115 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9118 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9119 (DIVWEUO): Likewise.
9121 (DIVDEUO): Likewise.
9122 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9123 DIVWEUO and DIVDEUO.
9124 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9125 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9126 (div_extend): Likewise.
9127 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9129 (__builtin_divweuo): Likewise.
9130 (__builtin_divdeo): Likewise.
9131 (__builtin_divdeuo): Likewise.
9133 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9136 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9137 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9139 PR tree-optimization/82004
9140 * gimple-match-head.c (optimize_pow_to_exp): New function.
9141 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9142 Don't fold to exp if optimize_pow_to_exp is false.
9144 2018-03-28 Martin Liska <mliska@suse.cz>
9147 * calls.c (initialize_argument_information): Fix trailing space.
9148 * common.opt: Fix typo and provide better explanation for
9149 -fsanitize-coverage option.
9150 * config/i386/i386.opt: Fix typo.
9152 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9153 Martin Liska <mliska@suse.cz>
9156 * gimplify.c (asan_poison_variable): Don't do the check for
9157 gimplify_omp_ctxp here.
9158 (gimplify_decl_expr): Do it here.
9159 (gimplify_target_expr): Likewise.
9161 2018-03-28 Martin Liska <mliska@suse.cz>
9164 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9165 chkp_type_bounds_count if MPX is not enabled.
9167 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9169 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9171 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9174 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9175 function to create the function decl for complex long double
9176 multiply and divide for -mabi=ieeelongdouble.
9177 (init_float128_ieee): Call it.
9179 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9182 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9183 -fcf-protection=branch -mibt.
9184 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9186 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9189 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9190 arm_word_relocations.
9192 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9195 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9196 extern array declarations.
9198 2018-03-27 Richard Biener <rguenther@suse.de>
9201 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9202 explicit single_use checks.
9204 2018-03-27 Richard Biener <rguenther@suse.de>
9206 PR tree-optimization/85082
9207 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9210 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9212 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9213 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9214 Turn on fasynchronous-unwind-tables and funwind-tables.
9216 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9219 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9220 (*bmi_blsr_<mode>_ccz): Ditto.
9222 2018-03-26 Tom de Vries <tom@codesourcery.com>
9224 PR tree-optimization/85063
9225 * omp-general.c (offloading_function_p): New function. Factor out
9227 * omp-offload.c (pass_omp_target_link::gate): ... here.
9228 * omp-general.h (offloading_function_p): Declare.
9229 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9230 with attribute omp declare target for offloading functions.
9232 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9234 PR tree-optimization/84005
9235 * tree-data-ref.h (get_base_for_alignment): Declare.
9236 * tree-data-ref.c (get_base_for_alignment_1): New function.
9237 (get_base_for_alignment): Likewise.
9238 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9239 get_base_for_alignment to find a suitable base object, instead
9240 of always using drb->base_address.
9242 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9245 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9246 known size by default.
9248 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9251 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9252 and non BLKmode operands.
9254 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9257 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9258 Clean up attributes.
9260 2018-03-23 Richard Biener <rguenther@suse.de>
9263 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9264 we are going to emit early debug for LTO.
9266 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9269 * function.c (match_asm_constraints_1): Don't optimize if input
9270 doesn't satisfy general_operand predicate for output's mode.
9273 * alias.c (write_dependence_p): Don't require for x_canonicalized
9274 non-VOIDmode if x has VOIDmode.
9277 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9278 just don't try to optimize it rather than assert it never happens.
9280 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9282 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9283 macro expansions for definition of ST_INTERNAL_<mode> and
9284 LD_INTERNAL_<mode> builtins.
9285 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9287 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9289 (altivec_expand_st_builtin): Likewise.
9290 (altivec_expand_builtin): Remove calls to deleted functions.
9291 (rs6000_address_for_altivec): Delete this function.
9292 * config/rs6000/vector.md: Remove expands for
9293 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9295 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9298 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9299 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9300 re-computing once computed.
9301 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9302 (arm_init_machine_status): Initialize
9303 machine->static_chain_stack_bytes.
9305 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9308 * doc/extend.texi: Add four new prototypes for vec_ld.
9309 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9310 definitions for more logical presentation.
9311 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9312 entries for V1TI variants of __builtin_altivec_ld builtin.
9313 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9314 handling of V1TI variant of LVX icode pattern.
9315 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9316 (rs6000_gimple_fold_builtin): Likewise.
9317 (altivec_init_builtins): Add code to define
9318 __builtin_altivec_lvx_v1ti function.
9320 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9323 * function.c (match_asm_constraints_1): Don't do the optimization
9324 if input isn't a REG, SUBREG, MEM or constant.
9326 2018-03-22 Tom de Vries <tom@codesourcery.com>
9328 PR tree-optimization/84956
9329 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9330 bb_has_abnormal_pred.
9332 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9335 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9336 DECL_INITIAL (decl) to decl at the end.
9337 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9340 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9342 * doc/extend.texi (__builtin_tgmath): Document when complex
9343 integer types are treated as _Complex _Float64.
9345 2018-03-21 Tom de Vries <tom@codesourcery.com>
9347 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9349 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9351 PR tree-optimization/84960
9352 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9353 if it is ENTRY block, move them into single succ of ENTRY in that case.
9355 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9357 PR tree-optimization/84811
9358 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9359 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9360 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9361 (known_subrange_p): Update accordingly. Cast each value involved
9362 in the size comparison, rather than casting the result of the
9365 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9367 PR tree-optimization/84982
9368 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9369 by flipping the least significant bit rather than all bits from
9370 bitpos to bitpos + bitsize - 1.
9372 2018-03-21 Nathan Sidwell <nathan@acm.org>
9374 * doc/extend.texi (Deprecated Features): Remove mention of
9375 long-deleted deprecations.
9377 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9380 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9381 * configure: Regenerate.
9383 2018-03-21 Tom de Vries <tom@codesourcery.com>
9385 PR tree-optimization/83126
9386 * tree-parloops.c (num_phis): New function.
9387 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9389 2018-03-21 Nathan Sidwell <nathan@acm.org>
9391 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9392 mention anon-struct/union members and trailing attributes.
9394 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9396 PR tree-optimization/84969
9397 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9398 builtin memset partitions if they set different rhs values.
9400 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9402 PR rtl-optimization/84989
9403 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9404 VEC_DUPLICATE with scalar result mode.
9406 2018-03-21 Martin Liska <mliska@suse.cz>
9409 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9410 not intended return statement.
9412 2018-03-21 Martin Liska <mliska@suse.cz>
9415 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9416 (chkp_find_bound_slots_1): Limit number of iterations.
9418 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9421 * Minor grammar fixes for x86 options.
9423 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9426 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9427 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9429 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9432 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9433 (altivec_lvx_<mode>_1op): Likewise.
9434 (altivec_stvx_<mode>_2op): Likewise.
9435 (altivec_stvx_<mode>_1op): Likewise.
9436 (altivec_lvx_<VM2:mode>): New define_expand.
9437 (altivec_stvx_<VM2:mode>): Likewise.
9438 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9439 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9440 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9441 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9442 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9443 (rs6000_gen_lvx): Likewise.
9444 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9445 (altivec_expand_stv_builtin): Likewise.
9446 (altivec_expand_builtin): Likewise.
9447 * config/rs6000/vector.md: Likewise.
9449 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9452 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9455 2018-03-20 Richard Biener <rguenther@suse.de>
9458 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9459 sign-conversions as zero, fall back to standard scalar_stmt
9462 2018-03-20 Martin Liska <mliska@suse.cz>
9465 * predict.c (rebuild_frequencies): Handle case when we have
9466 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9468 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9471 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9472 flag_section_anchors.
9473 * varasm.c (use_blocks_for_decl_p): Remove hack for
9474 dw2_force_const_mem.
9477 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9479 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9480 be created, use lowpart_subreg of operands[0] rather than operands[0]
9482 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9483 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9484 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9485 and n constraint instead of aarch64_shift_imm_di and Usd.
9486 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9487 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9489 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9492 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9493 to favor GPR over NEON registers.
9494 (<shift>di3_neon): Likewise.
9496 2018-03-20 Tom de Vries <tom@codesourcery.com>
9499 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9500 (nvptx_process_pars): Emit bar.sync asap and alap.
9502 2018-03-20 Tom de Vries <tom@codesourcery.com>
9505 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9506 seen_label if seen_label is already set.
9508 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9511 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9512 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9513 Use 1U instead of 1. Formatting fixes.
9516 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9517 instead of TREE_TYPE (s1) for the return value.
9519 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9521 PR tree-optimization/84946
9522 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9523 bitsize + bitsize in poly_uint64 rather than poly_int64.
9526 * dwarf2asm.c: Include fold-const.c.
9527 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9528 of decl rather than decl itself.
9530 PR rtl-optimization/84643
9531 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9533 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9536 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9537 calling assemble_variable.
9539 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9542 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9543 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9545 2018-03-19 Jim Wilson <jimw@sifive.com>
9548 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9549 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9550 (riscv_first_stack_step): Likewise.
9551 (riscv_option_override): Use STACK_BOUNDARY instead of
9553 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9555 (BIGGEST_ALIGNMENT): Set to 128.
9556 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9557 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9560 2018-03-19 Richard Biener <rguenther@suse.de>
9562 PR tree-optimization/84933
9563 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9564 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9566 2018-03-19 Richard Biener <rguenther@suse.de>
9568 PR tree-optimization/84859
9569 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9570 (cond_if_else_store_replacement): Perform sinking operation on
9571 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9572 Generalize what a BB with a single eligible store is.
9574 2018-03-19 Richard Biener <rguenther@suse.de>
9576 PR tree-optimization/84929
9577 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9578 chrec_is_positive against non-chrec arg.
9580 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9583 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9585 2018-03-18 Martin Liska <mliska@suse.cz>
9587 PR rtl-optimization/84635
9588 * regrename.c (build_def_use): Use matches_mode only when
9591 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9593 PR tree-optimization/84913
9594 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9595 vectorize chains of COND_EXPRs.
9597 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9599 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9601 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9603 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9605 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9607 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9609 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9610 Kito Cheng <kito.cheng@gmail.com>
9612 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9613 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9614 (nds32_adjust_reg_alloc_order): New function.
9615 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9617 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9619 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9620 nds32_print_operand, nds32_print_operand_address): Use
9621 HOST_WIDE_INT_PRINT_DEC instead.
9623 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9625 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9627 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9630 * config/i386/i386.c (initial_ix86_tune_features,
9631 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9633 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9634 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9635 rather than 1u << ix86_tune. Formatting fix.
9636 (ix86_option_override_internal): Change ix86_arch_mask from
9637 unsigned int to unsigned HOST_WIDE_INT, initialize to
9638 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9639 (ix86_function_specific_restore): Likewise.
9641 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9644 * postreload.c (reload_combine_recognize_pattern): Perform
9645 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9646 truncate_int_for_mode the result for the destination's mode.
9649 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9650 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9652 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9655 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9658 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9659 regno_allocno_class_array and sorted_pseudos.
9660 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9661 insns where regno is used.
9663 2018-03-16 Martin Liska <mliska@suse.cz>
9666 * multiple_target.c (create_dispatcher_calls): Redirect
9667 reference in the symbol table.
9669 2018-03-16 Martin Liska <mliska@suse.cz>
9672 * multiple_target.c (create_dispatcher_calls): Redirect also
9675 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9679 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9680 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9681 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9683 2018-03-16 Julia Koval <julia.koval@intel.com>
9685 * doc/invoke.texi (Skylake Server): Add CLWB.
9686 Cannonlake): Remove CLWB.
9688 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9690 PR tree-optimization/84841
9691 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9693 (FLOAT_ONE_CONST_TYPE): Define.
9694 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9695 (sort_by_operand_rank): Put entries with higher constant_type last
9696 rather than first to match comments.
9698 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9700 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9703 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9706 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9707 check and use address of target always.
9709 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9712 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9713 (indirect_thunk_bnd_needed): Likewise.
9714 (indirect_thunks_used): Likewise.
9715 (indirect_thunks_bnd_used): Likewise.
9716 (indirect_return_needed): New.
9717 (indirect_return_bnd_needed): Likewise.
9718 (output_indirect_thunk_function): Add a bool argument for
9720 (output_indirect_thunk_function): Don't generate alias for
9721 function return thunk.
9722 (ix86_code_end): Call output_indirect_thunk_function to generate
9723 function return thunks.
9724 (ix86_output_function_return): Set indirect_return_bnd_needed
9725 and indirect_return_needed instead of indirect_thunk_bnd_needed
9726 and indirect_thunk_needed.
9728 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9730 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9731 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9732 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9734 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9735 Paul Hua <paul.hua.gm@gmail.com>
9738 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9740 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9742 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9743 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9746 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9749 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9750 instead of GET_MODE_SIZE when comparing Units.
9752 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9755 * varasm.c (hash_section): Return an unchangeble hash value
9756 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9757 Return !aarch64_can_use_per_function_literal_pools_p ().
9759 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9762 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9763 rather than address of cmode as last argument to prepare_cmp_insn.
9765 2018-03-15 Julia Koval <julia.koval@intel.com>
9767 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9768 F_AVX512VNNI, F_AVX512BITALG): New.
9770 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9773 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9774 insn for floating-point loads and stores.
9776 2018-03-14 Carl Love <cel@us.ibm.com>
9778 * config/rs6000/rs6000-c.c: Add macro definitions for
9779 ALTIVEC_BUILTIN_VEC_PERMXOR.
9780 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9781 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9782 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9783 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9785 * config/doc/extend.texi: Add prototypes for vec_permxor.
9787 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9790 * diagnostic-show-locus.c (class layout_point): Convert m_line
9791 from int to linenum_type.
9792 (line_span::comparator): Use linenum "compare" function when
9793 comparing line numbers.
9794 (test_line_span): New function.
9795 (layout_range::contains_point): Convert param "row" from int to
9797 (layout_range::intersects_line_p): Likewise.
9798 (layout::will_show_line_p): Likewise.
9799 (layout::print_source_line): Likewise.
9800 (layout::should_print_annotation_line_p): Likewise.
9801 (layout::print_annotation_line): Likewise.
9802 (layout::print_leading_fixits): Likewise.
9803 (layout::annotation_line_showed_range_p): Likewise.
9804 (struct line_corrections): Likewise for field m_row.
9805 (line_corrections::line_corrections): Likewise for param "row".
9806 (layout::print_trailing_fixits): Likewise.
9807 (layout::get_state_at_point): Likewise.
9808 (layout::get_x_bound_for_row): Likewise.
9809 (layout::print_line): Likewise.
9810 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9811 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9812 * input.c (selftest::test_linenum_comparisons): New function.
9813 (selftest::input_c_tests): Call it.
9814 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9815 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9816 * selftest.h (ASSERT_GT): New macro.
9817 (ASSERT_GT_AT): New macro.
9818 (ASSERT_LT): New macro.
9819 (ASSERT_LT_AT): New macro.
9821 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9823 PR rtl-optimization/84780
9824 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9826 2018-03-14 Martin Liska <mliska@suse.cz>
9828 * tree.c (record_node_allocation_statistics): Use
9829 get_stats_node_kind.
9830 (get_stats_node_kind): New function extracted from
9831 record_node_allocation_statistics.
9832 (free_node): Use get_stats_node_kind.
9834 2018-03-14 Richard Biener <rguenther@suse.de>
9836 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9837 that the value-set of ANTIC_IN doesn't grow.
9840 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9842 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9843 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9845 2018-03-14 Julia Koval <julia.koval@intel.com>
9847 * config.gcc (icelake-client, icelake-server): New.
9849 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9850 (initial_ix86_arch_features): Ditto.
9851 (PTA_SKYLAKE): Add SGX.
9852 (PTA_ICELAKE): Remove.
9853 (PTA_ICELAKE_CLIENT): New.
9854 (PTA_ICELAKE_SERVER): New.
9855 (ix86_option_override_internal): Split up icelake on icelake client and
9857 (get_builtin_code_for_version): Ditto.
9858 (fold_builtin_cpu): Ditto.
9859 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9860 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9861 * config/i386/i386.h (processor_type): Ditto.
9862 * doc/invoke.texi: Ditto.
9864 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9867 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9868 INTEGER_CST offset, add it together with bitpos / 8 and
9869 sign extend based on POINTER_SIZE.
9873 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9876 * config/i386/constraints.md (Yc): New register constraint.
9877 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9878 Use Yc constraint for alternative 2 of operand 0. Remove
9879 preferred_for_speed attribute.
9881 2018-03-14 Richard Biener <rguenther@suse.de>
9883 PR tree-optimization/84830
9884 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9885 with the old one to avoid oscillations.
9887 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9890 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9892 (assign_by_spills): Return a flag of reload assignment failure.
9893 Do not process the reload assignment failures. Do not spill other
9894 reload pseudos if they has the same reg class. Update n if
9896 (lra_assign): Add a return arg. Set up from the result of
9897 assign_by_spills call.
9898 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9899 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9900 usage_insns if it is not NULL.
9901 (spill_hard_reg_in_range): New function.
9902 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9903 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9904 function prototypes.
9905 (lra_assign): Change prototype.
9906 * lra.c (lra): Add code to deal with fails by splitting hard reg
9909 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9911 * config/riscv/riscv.opt (mrelax): New option.
9912 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9913 "norelax" when riscv_mrelax is disabled.
9914 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9916 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9919 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9920 reassociation for int modes.
9922 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9924 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9925 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9927 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9928 * config/aarch64/aarch64-sve.md
9929 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9930 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9931 (*extend<mode><Vwide>2): Rename to...
9932 (aarch64_sve_extend<mode><Vwide>2): ...this.
9933 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9934 renaming the old pattern to...
9935 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9937 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9938 define_expand, renaming the old pattern to...
9939 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9940 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9941 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9942 account when deciding which SVE instruction the optab should use.
9943 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9945 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9947 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9948 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9949 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9950 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9951 (tlsdesc_small_<mode>): Turn a define_expand and use
9952 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9953 (tlsdesc_small_advsimd_<mode>): ...this.
9954 (tlsdesc_small_sve_<mode>): New pattern.
9956 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9958 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9959 (UNSPEC_UMUL_HIGHPART): New constants.
9960 (MUL_HIGHPART): New int iteraor.
9961 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9962 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9964 (*<su>mul<mode>3_highpart): New define_insn.
9966 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9969 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9972 2018-03-13 Martin Liska <mliska@suse.cz>
9975 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9977 (sem_item_optimizer::~sem_item_optimizer): Release it.
9978 (sem_item_optimizer::merge_classes): Register variable aliases.
9979 (sem_item_optimizer::fixup_pt_set): New function.
9980 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9981 * ipa-icf.h: Declare new variables and functions.
9983 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9986 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9987 integer_pow2p@2 and test integer_pow2p in condition.
9988 (A < 0 ? C : 0): Similarly for @1.
9991 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9992 characters starting at p contain '\0' character, don't look beyond
9996 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9997 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10000 * reg-stack.c (change_stack): Change update_end var from int to
10001 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10002 also call set_block_for_insn on the newly added insns and rescan.
10005 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10006 on the last operand.
10009 * tree.c (stabilize_reference_1): Return save_expr (e) for
10010 STATEMENT_LIST even if it doesn't have side-effects.
10012 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10014 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10016 2018-03-12 Renlin Li <renlin.li@arm.com>
10018 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10019 aarch64_output_scalar_simd_mov_immediate.
10021 2018-03-12 Martin Sebor <msebor@redhat.com>
10023 PR tree-optimization/83456
10024 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10025 for perfectly overlapping calls to memcpy.
10026 (gimple_fold_builtin_memory_chk): Same.
10027 (gimple_fold_builtin_strcpy): Handle no-warning.
10028 (gimple_fold_builtin_stxcpy_chk): Same.
10029 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10031 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10033 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10034 parameter. Use it for SFmode.
10035 (rs6000_function_arg_advance_1): Adjust.
10036 (rs6000_function_arg): Adjust.
10037 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10039 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10041 PR rtl-optimization/84169
10042 PR rtl-optimization/84780
10043 * combine.c (can_combine_p): Check for a 2-insn combination whether
10044 the destination register is used between the two insns, too.
10046 2018-03-12 Richard Biener <rguenther@suse.de>
10048 PR tree-optimization/84803
10049 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10050 for refs DR analysis didn't process.
10052 2018-03-12 Richard Biener <rguenther@suse.de>
10054 PR tree-optimization/84777
10055 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10056 force-vectorize loops ignore whether we are optimizing for size.
10058 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10060 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10061 (TARGET_MD_ASM_ADJUST): Define.
10063 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10064 Kito Cheng <kito.cheng@gmail.com>
10065 Chung-Ju Wu <jasonwucj@gmail.com>
10067 * config/nds32/nds32.c (nds32_compute_stack_frame,
10068 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10069 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10070 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10071 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10072 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10073 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10074 * config/nds32/nds32.md (prologue, epilogue): Use macro
10075 NDS32_V3PUSH_AVAILABLE_P to do checking.
10077 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10080 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10081 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10082 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10083 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10084 addition of most attributes on !orig_type_die or the attribute not
10085 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10087 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10088 Chung-Ju Wu <jasonwucj@gmail.com>
10090 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10091 __NDS32_VH__ macro.
10092 * config/nds32/nds32.opt (mvh): New option.
10094 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10095 Chung-Ju Wu <jasonwucj@gmail.com>
10097 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10099 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10100 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10103 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10104 Chung-Ju Wu <jasonwucj@gmail.com>
10106 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10108 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10109 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10111 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10112 Kito Cheng <kito.cheng@gmail.com>
10113 Chung-Ju Wu <jasonwucj@gmail.com>
10115 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10116 UNSPEC_FFMISM and UNSPEC_FLMISM.
10117 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10118 for ffb, ffmism and flmism.
10119 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10120 (unspec_ffmism): Ditto.
10121 (unspec_flmism): Ditto.
10122 (nds32_expand_builtin_impl): Check if string extension is available.
10123 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10124 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10126 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10129 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10132 * lra-assigns.c (assign_by_spills): Return a flag of reload
10133 assignment failure. Do not process the reload assignment
10134 failures. Do not spill other reload pseudos if they has the same
10136 (lra_assign): Add a return arg. Set up from the result of
10137 assign_by_spills call.
10138 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10139 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10140 usage_insns if it is not NULL.
10141 (spill_hard_reg_in_range): New function.
10142 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10143 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10144 function prototypes.
10145 (lra_assign): Change prototype.
10146 * lra.c (lra): Add code to deal with fails by splitting hard reg
10149 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10152 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10154 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10157 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10158 (dw_val_node): Add val_symbolic_view.
10159 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10160 (symview_upper_bound): New.
10161 (new_line_info_table): Initialize symviews_since_reset.
10162 (dwarf2out_source_line): Count symviews_since_reset and set
10163 symview_upper_bound.
10164 (dw_val_equal_p): Handle symview.
10165 (add_AT_symview): New.
10166 (print_dw_val): Handle symview.
10167 (attr_checksum, attr_checksum_ordered): Likewise.
10168 (same_dw_val_p, size_of_die): Likewise.
10169 (value_format, output_die): Likewise.
10170 (add_high_low_attributes): Use add_AT_symview for entry_view.
10171 (dwarf2out_finish): Reset symview_upper_bound, clear
10174 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10177 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10178 Add strict argument and use it.
10179 (rs6000_split_multireg_move): Update for new strict argument.
10180 (mem_operand_gpr): Disallow all non-offsettable addresses.
10181 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10183 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10186 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10187 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10188 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10191 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10192 decl, use remap_type if we want to use the type.
10194 2018-03-09 Martin Sebor <msebor@redhat.com>
10196 PR tree-optimization/84526
10197 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10199 (builtin_access::generic_overlap): Be prepared to handle non-array
10202 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10204 PR rtl-optimization/84682
10205 * lra-constraints.c (process_address_1): Check is_address flag
10206 for address constraints.
10207 (process_alt_operands): Likewise.
10208 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10209 preprocess_constraints.
10210 * recog.h (preprocess_constraints): Add oploc parameter.
10212 * recog.c (preprocess_constraints): Test address_operand for
10213 CT_ADDRESS constraints.
10215 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10218 * lra-assigns.c (assign_by_spills): Return a flag of reload
10219 assignment failure. Do not process the reload assignment
10220 failures. Do not spill other reload pseudos if they has the same
10222 (lra_assign): Add a return arg. Set up from the result of
10223 assign_by_spills call.
10224 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10225 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10226 usage_insns if it is not NULL.
10227 (spill_hard_reg_in_range): New function.
10228 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10229 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10230 function prototypes.
10231 (lra_assign): Change prototype.
10232 * lra.c (lra): Add code to deal with fails by splitting hard reg
10235 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10238 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10239 Accept complain bool parameter. Only emit errors if it is true.
10240 (arm_parse_cpu_option_name): Likewise.
10241 (arm_target_thumb_only): Adjust callers of the above.
10242 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10243 prototype to take a default true bool parameter.
10244 (arm_parse_arch_option_name): Likewise.
10246 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10247 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10251 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10252 * configure: Regenerate.
10253 * configure.ac ("linker --version-script option"): New.
10254 ("linker soname option"): New.
10256 2018-03-09 Richard Biener <rguenther@suse.de>
10258 PR tree-optimization/84775
10259 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10260 immediate uses of predicate stmts and mark them modified.
10263 PR tree-optimization/84178
10264 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10266 (version_loop_for_if_conversion): Delay update_ssa call.
10267 (tree_if_conversion): Delay update_ssa until after predicate
10270 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10273 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10274 when the function accesses prior frames.
10276 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10279 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10280 gen_llsym, otherwise call maybe_gen_llsym.
10282 PR inline-asm/84742
10283 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10284 has ',' character inside of it.
10286 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10289 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10290 as clobbering CC_REGNUM.
10292 2018-03-08 Richard Biener <rguenther@suse.de>
10294 PR middle-end/84552
10295 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10296 (follow_copies_to_constant): Do not follow SSA names registered
10299 2018-03-08 Richard Biener <rguenther@suse.de>
10301 PR tree-optimization/84178
10302 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10304 (version_loop_for_if_conversion): Delay update_ssa call.
10305 (tree_if_conversion): Delay update_ssa until after predicate
10308 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10310 PR tree-optimization/84178
10311 * tree-if-conv.c (release_bb_predicate): Remove the
10312 the assertion that the stmts have NULL use_ops.
10313 Discard the statements, asserting that they haven't
10314 yet been added to a BB.
10316 2018-03-08 Richard Biener <rguenther@suse.de>
10318 PR tree-optimization/84746
10319 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10320 (phi_translate): Pass in destination ANTIC_OUT set.
10321 (phi_translate_1): Likewise. For a simplified result lookup
10322 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10323 (phi_translate_set): Adjust.
10324 (do_pre_regular_insertion): Likewise.
10325 (do_pre_partial_partial_insertion): Likewise.
10327 2018-03-08 Martin Liska <mliska@suse.cz>
10329 PR gcov-profile/84735
10330 * doc/gcov.texi: Document usage of profile files.
10331 * gcov-io.h: Document changes in the format.
10333 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10337 * dwarf2out.c (struct dw_line_info_table): Update comments for
10339 (FORCE_RESET_NEXT_VIEW): New.
10340 (FORCE_RESETTING_VIEW_P): New.
10341 (RESETTING_VIEW_P): Check for -1 too.
10342 (ZERO_VIEW_P): Likewise.
10343 (new_line_info_table): Force-reset next view.
10344 (dwarf2out_begin_function): Likewise.
10345 (dwarf2out_source_line): Simplify zero_view_p initialization.
10346 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10347 view directly. Omit view when omitting .loc at line 0.
10349 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10351 PR tree-optimization/84740
10352 * tree-switch-conversion.c (process_switch): Call build_constructors
10353 only if info.phi_count is non-zero.
10355 PR tree-optimization/84739
10356 * tree-tailcall.c (find_tail_calls): Check call arguments against
10357 DECL_ARGUMENTS (current_function_decl) rather than
10358 DECL_ARGUMENTS (func) when checking for tail recursion.
10360 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10362 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10363 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10364 Volker Reichelt's entry and add entries for people that perform
10365 GCC fuzzy testing and report numerous bugs.
10367 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10370 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10371 readonly data in sdata, if that is disabled.
10372 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10373 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10374 -mreadonly-in-sdata option.
10376 2018-03-07 Martin Sebor <msebor@redhat.com>
10378 PR tree-optimization/84468
10379 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10380 basic block when looking for nul assignment.
10382 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10385 * except.h (output_function_exception_table): Adjust prototype.
10386 * except.c (output_function_exception_table): Remove FNNAME parameter
10387 and add SECTION parameter. Ouput one part of the table at a time.
10388 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10389 the first part of the exception table and emit unwind directives.
10390 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10391 (i386_pe_seh_cold_init): Likewise.
10392 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10393 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10394 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10395 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10396 * config/i386/winnt.c: Include except.h.
10397 (struct seh_frame_state): Add reg_offset, after_prologue and
10398 in_cold_section fields.
10399 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10400 (i386_pe_seh_cold_init): New function.
10401 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10402 to seh->in_cold_section.
10403 (seh_emit_push): Record the offset of the push.
10404 (seh_emit_save): Record the offet of the save.
10405 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10406 Test seh->after_prologue to disregard the epilogue.
10407 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10408 (i386_pe_end_cold_function): New function.
10410 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10413 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10414 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10417 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10419 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10420 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10422 PR middle-end/84723
10423 * multiple_target.c: Include tree-inline.h and intl.h.
10424 (expand_target_clones): Diagnose and fail if node->definition and
10425 !tree_versionable_function_p (node->decl).
10427 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10429 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10431 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10432 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10433 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10435 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10438 * combine.c (try_combine): Use reg_or_subregno instead of handling
10439 just paradoxical SUBREGs and REGs.
10441 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10443 * config/arc/arc.c (arc_finalize_pic): Remove function.
10444 (arc_must_save_register): We use single base PIC register, remove
10445 checks to save/restore the PIC register.
10446 (arc_expand_prologue): Likewise.
10447 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10449 (arc_verify_short): Likewise.
10450 (arc_attr_type): Likewise.
10451 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10452 (walk_stores): Likewise.
10453 (arc_address_cost): Make it static.
10454 (arc_verify_short): Likewise.
10455 (branch_dest): Likewise.
10456 (arc_attr_type): Likewise.
10457 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10458 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10459 (arc_final_prescan_insn): Remove inserting the nops due to
10460 hardware hazards. It is done in reorg step.
10461 (insn_length_variant_t): Remove.
10462 (insn_length_parameters_t): Likewise.
10463 (arc_insn_length_parameters): Likewise.
10464 (arc_get_insn_variants): Likewise.
10465 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10467 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10469 PR inline-asm/84683
10470 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10473 PR tree-optimization/84687
10474 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10476 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10478 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10480 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10481 Rename to ppc_speculation_barrier.
10482 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10483 __builtin_ppc_speculation_barrier.
10485 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10488 * combine.c (combine_simplify_rtx): Don't try to simplify if
10489 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10492 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10494 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10495 to 32 bytes when compiling for POWER9.
10497 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10500 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10501 regparm >= 3 with no arg reg available also for calls with
10502 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10505 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10507 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10509 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10512 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10514 2018-03-05 Richard Biener <rguenther@suse.de>
10516 PR tree-optimization/84486
10517 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10518 When inserting a __builtin_assume_aligned call set the LHS
10519 SSA name alignment info accordingly.
10521 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10523 PR tree-optimization/84114
10524 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10525 Avoid reassociation of FLOAT_MODE addition.
10527 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10529 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10530 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10531 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10532 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10533 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10534 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10535 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10537 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10538 __builtin_ia32_wbinvd): New builtins.
10539 (SPECIAL_ARGS2): New.
10540 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10541 (SPECIAL_ARGS2): New.
10542 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10543 (ix86_valid_target_attribute_inner_p): Ditto.
10544 (ix86_init_mmx_sse_builtins): Add special_args2.
10545 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10546 TARGET_WBNOINVD_P): New.
10547 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10548 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10549 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10550 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10551 * config/i386/pconfigintrin.h: New file.
10552 * config/i386/wbnoinvdintrin.h: Ditto.
10553 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10555 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10557 2018-03-05 Richard Biener <rguenther@suse.de>
10559 PR tree-optimization/84670
10560 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10562 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10563 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10564 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10565 doesn't grow if all successors (recursively) were visited at least
10568 2018-03-05 Richard Biener <rguenther@suse.de>
10570 PR tree-optimization/84650
10571 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10572 if executed in the loop pipeline.
10574 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10576 * doc/configfiles.texi (Configuration Files): Move info about
10577 conditionalizing $target-protos.h to...
10578 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10579 differs from $target-protos.h.
10581 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10582 Chung-Ju Wu <jasonwucj@gmail.com>
10584 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10585 * config/nds32/nds32-multiple.md (setmemsi): Define.
10586 * config/nds32/nds32-memory-manipulation.c
10587 (nds32_gen_dup_4_byte_to_word_value): New.
10588 (emit_setmem_word_loop): New.
10589 (emit_setmem_byte_loop): New.
10590 (nds32_expand_setmem_loop): New.
10591 (nds32_expand_setmem_loop_v3m): New.
10592 (nds32_expand_setmem_unroll): New.
10593 (nds32_expand_setmem): New.
10595 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10596 Chung-Ju Wu <jasonwucj@gmail.com>
10598 * config/nds32/nds32-memory-manipulation.c
10599 (nds32_emit_load_store): New.
10600 (nds32_emit_post_inc_load_store): New.
10601 (nds32_emit_mem_move): New.
10602 (nds32_emit_mem_move_block): New.
10603 (nds32_expand_movmemsi_loop_unknown_size): New.
10604 (nds32_expand_movmemsi_loop_known_size): New.
10605 (nds32_expand_movmemsi_loop): New.
10606 (nds32_expand_movmemsi_unroll): New.
10607 (nds32_expand_movmemqi): Rename ...
10608 (nds32_expand_movmemsi): ... to this.
10609 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10610 (movmemsi): ... to this.
10611 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10612 (nds32_expand_movmemsi): ... to this.
10614 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10615 Monk Chiang <sh.chiang04@gmail.com>
10616 Chung-Ju Wu <jasonwucj@gmail.com>
10618 * config/nds32/nds32-protos.h
10619 (nds32_expand_load_multiple): New arguments.
10620 (nds32_expand_store_multiple): Ditto.
10621 (nds32_valid_multiple_load_store): Rename ...
10622 (nds32_valid_multiple_load_store_p): ... to this.
10623 * config/nds32/nds32-memory-manipulation.c
10624 (nds32_expand_load_multiple): Refine implementation.
10625 (nds32_expand_store_multiple): Ditto.
10626 * config/nds32/nds32-multiple.md
10627 (load_multiple): Update nds32_expand_load_multiple interface.
10628 (store_multiple): Update nds32_expand_store_multiple interface.
10629 * config/nds32/nds32-predicates.c
10630 (nds32_valid_multiple_load_store): Rename ...
10631 (nds32_valid_multiple_load_store_p): ... to this and refine
10633 * config/nds32/predicates.md
10634 (nds32_load_multiple_and_update_address_operation): New predicate.
10635 (nds32_store_multiple_and_update_address_operation): New predicate.
10637 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10638 Chung-Ju Wu <jasonwucj@gmail.com>
10640 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10641 (combo): New attribute.
10642 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10644 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10646 * config/nds32/nds32.opt: Change -mcmodel= default value.
10648 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10649 Monk Chiang <sh.chiang04@gmail.com>
10650 Chung-Ju Wu <jasonwucj@gmail.com>
10652 * config/nds32/constants.md (unspec_element): New enum.
10653 * config/nds32/constraints.md (Umw): New constraint.
10654 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10655 * config/nds32/nds32-intrinsic.md: Likewise.
10656 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10657 (nds32_valid_smw_lwm_base_p): New.
10658 (nds32_output_smw_single_word): New.
10659 (nds32_output_lmw_single_word): New.
10660 (nds32_expand_unaligned_load): New.
10661 (nds32_expand_unaligned_store): New.
10662 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10663 (nds32_output_smw_single_word): Declare.
10664 (nds32_output_lmw_single_word): Declare.
10665 (nds32_expand_unaligned_load): Declare.
10666 (nds32_expand_unaligned_store): Declare.
10667 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10668 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10669 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10670 NDS32_BUILTIN_UASTORE_DW.
10671 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10674 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10675 Kito Cheng <kito.cheng@gmail.com>
10676 Chung-Ju Wu <jasonwucj@gmail.com>
10678 * config/nds32/nds32-intrinsic.c
10679 (nds32_expand_builtin_null_ftype_reg): Delete.
10680 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10681 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10682 (nds32_read_argument): New.
10683 (nds32_legitimize_target): Ditto.
10684 (nds32_legitimize_argument): Ditto.
10685 (nds32_check_constant_argument): Ditto.
10686 (nds32_expand_unop_builtin): Ditto.
10687 (nds32_expand_unopimm_builtin): Ditto.
10688 (nds32_expand_binop_builtin): Ditto.
10689 (nds32_builtin_decl_impl): Ditto.
10690 (builtin_description): Ditto.
10691 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10692 (nds32_init_builtins_impl): Ditto.
10693 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10694 (nds32_builtin_decl): New.
10695 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10696 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10698 2018-03-02 Jeff Law <law@redhat.com>
10700 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10701 (redundant_insn, fill_simple_delay_slots): Likewise.
10702 (fill_slots_from_thread): Likewise.
10703 * resource.c (mark_referenced_resources): Likewise.
10704 (mark_set_resources, find_dead_or_set_registers): Likewise.
10706 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10708 * substring-locations.h (format_warning_va): Formatting fix for
10709 ATTRIBUTE_GCC_DIAG.
10710 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10712 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10713 * substring-locations.c: Include intl.h.
10714 (format_warning_va): Turned into small wrapper around
10715 format_warning_n_va, renamed to ...
10716 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10717 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10719 (format_warning_at_substring_n): New function.
10720 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10721 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10722 format_warning_at_substring with just a shorter name instead of
10723 const function pointer.
10724 (fmtwarn_n): New function.
10725 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10726 appropriate, get rid of all the fmtstr temporaries, move conditionals
10727 with G_() wrapped string literals directly into fmtwarn arguments,
10728 cast dir.len to (int), formatting fixes.
10730 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10732 * doc/invoke.texi: Remove "Cilk Plus" references.
10734 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10735 Richard Biener <rguenther@suse.de>
10738 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10739 for error or warning attributes if CALL_FROM_THUNK_P is set.
10742 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10745 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10746 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10749 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10750 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10752 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10753 instead of -1U in last predictors element's probability member.
10755 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10758 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10759 arguments if they are comparable.
10761 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10763 PR tree-optimization/84634
10764 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10765 masks and masked_loop_p with a single loop_masks, making sure it's
10766 null for bb vectorization.
10768 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10770 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10771 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10772 loop->force_vectorize to check whether there is no alias.
10774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10777 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10779 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10781 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10782 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10783 instead of a loop around prev_real_insn.
10784 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10787 PR inline-asm/84625
10788 * config/i386/i386.c (ix86_print_operand): Use conditional
10789 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10792 2018-03-02 Richard Biener <rguenther@suse.de>
10794 PR tree-optimization/84427
10795 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10796 (bitmap_set_subtract_values): Rewrite to handle multiple
10799 (prune_clobbered_mems): Likewise.
10800 (phi_translate): Take edge instead of pred/phiblock.
10801 (phi_translate_1): Likewise.
10802 (phi_translate_set): Likewise. Insert all translated
10803 exprs for a value into the set, keeping possibly multiple
10804 expressions per value.
10805 (compute_antic_aux): Adjust for phi_translate changes.
10806 When intersecting union the expressions and prune those
10807 not in the final value set, keeping possibly multiple
10808 expressions per value. Do not use value-insertion
10809 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10810 all expressions. Add verification that the value-sets
10811 only shrink during iteration.
10812 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10813 (do_pre_regular_insertion): Likewise.
10814 (do_pre_partial_partial_insertion): Likewise.
10816 2018-03-02 Richard Biener <rguenther@suse.de>
10819 * config/darwin.c (saved_debug_info_level): New static global.
10820 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10821 (darwin_asm_lto_end): Restore debug info generation settings.
10823 2018-03-01 Martin Liska <mliska@suse.cz>
10826 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10827 volatile arguments.
10829 2018-03-01 Richard Biener <rguenther@suse.de>
10832 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10833 type annotation with LTO when debug was disabled at compile-time.
10835 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10837 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10839 (mips_final_postscan_insn): Likewise.
10841 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10843 PR rtl-optimization/84528
10844 * alias.c (init_alias_target): Add commentary.
10845 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10846 a unique base value if the frame pointer is not eliminated
10847 to the stack pointer.
10849 2018-03-01 Tom de Vries <tom@codesourcery.com>
10851 PR rtl-optimization/83327
10852 * lra-int.h (hard_regs_spilled_into): Declare.
10853 * lra.c (hard_regs_spilled_into): Define.
10854 (init_reg_info): Init hard_regs_spilled_into.
10855 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10856 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10857 (process_bb_lives): Handle hard_regs_spilled_into.
10858 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10859 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10861 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10863 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10864 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10865 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10866 * config/rs6000/aix72.h: New file.
10868 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10870 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10871 instead of warning_at with conditional singular and plural messages
10875 * stor-layout.c (update_alignment_for_field): For
10876 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10877 && !DECL_PACKED (field), do the alignment update, just use
10878 only desired_align instead of MAX (type_align, desired_align)
10880 (place_field): Don't do known_align < desired_align handling
10881 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10882 is non-NULL, instead do it after rli->prev_field handling and
10883 only if not within a bitfield word. For DECL_PACKED (field)
10884 use type_align of BITS_PER_UNIT.
10886 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10888 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10889 superfluous parentheses and trailing spaces.
10891 2018-02-28 Richard Biener <rguenther@suse.de>
10893 PR tree-optimization/84584
10894 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10895 SCoPs with fake exit edge.
10897 2018-02-28 Martin Liska <mliska@suse.cz>
10900 * timevar.c (timer::print): Fix format to properly print 100%
10903 2018-02-28 Richard Biener <rguenther@suse.de>
10905 PR middle-end/84607
10906 * genmatch.c (capture_info::walk_match): Do not mark
10907 captured expressions without operands as expr_p given
10908 they act more like predicates and should be subject to
10909 "lost tail" side-effect preserving.
10911 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10913 PR rtl-optimization/81611
10914 * auto-inc-dec.c (attempt_change): Move dead note from
10915 mem_insn if it's the next use of regno
10916 (find_address): Take address use of reg holding
10917 non-incremented value. Add parm to limit search to the named
10919 (merge_in_block): Attempt to use a mem insn that is the next
10920 use of the original regno.
10922 2018-02-27 Martin Sebor <msebor@redhat.com>
10925 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10926 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10928 2018-02-27 Martin Sebor <msebor@redhat.com>
10930 PR translation/84207
10931 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10932 n argument to unsigned HOST_WIDE_INT.
10933 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10934 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10935 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10936 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10938 2018-02-27 Richard Biener <rguenther@suse.de>
10940 PR tree-optimization/84512
10941 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10942 Do not use the estimate returned from record_stmt_cost for
10943 the scalar iteration cost but sum properly using add_stmt_cost.
10945 2018-02-27 Richard Biener <rguenther@suse.de>
10947 PR tree-optimization/84466
10948 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10949 Adjust last change to less strictly validate use operands.
10951 2018-02-27 Martin Liska <mliska@suse.cz>
10953 PR gcov-profile/84548
10954 * gcov.c (process_file): Allow partial overlap and consider it
10955 also as group functions.
10956 (output_lines): Properly calculate range of lines for a group.
10958 2018-02-27 Martin Liska <mliska@suse.cz>
10960 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10961 'ggc' suffixes. Change first column width.
10962 (timer::print): Fix formatting of the column.
10964 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10966 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10967 preserve inline entry blocks for the sake of debug inline
10968 entry point markers alone.
10969 (remove_unused_locals): Suggest in comments a better place to
10970 force the preservation of inline entry blocks that are
10971 otherwise unused, but do not preserve them.
10973 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10975 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10977 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10980 * config/i386/constraints.md (Bs): Replace
10981 ix86_indirect_branch_register with
10982 TARGET_INDIRECT_BRANCH_REGISTER.
10984 * config/i386/i386.md (indirect_jump): Likewise.
10985 (tablejump): Likewise.
10986 (*sibcall_memory): Likewise.
10987 (*sibcall_value_memory): Likewise.
10988 Peepholes of indirect call and jump via memory: Likewise.
10989 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10990 (*sibcall_value_GOT_32): Likewise.
10991 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10992 (GOT_memory_operand): Likewise.
10993 (call_insn_operand): Likewise.
10994 (sibcall_insn_operand): Likewise.
10995 (GOT32_symbol_operand): Likewise.
10996 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10998 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11000 PR rtl-optimization/83496
11001 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11002 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11003 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11004 redundant insn, if any.
11005 (relax_delay_slots): Likewise.
11006 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11008 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11010 PR tree-optimization/83965
11011 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11012 that grouped statements are part of a reduction chain. Return
11013 true if the statement is not marked as a reduction itself but
11014 is part of a group.
11015 (vect_recog_dot_prod_pattern): Don't check whether the statement
11016 is part of a group here.
11017 (vect_recog_sad_pattern): Likewise.
11018 (vect_recog_widen_sum_pattern): Likewise.
11020 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11023 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11025 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11028 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11030 (ix86_output_indirect_function_return): New prototype.
11031 (ix86_split_simple_return_pop_internal): Likewise.
11032 * config/i386/i386.c (indirect_return_via_cx): New.
11033 (indirect_return_via_cx_bnd): Likewise.
11034 (indirect_thunk_name): Handle return va CX_REG.
11035 (output_indirect_thunk_function): Create alias for
11036 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11037 (ix86_output_indirect_jmp): Remove the bool argument.
11038 (ix86_output_indirect_function_return): New function.
11039 (ix86_split_simple_return_pop_internal): Likewise.
11040 * config/i386/i386.md (*indirect_jump): Don't pass false
11041 to ix86_output_indirect_jmp.
11042 (*tablejump_1): Likewise.
11043 (simple_return_pop_internal): Change it to define_insn_and_split.
11044 Call ix86_split_simple_return_pop_internal to split it for
11045 -mfunction-return=.
11046 (simple_return_indirect_internal): Call
11047 ix86_output_indirect_function_return instead of
11048 ix86_output_indirect_jmp.
11050 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11053 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11054 memset and value initialization afterwards.
11056 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11058 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11060 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11063 * common/config/aarch64/aarch64-common.c
11064 (aarch_option_optimization_table[]): Switch
11065 off fomit-frame-pointer
11067 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11068 Chung-Ju Wu <jasonwucj@gmail.com>
11070 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11072 (store_multiple): Ditto.
11074 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11076 * config.gcc: Add --with-cpu support for nds32 target.
11077 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11078 * config/nds32/nds32.opt: Add -mcpu= option.
11080 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11082 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11083 isel=yes): Warn for these deprecated options.
11085 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11087 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11088 ISA_2_5_MASKS_EMBEDDED.
11090 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11092 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11093 p->max as pointers rather than using iterative_hash_expr.
11095 2018-02-23 Carl Love <cel@us.ibm.com>
11097 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11098 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11100 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11101 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11102 P8V_BUILTIN_VEC_VUNSIGNED2.
11104 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11107 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11108 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11110 * lra-constraints.c (curr_insn_transform): Set up
11111 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11113 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11114 * lra-eliminations.c (spill_pseudos): Ditto.
11115 (process_insn_for_elimination): Ditto.
11116 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11117 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11118 LRA_NON_CLOBBERED_ALT.
11120 2018-02-22 Martin Sebor <msebor@redhat.com>
11122 PR tree-optimization/84480
11123 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11124 to maybe_diag_stxncpy_trunc. Call it.
11125 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11126 from gimple_fold_builtin_strcpy. Print inlining stack.
11127 (handle_builtin_stxncpy): Print inlining stack.
11128 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11130 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11133 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11134 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11135 and -fcheck-pointer-bounds are used together.
11136 (indirect_thunk_prefix): New enum.
11137 (indirect_thunk_need_prefix): New function.
11138 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11139 "_nt" instead of "_bnd" for NOTRACK prefix.
11140 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11141 (output_indirect_thunk_function): Likewise.
11143 (ix86_code_end): Update output_indirect_thunk_function calls.
11144 (ix86_output_indirect_branch_via_reg): Replace
11145 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11146 (ix86_output_indirect_branch_via_push): Likewise.
11147 (ix86_output_function_return): Likewise.
11148 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11149 incompatible with -fcf-protection=branch and
11150 -fcheck-pointer-bounds.
11152 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11155 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11156 Change gcc_assert call to output_operand_lossage.
11158 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11160 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11162 2018-02-22 DJ Delorie <dj@redhat.com>
11163 Sebastian Perta <sebastian.perta@renesas.com>
11164 Oleg Endo <olegendo@gcc.gnu.org>
11166 * config/rx/rx.c (rx_rtx_costs): New function.
11167 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11169 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11171 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11173 2018-02-22 Martin Liska <mliska@suse.cz>
11176 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11177 Add "native" as a possible value.
11179 2018-02-22 Martin Liska <mliska@suse.cz>
11182 * config/i386/i386.c (ix86_option_override_internal):
11183 Add "native" as a possible value for -march and -mtune.
11185 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11188 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11189 to all type variants.
11191 PR tree-optimization/84503
11192 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11193 width as info->bitpos + info->bitsize - start.
11194 (merged_store_group::merge_overlapping): Simplify width computation.
11195 (check_no_overlap): New function.
11196 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11197 start + width and last_order of the group, fail if check_no_overlap
11199 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11200 to group if check_no_overlap fails.
11202 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11204 * config/rs6000/altivec.md: Delete contraint arguments to
11205 define_expand, define_split, and define_peephole2, and in
11206 define_insn_and_split if always unused.
11207 * config/rs6000/darwin.md: Ditto.
11208 * config/rs6000/dfp.md: Ditto.
11209 * config/rs6000/rs6000.md: Ditto.
11210 * config/rs6000/sync.md: Ditto.
11211 * config/rs6000/vector.md: Ditto.
11212 * config/rs6000/vsx.md: Ditto.
11214 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11216 * config/rs6000/altivec.md: Write output control strings as braced
11217 blocks instead of double-quoted strings.
11218 * config/rs6000/darwin.md: Ditto.
11219 * config/rs6000/rs6000.md: Ditto.
11220 * config/rs6000/vector.md: Ditto.
11221 * config/rs6000/vsx.md: Ditto.
11223 2018-02-21 Jason Merrill <jason@redhat.com>
11225 PR c++/84314 - ICE with templates and fastcall attribute.
11226 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11228 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11230 * ipa-cp.c (determine_versionability): Fix comment typos.
11232 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11235 * ipa-cp.c (determine_versionability): Do not version functions caling
11238 2018-02-21 Martin Liska <mliska@suse.cz>
11241 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11242 Add "native" as a possible value.
11243 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11244 the macro when native cpu detection is available.
11246 2018-02-21 Martin Liska <mliska@suse.cz>
11249 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11250 Add "native" as a possible value.
11251 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11252 when native cpu detection is available.
11254 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11255 Martin Sebor <msebor@redhat.com>
11257 PR tree-optimization/84478
11258 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11260 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11261 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11262 support which is conservatively correct, for 2 only stay conservative
11263 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11264 argument to the 2 argument get_range_strlen, adjust 6 arg
11265 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11267 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11268 (gimple_fold_builtin_strlen): Pass true as last argument to
11271 2018-02-20 Martin Sebor <msebor@redhat.com>
11273 PR middle-end/84095
11274 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11275 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11276 (builtin_memref::builtin_memref): Factor out parts into
11277 set_base_and_offset and call it.
11279 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11281 PR middle-end/84406
11282 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11283 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11284 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11285 search at the associated MODE_INT.
11287 2018-02-20 Jeff Law <law@redhat.com>
11289 PR middle-end/82123
11290 PR tree-optimization/81592
11291 PR middle-end/79257
11292 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11293 for range data rather than using global data.
11294 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11295 range data rather than using global data.
11296 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11297 pass it to children as needed.
11298 (struct directive::fmtresult): Similarly.
11299 (struct directive::set_width): Similarly.
11300 (struct directive::set_precision): Similarly.
11301 (format_integer, format_directive, parse_directive): Similarly.
11302 (format_none): Accept unnamed vr_values parameter.
11303 (format_percent, format_floating, format_character): Similarly.
11304 (format_string, format_plain): Similarly.
11305 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11306 the EVRP range analyzer for range data rather than using global data.
11307 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11308 gimple-ssa-evrp-analyze.h
11309 (class sprintf_dom_walker): Add after_dom_children member function.
11310 Add evrp_range_analyzer member.
11311 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11312 range analyzer as needed.
11313 (sprintf_dom_walker::after_dom_children): New member function.
11314 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11316 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11317 (evrp_range_analyzer::pop_to_marker): Likewise.
11319 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11321 PR tree-optimization/84419
11322 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11323 with the required type if its current type is compatible but
11326 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11328 PR middle-end/82004
11329 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11330 after vectorization.
11332 2018-02-20 Martin Liska <mliska@suse.cz>
11335 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11336 possible values if we don't have a hint.
11338 2018-02-20 Martin Liska <mliska@suse.cz>
11342 * final.c (shorten_branches): Build align_tab array with one
11344 * opts.c (finish_options): Add alignment option limit check.
11345 (MAX_CODE_ALIGN): Likewise.
11346 (MAX_CODE_ALIGN_VALUE): Likewise.
11347 * doc/invoke.texi: Document maximum allowed option value for
11348 all -falign-* options.
11350 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11353 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11354 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11355 * var-tracking.c (emit_note_insn_var_location): Remove all references
11356 to NOTE_INSN_CALL_ARG_LOCATION.
11357 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11358 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11359 Use copy_rtx_if_shared.
11360 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11361 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11362 (dwarf2out_var_location): Remove handling of
11363 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11365 * final.c (final_scan_insn): Remove all references to
11366 NOTE_INSN_CALL_ARG_LOCATION.
11367 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11368 before dumping final insns.
11369 * except.c (emit_note_eh_region_end): Remove all references to
11370 NOTE_INSN_CALL_ARG_LOCATION.
11371 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11372 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11373 * config/arc/arc.c (hwloop_optimize): Likewise.
11374 * config/arm/arm.c (create_fix_barrier): Likewise.
11375 * config/s390/s390.c (s390_chunkify_start): Likewise.
11376 * config/sh/sh.c (find_barrier): Likewise.
11377 * config/i386/i386.c (rest_of_insert_endbranch,
11378 ix86_seh_fixup_eh_fallthru): Likewise.
11379 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11380 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11381 * config/frv/frv.c (frv_function_prologue): Likewise.
11382 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11384 (note_outside_basic_block_p): Remove all references to
11385 NOTE_INSN_CALL_ARG_LOCATION.
11386 * gengtype.c (adjust_field_rtx_def): Likewise.
11387 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11389 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11390 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11393 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11396 PR tree-optimization/84452
11397 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11398 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11401 2018-02-19 Martin Liska <mliska@suse.cz>
11404 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11406 2018-02-19 Martin Liska <mliska@suse.cz>
11407 Richard Sandiford <richard.sandiford@linaro.org>
11409 PR tree-optimization/82491
11410 * gimple-fold.c (get_base_constructor): Make earlier bail out
11413 2018-02-19 Carl Love <cel@us.ibm.com>
11415 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11416 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11418 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11419 P8V_BUILTIN_VEC_NEG.
11421 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11423 * config/rl78/rl78.md (movdf): New define expand.
11425 2018-02-19 Martin Liska <mliska@suse.cz>
11428 * doc/invoke.texi: Fix typo.
11429 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11431 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11433 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11434 handle rs6000_single_float and rs6000_double_float specially for
11437 2018-02-16 Jeff Law <law@redhat.com>
11439 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11440 (pop_regs): Use it.
11442 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11445 * ipa-inline.c (inline_small_functions): Fix a typo.
11447 2018-02-16 Nathan Sidwell <nathan@acm.org>
11449 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11451 2018-02-16 Carl Love <cel@us.ibm.com>
11453 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11454 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11455 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11456 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11457 expansion to P8V_BUILTIN_VEC_FLOAT2.
11459 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11461 PR rtl-optimization/70023
11462 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11463 src_regno into account.
11465 2018-02-16 Carl Love <cel@us.ibm.com>
11467 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11468 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11469 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11470 * config/rs6000/rs6000.c: Remove case statements for
11471 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11472 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11473 and P9V_BUILTIN_VEC_VINSERT4B.
11474 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11475 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11476 * config/rs6000/vsx.md:
11477 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11480 2018-02-16 Carl Love <cel@us.ibm.com>
11482 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11484 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11486 * config/rs6000/rs6000-c.c: Add the definitions for
11487 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11488 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11489 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11490 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11491 definition for insert4b and define insn *insert3b_internal.
11492 * doc/extend.texi: Add documentation for vec_extract4b.
11494 2018-02-16 Nathan Sidwell <nathan@acm.org>
11496 * doc/extend.texi (Backwards Compatibility): Mention friend
11497 injection. Note for-scope is deprecated.
11498 * doc/invoke.texi (-ffriend-injection): Deprecate.
11500 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11502 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11503 that moved to I2, also allow destinations that are a paradoxical
11504 subreg (instead of a normal reg).
11506 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11509 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11512 2018-02-16 Richard Biener <rguenther@suse.de>
11514 PR tree-optimization/84037
11515 PR tree-optimization/84016
11517 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11518 Adjust vec_construct for the fact we need additional higher latency
11519 128bit inserts for AVX256 and AVX512 vector builds.
11520 (ix86_add_stmt_cost): Scale vector construction cost for
11523 2018-02-16 Richard Biener <rguenther@suse.de>
11525 PR tree-optimization/84417
11526 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11527 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11528 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11530 2018-02-16 Martin Liska <mliska@suse.cz>
11533 * internal-fn.def (ASAN_CHECK): Set proper flags.
11534 (ASAN_MARK): Likewise.
11536 2018-02-16 Julia Koval <julia.koval@intel.com>
11538 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11539 from PTA_CANNONLAKE.
11541 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11544 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11545 Use ++iter rather than iter++ for std::list iterators.
11546 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11547 defer deleting them until all nodes in the forest are processed. Do
11548 free even leaf nodes. Change to_process into auto_vec.
11551 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11552 * vec.h (vec_default_construct): Use memset instead of placement new
11553 if BROKEN_VALUE_INITIALIZATION is defined.
11554 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11555 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11558 PR rtl-optimization/83723
11559 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11560 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11561 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11563 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11565 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11567 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11569 PR rtl-optimization/81443
11570 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11571 from inner REGs to paradoxical SUBREGs.
11573 2018-02-16 Richard Biener <rguenther@suse.de>
11575 PR tree-optimization/84399
11576 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11577 For operands we can analyze at their definition make sure we can
11578 analyze them at each use as well.
11580 2018-02-16 Richard Biener <rguenther@suse.de>
11582 PR tree-optimization/84190
11583 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11584 volatile accesses if the decl isn't volatile.
11586 2018-02-15 Jason Merrill <jason@redhat.com>
11588 PR c++/84314 - ICE with templates and fastcall attribute.
11589 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11590 TYPE_CANONICAL on an existing type.
11592 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11594 PR tree-optimization/84383
11595 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11596 dstoff nor call operand_equal_p if dstbase is NULL.
11598 PR tree-optimization/84334
11599 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11600 also a CONSTANT_CLASS_P, punt.
11602 2018-02-14 Jim Wilson <jimw@sifive.com>
11604 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11605 first SMALL_OPERAND check. New local min_second_step. Move assert
11606 to where locals are set. Add TARGET_RVC support.
11607 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11609 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11611 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11613 2018-02-14 Martin Sebor <msebor@redhat.com>
11615 PR tree-optimization/83698
11616 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11617 arrays constrain the offset range to their bounds.
11618 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11619 (builtin_access::overlap): Avoid setting the size of overlap if it's
11621 (maybe_diag_overlap): Also consider arrays when deciding what values
11622 of offsets to include in diagnostics.
11624 2018-02-14 Martin Sebor <msebor@redhat.com>
11627 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11628 that correspond to the kind of a declaration.
11630 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11633 * config/pa/pa.md: Load address of PIC label using the linkage table
11634 if the label is nonlocal.
11636 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11638 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11639 warning message if user requests -maltivec=be.
11640 * doc/invoke.texi: Document deprecation of -maltivec=be.
11642 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11645 * config/rs6000/rs6000-c.c: Update definitions for
11646 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11647 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11649 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11652 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11653 add _get_ssp intrinsics. Remove argument from
11654 __builtin_ia32_rdssp[d|q].
11655 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11656 * config/i386/i386-builtin.def: Remove argument from
11657 __builtin_ia32_rdssp[d|q].
11658 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11659 ix86_expand_special_args_builtin for _rdssp[d|q].
11660 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11661 Clear register before usage.
11662 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11663 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11665 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11667 PR tree-optimization/84357
11668 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11669 operand 1 of an ARRAY_REF too.
11671 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11674 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11675 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11677 (set_of_reg): New struct.
11678 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11679 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11680 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11682 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11683 Split into bitclr, bitset, bitinvert patterns if appropriate.
11684 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11685 use rx_fuse_in_memory_bitop.
11686 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11687 to named insn, correct maximum insn length.
11689 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11692 * machmode.def: Define a complex mode for PARTIAL_INT.
11693 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11695 * doc/rtl.texi: Document CSPImode.
11696 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11698 (msp430_hard_regno_nregs_with_padding): Likewise.
11700 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11703 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11705 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11707 PR rtl-optimization/84169
11708 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11709 we generated a parallel as new i3 and we split that to new i2 and i3
11710 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11711 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11712 those to i2, not i1. Partially rewrite this scan code.
11714 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11717 * stor-layout.c (place_field): For variable length fields, adjust
11718 offset_align afterwards not just based on the field's alignment,
11719 but also on the size.
11721 PR middle-end/84309
11722 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11723 of exps and logs in the use_exp2 case.
11725 2018-02-13 Jeff Law <law@redhat.com>
11727 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11728 entry for "vector".
11730 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11733 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11737 * common.opt (gas-loc-support, gas-locview-support): New.
11738 (ginline-points, ginternal-reset-location-views): New.
11739 * doc/invoke.texi: Document them. Use @itemx where intended.
11740 (gvariable-location-views): Adjust.
11741 * target.def (reset_location_view): New.
11742 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11743 (TARGET_RESET_LOCATION_VIEW): New.
11744 * doc/tm.texi: Rebuilt.
11745 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11746 (dwarf2out_default_as_locview_support): New.
11747 (output_asm_line_debug_info): Use option variables.
11748 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11749 (output_loc_list): Likewise.
11750 (add_high_low_attributes): Check option variables.
11751 Don't output entry view attribute in strict mode.
11752 (gen_inlined_subroutine_die): Check option variables.
11753 (dwarf2out_inline_entry): Likewise.
11754 (init_sections_and_labels): Likewise.
11755 (dwarf2out_early_finish): Likewise.
11756 (maybe_reset_location_view): New, from...
11757 (dwarf2out_var_location): ... here. Call it.
11758 * debug.h (dwarf2out_default_as_loc_support): Declare.
11759 (dwarf2out_default_as_locview_support): Declare.
11760 * hooks.c (hook_int_rtx_insn_0): New.
11761 * hooks.h (hook_int_rtx_insn_0): Declare.
11762 * toplev.c (process_options): Take -gas-loc-support and
11763 -gas-locview-support from dwarf2out. Enable
11764 -gvariable-location-views by default only with locview
11765 assembler support. Enable -ginternal-reset-location-views by
11766 default only if the target defines the corresponding hook.
11767 Enable -ginline-points by default if location views are
11768 enabled; force it disabled if statement frontiers are
11770 * tree-inline.c (expand_call_inline): Check option variables.
11771 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11773 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11775 PR tree-optimization/84321
11776 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11777 handling. Also check whether the anti-range contains any values
11778 that satisfy the mask; switch to a VR_RANGE if not.
11780 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11783 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11785 2018-02-13 Martin Jambor <mjambor@suse.cz>
11788 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11789 of call statements, also set location of a load to a temporary.
11791 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11793 * config/rl78/rl78.c (add_vector_labels): New function.
11794 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11795 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11796 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11797 which checks that no arguments are passed.
11798 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11799 * doc/extend.texi: Documentation for the new attribute.
11801 2018-02-13 Andreas Schwab <schwab@suse.de>
11803 * config/riscv/linux.h (CPP_SPEC): Define.
11805 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11808 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11809 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11810 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11811 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11812 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11813 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11814 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11815 temporarily for AES and PCLMUL builtins.
11817 PR tree-optimization/84339
11818 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11819 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11822 PR middle-end/84309
11823 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11824 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11825 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11827 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11829 * omp-simd-clone.h: New file.
11830 * omp-simd-clone.c: Include omp-simd-clone.h.
11831 (expand_simd_clones): No longer static.
11832 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11833 cgraph.h and omp-simd-clone.h.
11834 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11835 (vect_recog_widen_shift_pattern): Formatting fix.
11836 (vect_pattern_recog_1): Don't check optab for calls.
11839 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11840 operands[2] into a REG before using gen_lowpart on it.
11842 2018-02-12 Jeff Law <law@redhat.com>
11845 * config/sh/sh.c (find_barrier): Consider a sibling call
11848 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11849 successfully back substituting a reg.
11851 2018-02-12 Richard Biener <rguenther@suse.de>
11853 PR tree-optimization/84037
11854 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11855 parameter, move visited init to caller.
11856 (vect_slp_analyze_operations): Separate cost from validity
11857 check, initialize visited once for all instances.
11858 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11860 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11862 (vect_model_promotion_demotion_cost): Likewise.
11863 (vectorizable_bswap): Guard cost modeling with !slp_node
11864 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11866 (vectorizable_call): Likewise.
11867 (vectorizable_conversion): Likewise.
11868 (vectorizable_assignment): Likewise.
11869 (vectorizable_shift): Likewise.
11870 (vectorizable_operation): Likewise.
11871 (vectorizable_store): Likewise.
11872 (vectorizable_load): Likewise.
11873 (vectorizable_condition): Likewise.
11874 (vectorizable_comparison): Likewise.
11876 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11879 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11880 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11881 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11883 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11885 PR middle-end/83665
11886 * params.def (inline-min-speedup): Increase from 8 to 15.
11887 (max-inline-insns-auto): Decrease from 40 to 30.
11888 * ipa-split.c (consider_split): Add some buffer for function to
11889 be considered inlining candidate.
11890 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11893 2018-02-12 Richard Biener <rguenther@suse.de>
11895 PR tree-optimization/84037
11896 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11897 matched stmts if we cannot swap the non-matched ones.
11899 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11901 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11902 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11903 _mm_maskz_scalef_round_ss): New intrinsics.
11904 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11905 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11906 __builtin_ia32_scalefss_round): Remove.
11907 (__builtin_ia32_scalefsd_mask_round,
11908 __builtin_ia32_scalefss_mask_round): New intrinsics.
11909 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11910 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11911 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11912 "<round_constraint>")): Changed to ...
11913 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11914 "<round_scalar_constraint>")): ... this.
11915 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11916 %0, %1, %2<round_op3>}"): Changed to ...
11917 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11918 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11919 %2<round_scalar_mask_op3>}"): ... this.
11920 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11922 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11924 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11925 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11926 (_mm_maskz_sqrt_round_ss): New intrinsics.
11927 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11928 (__builtin_ia32_sqrtsd_mask_round)
11929 (__builtin_ia32_sqrtss_mask_round): New builtins.
11930 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11931 (__builtin_ia32_sqrtss_round): Remove.
11932 (__builtin_ia32_sqrtsd_mask_round)
11933 (__builtin_ia32_sqrtss_mask_round): New builtins.
11934 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11935 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11936 ((match_operand:VF_128 1 "vector_operand"
11937 "xBm,<round_constraint>")): Changed to ...
11938 ((match_operand:VF_128 1 "vector_operand"
11939 "xBm,<round_scalar_constraint>")): ... this.
11940 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11941 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11942 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11943 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11944 %<iptr>1<round_scalar_mask_op3>}): ... this.
11945 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11946 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11948 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11951 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11952 Cast vec_cmpeq result to correct type.
11953 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11954 Cast vec_cmpgt result to correct type.
11956 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11958 * final.c (final_scan_insn_1): Renamed from...
11959 (final_scan_insn): ... this. New wrapper, to recover
11960 seen from the outermost call in recursive ones.
11961 * config/sparc/sparc.c (output_return): Drop seen from call.
11962 (output_sibcall): Likewise.
11963 * config/visium/visium.c (output_branch): Likewise.
11965 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11967 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11970 2018-02-10 Alan Modra <amodra@gmail.com>
11973 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11974 Specify LR as an input.
11976 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11979 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11980 remove_member_access_dummy_vars): New functions.
11981 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11982 lower_omp_1, execute_lower_omp): Use them.
11984 PR rtl-optimization/84308
11985 * shrink-wrap.c (spread_components): Release todo vector.
11987 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11989 PR rtl-optimization/57193
11990 * ira-color.c (struct allocno_color_data): Add member
11991 conflict_allocno_hard_prefs.
11992 (update_conflict_allocno_hard_prefs): New.
11993 (bucket_allocno_compare_func): Add a preference based on
11994 conflict_allocno_hard_prefs.
11995 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11996 (color_allocnos): Remove a dead code. Initiate
11997 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11999 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12002 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12003 constraint from =wa to wa. Avoid a subreg on the output operand,
12004 instead use a pseudo and subreg it in a move.
12005 (p9_xxbrd_<mode>): Changed to ...
12006 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12007 (p9_xxbrd_v2df): New expander.
12008 (p9_xxbrw_<mode>): Changed to ...
12009 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12010 (p9_xxbrw_v4sf): New expander.
12012 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12014 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12016 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12019 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12020 multiply in 32-bit mode.
12021 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12022 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12025 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12027 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12028 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12029 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12030 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12032 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12035 * dwarf2out.c (is_trivial_indirect_ref): New function.
12036 (dwarf2out_late_global_decl): Do not generate a location
12037 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12038 and that are not defined in the current unit.
12040 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12042 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12043 instead of a libcall for UNORDERED.
12045 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12048 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12049 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12051 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12054 * config/s390/s390.c (s390_set_current_function): Invoke
12055 s390_indirect_branch_settings also if fndecl didn't change.
12057 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12059 * config/rs6000/rs6000.md (blockage): Set length to zero.
12061 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12063 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12065 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12068 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12069 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12073 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12074 PARALLEL incoming that failed vt_get_decl_and_offset check.
12076 PR middle-end/84237
12077 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12078 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12080 (get_variable_section): For decls in named .bss* sections pass true as
12081 second argument to bss_initializer_p.
12083 2018-02-09 Marek Polacek <polacek@redhat.com>
12084 Jakub Jelinek <jakub@redhat.com>
12087 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12088 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12089 Sync some changes from cxx_fold_indirect_ref.
12091 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12093 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12095 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12096 (BLOCK_INLINE_ENTRY_LABEL): New.
12097 (dwarf2out_var_location): Disregard inline entry markers.
12098 (inline_entry_data): New struct.
12099 (inline_entry_data_hasher): New hashtable type.
12100 (inline_entry_data_hasher::hash): New.
12101 (inline_entry_data_hasher::equal): New.
12102 (inline_entry_data_table): New variable.
12103 (add_high_low_attributes): Add DW_AT_entry_pc and
12104 DW_AT_GNU_entry_view attributes if a pending entry is found
12105 in inline_entry_data_table. Add old entry_pc attribute only
12106 if debug nonbinding markers are disabled.
12107 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12108 markers are enabled.
12109 (block_within_block_p, dwarf2out_inline_entry): New.
12110 (dwarf2out_finish): Check that no entries remained in
12111 inline_entry_data_table.
12112 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12113 (final_scan_insn, notice_source_line): Likewise.
12114 (rest_of_clean_state): Skip inline entry markers.
12115 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12117 * gimple.c (gimple_build_debug_inline_entry): New.
12118 * gimple.h (enum gimple_debug_subcode): Add
12119 GIMPLE_DEBUG_INLINE_ENTRY.
12120 (gimple_build_debug_inline_entry): Declare.
12121 (gimple_debug_inline_entry_p): New.
12122 (gimple_debug_nonbind_marker_p): Adjust.
12123 * insn-notes.def (INLINE_ENTRY): New.
12124 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12125 inline entry marker notes.
12126 (print_insn): Likewise.
12127 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12128 (INSN_DEBUG_MARKER_KIND): Likewise.
12129 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12130 * tree-inline.c (expand_call_inline): Build and insert
12131 debug_inline_entry stmt.
12132 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12133 inline entry blocks early, if nonbind markers are enabled.
12134 (dump_scope_block): Dump fragment info.
12135 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12136 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12137 (gimple_build_debug_inline_entry): New.
12138 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12139 Enable/disable inline entry points too.
12140 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12141 (DEBUG_INSN): Describe inline entry markers.
12143 * common.opt (gvariable-location-views): New.
12144 (gvariable-location-views=incompat5): New.
12145 * config.in: Rebuilt.
12146 * configure: Rebuilt.
12147 * configure.ac: Test assembler for view support.
12148 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12149 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12150 * dwarf2out.c (var_loc_view): New typedef.
12151 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12152 (dwarf2out_locviews_in_attribute): New.
12153 (dwarf2out_locviews_in_loclist): New.
12154 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12155 (enum dw_line_info_opcode): Add LI_adv_address.
12156 (struct dw_line_info_table): Add view.
12157 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12158 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12159 (zero_view_p): New variable.
12160 (ZERO_VIEW_P): New macro.
12161 (output_asm_line_debug_info): New.
12162 (struct var_loc_node): Add view.
12163 (add_AT_view_list, AT_loc_list): New.
12164 (add_var_loc_to_decl): Add view param. Test it against last.
12165 (new_loc_list): Add view params. Record them.
12166 (AT_loc_list_ptr): Handle loc and view lists.
12167 (view_list_to_loc_list_val_node): New.
12168 (print_dw_val): Handle dw_val_class_view_list.
12169 (size_of_die): Likewise.
12170 (value_format): Likewise.
12171 (loc_list_has_views): New.
12172 (gen_llsym): Set vl_symbol too.
12173 (maybe_gen_llsym, skip_loc_list_entry): New.
12174 (dwarf2out_maybe_output_loclist_view_pair): New.
12175 (output_loc_list): Output view list or entries too.
12176 (output_view_list_offset): New.
12177 (output_die): Handle dw_val_class_view_list.
12178 (output_dwarf_version): New.
12179 (output_compilation_unit_header): Use it.
12180 (output_skeleton_debug_sections): Likewise.
12181 (output_rnglists, output_line_info): Likewise.
12182 (output_pubnames, output_aranges): Update version comments.
12183 (output_one_line_info_table): Output view numbers in asm comments.
12184 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12185 Call maybe_gen_llsym.
12186 (loc_list_from_tree_1): Adjust.
12187 (add_AT_location_description): Create view list attribute if
12188 needed, check it's absent otherwise.
12189 (convert_cfa_to_fb_loc_list): Adjust.
12190 (maybe_emit_file): Call output_asm_line_debug_info for test.
12191 (dwarf2out_var_location): Reset views as needed. Precompute
12192 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12193 attribute. Set view.
12194 (new_line_info_table): Reset next view.
12195 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12196 (dwarf2out_source_line): Likewise. Output view resets and labels to
12197 the assembler, or select appropriate line info opcodes.
12198 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12199 (optimize_string_length): Catch it. Adjust.
12200 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12201 dw_val_class_view_list, and remove it if no longer needed.
12202 (hash_loc_list): Hash view numbers.
12203 (loc_list_hasher::equal): Compare them.
12204 (optimize_location_lists): Check whether a view list symbol is
12205 needed, and whether the locview attribute is present, and
12206 whether they match. Remove the locview attribute if no longer
12208 (index_location_lists): Call skip_loc_list_entry for test.
12209 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12210 Use output_dwarf_version.
12211 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12212 (struct dw_val_node): Add val_view_list.
12213 * final.c (SEEN_NEXT_VIEW): New.
12214 (set_next_view_needed): New.
12215 (clear_next_view_needed): New.
12216 (maybe_output_next_view): New.
12217 (final_start_function): Rename to...
12218 (final_start_function_1): ... this. Take pointer to FIRST,
12219 add SEEN parameter. Emit param bindings in the initial view.
12220 (final_start_function): Reintroduce SEEN-less interface.
12221 (final): Rename to...
12222 (final_1): ... this. Take SEEN parameter. Output final pending
12223 next view at the end.
12224 (final): Reintroduce seen-less interface.
12225 (final_scan_insn): Output pending next view before switching
12226 sections or ending a block. Mark the next view as needed when
12227 outputting variable locations. Notify debug backend of section
12228 changes, and of location view changes.
12229 (rest_of_handle_final): Adjust.
12230 * toplev.c (process_options): Autodetect value for debug variable
12231 location views option. Warn on incompat5 without -gdwarf-5.
12232 * doc/invoke.texi (gvariable-location-views): New.
12233 (gvariable-location-views=incompat5): New.
12234 (gno-variable-location-views): New.
12236 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12238 PR tree-optimization/84136
12239 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12240 that the result of find_edge is non-NULL.
12242 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12245 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12246 storing integer register in SImode. Fix cost of 256 and 512
12247 byte aligned SSE register store.
12249 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12251 * config/i386/i386.c (ix86_multiplication_cost): Fix
12252 multiplication cost for TARGET_AVX512DQ.
12254 2018-02-08 Marek Polacek <polacek@redhat.com>
12256 PR tree-optimization/84238
12257 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12260 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12262 PR tree-optimization/84265
12263 * tree-vect-stmts.c (vectorizable_store): Don't treat
12264 VMAT_CONTIGUOUS accesses as grouped.
12265 (vectorizable_load): Likewise.
12267 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12269 PR tree-optimization/81635
12270 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12271 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12272 (test_round_for_mask): New functions.
12273 (wide_int_cc_tests): Call test_round_for_mask.
12274 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12275 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12276 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12277 range returned by get_range_info.
12279 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12282 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12283 * symtab.c: Include builtins.h
12284 (symtab_node::output_to_lto_symbol_table_p): Move here
12285 from lto-streamer-out.c:output_symbol_p.
12286 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12287 (output_symbol_p): Move all logic to symtab.c
12288 (produce_symtab): Update.
12290 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12292 * config/s390/s390-opts.h (enum indirect_branch): Define.
12293 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12294 (s390_indirect_branch_via_thunk)
12295 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12296 (enum s390_indirect_branch_type): Define.
12297 * config/s390/s390.c (struct s390_frame_layout, struct
12298 machine_function): Remove.
12299 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12300 (indirect_branch_table_label_no, indirect_branch_table_name):
12302 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12303 (enum s390_indirect_branch_option): Define.
12304 (s390_return_addr_from_memory): New function.
12305 (s390_handle_string_attribute): New function.
12306 (s390_attribute_table): Add new attribute handler.
12307 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12308 (s390_indirect_branch_via_thunk): New function.
12309 (s390_indirect_branch_via_inline_thunk): New function.
12310 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12311 sibling call optimization for non z10 compiles.
12312 (s390_emit_call): Force indirect branch target to be a single
12313 register. Add r1 clobber for non-z10 compiles.
12314 (s390_emit_epilogue): Emit return jump via return_use expander.
12315 (s390_reorg): Handle JUMP_INSNs as execute targets.
12316 (s390_option_override_internal): Perform validity checks for the
12317 new command line options.
12318 (s390_indirect_branch_attrvalue): New function.
12319 (s390_indirect_branch_settings): New function.
12320 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12321 (s390_output_indirect_thunk_function): New function.
12322 (s390_code_end): Implement target hook.
12323 (s390_case_values_threshold): Implement target hook.
12324 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12326 * config/s390/s390.h (struct s390_frame_layout)
12327 (struct machine_function): Move here from s390.c.
12328 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12329 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12330 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12331 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12332 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12333 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12334 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12335 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12336 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12337 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12338 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12339 (mnemonic attribute): Add values which aren't recognized
12341 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12342 pattern for branch conversion. Fix mnemonic attribute.
12343 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12344 indirect branch via thunk if requested.
12345 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12346 ("*indirect_jump"): Disable for branch conversion using out of
12348 ("indirect_jump_via_thunk<mode>_z10")
12349 ("indirect_jump_via_thunk<mode>")
12350 ("indirect_jump_via_inlinethunk<mode>_z10")
12351 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12352 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12353 ("casesi_jump_via_inlinethunk<mode>_z10")
12354 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12355 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12356 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12357 ("*indirect2_jump"): Disable for branch conversion.
12358 ("casesi_jump"): Turn into expander and expand patterns for branch
12360 ("return_use"): New expander.
12361 ("*return"): Emit return via thunk and rename it to ...
12362 ("*return<mode>"): ... this one.
12363 * config/s390/s390.opt: Add new options and and enum for the
12366 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12368 * lra-constraints.c (match_reload): Unconditionally use
12369 gen_lowpart_SUBREG, rather than selecting between that
12370 and equivalent gen_rtx_SUBREG code.
12372 2018-02-08 Richard Biener <rguenther@suse.de>
12374 PR tree-optimization/84233
12375 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12376 changed flag instead of boguously re-using phi_inserted.
12378 2018-02-08 Martin Jambor <mjambor@suse.cz>
12380 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12381 static local variables.
12383 2018-02-08 Richard Biener <rguenther@suse.de>
12385 PR tree-optimization/84278
12386 * tree-vect-stmts.c (vectorizable_store): When looking for
12387 smaller vector types to perform grouped strided loads/stores
12388 make sure the mode is supported by the target.
12389 (vectorizable_load): Likewise.
12391 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12393 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12394 Increase LDP/STP opportunities by adding adjacent callee-saves.
12396 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12398 PR rtl-optimization/84068
12399 PR rtl-optimization/83459
12400 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12402 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12404 PR tree-optimization/84224
12405 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12406 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12407 non-zero arguments.
12409 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12412 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12413 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12414 count, remove one USE.
12416 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12418 * doc/install.texi (Configuration): Document the
12419 --with-long-double-format={ibm,ieee} PowerPC configuration
12423 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12424 Convert from define_expand to be define_insn_and_split. Rework
12425 float/double/_Float128 conversions to QI/HI/SImode to work with
12426 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12427 conversions to QI/HImode types did a store and then a load to
12428 truncate the value. For conversions to VSX registers, don't split
12429 the insn, instead emit the code directly. Use the code iterator
12430 any_fix to combine signed and unsigned conversions.
12431 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12432 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12433 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12434 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12435 (fix_<mode>di2_hw): Likewise.
12436 (fixuns_<mode>di2_hw): Likewise.
12437 (fix_<mode>si2_hw): Likewise.
12438 (fixuns_<mode>si2_hw): Likewise.
12439 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12440 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12441 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12442 fix<uns>_trunc<SFDF:mode>si2_p8.
12443 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12444 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12445 (fix<uns>_<mode>_mem): Likewise.
12446 (fctiw<u>z_<mode>_mem): Likewise.
12447 (fix<uns>_<mode>_mem): Likewise.
12448 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12449 the register allocator from doing a direct move to the GPRs to do
12450 a store, and instead use the ISA 3.0 store byte/half-word from
12451 vector register instruction. For IEEE 128-bit floating point,
12452 also optimize stores of 32-bit ints.
12453 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12455 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12457 * genextract.c (push_pathstr_operand): New function to support
12459 (walk_rtx): Call push_pathstr_operand.
12460 (print_path): Support [a-zA-Z].
12462 2018-02-07 Richard Biener <rguenther@suse.de>
12464 PR tree-optimization/84037
12465 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12466 (cse_and_gimplify_to_preheader): Declare.
12467 (vect_get_place_in_interleaving_chain): Likewise.
12468 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12470 (_loop_vec_info::~_loop_vec_info): Delete it.
12471 (cse_and_gimplify_to_preheader): New function.
12472 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12473 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12474 (vectorizable_load): Likewise. For grouped stores always base
12475 the IV on the first element.
12476 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12477 condition before gimplifying.
12479 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12481 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12482 *DIV_EXPR and *MOD_EXPR.
12484 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12487 * config/i386/i386.c (ix86_option_override_internal): Mask out
12488 the CF_SET bit when checking -fcf-protection.
12490 2018-02-07 Tom de Vries <tom@codesourcery.com>
12493 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12496 2018-02-07 Richard Biener <rguenther@suse.de>
12498 PR tree-optimization/84204
12499 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12502 PR tree-optimization/84205
12503 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12504 special-case isl_ast_op_zdiv_r.
12506 PR tree-optimization/84223
12507 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12508 Only add conditions from within the region.
12509 (gather_bbs::after_dom_children): Adjust.
12511 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12514 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12515 * config/avr/avr.md: Only post-reload split REG-REG moves if
12516 either register is GENERAL_REG_P.
12518 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12520 PR tree-optimization/84235
12521 * tree-ssa-scopedtables.c
12522 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12523 if the subtraction is performed in floating point type where NaNs are
12524 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12525 build 1. Formatting fix.
12527 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12530 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12531 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12532 and skip it regardless of bb boundaries. Use CALL_P macro,
12533 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12534 unnecessarily, formatting fix.
12536 2018-02-06 Michael Collison <michael.collison@arm.com>
12538 * config/arm/thumb2.md:
12539 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12540 (*thumb_mov_notscc): Ditto.
12542 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12545 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12546 unsigned_fix, not "s".
12548 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12550 * configure.ac (gcc_fn_eh_frame_ro): New function.
12551 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12552 correct .eh_frame permissions.
12553 * configure: Regenerate.
12555 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12557 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12558 irrelevant options.
12560 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12562 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12563 Display warning message for -mno-speculate-indirect-jumps.
12565 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12567 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12569 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12571 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12573 PR tree-optimization/84225
12574 * tree-eh.c (find_trapping_overflow): Only call
12575 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12577 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12580 * config/i386/i386.c: Reimplement the check of possible options
12581 -mibt/-mshstk conbination. Change error messages.
12582 * doc/invoke.texi: Fix a typo: remove extra '='.
12584 2018-02-06 Marek Polacek <polacek@redhat.com>
12586 PR tree-optimization/84228
12587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12589 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12592 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12593 emitted arch directives.
12594 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12595 __ARM_FEATURE_COPROC before changing architectures.
12597 2018-02-06 Richard Biener <rguenther@suse.de>
12599 * config/i386/i386.c (print_reg): Fix typo.
12600 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12602 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12604 * configure: Regenerate.
12606 2018-02-05 Martin Sebor <msebor@redhat.com>
12608 PR tree-optimization/83369
12609 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12612 2018-02-05 Martin Liska <mliska@suse.cz>
12614 * doc/invoke.texi: Cherry-pick upstream r323995.
12616 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12618 * ira.c (ira_init_register_move_cost): Adjust comment.
12620 2018-02-05 Martin Liska <mliska@suse.cz>
12622 PR gcov-profile/84137
12623 * doc/gcov.texi: Fix typo in documentation.
12625 2018-02-05 Martin Liska <mliska@suse.cz>
12627 PR gcov-profile/83879
12628 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12629 using dlopen functionality.
12631 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12633 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12634 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12635 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12636 _mm_maskz_range_round_ss): New intrinsics.
12637 (__builtin_ia32_rangesd128_round)
12638 (__builtin_ia32_rangess128_round): Remove.
12639 (__builtin_ia32_rangesd128_mask_round,
12640 __builtin_ia32_rangess128_mask_round): New builtins.
12641 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12642 __builtin_ia32_rangess128_round): Remove.
12643 (__builtin_ia32_rangesd128_mask_round,
12644 __builtin_ia32_rangess128_mask_round): New builtins.
12645 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12646 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12647 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12648 "<round_saeonly_constraint>")): Changed to ...
12649 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12650 "<round_saeonly_scalar_constraint>")): ... this.
12651 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12652 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12653 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12654 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12655 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12657 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12659 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12661 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12662 Remove all values except native, 8540 and 8548.
12664 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12666 * config/i386/i386.c (ix86_output_function_return): Pass
12667 INVALID_REGNUM, instead of -1, as invalid register number to
12668 indirect_thunk_name and output_indirect_thunk.
12670 2018-02-02 Julia Koval <julia.koval@intel.com>
12672 * config.gcc: Add -march=icelake.
12673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12674 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12675 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12676 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12677 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12678 (processor_target_table): Add icelake.
12679 (ix86_option_override_internal): Handle new PTAs.
12680 (get_builtin_code_for_version): Handle icelake.
12681 (M_INTEL_COREI7_ICELAKE): New.
12682 (fold_builtin_cpu): Handle icelake.
12683 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12684 * doc/invoke.texi: Add -march=icelake.
12686 2018-02-02 Julia Koval <julia.koval@intel.com>
12688 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12689 to wide_int_bitmask.
12690 * wide-int-bitmask.h: New.
12692 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12695 * config/i386/i386.md: Replace Pmode with word_mode in
12696 builtin_setjmp_setup and builtin_longjmp to support x32.
12698 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12702 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12704 (rs6000_supported_cpu_names): New static variable.
12705 (linux_cpu_translation_table): Likewise.
12706 (elf_platform) <cpu>: Define new static variable and use it.
12707 Translate kernel AT_PLATFORM name to canonical name if needed.
12708 Error if platform name is unknown.
12710 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12713 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12715 2018-02-01 Jeff Law <law@redhat.com>
12718 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12719 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12720 the scratch if RELEASE_VIA_POP is false.
12721 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12722 If we have to save a temporary register, decrement SIZE appropriately.
12723 Pass new arguments to release_scratch_register_on_entry.
12724 (ix86_adjust_stack_and_probe): Likewise.
12725 (ix86_emit_probe_stack_range): Pass new arguments to
12726 release_scratch_register_on_entry.
12728 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12730 PR rtl-optimization/84157
12731 * combine.c (change_zero_ext): Use REG_P predicate in
12732 front of HARD_REGISTER_P predicate.
12734 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12736 * config/avr/avr.c (avr_option_override): Move disabling of
12737 -fdelete-null-pointer-checks to...
12738 * common/config/avr/avr-common.c (avr_option_optimization_table):
12741 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12743 PR tree-optimization/81635
12744 * tree-data-ref.c (split_constant_offset_1): For types that
12745 wrap on overflow, try to use range info to prove that wrapping
12748 2018-02-01 Renlin Li <renlin.li@arm.com>
12751 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12752 TAILCALL_ADDR_REGS.
12753 (aarch64_register_move_cost): Likewise.
12754 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12755 TAILCALL_ADDR_REGS.
12756 (REG_CLASS_NAMES): Likewise.
12757 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12758 TAILCALL_ADDR_REGS. Remove IP registers.
12759 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12761 2018-02-01 Richard Biener <rguenther@suse.de>
12763 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12764 for specifying RPO order and allow NULL for that.
12765 * domwalk.c (dom_walker::dom_walker): Likewise.
12766 (dom_walker::walk): Handle NULL RPO order.
12767 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12769 (rewrite_update_dom_walker): Likewise.
12770 (mark_def_dom_walker): Likewise.
12772 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12774 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12775 (aarch64_maybe_expand_sve_subreg_move): Declare.
12776 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12777 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12779 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12780 that are semantically a reverse operation.
12781 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12782 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12783 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12785 (aarch64_can_change_mode_class): For big-endian, forbid changes
12786 between two SVE modes if they have different element sizes.
12788 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12790 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12791 the TImode handling for big-endian targets.
12793 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12795 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12796 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12798 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12799 Remove BSWAP handing for big-endian targets and use the form of
12800 LD1RQ appropariate for the mode.
12802 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12804 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12805 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12806 duplicated element.
12808 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12811 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12812 check for operands that need to go through aarch64_sve_reload_be.
12814 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12816 PR tree-optimization/81661
12817 PR tree-optimization/84117
12818 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12819 * tree-eh.c: Include gimplify.h.
12820 (find_trapping_overflow, replace_trapping_overflow,
12821 rewrite_to_non_trapping_overflow): New functions.
12822 * tree-vect-loop.c: Include tree-eh.h.
12823 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12824 * tree-data-ref.c: Include tree-eh.h.
12825 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12827 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12829 PR rtl-optimization/84123
12830 * combine.c (change_zero_ext): Check if hard register satisfies
12831 can_change_dest_mode before calling gen_lowpart_SUBREG.
12833 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12836 * ira.c (ira_init_register_move_cost): Remove assert.
12838 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12840 PR rtl-optimization/84071
12841 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12842 * doc/tm.texi: Regenerate.
12844 2018-01-31 Richard Biener <rguenther@suse.de>
12846 PR tree-optimization/84132
12847 * tree-data-ref.c (analyze_miv_subscript): Properly
12848 check whether evolution_function_is_affine_multivariate_p
12849 before calling gcd_of_steps_may_divide_p.
12851 2018-01-31 Julia Koval <julia.koval@intel.com>
12854 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12855 * config/i386/i386.md (rdpid_rex64) New.
12856 (rdpid): Make 32bit only.
12858 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12861 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12862 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12864 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12867 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12869 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12871 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12873 PR rtl-optimization/84071
12874 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12875 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12877 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12879 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12880 (arc_attribute_table): Add 'aux' attribute.
12881 (arc_in_small_data_p): Consider aux like variables.
12882 (arc_is_aux_reg_p): New function.
12883 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12884 (arc_get_aux_arg): New function.
12885 (prepare_move_operands): Handle aux-register access.
12886 (arc_handle_aux_attribute): New function.
12887 * doc/extend.texi (ARC Variable attributes): Add subsection.
12889 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12891 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12892 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12893 (arc_attribute_table): Add 'uncached' attribute.
12894 (arc_print_operand): Print '.di' flag for uncached memory
12896 (arc_in_small_data_p): Do not consider for small data the uncached
12898 (arc_is_uncached_mem_p): New function.
12899 * config/arc/predicates.md (compact_store_memory_operand): Check
12900 for uncached memory accesses.
12901 (nonvol_nonimm_operand): Likewise.
12902 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12904 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12907 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12908 falign-loops=): Add Optimization flag.
12910 2018-01-30 Jeff Law <law@redhat.com>
12913 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12914 INT_REGISTERS_SAVED. Check it prior to calling
12915 get_scratch_register_on_entry.
12916 (ix86_adjust_stack_and_probe): Similarly.
12917 (ix86_emit_probe_stack_range): Similarly.
12918 (ix86_expand_prologue): Corresponding changes.
12920 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12923 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12924 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12926 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12929 * lra-constraints.c (curr_insn_transform): Process AND in the
12932 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12934 PR rtl-optimization/83986
12935 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12936 dependence against last_pending_memory_flush in addition to
12937 pending_jump_insns.
12939 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12941 PR tree-optimization/81611
12942 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12945 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12948 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12951 2018-01-30 Richard Biener <rguenther@suse.de>
12952 Jakub Jelinek <jakub@redhat.com>
12954 PR tree-optimization/84111
12955 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12956 inner loops added during recursion, as they don't have up-to-date
12959 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12962 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12963 (can_inline_edge_by_limits_p): ... here.
12964 (can_early_inline_edge_p, check_callers,
12965 update_caller_keys, update_callee_keys, recursive_inlining,
12966 add_new_edges_to_heap, speculation_useful_p,
12967 inline_small_functions,
12968 inline_small_functions, flatten_function,
12969 inline_to_all_callers_1): Update.
12971 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12973 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12976 2018-01-30 Richard Biener <rguenther@suse.de>
12978 PR tree-optimization/83008
12979 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12980 invariant and constant vector uses in stmts when they need
12981 more than one stmt.
12983 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12986 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12987 * configure: Regenerate.
12989 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12991 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12993 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12994 Use gen_rtx_REG rather than gen_lowpart.
12996 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12998 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12999 rather than 0 when creating partial subregs.
13001 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13003 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13006 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13009 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13010 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13011 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13012 flags. This restores the settings used before the 2017-07-24.
13013 Turning off pre increment/decrement/modify allows IVOPTS to
13014 optimize DF/SF loops where the index is an int.
13016 2018-01-29 Richard Biener <rguenther@suse.de>
13017 Kelvin Nilsen <kelvin@gcc.gnu.org>
13020 * tree-vect-stmts.c (vectorizable_call): Don't call
13021 targetm.vectorize_builtin_md_vectorized_function if callee is
13024 2018-01-22 Carl Love <cel@us.ibm.com>
13026 * doc/extend.tex: Fix typo in second arg in
13027 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13029 2018-01-29 Richard Biener <rguenther@suse.de>
13031 PR tree-optimization/84086
13032 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13033 (flush_ssaname_freelist): When SSA names were released reset
13034 the SCEV hash table.
13036 2018-01-29 Richard Biener <rguenther@suse.de>
13038 PR tree-optimization/84057
13039 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13040 removed paths when removing edges.
13042 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13044 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13045 -mfunction-return=@var{choice}.
13047 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13049 PR diagnostic/84034
13050 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13051 Handle CR like TAB.
13052 (layout::print_source_line): Likewise.
13053 (test_get_line_width_without_trailing_whitespace): Add test cases.
13055 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13057 PR middle-end/84040
13058 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13061 2018-01-26 Jim Wilson <jimw@sifive.com>
13063 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13065 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13068 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13070 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13071 and CMP + SUB-immediate -> SUBS.
13073 2018-01-26 Martin Sebor <msebor@redhat.com>
13075 PR tree-optimization/83896
13076 * tree-ssa-strlen.c (get_string_len): Rename...
13077 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13078 Avoid assuming length is constant.
13079 (handle_char_store): Use HOST_WIDE_INT for string length.
13081 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13084 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13085 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13087 2018-01-26 Richard Biener <rguenther@suse.de>
13089 PR rtl-optimization/84003
13090 * dse.c (record_store): Only record redundant stores when
13091 the earlier store aliases at least all accesses the later one does.
13093 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13095 PR rtl-optimization/83985
13096 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13097 REG_CFA_RESTORE insns.
13098 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13099 REG_CFA_RESTORE insns here.
13102 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13103 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13105 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13107 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13108 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13109 (arc_init): Likewise.
13110 (arc_override_options): Likewise.
13111 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13113 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13115 * config/arc/arc.h (TARGET_DBNZ): Define.
13116 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13117 properly set the tune attribute.
13118 (dbnz): Use TARGET_DBNZ guard.
13119 * config/arc/arc.opt (mtune): Add core3 option.
13121 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13123 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13124 recognize new pic like addresses.
13125 (arc_delegitimize_address): Clean up.
13127 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13129 * config/arc/arc-arches.def: Option mrf16 valid for all
13131 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13132 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13133 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13134 * config/arc/arc-tables.opt: Regenerate.
13135 * config/arc/arc.c (arc_conditional_register_usage): Handle
13136 reduced register file case.
13137 (arc_file_start): Set must have build attributes.
13138 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13139 mrf16 option value.
13140 * config/arc/arc.opt (mrf16): Add new option.
13141 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13142 * config/arc/genmultilib.awk: Handle new mrf16 option.
13143 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13144 * config/arc/t-multilib: Regenerate.
13145 * doc/invoke.texi (ARC Options): Document mrf16 option.
13147 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13149 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13150 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13151 (arc_attribute_table): Add 'secure_call' attribute.
13152 (arc_print_operand): Print secure call operand.
13153 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13155 (arc_is_secure_call_p): New function. * config/arc/arc.md
13156 (call_i): Add support for sjli instruction.
13157 (call_value_i): Likewise.
13158 * config/arc/constraints.md (Csc): New constraint.
13160 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13161 John Eric Martin <John.Martin@emmicro-us.com>
13163 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13164 * config/arc/arc.c (_arc_jli_section): New struct.
13165 (arc_jli_section): New type.
13166 (rc_jli_sections): New static variable.
13167 (arc_handle_jli_attribute): New function.
13168 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13169 (arc_file_end): New function.
13170 (TARGET_ASM_FILE_END): Define.
13171 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13172 (arc_add_jli_section): New function.
13173 (jli_call_scan): Likewise.
13174 (arc_reorg): Call jli_call_scan.
13175 (arc_output_addsi): Remove 'S' from printing asm operand.
13176 (arc_is_jli_call_p): New function.
13177 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13179 (movhi_insn): Likewise.
13180 (movsi_insn): Likewise.
13181 (movsi_set_cc_insn): Likewise.
13182 (loadqi_update): Likewise.
13183 (load_zeroextendqisi_update): Likewise.
13184 (load_signextendqisi_update): Likewise.
13185 (loadhi_update): Likewise.
13186 (load_zeroextendhisi_update): Likewise.
13187 (load_signextendhisi_update): Likewise.
13188 (loadsi_update): Likewise.
13189 (loadsf_update): Likewise.
13190 (movsicc_insn): Likewise.
13191 (bset_insn): Likewise.
13192 (bxor_insn): Likewise.
13193 (bclr_insn): Likewise.
13194 (bmsk_insn): Likewise.
13195 (bicsi3_insn): Likewise.
13196 (cmpsi_cc_c_insn): Likewise.
13197 (movsi_ne): Likewise.
13198 (movsi_cond_exec): Likewise.
13199 (clrsbsi2): Likewise.
13200 (norm_f): Likewise.
13210 (call_i): Remove 'S' asm letter, add jli instruction.
13211 (call_value_i): Likewise.
13212 * config/arc/arc.op (mjli-always): New option.
13213 * config/arc/constraints.md (Cji): New constraint.
13214 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13216 (subsf3_fpx): Likewise.
13217 (mulsf3_fpx): Likewise.
13218 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13220 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13221 function attrbutes.
13222 * doc/invoke.texi (ARC): Document mjli-always option.
13224 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13226 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13227 avoid addition with 0 and use incw and decw where possible.
13229 2018-01-26 Richard Biener <rguenther@suse.de>
13231 PR tree-optimization/81082
13232 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13233 association if it requires casting to unsigned.
13234 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13235 from fold_plusminus_mult_expr to catch important cases late when
13236 range info is available.
13238 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13240 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13241 * configure.ac (hidden_linkonce): New test.
13242 * configure: Regenerate.
13243 * config.in: Regenerate.
13245 2018-01-26 Julia Koval <julia.koval@intel.com>
13247 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13248 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13249 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13250 _mm_mask_bitshuffle_epi64_mask): Fix type.
13251 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13252 USI_FTYPE_V4DI_V4DI_USI): Remove.
13253 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13254 __builtin_ia32_vpshufbitqmb256_mask,
13255 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13256 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13257 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13259 2018-01-26 Alan Modra <amodra@gmail.com>
13262 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13263 UNSPEC_VBPERMQ. Sort other unspecs.
13265 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13267 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13269 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13271 PR middle-end/83055
13272 * predict.c (drop_profile): Do not push/pop cfun; update also
13274 (handle_missing_profiles): Fix logic looking for zero profiles.
13276 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13278 PR middle-end/83977
13279 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13280 on functions with #pragma omp declare simd or functions with simd
13282 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13283 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13284 Remove trailing \n from warning_at calls.
13286 2018-01-25 Tom de Vries <tom@codesourcery.com>
13289 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13290 for neutered workers.
13292 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13295 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13296 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13298 2018-01-24 Jeff Law <law@redhat.com>
13301 * i386.c (get_probe_interval): Move to earlier point.
13302 (ix86_compute_frame_layout): If -fstack-clash-protection and
13303 the frame is larger than the probe interval, then use pushes
13304 to save registers rather than reg->mem moves.
13305 (ix86_expand_prologue): Remove conditional for int_registers_saved
13308 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13311 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13312 min/max for never referenced object.
13314 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13316 PR middle-end/83977
13317 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13319 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13320 attributes from DECL_ATTRIBUTES (decl) without affecting
13321 DECL_ATTRIBUTES (current_function_decl).
13322 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13323 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13325 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13327 PR tree-optimization/83979
13328 * fold-const.c (fold_comparison): Use constant_boolean_node
13329 instead of boolean_{true,false}_node.
13331 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13333 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13336 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13338 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13339 Simplify the clause that sets the length attribute.
13340 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13341 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13342 clause that sets the length attribute.
13343 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13345 2018-01-24 Tom de Vries <tom@codesourcery.com>
13348 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13349 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13350 Add strict parameter.
13351 (prevent_branch_around_nothing): Insert dummy insn between branch to
13352 label and label with no ptx insn inbetween.
13353 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13355 2018-01-24 Tom de Vries <tom@codesourcery.com>
13358 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13359 for neutered threads in warp.
13360 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13362 2018-01-24 Richard Biener <rguenther@suse.de>
13364 PR tree-optimization/83176
13365 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13368 2018-01-24 Richard Biener <rguenther@suse.de>
13370 PR tree-optimization/82819
13371 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13372 code generating pluses that are no-ops in the target precision.
13374 2018-01-24 Richard Biener <rguenther@suse.de>
13376 PR middle-end/84000
13377 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13379 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13381 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13382 to merge probabilities.
13383 * predict.c (probably_never_executed): Also mark as cold functions
13384 with global 0 profile and guessed local profile.
13385 * profile-count.c (profile_probability::combine_with_count): New
13387 * profile-count.h (profile_probability::operator*,
13388 profile_probability::operator*=, profile_probability::operator/,
13389 profile_probability::operator/=): Reduce precision to adjusted
13390 and set value to guessed on contradictory divisions.
13391 (profile_probability::combine_with_freq): Remove.
13392 (profile_probability::combine_wiht_count): Declare.
13393 (profile_count::force_nonzero):: Set to adjusted.
13394 (profile_count::probability_in):: Set quality to adjusted.
13395 * tree-ssa-tail-merge.c (replace_block_by): Use
13396 combine_with_count.
13398 2018-01-23 Andrew Waterman <andrew@sifive.com>
13399 Jim Wilson <jimw@sifive.com>
13401 * config/riscv/riscv.c (riscv_stack_boundary): New.
13402 (riscv_option_override): Set riscv_stack_boundary. Handle
13403 riscv_preferred_stack_boundary_arg.
13404 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13405 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13406 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13407 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13408 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13409 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13411 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13414 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13415 of struct ix86_frame.
13416 (ix86_expand_epilogue): Likewise. Add a local variable for
13417 the reg_save_offset field in struct ix86_frame.
13419 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13421 PR tree-optimization/82604
13422 * tree-loop-distribution.c (enum partition_kind): New enum item
13423 PKIND_PARTIAL_MEMSET.
13424 (partition_builtin_p): Support above new enum item.
13425 (generate_code_for_partition): Ditto.
13426 (compute_access_range): Differentiate cases that equality can be
13427 proven at all loops, the innermost loops or no loops.
13428 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13429 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13430 (finalize_partitions, distribute_loop): Don't fuse partition of
13431 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13432 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13433 parloop is enabled.
13435 2018-01-23 Martin Liska <mliska@suse.cz>
13437 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13438 order to ignore the predictor.
13439 (PRED_POLYMORPHIC_CALL): Likewise.
13440 (PRED_RECURSIVE_CALL): Likewise.
13442 2018-01-23 Martin Liska <mliska@suse.cz>
13444 * tree-profile.c (tree_profiling): Print function header to
13445 aware reader which function we are working on.
13446 * value-prof.c (gimple_find_values_to_profile): Do not print
13447 not interesting value histograms.
13449 2018-01-23 Martin Liska <mliska@suse.cz>
13451 * profile-count.h (enum profile_quality): Add
13452 profile_uninitialized as the first value. Do not number values
13453 as they are zero based.
13454 (profile_count::verify): Update sanity check.
13455 (profile_probability::verify): Likewise.
13457 2018-01-23 Nathan Sidwell <nathan@acm.org>
13459 * doc/invoke.texi (ffor-scope): Deprecate.
13461 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13463 PR tree-optimization/83510
13464 * domwalk.c (set_all_edges_as_executable): New function.
13465 (dom_walker::dom_walker): Convert bool param
13466 "skip_unreachable_blocks" to enum reachability. Move setup of
13467 edge flags to set_all_edges_as_executable and only do it when
13468 reachability is REACHABLE_BLOCKS.
13469 * domwalk.h (enum dom_walker::reachability): New enum.
13470 (dom_walker::dom_walker): Convert bool param
13471 "skip_unreachable_blocks" to enum reachability.
13472 (set_all_edges_as_executable): New decl.
13473 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13474 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13476 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13477 but converting true to REACHABLE_BLOCKS.
13478 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13480 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13481 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13482 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13484 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13485 if check_all_array_refs will be called.
13487 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13489 * tree.c (selftest::test_location_wrappers): Add more test
13492 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13494 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13495 (selftest::test_bit_in_range): Likewise.
13497 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13500 * doc/sourcebuild.texi (vect_float): Say that the selector
13501 only describes the situation when -funsafe-math-optimizations is on.
13502 (vect_float_strict): Document.
13504 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13506 PR tree-optimization/83965
13507 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13508 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13509 instead of checking only for a reduction.
13510 (vect_recog_widen_sum_pattern): Likewise.
13512 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13514 * predict.c (probably_never_executed): Only use precise profile info.
13515 (compute_function_frequency): Skip after inlining hack since we now
13516 have quality checking.
13518 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13520 * profile-count.h (profile_probability::very_unlikely,
13521 profile_probability::unlikely, profile_probability::even): Set
13522 precision to guessed.
13524 2018-01-23 Richard Biener <rguenther@suse.de>
13526 PR tree-optimization/83963
13527 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13528 Properly terminate dominator walk when crossing the exit edge not
13529 when visiting its source block.
13531 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13534 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13535 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13537 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13539 PR tree-optimization/83957
13540 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13541 semicolon after for body surrounded by braces.
13543 PR tree-optimization/83081
13544 * profile-count.h (profile_probability::split): New method.
13545 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13546 Use profile_probability::split.
13547 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13548 when splitting a single conditional jump into 2.
13550 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13552 PR tree-optimization/69452
13553 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13556 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13558 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13559 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13560 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13562 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13564 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13566 * config/rl78/rl78.md (movdi): New define_expand.
13567 * config/rl78/rl78.c (rl78_split_movdi): New function.
13569 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13572 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13574 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13575 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13576 128-bit to produce an UNSPEC move to get the double word with the
13577 signbit and then a shift directly to do signbit.
13578 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13579 implementation with a new version that just does either a direct
13580 move or a regular move. Move memory interface to separate insns.
13581 Move insns so they are next to the expander.
13582 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13583 with signbit move. Split big and little endian case.
13584 (signbit<mode>2_dm_mem_le): Likewise.
13585 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13586 (signbit<mode>2_dm2): Likewise.
13588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13590 * config/rl78/rl78.md (anddi3): New define_expand.
13592 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13594 * config/rl78/rl78.md (umindi3): New define_expand.
13596 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13598 * config/rl78/rl78.md (smindi3): New define_expand.
13600 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13602 * config/rl78/rl78.md (smaxdi3): New define_expand.
13604 2018-01-22 Carl Love <cel@us.ibm.com>
13606 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13607 LVX_V1TI): Add macro expansion.
13608 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13609 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13610 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13611 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13612 Change check to determine if the instruction is a byte reversing
13613 entry. Fix typo in comment.
13614 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13615 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13616 Add def_builtin calls for new builtins.
13617 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13618 Add define_insn expansion.
13620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13622 * config/rl78/rl78.md (umaxdi3): New define_expand.
13624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13626 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13627 for non-QImode registers.
13629 2018-01-22 Richard Biener <rguenther@suse.de>
13631 PR tree-optimization/83963
13632 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13633 including the loop exit block.
13634 (scop_detection::merge_sese): Likewise.
13635 (scop_detection::add_scop): Do it here instead.
13637 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13639 * doc/sourcebuild.texi (arm_softfloat): Document.
13641 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13644 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13645 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13646 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13648 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13649 David Edelsohn <dje.gcc@gmail.com>
13652 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13653 Change "crset eq" to "crset 2".
13654 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13655 (*call_indirect_aix<mode>_nospec): Likewise.
13656 (*call_value_indirect_aix<mode>_nospec): Likewise.
13657 (*call_indirect_elfv2<mode>_nospec): Likewise.
13658 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13659 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13660 change assembly output from . to $.
13661 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13662 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13663 (*tablejump<mode>_internal1_nospec): Likewise.
13665 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13668 * config/sh/sh_optimize_sett_clrt.cc:
13669 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13671 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13673 PR tree-optimization/83940
13674 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13675 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13676 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13677 use it to pass back the definition type.
13678 (vect_check_store_rhs): Likewise rhs_dt_out.
13679 (vect_build_gather_load_calls): Add a mask_dt argument and use
13680 it instead of a call to vect_is_simple_use.
13681 (vectorizable_store): Update calls to vect_check_load_store_mask
13682 and vect_check_store_rhs. Use the dt returned by the latter instead
13683 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13684 instead of calls to vect_is_simple_use. Pass the scalar rather
13685 than the vector operand to vect_is_simple_use when handling
13686 second and subsequent copies of an rhs value.
13687 (vectorizable_load): Update calls to vect_check_load_store_mask
13688 and vect_build_gather_load_calls. Use the cached mask_dt and
13689 gs_info.offset_dt instead of calls to vect_is_simple_use.
13691 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13693 PR middle-end/83945
13694 * tree-emutls.c: Include gimplify.h.
13695 (lower_emutls_2): New function.
13696 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13697 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13698 it before further processing.
13701 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13702 UINTVAL (trueop1) instead of INTVAL (op1).
13704 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13708 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13709 INCOMING_FRAME_SP_OFFSET if not defined.
13710 (scan_trace): Add ENTRY argument. If true and
13711 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13712 emit a note to adjust the CFA offset.
13713 (create_cfi_notes): Adjust scan_trace callers.
13714 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13715 INCOMING_FRAME_SP_OFFSET in the CIE.
13716 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13717 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13719 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13720 * doc/tm.texi: Regenerated.
13722 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13724 PR rtl-optimization/83147
13725 * lra-constraints.c (remove_inheritance_pseudos): Use
13726 lra_substitute_pseudo_within_insn.
13728 2018-01-19 Tom de Vries <tom@codesourcery.com>
13729 Cesar Philippidis <cesar@codesourcery.com>
13732 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13734 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13737 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13738 spaces for function labels.
13740 2018-01-19 Martin Liska <mliska@suse.cz>
13742 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13743 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13744 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13745 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13746 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13747 (PRED_CONST_RETURN): Change from 69 to 65.
13748 (PRED_NULL_RETURN): Change from 91 to 71.
13749 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13750 (PRED_LOOP_GUARD): Change from 66 to 73.
13752 2018-01-19 Martin Liska <mliska@suse.cz>
13754 * predict.c (predict_insn_def): Add new assert.
13755 (struct branch_predictor): Change type to signed integer.
13756 (test_prediction_value_range): Amend test to cover
13757 PROB_UNINITIALIZED.
13758 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13759 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13760 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13761 (PRED_LOOP_IV_COMPARE): Likewise.
13762 * predict.h (PROB_UNINITIALIZED): Define new constant.
13764 2018-01-19 Martin Liska <mliska@suse.cz>
13766 * predict.c (dump_prediction): Add new format for
13767 analyze_brprob.py script which is enabled with -details
13769 * profile-count.h (precise_p): New function.
13771 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13773 PR tree-optimization/83922
13774 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13775 there are no statements that need masking.
13776 (vect_active_double_reduction_p): New function.
13777 (vect_analyze_loop_operations): Use it when handling phis that
13778 are not in the loop header.
13780 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13782 PR tree-optimization/83914
13783 * tree-vect-loop.c (vectorizable_induction): Don't convert
13784 init_expr or apply the peeling adjustment for inductions
13785 that are nested within the vectorized loop.
13787 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13789 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13792 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13796 * function.h (gimplify_parameters): Add gimple_seq * argument.
13797 * function.c: Include gimple.h and options.h.
13798 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13799 for the added local temporaries if needed.
13800 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13801 if there are any parameter cleanups, wrap whole body into a
13802 try/finally with the cleanups.
13804 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13807 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13808 Use GET_MODE_CLASS for scalar floating point.
13810 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13814 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13815 Fix call of call_cgraph_insertion_hooks.
13817 2018-01-18 Martin Sebor <msebor@redhat.com>
13819 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13821 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13824 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13827 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13830 * common.opt: (-ffile-prefix-map): New option.
13831 * opts.c (common_handle_option): Defer it.
13832 * opts-global.c (handle_common_deferred_options): Handle it.
13833 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13834 * file-prefix-map.h: New file.
13835 (remap_debug_filename, add_debug_prefix_map): ...here.
13836 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13837 * final.c (debug_prefix_map, add_debug_prefix_map
13838 remap_debug_filename): Move to...
13839 * file-prefix-map.c: New file.
13840 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13841 generalize, get rid of alloca(), use strrchr() instead of strchr().
13842 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13843 Implement in terms of add_prefix_map().
13844 (remap_macro_filename, remap_debug_filename): Implement in term of
13846 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13847 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13848 * dbxout.c: Include file-prefix-map.h.
13849 * varasm.c: Likewise.
13850 * vmsdbgout.c: Likewise.
13851 * xcoffout.c: Likewise.
13852 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13853 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13854 * doc/invoke.texi (-ffile-prefix-map): Document.
13855 (-fdebug-prefix-map): Update description.
13857 2018-01-18 Martin Liska <mliska@suse.cz>
13859 * config/i386/i386.c (indirect_thunk_name): Document that also
13861 (output_indirect_thunk): Document why both instructions
13862 (pause and lfence) are generated.
13864 2018-01-18 Richard Biener <rguenther@suse.de>
13866 PR tree-optimization/83887
13867 * graphite-scop-detection.c
13868 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13869 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13870 (scop_detection::merge_sese): Re-implement with a flood-fill
13871 algorithm that properly finds a SESE region if it exists.
13873 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13876 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13877 pointer_diff optimizations use view_convert instead of convert.
13879 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13881 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13882 Generate different code for -mno-speculate-indirect-jumps.
13883 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13884 (*call_indirect_aix<mode>): Disable for
13885 -mno-speculate-indirect-jumps.
13886 (*call_indirect_aix<mode>_nospec): New define_insn.
13887 (*call_value_indirect_aix<mode>): Disable for
13888 -mno-speculate-indirect-jumps.
13889 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13890 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13891 -mno-speculate-indirect-jumps.
13892 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13894 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13896 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13897 long double type, set the flags for noting the default long double
13898 type, even if we don't pass or return a long double type.
13900 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13903 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13906 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13908 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13909 support for merge[hl].
13910 (fold_mergehl_helper): New helper function.
13911 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13912 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13913 (altivec_vmrglw_direct): Add xxmrglw insn.
13915 2018-01-17 Andrew Waterman <andrew@sifive.com>
13917 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13918 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13920 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13923 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13924 call the lto_location_cache before reading the
13925 DECL_SOURCE_LOCATION of the types.
13927 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13928 Richard Sandiford <richard.sandiford@linaro.org>
13930 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13931 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13932 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13933 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13934 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13936 * config/aarch64/constraints.md (aarch64_movti_operand):
13938 * config/aarch64/predicates.md (Uti): Add new constraint.
13940 2018-01-17 Carl Love <cel@us.ibm.com>
13942 * config/rs6000/vsx.md (define_expand xl_len_r,
13943 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13944 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13946 (define_expand, define_insn): Move the shift left from the
13947 define_insn to the define_expand for lxvl and stxvl instructions.
13948 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13949 and XL_LEN_R definitions to PURE.
13951 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13953 * config/i386/i386.c (indirect_thunk_name): Declare regno
13954 as unsigned int. Compare regno with INVALID_REGNUM.
13955 (output_indirect_thunk): Ditto.
13956 (output_indirect_thunk_function): Ditto.
13957 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13958 in the call to output_indirect_thunk_function.
13960 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13962 PR middle-end/83884
13963 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13964 rather than the size of inner_type to determine the stack slot size
13965 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13967 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13970 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13973 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13975 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13976 endian Linux systems to optionally enable multilibs for selecting
13977 the long double type if the user configured an explicit type.
13978 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13979 have no long double multilibs if not defined.
13980 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13981 warn if the user used -mabi={ieee,ibm}longdouble and we built
13982 multilibs for long double.
13983 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13984 appropriate multilib option.
13985 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13987 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13988 for building long double multilibs.
13989 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13991 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13993 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13996 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13998 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14001 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14004 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14007 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14009 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14010 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14012 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14014 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14015 different rtl trees depending on TARGET_64BIT.
14016 (rs6000_gen_lvx): Likewise.
14018 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14020 * config/visium/visium.md (nop): Tweak comment.
14021 (hazard_nop): Likewise.
14023 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14025 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14026 -mspeculate-indirect-jumps.
14027 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14028 for -mno-speculate-indirect-jumps.
14029 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14030 (*call_value_indirect_elfv2<mode>): Disable for
14031 -mno-speculate-indirect-jumps.
14032 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14033 (indirect_jump): Emit different RTL for
14034 -mno-speculate-indirect-jumps.
14035 (*indirect_jump<mode>): Disable for
14036 -mno-speculate-indirect-jumps.
14037 (*indirect_jump<mode>_nospec): New define_insn.
14038 (tablejump): Emit different RTL for
14039 -mno-speculate-indirect-jumps.
14040 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14041 (tablejumpsi_nospec): New define_expand.
14042 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14043 (tablejumpdi_nospec): New define_expand.
14044 (*tablejump<mode>_internal1): Disable for
14045 -mno-speculate-indirect-jumps.
14046 (*tablejump<mode>_internal1_nospec): New define_insn.
14047 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14050 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14052 * caller-save.c (insert_save): Drop unnecessary parameter. All
14055 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14056 Richard Biener <rguenth@suse.de>
14059 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14060 return early, inline manually is_gimple_sizepos. Make sure if we
14061 call gimplify_expr we don't end up with a gimple constant.
14062 * tree.c (variably_modified_type_p): Don't return true for
14063 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14064 * gimplify.h (is_gimple_sizepos): Remove.
14066 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14068 PR tree-optimization/83857
14069 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14070 vectorizable_live_operation for pure SLP statements.
14071 (vectorizable_live_operation): Handle PHIs.
14073 2018-01-16 Richard Biener <rguenther@suse.de>
14075 PR tree-optimization/83867
14076 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14077 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14079 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14082 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14083 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14084 If off is not INTEGER_CST, issue a may not be aligned warning
14085 rather than isn't aligned. Use isn%'t rather than isn't.
14086 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14088 <case MULT_EXPR>: Improve the case when bottom and one of the
14089 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14090 operand, in that case check if the other operand is multiple of
14091 bottom divided by the INTEGER_CST operand.
14093 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14096 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14097 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14098 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14099 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14100 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14101 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14102 (pa_function_arg_size): New function.
14104 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14106 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14107 in a separate statement.
14109 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14111 PR tree-optimization/83847
14112 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14113 group gathers and scatters.
14115 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14117 PR rtl-optimization/86620
14118 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14120 PR rtl-optimization/83213
14121 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14122 to last if both are JUMP_INSNs.
14124 PR tree-optimization/83843
14125 * gimple-ssa-store-merging.c
14126 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14127 store_immediate_info for bswap/nop orig_stores.
14129 2018-01-15 Andrew Waterman <andrew@sifive.com>
14131 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14133 <UDIV>: Increase cost if !TARGET_DIV.
14135 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14137 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14138 (define_attr "cr_logical_3op"): New.
14139 (cceq_ior_compare): Adjust.
14140 (cceq_ior_compare_complement): Adjust.
14141 (*cceq_rev_compare): Adjust.
14142 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14143 (is_cracked_insn): Adjust.
14144 (insn_must_be_first_in_group): Adjust.
14145 * config/rs6000/40x.md: Adjust.
14146 * config/rs6000/440.md: Adjust.
14147 * config/rs6000/476.md: Adjust.
14148 * config/rs6000/601.md: Adjust.
14149 * config/rs6000/603.md: Adjust.
14150 * config/rs6000/6xx.md: Adjust.
14151 * config/rs6000/7450.md: Adjust.
14152 * config/rs6000/7xx.md: Adjust.
14153 * config/rs6000/8540.md: Adjust.
14154 * config/rs6000/cell.md: Adjust.
14155 * config/rs6000/e300c2c3.md: Adjust.
14156 * config/rs6000/e500mc.md: Adjust.
14157 * config/rs6000/e500mc64.md: Adjust.
14158 * config/rs6000/e5500.md: Adjust.
14159 * config/rs6000/e6500.md: Adjust.
14160 * config/rs6000/mpc.md: Adjust.
14161 * config/rs6000/power4.md: Adjust.
14162 * config/rs6000/power5.md: Adjust.
14163 * config/rs6000/power6.md: Adjust.
14164 * config/rs6000/power7.md: Adjust.
14165 * config/rs6000/power8.md: Adjust.
14166 * config/rs6000/power9.md: Adjust.
14167 * config/rs6000/rs64.md: Adjust.
14168 * config/rs6000/titan.md: Adjust.
14170 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14172 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14173 ix86_indirect_branch_register logic.
14175 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14177 * config/i386/constraints.md (Bs): Update
14178 ix86_indirect_branch_register check. Don't check
14179 ix86_indirect_branch_register with GOT_memory_operand.
14181 * config/i386/predicates.md (GOT_memory_operand): Don't check
14182 ix86_indirect_branch_register here.
14183 (GOT32_symbol_operand): Likewise.
14185 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14187 * config/i386/predicates.md (constant_call_address_operand):
14188 Rewrite ix86_indirect_branch_register logic.
14189 (sibcall_insn_operand): Likewise.
14191 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14193 * config/i386/constraints.md (Bs): Replace
14194 ix86_indirect_branch_thunk_register with
14195 ix86_indirect_branch_register.
14197 * config/i386/i386.md (indirect_jump): Likewise.
14198 (tablejump): Likewise.
14199 (*sibcall_memory): Likewise.
14200 (*sibcall_value_memory): Likewise.
14201 Peepholes of indirect call and jump via memory: Likewise.
14202 * config/i386/i386.opt: Likewise.
14203 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14204 (GOT_memory_operand): Likewise.
14205 (call_insn_operand): Likewise.
14206 (sibcall_insn_operand): Likewise.
14207 (GOT32_symbol_operand): Likewise.
14209 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14211 PR middle-end/83837
14212 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14213 type rather than type addr's type points to.
14214 (expand_omp_atomic_mutex): Likewise.
14215 (expand_omp_atomic): Likewise.
14217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14220 * config/i386/i386.c (output_indirect_thunk_function): Use
14221 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14222 for __x86_return_thunk.
14224 2018-01-15 Richard Biener <rguenther@suse.de>
14226 PR middle-end/83850
14227 * expmed.c (extract_bit_field_1): Fix typo.
14229 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14232 * config/arm/iterators.md (VF): New mode iterator.
14233 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14234 Remove integer-related logic from pattern.
14235 (neon_vabd<mode>_3): Likewise.
14237 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14239 PR middle-end/82694
14240 * common.opt (fstrict-overflow): No longer an alias.
14241 (fwrapv-pointer): New option.
14242 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14243 also for pointer types based on flag_wrapv_pointer.
14244 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14245 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14246 opts->x_flag_wrapv got set.
14247 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14248 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14249 POINTER_TYPE_OVERFLOW_UNDEFINED.
14250 * match.pd: Likewise in address comparison pattern.
14251 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14253 2018-01-15 Richard Biener <rguenther@suse.de>
14256 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14257 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14258 Reset type names to their identifier if their TYPE_DECL doesn't
14259 have linkage (and thus is used for ODR and devirt).
14260 (save_debug_info_for_decl): Remove.
14261 (save_debug_info_for_type): Likewise.
14262 (add_tree_to_fld_list): Adjust.
14263 * tree-pretty-print.c (dump_generic_node): Make dumping of
14264 type names more robust.
14266 2018-01-15 Richard Biener <rguenther@suse.de>
14268 * BASE-VER: Bump to 8.0.1.
14270 2018-01-14 Martin Sebor <msebor@redhat.com>
14273 * builtins.c (check_access): Avoid warning when the no-warning bit
14276 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14278 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14279 * ira-color (allocno_hard_regs_compare): Likewise.
14281 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14284 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14285 Use .pushsection/.popsection.
14287 2018-01-14 Martin Sebor <msebor@redhat.com>
14290 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14292 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14294 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14295 entry from extra_headers.
14296 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14297 extra_headers, make the list bitwise identical to the i?86-*-* one.
14299 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14301 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14302 -mcmodel=large with -mindirect-branch=thunk,
14303 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14304 -mfunction-return=thunk-extern.
14305 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14306 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14307 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14309 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14311 * config/i386/i386.c (print_reg): Print the name of the full
14312 integer register without '%'.
14313 (ix86_print_operand): Handle 'V'.
14314 * doc/extend.texi: Document 'V' modifier.
14316 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14318 * config/i386/constraints.md (Bs): Disallow memory operand for
14319 -mindirect-branch-register.
14321 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14322 (GOT_memory_operand): Likewise.
14323 (call_insn_operand): Likewise.
14324 (sibcall_insn_operand): Likewise.
14325 (GOT32_symbol_operand): Likewise.
14326 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14327 for -mindirect-branch-register.
14328 (tablejump): Likewise.
14329 (*sibcall_memory): Likewise.
14330 (*sibcall_value_memory): Likewise.
14331 Disallow peepholes of indirect call and jump via memory for
14332 -mindirect-branch-register.
14333 (*call_pop): Replace m with Bw.
14334 (*call_value_pop): Likewise.
14335 (*sibcall_pop_memory): Replace m with Bs.
14336 * config/i386/i386.opt (mindirect-branch-register): New option.
14337 * doc/invoke.texi: Document -mindirect-branch-register option.
14339 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14341 * config/i386/i386-protos.h (ix86_output_function_return): New.
14342 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14343 set function_return_type.
14344 (indirect_thunk_name): Add ret_p to indicate thunk for function
14346 (output_indirect_thunk_function): Pass false to
14347 indirect_thunk_name.
14348 (ix86_output_indirect_branch_via_reg): Likewise.
14349 (ix86_output_indirect_branch_via_push): Likewise.
14350 (output_indirect_thunk_function): Create alias for function
14351 return thunk if regno < 0.
14352 (ix86_output_function_return): New function.
14353 (ix86_handle_fndecl_attribute): Handle function_return.
14354 (ix86_attribute_table): Add function_return.
14355 * config/i386/i386.h (machine_function): Add
14356 function_return_type.
14357 * config/i386/i386.md (simple_return_internal): Use
14358 ix86_output_function_return.
14359 (simple_return_internal_long): Likewise.
14360 * config/i386/i386.opt (mfunction-return=): New option.
14361 (indirect_branch): Mention -mfunction-return=.
14362 * doc/extend.texi: Document function_return function attribute.
14363 * doc/invoke.texi: Document -mfunction-return= option.
14365 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14367 * config/i386/i386-opts.h (indirect_branch): New.
14368 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14369 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14370 with local indirect jump when converting indirect call and jump.
14371 (ix86_set_indirect_branch_type): New.
14372 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14373 (indirectlabelno): New.
14374 (indirect_thunk_needed): Likewise.
14375 (indirect_thunk_bnd_needed): Likewise.
14376 (indirect_thunks_used): Likewise.
14377 (indirect_thunks_bnd_used): Likewise.
14378 (INDIRECT_LABEL): Likewise.
14379 (indirect_thunk_name): Likewise.
14380 (output_indirect_thunk): Likewise.
14381 (output_indirect_thunk_function): Likewise.
14382 (ix86_output_indirect_branch_via_reg): Likewise.
14383 (ix86_output_indirect_branch_via_push): Likewise.
14384 (ix86_output_indirect_branch): Likewise.
14385 (ix86_output_indirect_jmp): Likewise.
14386 (ix86_code_end): Call output_indirect_thunk_function if needed.
14387 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14389 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14390 (ix86_attribute_table): Add indirect_branch.
14391 * config/i386/i386.h (machine_function): Add indirect_branch_type
14392 and has_local_indirect_jump.
14393 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14395 (tablejump): Likewise.
14396 (*indirect_jump): Use ix86_output_indirect_jmp.
14397 (*tablejump_1): Likewise.
14398 (simple_return_indirect_internal): Likewise.
14399 * config/i386/i386.opt (mindirect-branch=): New option.
14400 (indirect_branch): New.
14403 (thunk-inline): Likewise.
14404 (thunk-extern): Likewise.
14405 * doc/extend.texi: Document indirect_branch function attribute.
14406 * doc/invoke.texi: Document -mindirect-branch= option.
14408 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14411 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14413 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14415 * ipa-inline.c (want_inline_small_function_p): Return false if
14416 inlining has already failed with CIF_FINAL_ERROR.
14417 (update_caller_keys): Call want_inline_small_function_p before
14419 (update_callee_keys): Likewise.
14421 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14423 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14425 (rs6000_quadword_masked_address_p): Likewise.
14426 (quad_aligned_load_p): Likewise.
14427 (quad_aligned_store_p): Likewise.
14428 (const_load_sequence_p): Add comment to describe the outer-most loop.
14429 (mimic_memory_attributes_and_flags): New function.
14430 (rs6000_gen_stvx): Likewise.
14431 (replace_swapped_aligned_store): Likewise.
14432 (rs6000_gen_lvx): Likewise.
14433 (replace_swapped_aligned_load): Likewise.
14434 (replace_swapped_load_constant): Capitalize argument name in
14435 comment describing this function.
14436 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14437 and stores that access quad-word aligned addresses and replace
14438 with stvx or lvx instructions when appropriate.
14439 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14440 New function prototype.
14441 (rs6000_quadword_masked_address_p): Likewise.
14442 (rs6000_gen_lvx): Likewise.
14443 (rs6000_gen_stvx): Likewise.
14444 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14445 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14446 when memory address is aligned.
14447 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14448 this split to select lvx instruction when memory address is aligned.
14449 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14450 instruction when memory address is aligned.
14451 (*vsx_le_perm_load_v16qi): Likewise.
14452 (four unnamed splitters): Modify to select the stvx instruction
14453 when memory is aligned.
14455 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14457 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14458 which appears in the queue multiple times.
14460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14461 Alan Hayward <alan.hayward@arm.com>
14462 David Sherwood <david.sherwood@arm.com>
14464 * tree-vectorizer.h (vec_lower_bound): New structure.
14465 (_loop_vec_info): Add check_nonzero and lower_bounds.
14466 (LOOP_VINFO_CHECK_NONZERO): New macro.
14467 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14468 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14469 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14470 fields. Make seg_len the distance travelled, not including the
14472 (dr_direction_indicator): Declare.
14473 (dr_zero_step_indicator): Likewise.
14474 (dr_known_forward_stride_p): Likewise.
14475 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14477 (runtime_alias_check_p): Allow runtime alias checks with
14479 (operator ==): Compare access_size and align.
14480 (prune_runtime_alias_test_list): Rework for new distinction between
14481 the access_size and seg_len.
14482 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14484 (get_segment_min_max): New function.
14485 (create_intersect_range_checks): Use it.
14486 (dr_step_indicator): New function.
14487 (dr_direction_indicator): Likewise.
14488 (dr_zero_step_indicator): Likewise.
14489 (dr_known_forward_stride_p): Likewise.
14490 * tree-loop-distribution.c (data_ref_segment_size): Return
14491 DR_STEP * (niters - 1).
14492 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14494 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14495 (vect_preserves_scalar_order_p): New function, split out from...
14496 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14497 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14498 (vect_vfa_access_size): New function.
14499 (vect_vfa_align): Likewise.
14500 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14501 (dump_lower_bound): New function.
14502 (vect_check_lower_bound): Likewise.
14503 (vect_small_gap_p): Likewise.
14504 (vectorizable_with_step_bound_p): Likewise.
14505 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14506 depencies if the vectorization factor is 1. Convert the checks
14507 for nonzero steps into checks on the bounds of DR_STEP. Try using
14508 a bunds check for variable steps if the minimum required step is
14509 relatively small. Update calls to the dr_with_seg_len
14510 constructor and to vect_compile_time_alias.
14511 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14513 (vect_loop_versioning): Call it.
14514 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14516 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14519 Alan Hayward <alan.hayward@arm.com>
14520 David Sherwood <david.sherwood@arm.com>
14522 * doc/sourcebuild.texi (vect_scatter_store): Document.
14523 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14525 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14527 * genopinit.c (main): Add supports_vec_scatter_store and
14528 supports_vec_scatter_store_cached to target_optabs.
14529 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14530 IFN_MASK_SCATTER_STORE.
14531 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14533 * internal-fn.h (internal_store_fn_p): Declare.
14534 (internal_fn_stored_value_index): Likewise.
14535 * internal-fn.c (scatter_store_direct): New macro.
14536 (expand_scatter_store_optab_fn): New function.
14537 (direct_scatter_store_optab_supported_p): New macro.
14538 (internal_store_fn_p): New function.
14539 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14540 IFN_MASK_SCATTER_STORE.
14541 (internal_fn_mask_index): Likewise.
14542 (internal_fn_stored_value_index): New function.
14543 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14544 for scatter stores.
14545 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14546 * optabs-query.c (supports_vec_scatter_store_p): New function.
14547 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14548 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14549 true for scatter stores.
14550 (vect_gather_scatter_fn_p): Handle scatter stores too.
14551 (vect_check_gather_scatter): Consider using scatter stores if
14552 supports_vec_scatter_store_p.
14553 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14554 scatter stores too.
14555 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14556 internal_fn_stored_value_index.
14557 (check_load_store_masking): Handle scatter stores too.
14558 (vect_get_store_rhs): Make public.
14559 (vectorizable_call): Use internal_store_fn_p.
14560 (vectorizable_store): Handle scatter store internal functions.
14561 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14562 when deciding whether the end of the group has been reached.
14563 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14564 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14565 (mask_scatter_store<mode>): New insns.
14567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14568 Alan Hayward <alan.hayward@arm.com>
14569 David Sherwood <david.sherwood@arm.com>
14571 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14572 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14573 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14575 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14576 Use vect_truncate_gather_scatter_offset if we can't treat the
14577 operation as a normal gather load or scatter store.
14578 (get_group_load_store_type): Take the gather_scatter_info
14579 as argument. Try using a gather load or scatter store for
14580 single-element groups.
14581 (get_load_store_type): Update calls to get_group_load_store_type
14582 and vect_use_strided_gather_scatters_p.
14584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14585 Alan Hayward <alan.hayward@arm.com>
14586 David Sherwood <david.sherwood@arm.com>
14588 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14589 optional tree argument.
14590 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14592 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14593 but continue to use the current value as a fallback.
14594 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14595 to compare the updates.
14596 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14597 (get_load_store_type): Use it when handling a strided access.
14598 (vect_get_strided_load_store_ops): New function.
14599 (vect_get_data_ptr_increment): Likewise.
14600 (vectorizable_load): Handle strided gather loads. Always pass
14601 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14604 Alan Hayward <alan.hayward@arm.com>
14605 David Sherwood <david.sherwood@arm.com>
14607 * doc/md.texi (gather_load@var{m}): Document.
14608 (mask_gather_load@var{m}): Likewise.
14609 * genopinit.c (main): Add supports_vec_gather_load and
14610 supports_vec_gather_load_cached to target_optabs.
14611 * optabs-tree.c (init_tree_optimization_optabs): Use
14612 ggc_cleared_alloc to allocate target_optabs.
14613 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14614 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14616 * internal-fn.h (internal_load_fn_p): Declare.
14617 (internal_gather_scatter_fn_p): Likewise.
14618 (internal_fn_mask_index): Likewise.
14619 (internal_gather_scatter_fn_supported_p): Likewise.
14620 * internal-fn.c (gather_load_direct): New macro.
14621 (expand_gather_load_optab_fn): New function.
14622 (direct_gather_load_optab_supported_p): New macro.
14623 (direct_internal_fn_optab): New function.
14624 (internal_load_fn_p): Likewise.
14625 (internal_gather_scatter_fn_p): Likewise.
14626 (internal_fn_mask_index): Likewise.
14627 (internal_gather_scatter_fn_supported_p): Likewise.
14628 * optabs-query.c (supports_at_least_one_mode_p): New function.
14629 (supports_vec_gather_load_p): Likewise.
14630 * optabs-query.h (supports_vec_gather_load_p): Declare.
14631 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14632 and memory_type field.
14633 (NUM_PATTERNS): Bump to 15.
14634 * tree-vect-data-refs.c: Include internal-fn.h.
14635 (vect_gather_scatter_fn_p): New function.
14636 (vect_describe_gather_scatter_call): Likewise.
14637 (vect_check_gather_scatter): Try using internal functions for
14638 gather loads. Recognize existing calls to a gather load function.
14639 (vect_analyze_data_refs): Consider using gather loads if
14640 supports_vec_gather_load_p.
14641 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14642 (vect_get_gather_scatter_offset_type): Likewise.
14643 (vect_convert_mask_for_vectype): Likewise.
14644 (vect_add_conversion_to_patterm): Likewise.
14645 (vect_try_gather_scatter_pattern): Likewise.
14646 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14647 (vect_vect_recog_func_ptrs): Add it.
14648 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14649 internal_fn_mask_index and internal_gather_scatter_fn_p.
14650 (check_load_store_masking): Take the gather_scatter_info as an
14651 argument and handle gather loads.
14652 (vect_get_gather_scatter_ops): New function.
14653 (vectorizable_call): Check internal_load_fn_p.
14654 (vectorizable_load): Likewise. Handle gather load internal
14656 (vectorizable_store): Update call to check_load_store_masking.
14657 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14658 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14659 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14660 (aarch64_gather_scale_operand_d): New predicates.
14661 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14662 (mask_gather_load<mode>): New insns.
14664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14665 Alan Hayward <alan.hayward@arm.com>
14666 David Sherwood <david.sherwood@arm.com>
14668 * optabs.def (fold_left_plus_optab): New optab.
14669 * doc/md.texi (fold_left_plus_@var{m}): Document.
14670 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14671 * internal-fn.c (fold_left_direct): Define.
14672 (expand_fold_left_optab_fn): Likewise.
14673 (direct_fold_left_optab_supported_p): Likewise.
14674 * fold-const-call.c (fold_const_fold_left): New function.
14675 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14676 * tree-parloops.c (valid_reduction_p): New function.
14677 (gather_scalar_reductions): Use it.
14678 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14679 (vect_finish_replace_stmt): Declare.
14680 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14681 (needs_fold_left_reduction_p): New function, split out from...
14682 (vect_is_simple_reduction): ...here. Accept reductions that
14683 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14684 (vect_force_simple_reduction): Also store the reduction type in
14685 the assignment's STMT_VINFO_REDUC_TYPE.
14686 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14687 (merge_with_identity): New function.
14688 (vect_expand_fold_left): Likewise.
14689 (vectorize_fold_left_reduction): Likewise.
14690 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14691 scalar phi in place for it. Check for target support and reject
14692 cases that would reassociate the operation. Defer the transform
14693 phase to vectorize_fold_left_reduction.
14694 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14695 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14696 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14700 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14701 call to ifc_temp_var.
14703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14704 Alan Hayward <alan.hayward@arm.com>
14705 David Sherwood <david.sherwood@arm.com>
14707 * target.def (legitimize_address_displacement): Take the original
14708 offset as a poly_int.
14709 * targhooks.h (default_legitimize_address_displacement): Update
14711 * targhooks.c (default_legitimize_address_displacement): Likewise.
14712 * doc/tm.texi: Regenerate.
14713 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14714 as an argument, moving assert of ad->disp == ad->disp_term to...
14715 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14716 Try calling targetm.legitimize_address_displacement before expanding
14717 the address rather than afterwards, and adjust for the new interface.
14718 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14719 Match the new hook interface. Handle SVE addresses.
14720 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14721 new hook interface.
14723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14725 * Makefile.in (OBJS): Add early-remat.o.
14726 * target.def (select_early_remat_modes): New hook.
14727 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14728 * doc/tm.texi: Regenerate.
14729 * targhooks.h (default_select_early_remat_modes): Declare.
14730 * targhooks.c (default_select_early_remat_modes): New function.
14731 * timevar.def (TV_EARLY_REMAT): New timevar.
14732 * passes.def (pass_early_remat): New pass.
14733 * tree-pass.h (make_pass_early_remat): Declare.
14734 * early-remat.c: New file.
14735 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14737 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14740 Alan Hayward <alan.hayward@arm.com>
14741 David Sherwood <david.sherwood@arm.com>
14743 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14744 vfm1 with a bound_epilog parameter.
14745 (vect_do_peeling): Update calls accordingly, and move the prologue
14746 call earlier in the function. Treat the base bound_epilog as 0 for
14747 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14748 this base when peeling for gaps.
14749 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14750 with fully-masked loops.
14751 (vect_estimate_min_profitable_iters): Handle the single peeled
14752 iteration in that case.
14754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14755 Alan Hayward <alan.hayward@arm.com>
14756 David Sherwood <david.sherwood@arm.com>
14758 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14759 single-element interleaving even if the size is not a power of 2.
14760 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14761 accesses for single-element interleaving if the group size is
14764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14765 Alan Hayward <alan.hayward@arm.com>
14766 David Sherwood <david.sherwood@arm.com>
14768 * doc/md.texi (fold_extract_last_@var{m}): Document.
14769 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14770 * optabs.def (fold_extract_last_optab): New optab.
14771 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14772 * internal-fn.c (fold_extract_direct): New macro.
14773 (expand_fold_extract_optab_fn): Likewise.
14774 (direct_fold_extract_optab_supported_p): Likewise.
14775 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14776 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14777 EXTRACT_LAST_REDUCTION.
14778 (get_initial_def_for_reduction): Do not create an initial vector
14779 for EXTRACT_LAST_REDUCTION reductions.
14780 (vectorizable_reduction): Leave the scalar phi in place for
14781 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14782 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14783 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14784 transform phase to vectorizable_condition.
14785 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14787 (vect_finish_stmt_generation): ...here.
14788 (vect_finish_replace_stmt): New function.
14789 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14790 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14792 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14795 Alan Hayward <alan.hayward@arm.com>
14796 David Sherwood <david.sherwood@arm.com>
14798 * doc/md.texi (extract_last_@var{m}): Document.
14799 * optabs.def (extract_last_optab): New optab.
14800 * internal-fn.def (EXTRACT_LAST): New internal function.
14801 * internal-fn.c (cond_unary_direct): New macro.
14802 (expand_cond_unary_optab_fn): Likewise.
14803 (direct_cond_unary_optab_supported_p): Likewise.
14804 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14805 loops using EXTRACT_LAST.
14806 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14807 (extract_last_<mode>): ...this optab.
14808 (vec_extract<mode><Vel>): Update accordingly.
14810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14811 Alan Hayward <alan.hayward@arm.com>
14812 David Sherwood <david.sherwood@arm.com>
14814 * target.def (empty_mask_is_expensive): New hook.
14815 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14816 * doc/tm.texi: Regenerate.
14817 * targhooks.h (default_empty_mask_is_expensive): Declare.
14818 * targhooks.c (default_empty_mask_is_expensive): New function.
14819 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14820 if the target says that empty masks are expensive.
14821 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14823 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14826 Alan Hayward <alan.hayward@arm.com>
14827 David Sherwood <david.sherwood@arm.com>
14829 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14830 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14831 (vect_use_loop_mask_for_alignment_p): New function.
14832 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14833 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14834 niters_skip argument. Make sure that the first niters_skip elements
14835 of the first iteration are inactive.
14836 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14837 Update call to vect_set_loop_masks_directly.
14838 (get_misalign_in_elems): New function, split out from...
14839 (vect_gen_prolog_loop_niters): ...here.
14840 (vect_update_init_of_dr): Take a code argument that specifies whether
14841 the adjustment should be added or subtracted.
14842 (vect_update_init_of_drs): Likewise.
14843 (vect_prepare_for_masked_peels): New function.
14844 (vect_do_peeling): Skip prologue peeling if we're using a mask
14845 instead. Update call to vect_update_inits_of_drs.
14846 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14848 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14849 alignment. Do not include the number of peeled iterations in
14850 the minimum threshold in that case.
14851 (vectorizable_induction): Adjust the start value down by
14852 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14853 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14854 Take the number of skipped iterations into account when calculating
14856 * tree-vect-stmts.c (vect_gen_while_not): New function.
14858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14859 Alan Hayward <alan.hayward@arm.com>
14860 David Sherwood <david.sherwood@arm.com>
14862 * doc/sourcebuild.texi (vect_fully_masked): Document.
14863 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14864 default value to 0.
14865 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14867 (vect_analyze_loop_2): ...here. Don't check the vectorization
14868 factor against the number of loop iterations if the loop is
14871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14872 Alan Hayward <alan.hayward@arm.com>
14873 David Sherwood <david.sherwood@arm.com>
14875 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14876 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14877 (dump_groups): Update accordingly.
14878 (iv_use::mem_type): New member variable.
14879 (address_p): New function.
14880 (record_use): Add a mem_type argument and initialize the new
14882 (record_group_use): Add a mem_type argument. Use address_p.
14883 Remove obsolete null checks of base_object. Update call to record_use.
14884 (find_interesting_uses_op): Update call to record_group_use.
14885 (find_interesting_uses_cond): Likewise.
14886 (find_interesting_uses_address): Likewise.
14887 (get_mem_type_for_internal_fn): New function.
14888 (find_address_like_use): Likewise.
14889 (find_interesting_uses_stmt): Try find_address_like_use before
14890 calling find_interesting_uses_op.
14891 (addr_offset_valid_p): Use the iv mem_type field as the type
14892 of the addressed memory.
14893 (add_autoinc_candidates): Likewise.
14894 (get_address_cost): Likewise.
14895 (split_small_address_groups_p): Use address_p.
14896 (split_address_groups): Likewise.
14897 (add_iv_candidate_for_use): Likewise.
14898 (autoinc_possible_for_pair): Likewise.
14899 (rewrite_groups): Likewise.
14900 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14901 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14902 (get_alias_ptr_type_for_ptr_address): New function.
14903 (rewrite_use_address): Rewrite address uses in calls that were
14904 identified by find_address_like_use.
14906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14907 Alan Hayward <alan.hayward@arm.com>
14908 David Sherwood <david.sherwood@arm.com>
14910 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14912 * gimple-expr.h (is_gimple_addressable: Likewise.
14913 * gimple-expr.c (is_gimple_address): Likewise.
14914 * internal-fn.c (expand_call_mem_ref): New function.
14915 (expand_mask_load_optab_fn): Use it.
14916 (expand_mask_store_optab_fn): Likewise.
14918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14919 Alan Hayward <alan.hayward@arm.com>
14920 David Sherwood <david.sherwood@arm.com>
14922 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14923 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14924 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14925 (cond_umax@var{mode}): Document.
14926 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14927 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14928 (cond_umin_optab, cond_umax_optab): New optabs.
14929 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14930 (COND_IOR, COND_XOR): New internal functions.
14931 * internal-fn.h (get_conditional_internal_fn): Declare.
14932 * internal-fn.c (cond_binary_direct): New macro.
14933 (expand_cond_binary_optab_fn): Likewise.
14934 (direct_cond_binary_optab_supported_p): Likewise.
14935 (get_conditional_internal_fn): New function.
14936 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14937 Cope with reduction statements that are vectorized as calls rather
14939 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14940 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14941 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14942 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14943 (UNSPEC_COND_EOR): New unspecs.
14944 (optab): Add mappings for them.
14945 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14946 (sve_int_op, sve_fp_op): New int attributes.
14948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14949 Alan Hayward <alan.hayward@arm.com>
14950 David Sherwood <david.sherwood@arm.com>
14952 * optabs.def (while_ult_optab): New optab.
14953 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14954 * internal-fn.def (WHILE_ULT): New internal function.
14955 * internal-fn.h (direct_internal_fn_supported_p): New override
14956 that takes two types as argument.
14957 * internal-fn.c (while_direct): New macro.
14958 (expand_while_optab_fn): New function.
14959 (convert_optab_supported_p): Likewise.
14960 (direct_while_optab_supported_p): New macro.
14961 * wide-int.h (wi::udiv_ceil): New function.
14962 * tree-vectorizer.h (rgroup_masks): New structure.
14963 (vec_loop_masks): New typedef.
14964 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14965 and fully_masked_p.
14966 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14967 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14968 (vect_max_vf): New function.
14969 (slpeel_make_loop_iterate_ntimes): Delete.
14970 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14971 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14972 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14973 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14974 internal-fn.h, stor-layout.h and optabs-query.h.
14975 (vect_set_loop_mask): New function.
14976 (add_preheader_seq): Likewise.
14977 (add_header_seq): Likewise.
14978 (interleave_supported_p): Likewise.
14979 (vect_maybe_permute_loop_masks): Likewise.
14980 (vect_set_loop_masks_directly): Likewise.
14981 (vect_set_loop_condition_masked): Likewise.
14982 (vect_set_loop_condition_unmasked): New function, split out from
14983 slpeel_make_loop_iterate_ntimes.
14984 (slpeel_make_loop_iterate_ntimes): Rename to..
14985 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14986 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14987 (vect_do_peeling): Update call accordingly.
14988 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14990 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14991 mask_compare_type, can_fully_mask_p and fully_masked_p.
14992 (release_vec_loop_masks): New function.
14993 (_loop_vec_info): Use it to free the loop masks.
14994 (can_produce_all_loop_masks_p): New function.
14995 (vect_get_max_nscalars_per_iter): Likewise.
14996 (vect_verify_full_masking): Likewise.
14997 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14998 retries, and free the mask rgroups before retrying. Check loop-wide
14999 reasons for disallowing fully-masked loops. Make the final decision
15000 about whether use a fully-masked loop or not.
15001 (vect_estimate_min_profitable_iters): Do not assume that peeling
15002 for the number of iterations will be needed for fully-masked loops.
15003 (vectorizable_reduction): Disable fully-masked loops.
15004 (vectorizable_live_operation): Likewise.
15005 (vect_halve_mask_nunits): New function.
15006 (vect_double_mask_nunits): Likewise.
15007 (vect_record_loop_mask): Likewise.
15008 (vect_get_loop_mask): Likewise.
15009 (vect_transform_loop): Handle the case in which the final loop
15010 iteration might handle a partial vector. Call vect_set_loop_condition
15011 instead of slpeel_make_loop_iterate_ntimes.
15012 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15013 (check_load_store_masking): New function.
15014 (prepare_load_store_mask): Likewise.
15015 (vectorizable_store): Handle fully-masked loops.
15016 (vectorizable_load): Likewise.
15017 (supportable_widening_operation): Use vect_halve_mask_nunits for
15019 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15020 (vect_gen_while): New function.
15021 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15022 (aarch64_uqdec<mode>): New insn.
15024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15025 Alan Hayward <alan.hayward@arm.com>
15026 David Sherwood <david.sherwood@arm.com>
15028 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15029 (reduc_xor_scal_optab): New optabs.
15030 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15031 (reduc_xor_scal_@var{m}): Document.
15032 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15033 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15034 internal functions.
15035 * fold-const-call.c (fold_const_call): Handle them.
15036 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15037 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15038 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15039 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15040 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15041 (UNSPEC_XORV): New unspecs.
15042 (optab): Add entries for them.
15043 (BITWISEV): New int iterator.
15044 (bit_reduc_op): New int attributes.
15046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15047 Alan Hayward <alan.hayward@arm.com>
15048 David Sherwood <david.sherwood@arm.com>
15050 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15051 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15052 * optabs.def (vec_shl_insert_optab): New optab.
15053 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15054 (duplicate_and_interleave): Likewise.
15055 * tree-vect-loop.c: Include internal-fn.h.
15056 (neutral_op_for_slp_reduction): New function, split out from
15057 get_initial_defs_for_reduction.
15058 (get_initial_def_for_reduction): Handle option 2 for variable-length
15059 vectors by loading the neutral value into a vector and then shifting
15060 the initial value into element 0.
15061 (get_initial_defs_for_reduction): Replace the code argument with
15062 the neutral value calculated by neutral_op_for_slp_reduction.
15063 Use gimple_build_vector for constant-length vectors.
15064 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15065 but the first group_size elements have a neutral value.
15066 Use duplicate_and_interleave otherwise.
15067 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15068 Update call to get_initial_defs_for_reduction. Handle SLP
15069 reductions for variable-length vectors by creating one vector
15070 result for each scalar result, with the elements associated
15071 with other scalar results stubbed out with the neutral value.
15072 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15073 Require IFN_VEC_SHL_INSERT for double reductions on
15074 variable-length vectors, or SLP reductions that have
15075 a neutral value. Require can_duplicate_and_interleave_p
15076 support for variable-length unchained SLP reductions if there
15077 is no neutral value, such as for MIN/MAX reductions. Also require
15078 the number of vector elements to be a multiple of the number of
15079 SLP statements when doing variable-length unchained SLP reductions.
15080 Update call to vect_create_epilog_for_reduction.
15081 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15082 and remove initial values.
15083 (duplicate_and_interleave): Make public.
15084 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15085 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15088 Alan Hayward <alan.hayward@arm.com>
15089 David Sherwood <david.sherwood@arm.com>
15091 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15092 (can_duplicate_and_interleave_p): New function.
15093 (vect_get_and_check_slp_defs): Take the vector of statements
15094 rather than just the current one. Remove excess parentheses.
15095 Restriction rejectinon of vect_constant_def and vect_external_def
15096 for variable-length vectors to boolean types, or types for which
15097 can_duplicate_and_interleave_p is false.
15098 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15099 (duplicate_and_interleave): New function.
15100 (vect_get_constant_vectors): Use gimple_build_vector for
15101 constant-length vectors and suitable variable-length constant
15102 vectors. Use duplicate_and_interleave for other variable-length
15103 vectors. Don't defer the update when inserting new statements.
15105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15106 Alan Hayward <alan.hayward@arm.com>
15107 David Sherwood <david.sherwood@arm.com>
15109 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15110 min_profitable_iters doesn't go negative.
15112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15113 Alan Hayward <alan.hayward@arm.com>
15114 David Sherwood <david.sherwood@arm.com>
15116 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15117 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15118 * optabs.def (vec_mask_load_lanes_optab): New optab.
15119 (vec_mask_store_lanes_optab): Likewise.
15120 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15121 (MASK_STORE_LANES): Likewise.
15122 * internal-fn.c (mask_load_lanes_direct): New macro.
15123 (mask_store_lanes_direct): Likewise.
15124 (expand_mask_load_optab_fn): Handle masked operations.
15125 (expand_mask_load_lanes_optab_fn): New macro.
15126 (expand_mask_store_optab_fn): Handle masked operations.
15127 (expand_mask_store_lanes_optab_fn): New macro.
15128 (direct_mask_load_lanes_optab_supported_p): Likewise.
15129 (direct_mask_store_lanes_optab_supported_p): Likewise.
15130 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15132 (vect_load_lanes_supported): Likewise.
15133 * tree-vect-data-refs.c (strip_conversion): New function.
15134 (can_group_stmts_p): Likewise.
15135 (vect_analyze_data_ref_accesses): Use it instead of checking
15136 for a pair of assignments.
15137 (vect_store_lanes_supported): Take a masked_p parameter.
15138 (vect_load_lanes_supported): Likewise.
15139 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15140 vect_store_lanes_supported and vect_load_lanes_supported.
15141 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15142 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15143 parameter. Don't allow gaps for masked accesses.
15144 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15145 and vect_load_lanes_supported.
15146 (get_load_store_type): Take a masked_p parameter and update
15147 call to get_group_load_store_type.
15148 (vectorizable_store): Update call to get_load_store_type.
15149 Handle IFN_MASK_STORE_LANES.
15150 (vectorizable_load): Update call to get_load_store_type.
15151 Handle IFN_MASK_LOAD_LANES.
15153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15154 Alan Hayward <alan.hayward@arm.com>
15155 David Sherwood <david.sherwood@arm.com>
15157 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15159 * config/aarch64/aarch64-protos.h
15160 (aarch64_sve_struct_memory_operand_p): Declare.
15161 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15162 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15163 (VPRED, vpred): Handle SVE structure modes.
15164 * config/aarch64/constraints.md (Utx): New constraint.
15165 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15166 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15167 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15168 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15169 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15170 structure modes. Split into pieces after RA.
15171 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15172 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15174 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15175 SVE structure modes.
15176 (aarch64_classify_address): Likewise.
15177 (sizetochar): Move earlier in file.
15178 (aarch64_print_operand): Handle SVE register lists.
15179 (aarch64_array_mode): New function.
15180 (aarch64_sve_struct_memory_operand_p): Likewise.
15181 (TARGET_ARRAY_MODE): Redefine.
15183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15184 Alan Hayward <alan.hayward@arm.com>
15185 David Sherwood <david.sherwood@arm.com>
15187 * target.def (array_mode): New target hook.
15188 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15189 * doc/tm.texi: Regenerate.
15190 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15191 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15192 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15193 targetm.array_mode.
15194 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15198 Alan Hayward <alan.hayward@arm.com>
15199 David Sherwood <david.sherwood@arm.com>
15201 * fold-const.c (fold_binary_loc): Check the argument types
15202 rather than the result type when testing for a vector operation.
15204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15206 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15207 * doc/tm.texi: Regenerate.
15209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15210 Alan Hayward <alan.hayward@arm.com>
15211 David Sherwood <david.sherwood@arm.com>
15213 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15214 (sve): Document new AArch64 extension.
15215 * doc/md.texi (w): Extend the description of the AArch64
15216 constraint to include SVE vectors.
15217 (Upl, Upa): Document new AArch64 predicate constraints.
15218 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15220 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15221 (msve-vector-bits=): New option.
15222 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15223 SVE when these are disabled.
15224 (sve): New extension.
15225 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15226 modes. Adjust their number of units based on aarch64_sve_vg.
15227 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15228 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15229 aarch64_addr_query_type.
15230 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15231 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15232 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15233 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15234 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15235 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15236 (aarch64_simd_imm_zero_p): Delete.
15237 (aarch64_check_zero_based_sve_index_immediate): Declare.
15238 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15239 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15240 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15241 (aarch64_sve_float_mul_immediate_p): Likewise.
15242 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15243 rather than an rtx.
15244 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15245 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15246 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15247 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15248 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15249 (aarch64_regmode_natural_size): Likewise.
15250 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15251 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15253 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15254 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15255 for VG and the SVE predicate registers.
15256 (V_ALIASES): Add a "z"-prefixed alias.
15257 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15258 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15259 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15260 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15261 (REG_CLASS_NAMES): Add entries for them.
15262 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15263 and the predicate registers.
15264 (aarch64_sve_vg): Declare.
15265 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15266 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15267 (REGMODE_NATURAL_SIZE): Define.
15268 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15270 * config/aarch64/aarch64.c: Include cfgrtl.h.
15271 (simd_immediate_info): Add a constructor for series vectors,
15272 and an associated step field.
15273 (aarch64_sve_vg): New variable.
15274 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15275 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15276 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15277 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15278 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15279 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15280 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15281 (aarch64_get_mask_mode): New functions.
15282 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15283 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15284 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15285 predicate modes and predicate registers. Explicitly restrict
15286 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15287 to store a vector mode if it is recognized by
15288 aarch64_classify_vector_mode.
15289 (aarch64_regmode_natural_size): New function.
15290 (aarch64_hard_regno_caller_save_mode): Return the original mode
15292 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15293 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15294 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15295 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15297 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15298 does not overlap dest if the function is frame-related. Handle
15300 (aarch64_split_add_offset): New function.
15301 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15302 them aarch64_add_offset.
15303 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15304 and update call to aarch64_sub_sp.
15305 (aarch64_add_cfa_expression): New function.
15306 (aarch64_expand_prologue): Pass extra temporary registers to the
15307 functions above. Handle the case in which we need to emit new
15308 DW_CFA_expressions for registers that were originally saved
15309 relative to the stack pointer, but now have to be expressed
15310 relative to the frame pointer.
15311 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15313 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15314 IP0 and IP1 values for SVE frames.
15315 (aarch64_expand_vec_series): New function.
15316 (aarch64_expand_sve_widened_duplicate): Likewise.
15317 (aarch64_expand_sve_const_vector): Likewise.
15318 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15319 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15320 into the register, rather than emitting a SET directly.
15321 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15322 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15323 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15324 (offset_9bit_signed_scaled_p): New functions.
15325 (aarch64_replicate_bitmask_imm): New function.
15326 (aarch64_bitmask_imm): Use it.
15327 (aarch64_cannot_force_const_mem): Reject expressions involving
15328 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15329 (aarch64_classify_index): Handle SVE indices, by requiring
15330 a plain register index with a scale that matches the element size.
15331 (aarch64_classify_address): Handle SVE addresses. Assert that
15332 the mode of the address is VOIDmode or an integer mode.
15333 Update call to aarch64_classify_symbol.
15334 (aarch64_classify_symbolic_expression): Update call to
15335 aarch64_classify_symbol.
15336 (aarch64_const_vec_all_in_range_p): New function.
15337 (aarch64_print_vector_float_operand): Likewise.
15338 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15339 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15340 and the FP immediates 1.0 and 0.5.
15341 (aarch64_print_address_internal): Handle SVE addresses.
15342 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15343 (aarch64_regno_regclass): Handle predicate registers.
15344 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15346 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15347 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15348 (aarch64_convert_sve_vector_bits): New function.
15349 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15350 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15351 rather than an rtx.
15352 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15353 Handle SVE vector and predicate modes. Accept VL-based constants
15354 that need only one temporary register, and VL offsets that require
15355 no temporary registers.
15356 (aarch64_conditional_register_usage): Mark the predicate registers
15357 as fixed if SVE isn't available.
15358 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15359 Return true for SVE vector and predicate modes.
15360 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15361 rather than an unsigned int. Handle SVE modes.
15362 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15364 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15366 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15367 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15368 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15369 (aarch64_sve_float_mul_immediate_p): New functions.
15370 (aarch64_sve_valid_immediate): New function.
15371 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15372 Explicitly reject structure modes. Check for INDEX constants.
15373 Handle PTRUE and PFALSE constants.
15374 (aarch64_check_zero_based_sve_index_immediate): New function.
15375 (aarch64_simd_imm_zero_p): Delete.
15376 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15377 vector modes. Accept constants in the range of CNT[BHWD].
15378 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15379 ask for an Advanced SIMD mode.
15380 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15381 (aarch64_simd_vector_alignment): Handle SVE predicates.
15382 (aarch64_vectorize_preferred_vector_alignment): New function.
15383 (aarch64_simd_vector_alignment_reachable): Use it instead of
15385 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15386 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15388 (MAX_VECT_LEN): Delete.
15389 (expand_vec_perm_d): Add a vec_flags field.
15390 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15391 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15392 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15394 (aarch64_evpc_rev): Rename to...
15395 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15396 (aarch64_evpc_rev_global): New function.
15397 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15398 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15400 (aarch64_evpc_sve_tbl): New function.
15401 (aarch64_expand_vec_perm_const_1): Update after rename of
15402 aarch64_evpc_rev. Handle SVE permutes too, trying
15403 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15404 than aarch64_evpc_tbl.
15405 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15406 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15407 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15408 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15409 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15410 (aarch64_expand_sve_vcond): New functions.
15411 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15412 of aarch64_vector_mode_p.
15413 (aarch64_dwarf_poly_indeterminate_value): New function.
15414 (aarch64_compute_pressure_classes): Likewise.
15415 (aarch64_can_change_mode_class): Likewise.
15416 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15417 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15418 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15419 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15420 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15421 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15422 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15423 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15425 (Dn, Dl, Dr): Accept const as well as const_vector.
15426 (Dz): Likewise. Compare against CONST0_RTX.
15427 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15428 of "vector" where appropriate.
15429 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15430 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15431 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15432 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15433 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15434 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15435 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15436 (v_int_equiv): Extend to SVE modes.
15437 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15439 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15440 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15441 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15442 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15443 (SVE_COND_FP_CMP): New int iterators.
15444 (perm_hilo): Handle the new unpack unspecs.
15445 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15447 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15448 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15449 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15450 (aarch64_equality_operator, aarch64_constant_vector_operand)
15451 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15452 (aarch64_sve_nonimmediate_operand): Likewise.
15453 (aarch64_sve_general_operand): Likewise.
15454 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15455 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15456 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15457 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15458 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15459 (aarch64_sve_float_arith_immediate): Likewise.
15460 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15461 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15462 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15463 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15464 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15465 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15466 (aarch64_sve_float_arith_operand): Likewise.
15467 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15468 (aarch64_sve_float_mul_operand): Likewise.
15469 (aarch64_sve_vec_perm_operand): Likewise.
15470 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15471 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15472 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15473 as well as const_vector.
15474 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15475 in file. Use CONST0_RTX and CONSTM1_RTX.
15476 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15477 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15478 Use aarch64_simd_imm_zero.
15479 * config/aarch64/aarch64-sve.md: New file.
15480 * config/aarch64/aarch64.md: Include it.
15481 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15482 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15483 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15484 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15485 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15486 (sve): New attribute.
15487 (enabled): Disable instructions with the sve attribute unless
15489 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15490 aarch64_expand_mov_immediate.
15491 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15492 CNT[BHSD] immediates.
15493 (movti): Split CONST_POLY_INT moves into two halves.
15494 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15495 Split additions that need a temporary here if the destination
15496 is the stack pointer.
15497 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15498 (*add<mode>3_poly_1): New instruction.
15499 (set_clobber_cc): New expander.
15501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15503 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15504 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15505 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15506 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15507 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15508 Change innermode from fixed_mode_size to machine_mode.
15509 (simplify_subreg): Update call accordingly. Handle a constant-sized
15510 subreg of a variable-length CONST_VECTOR.
15512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15513 Alan Hayward <alan.hayward@arm.com>
15514 David Sherwood <david.sherwood@arm.com>
15516 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15517 (add_offset_to_base): New function, split out from...
15518 (create_mem_ref): ...here. When handling a scale other than 1,
15519 check first whether the address is valid without the offset.
15520 Add it into the base if so, leaving the index and scale as-is.
15522 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15525 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15526 fold_for_warn before checking if arg2 is INTEGER_CST.
15528 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15530 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15531 (store_multiple_operation): Delete.
15532 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15533 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15534 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15535 guarded by TARGET_STRING.
15536 (rs6000_output_load_multiple): Delete.
15537 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15538 OPTION_MASK_STRING / TARGET_STRING handling.
15539 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15540 (const rs6000_opt_masks) <"string">: Change mask to 0.
15541 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15542 (MASK_STRING): Delete.
15543 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15545 (load_multiple): Delete.
15552 (store_multiple): Delete.
15559 (movmemsi_8reg): Delete.
15560 (corresponding unnamed define_insn): Delete.
15561 (movmemsi_6reg): Delete.
15562 (corresponding unnamed define_insn): Delete.
15563 (movmemsi_4reg): Delete.
15564 (corresponding unnamed define_insn): Delete.
15565 (movmemsi_2reg): Delete.
15566 (corresponding unnamed define_insn): Delete.
15567 (movmemsi_1reg): Delete.
15568 (corresponding unnamed define_insn): Delete.
15569 * config/rs6000/rs6000.opt (mno-string): New.
15570 (mstring): Replace by deprecation warning stub.
15571 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15573 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15575 * regrename.c (regrename_do_replace): If replacing the same
15576 reg multiple times, try to reuse last created gen_raw_REG.
15579 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15580 main to workaround a bug in GDB.
15582 2018-01-12 Tom de Vries <tom@codesourcery.com>
15585 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15587 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15589 PR rtl-optimization/80481
15590 * ira-color.c (get_cap_member): New function.
15591 (allocnos_conflict_by_live_ranges_p): Use it.
15592 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15593 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15595 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15598 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15599 (*saddl_se_1): Ditto.
15600 (*ssubsi_1): Ditto.
15601 (*ssubl_se_1): Ditto.
15603 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15605 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15606 rather than wi::to_widest for DR_INITs.
15607 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15608 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15609 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15611 (vect_analyze_group_access_1): Note that here.
15613 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15615 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15616 polynomial type sizes.
15618 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15620 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15621 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15622 (gimple_add_tmp_var): Likewise.
15624 2018-01-12 Martin Liska <mliska@suse.cz>
15626 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15627 (gimple_alloc_sizes): Likewise.
15628 (dump_gimple_statistics): Use PRIu64 in printf format.
15629 * gimple.h: Change uint64_t to int.
15631 2018-01-12 Martin Liska <mliska@suse.cz>
15633 * tree-core.h: Use uint64_t instead of int.
15634 * tree.c (tree_node_counts): Likewise.
15635 (tree_node_sizes): Likewise.
15636 (dump_tree_statistics): Use PRIu64 in printf format.
15638 2018-01-12 Martin Liska <mliska@suse.cz>
15640 * Makefile.in: As qsort_chk is implemented in vec.c, add
15641 vec.o to linkage of gencfn-macros.
15642 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15643 passing the info to record_node_allocation_statistics.
15644 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15646 * ggc-common.c (struct ggc_usage): Add operator== and use
15647 it in operator< and compare function.
15648 * mem-stats.h (struct mem_usage): Likewise.
15649 * vec.c (struct vec_usage): Remove operator< and compare
15650 function. Can be simply inherited.
15652 2018-01-12 Martin Jambor <mjambor@suse.cz>
15655 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15656 * tree-ssa-math-opts.c: Include domwalk.h.
15657 (convert_mult_to_fma_1): New function.
15658 (fma_transformation_info): New type.
15659 (fma_deferring_state): Likewise.
15660 (cancel_fma_deferring): New function.
15661 (result_of_phi): Likewise.
15662 (last_fma_candidate_feeds_initial_phi): Likewise.
15663 (convert_mult_to_fma): Added deferring logic, split actual
15664 transformation to convert_mult_to_fma_1.
15665 (math_opts_dom_walker): New type.
15666 (math_opts_dom_walker::after_dom_children): New method, body moved
15667 here from pass_optimize_widening_mul::execute, added deferring logic
15669 (pass_optimize_widening_mul::execute): Moved most of code to
15670 math_opts_dom_walker::after_dom_children.
15671 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15672 * config/i386/i386.c (ix86_option_override_internal): Added
15673 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15675 2018-01-12 Richard Biener <rguenther@suse.de>
15678 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15679 inline instance vars.
15681 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15684 * config/rx/rx.c (rx_is_restricted_memory_address):
15685 Handle SUBREG case.
15687 2018-01-12 Richard Biener <rguenther@suse.de>
15689 PR tree-optimization/80846
15690 * target.def (split_reduction): New target hook.
15691 * targhooks.c (default_split_reduction): New function.
15692 * targhooks.h (default_split_reduction): Declare.
15693 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15694 target requests first reduce vectors by combining low and high
15696 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15697 (get_vectype_for_scalar_type_and_size): Export.
15698 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15699 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15700 * doc/tm.texi: Regenerate.
15701 * config/i386/i386.c (ix86_split_reduction): Implement
15702 TARGET_VECTORIZE_SPLIT_REDUCTION.
15704 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15707 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15708 in PIC mode except for TARGET_VXWORKS_RTP.
15709 * config/sparc/sparc.c: Include cfgrtl.h.
15710 (TARGET_INIT_PIC_REG): Define.
15711 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15712 (sparc_pic_register_p): New predicate.
15713 (sparc_legitimate_address_p): Use it.
15714 (sparc_legitimize_pic_address): Likewise.
15715 (sparc_delegitimize_address): Likewise.
15716 (sparc_mode_dependent_address_p): Likewise.
15717 (gen_load_pcrel_sym): Remove 4th parameter.
15718 (load_got_register): Adjust call to above. Remove obsolete stuff.
15719 (sparc_expand_prologue): Do not call load_got_register here.
15720 (sparc_flat_expand_prologue): Likewise.
15721 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15722 (sparc_use_pseudo_pic_reg): New function.
15723 (sparc_init_pic_reg): Likewise.
15724 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15725 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15727 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15729 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15730 Add item for branch_cost.
15732 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15734 PR rtl-optimization/83565
15735 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15736 not extend the result to a larger mode for rotate operations.
15737 (num_sign_bit_copies1): Likewise.
15739 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15742 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15744 Use values-Xc.o for -pedantic.
15745 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15747 2018-01-12 Martin Liska <mliska@suse.cz>
15750 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15752 (possible_polymorphic_call_targets): Use it.
15753 (ipa_devirt): Likewise.
15755 2018-01-12 Martin Liska <mliska@suse.cz>
15757 * profile-count.h (enum profile_quality): Use 0 as invalid
15758 enum value of profile_quality.
15760 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15762 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15763 -mext-string options.
15765 2018-01-12 Richard Biener <rguenther@suse.de>
15767 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15768 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15769 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15771 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15773 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15775 * configure.ac (--with-long-double-format): Add support for the
15776 configuration option to change the default long double format on
15778 * config.gcc (powerpc*-linux*-*): Likewise.
15779 * configure: Regenerate.
15780 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15781 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15782 used without modification.
15784 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15786 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15787 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15788 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15789 MISC_BUILTIN_SPEC_BARRIER.
15790 (rs6000_init_builtins): Likewise.
15791 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15793 (speculation_barrier): New define_insn.
15794 * doc/extend.texi: Document __builtin_speculation_barrier.
15796 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15799 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15800 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15801 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15803 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15804 integral modes instead of "ss" and "sd".
15805 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15806 vectors with 32-bit and 64-bit elements.
15807 (vecdupssescalarmodesuffix): New mode attribute.
15808 (vec_dup<mode>): Use it.
15810 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15813 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15814 frame if argument is passed on stack.
15816 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15819 * ree.c (combine_reaching_defs): Optimize also
15820 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15821 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15823 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15825 PR middle-end/83189
15826 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15828 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15830 PR middle-end/83718
15831 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15832 after they are computed.
15834 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15836 PR tree-optimization/83695
15837 * gimple-loop-linterchange.cc
15838 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15839 reset cached scev information after interchange.
15840 (pass_linterchange::execute): Remove call to scev_reset_htab.
15842 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15844 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15845 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15846 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15847 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15848 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15849 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15850 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15851 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15852 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15853 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15854 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15855 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15856 (V_lane_reg): Likewise.
15857 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15859 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15860 (vfmal_lane_low<mode>_intrinsic,
15861 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15862 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15863 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15864 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15865 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15866 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15868 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15870 * config/arm/arm-cpus.in (fp16fml): New feature.
15871 (ALL_SIMD): Add fp16fml.
15872 (armv8.2-a): Add fp16fml as an option.
15873 (armv8.3-a): Likewise.
15874 (armv8.4-a): Add fp16fml as part of fp16.
15875 * config/arm/arm.h (TARGET_FP16FML): Define.
15876 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15878 * config/arm/arm-modes.def (V2HF): Define.
15879 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15880 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15881 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15882 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15883 vfmsl_low, vfmsl_high): New set of builtins.
15884 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15885 (vfml_op): New code attribute.
15886 (VFMLHALVES): New int iterator.
15887 (VFML, VFMLSEL): New mode attributes.
15888 (V_reg): Define mapping for V2HF.
15889 (V_hi, V_lo): New mode attributes.
15890 (VF_constraint): Likewise.
15891 (vfml_half, vfml_half_selector): New int attributes.
15892 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15894 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15895 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15897 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15898 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15899 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15900 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15902 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15903 Document new effective target and option set.
15905 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15907 * config/arm/arm-cpus.in (armv8_4): New feature.
15908 (ARMv8_4a): New fgroup.
15909 (armv8.4-a): New arch.
15910 * config/arm/arm-tables.opt: Regenerate.
15911 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15912 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15913 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15914 Add matching rules for -march=armv8.4-a and extensions.
15915 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15917 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15920 * config/rx/rx.md (BW): New mode attribute.
15921 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15923 2018-01-11 Richard Biener <rguenther@suse.de>
15925 PR tree-optimization/83435
15926 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15927 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15928 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15930 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15931 Alan Hayward <alan.hayward@arm.com>
15932 David Sherwood <david.sherwood@arm.com>
15934 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15936 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15937 (aarch64_print_address_internal): Use it to check for a zero offset.
15939 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15940 Alan Hayward <alan.hayward@arm.com>
15941 David Sherwood <david.sherwood@arm.com>
15943 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15944 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15945 Return a poly_int64 rather than a HOST_WIDE_INT.
15946 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15947 rather than a HOST_WIDE_INT.
15948 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15949 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15950 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15951 final_offset from HOST_WIDE_INT to poly_int64.
15952 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15953 to_constant when getting the number of units in an Advanced SIMD
15955 (aarch64_builtin_vectorized_function): Check for a constant number
15957 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15959 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15960 attribute instead of GET_MODE_NUNITS.
15961 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15962 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15963 GET_MODE_SIZE for fixed-size registers.
15964 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15965 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15966 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15967 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15968 (aarch64_print_operand, aarch64_print_address_internal)
15969 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15970 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15971 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15972 Handle polynomial GET_MODE_SIZE.
15973 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15974 wider than SImode without modification.
15975 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15976 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15977 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15978 passing and returning SVE modes.
15979 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15980 rather than GEN_INT.
15981 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15982 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15983 (aarch64_allocate_and_probe_stack_space): Likewise.
15984 (aarch64_layout_frame): Cope with polynomial offsets.
15985 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15986 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15987 polynomial offsets.
15988 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15989 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15990 poly_int64 rather than a HOST_WIDE_INT.
15991 (aarch64_get_separate_components, aarch64_process_components)
15992 (aarch64_expand_prologue, aarch64_expand_epilogue)
15993 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15994 (aarch64_anchor_offset): New function, split out from...
15995 (aarch64_legitimize_address): ...here.
15996 (aarch64_builtin_vectorization_cost): Handle polynomial
15997 TYPE_VECTOR_SUBPARTS.
15998 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16000 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16001 number of elements from the PARALLEL rather than the mode.
16002 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16003 rather than GET_MODE_BITSIZE.
16004 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16005 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16006 (aarch64_expand_vec_perm_const_1): Handle polynomial
16007 d->perm.length () and d->perm elements.
16008 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16009 Apply to_constant to d->perm elements.
16010 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16011 polynomial CONST_VECTOR_NUNITS.
16012 (aarch64_move_pointer): Take amount as a poly_int64 rather
16014 (aarch64_progress_pointer): Avoid temporary variable.
16015 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16016 the mode attribute instead of GET_MODE.
16018 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16019 Alan Hayward <alan.hayward@arm.com>
16020 David Sherwood <david.sherwood@arm.com>
16022 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16023 x exists before using it.
16024 (aarch64_add_constant_internal): Rename to...
16025 (aarch64_add_offset_1): ...this. Replace regnum with separate
16026 src and dest rtxes. Handle the case in which they're different,
16027 including when the offset is zero. Replace scratchreg with an rtx.
16028 Use 2 additions if there is no spare register into which we can
16029 move a 16-bit constant.
16030 (aarch64_add_constant): Delete.
16031 (aarch64_add_offset): Replace reg with separate src and dest
16032 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16033 Use aarch64_add_offset_1.
16034 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16035 an rtx rather than an int. Take the delta as a poly_int64
16036 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16037 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16038 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16039 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16040 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16041 and aarch64_add_sp.
16042 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16043 aarch64_add_constant.
16045 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16047 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16048 Use scalar_float_mode.
16050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16052 * config/aarch64/aarch64-simd.md
16053 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16054 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16055 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16056 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16057 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16058 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16059 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16060 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16061 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16062 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16064 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16067 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16068 targ_options->x_arm_arch_string is non NULL.
16070 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16072 * config/aarch64/aarch64.h
16073 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16075 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16078 * expmed.c (emit_store_flag_force): Swap if const op0
16079 and change VOIDmode to mode of op0.
16081 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16083 PR rtl-optimization/83761
16084 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16085 than bytes to mode_for_size.
16087 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16089 PR middle-end/83189
16090 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16091 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16094 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16096 PR middle-end/83575
16097 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16098 when in layout mode.
16099 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16100 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16103 2018-01-10 Michael Collison <michael.collison@arm.com>
16105 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16106 * config/aarch64/aarch64-option-extension.def: Add
16107 AARCH64_OPT_EXTENSION of 'fp16fml'.
16108 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16109 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16110 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16111 * config/aarch64/constraints.md (Ui7): New constraint.
16112 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16113 (VFMLA_SEL_W): Ditto.
16116 (VFMLA16_LOW): New int iterator.
16117 (VFMLA16_HIGH): Ditto.
16118 (UNSPEC_FMLAL): New unspec.
16119 (UNSPEC_FMLSL): Ditto.
16120 (UNSPEC_FMLAL2): Ditto.
16121 (UNSPEC_FMLSL2): Ditto.
16122 (f16mac): New code attribute.
16123 * config/aarch64/aarch64-simd-builtins.def
16124 (aarch64_fmlal_lowv2sf): Ditto.
16125 (aarch64_fmlsl_lowv2sf): Ditto.
16126 (aarch64_fmlalq_lowv4sf): Ditto.
16127 (aarch64_fmlslq_lowv4sf): Ditto.
16128 (aarch64_fmlal_highv2sf): Ditto.
16129 (aarch64_fmlsl_highv2sf): Ditto.
16130 (aarch64_fmlalq_highv4sf): Ditto.
16131 (aarch64_fmlslq_highv4sf): Ditto.
16132 (aarch64_fmlal_lane_lowv2sf): Ditto.
16133 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16134 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16135 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16136 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16137 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16138 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16139 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16140 (aarch64_fmlal_lane_highv2sf): Ditto.
16141 (aarch64_fmlsl_lane_highv2sf): Ditto.
16142 (aarch64_fmlal_laneq_highv2sf): Ditto.
16143 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16144 (aarch64_fmlalq_lane_highv4sf): Ditto.
16145 (aarch64_fmlsl_lane_highv4sf): Ditto.
16146 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16147 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16148 * config/aarch64/aarch64-simd.md:
16149 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16150 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16151 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16152 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16153 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16154 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16155 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16156 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16157 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16158 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16159 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16160 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16161 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16162 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16163 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16164 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16165 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16166 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16167 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16168 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16169 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16170 (vfmlsl_low_u32): Ditto.
16171 (vfmlalq_low_u32): Ditto.
16172 (vfmlslq_low_u32): Ditto.
16173 (vfmlal_high_u32): Ditto.
16174 (vfmlsl_high_u32): Ditto.
16175 (vfmlalq_high_u32): Ditto.
16176 (vfmlslq_high_u32): Ditto.
16177 (vfmlal_lane_low_u32): Ditto.
16178 (vfmlsl_lane_low_u32): Ditto.
16179 (vfmlal_laneq_low_u32): Ditto.
16180 (vfmlsl_laneq_low_u32): Ditto.
16181 (vfmlalq_lane_low_u32): Ditto.
16182 (vfmlslq_lane_low_u32): Ditto.
16183 (vfmlalq_laneq_low_u32): Ditto.
16184 (vfmlslq_laneq_low_u32): Ditto.
16185 (vfmlal_lane_high_u32): Ditto.
16186 (vfmlsl_lane_high_u32): Ditto.
16187 (vfmlal_laneq_high_u32): Ditto.
16188 (vfmlsl_laneq_high_u32): Ditto.
16189 (vfmlalq_lane_high_u32): Ditto.
16190 (vfmlslq_lane_high_u32): Ditto.
16191 (vfmlalq_laneq_high_u32): Ditto.
16192 (vfmlslq_laneq_high_u32): Ditto.
16193 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16194 (AARCH64_FL_FOR_ARCH8_4): New.
16195 (AARCH64_ISA_F16FML): New ISA flag.
16196 (TARGET_F16FML): New feature flag for fp16fml.
16197 (doc/invoke.texi): Document new fp16fml option.
16199 2018-01-10 Michael Collison <michael.collison@arm.com>
16201 * config/aarch64/aarch64-builtins.c:
16202 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16203 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16204 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16205 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16206 (AARCH64_ISA_SHA3): New ISA flag.
16207 (TARGET_SHA3): New feature flag for sha3.
16208 * config/aarch64/iterators.md (sha512_op): New int attribute.
16209 (CRYPTO_SHA512): New int iterator.
16210 (UNSPEC_SHA512H): New unspec.
16211 (UNSPEC_SHA512H2): Ditto.
16212 (UNSPEC_SHA512SU0): Ditto.
16213 (UNSPEC_SHA512SU1): Ditto.
16214 * config/aarch64/aarch64-simd-builtins.def
16215 (aarch64_crypto_sha512hqv2di): New builtin.
16216 (aarch64_crypto_sha512h2qv2di): Ditto.
16217 (aarch64_crypto_sha512su0qv2di): Ditto.
16218 (aarch64_crypto_sha512su1qv2di): Ditto.
16219 (aarch64_eor3qv8hi): Ditto.
16220 (aarch64_rax1qv2di): Ditto.
16221 (aarch64_xarqv2di): Ditto.
16222 (aarch64_bcaxqv8hi): Ditto.
16223 * config/aarch64/aarch64-simd.md:
16224 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16225 (aarch64_crypto_sha512su0qv2di): Ditto.
16226 (aarch64_crypto_sha512su1qv2di): Ditto.
16227 (aarch64_eor3qv8hi): Ditto.
16228 (aarch64_rax1qv2di): Ditto.
16229 (aarch64_xarqv2di): Ditto.
16230 (aarch64_bcaxqv8hi): Ditto.
16231 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16232 (vsha512h2q_u64): Ditto.
16233 (vsha512su0q_u64): Ditto.
16234 (vsha512su1q_u64): Ditto.
16235 (veor3q_u16): Ditto.
16236 (vrax1q_u64): Ditto.
16237 (vxarq_u64): Ditto.
16238 (vbcaxq_u16): Ditto.
16239 * config/arm/types.md (crypto_sha512): New type attribute.
16240 (crypto_sha3): Ditto.
16241 (doc/invoke.texi): Document new sha3 option.
16243 2018-01-10 Michael Collison <michael.collison@arm.com>
16245 * config/aarch64/aarch64-builtins.c:
16246 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16248 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16249 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16250 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16251 (AARCH64_ISA_SM4): New ISA flag.
16252 (TARGET_SM4): New feature flag for sm4.
16253 * config/aarch64/aarch64-simd-builtins.def
16254 (aarch64_sm3ss1qv4si): Ditto.
16255 (aarch64_sm3tt1aq4si): Ditto.
16256 (aarch64_sm3tt1bq4si): Ditto.
16257 (aarch64_sm3tt2aq4si): Ditto.
16258 (aarch64_sm3tt2bq4si): Ditto.
16259 (aarch64_sm3partw1qv4si): Ditto.
16260 (aarch64_sm3partw2qv4si): Ditto.
16261 (aarch64_sm4eqv4si): Ditto.
16262 (aarch64_sm4ekeyqv4si): Ditto.
16263 * config/aarch64/aarch64-simd.md:
16264 (aarch64_sm3ss1qv4si): Ditto.
16265 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16266 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16267 (aarch64_sm4eqv4si): Ditto.
16268 (aarch64_sm4ekeyqv4si): Ditto.
16269 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16270 (sm3part_op): Ditto.
16271 (CRYPTO_SM3TT): Ditto.
16272 (CRYPTO_SM3PART): Ditto.
16273 (UNSPEC_SM3SS1): New unspec.
16274 (UNSPEC_SM3TT1A): Ditto.
16275 (UNSPEC_SM3TT1B): Ditto.
16276 (UNSPEC_SM3TT2A): Ditto.
16277 (UNSPEC_SM3TT2B): Ditto.
16278 (UNSPEC_SM3PARTW1): Ditto.
16279 (UNSPEC_SM3PARTW2): Ditto.
16280 (UNSPEC_SM4E): Ditto.
16281 (UNSPEC_SM4EKEY): Ditto.
16282 * config/aarch64/constraints.md (Ui2): New constraint.
16283 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16284 * config/arm/types.md (crypto_sm3): New type attribute.
16285 (crypto_sm4): Ditto.
16286 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16287 (vsm3tt1aq_u32): Ditto.
16288 (vsm3tt1bq_u32): Ditto.
16289 (vsm3tt2aq_u32): Ditto.
16290 (vsm3tt2bq_u32): Ditto.
16291 (vsm3partw1q_u32): Ditto.
16292 (vsm3partw2q_u32): Ditto.
16293 (vsm4eq_u32): Ditto.
16294 (vsm4ekeyq_u32): Ditto.
16295 (doc/invoke.texi): Document new sm4 option.
16297 2018-01-10 Michael Collison <michael.collison@arm.com>
16299 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16300 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16301 (AARCH64_FL_FOR_ARCH8_4): New.
16302 (AARCH64_FL_V8_4): New flag.
16303 (doc/invoke.texi): Document new armv8.4-a option.
16305 2018-01-10 Michael Collison <michael.collison@arm.com>
16307 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16308 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16309 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16310 * config/aarch64/aarch64-option-extension.def: Add
16311 AARCH64_OPT_EXTENSION of 'sha2'.
16312 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16313 (crypto): Disable sha2 and aes if crypto disabled.
16314 (crypto): Enable aes and sha2 if enabled.
16315 (simd): Disable sha2 and aes if simd disabled.
16316 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16318 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16319 (TARGET_SHA2): New feature flag for sha2.
16320 (TARGET_AES): New feature flag for aes.
16321 * config/aarch64/aarch64-simd.md:
16322 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16323 conditional on TARGET_AES.
16324 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16325 (aarch64_crypto_sha1hsi): Make pattern conditional
16327 (aarch64_crypto_sha1hv4si): Ditto.
16328 (aarch64_be_crypto_sha1hv4si): Ditto.
16329 (aarch64_crypto_sha1su1v4si): Ditto.
16330 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16331 (aarch64_crypto_sha1su0v4si): Ditto.
16332 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16333 (aarch64_crypto_sha256su0v4si): Ditto.
16334 (aarch64_crypto_sha256su1v4si): Ditto.
16335 (doc/invoke.texi): Document new aes and sha2 options.
16337 2018-01-10 Martin Sebor <msebor@redhat.com>
16339 PR tree-optimization/83781
16340 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16343 2018-01-11 Martin Sebor <msebor@gmail.com>
16344 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16346 PR tree-optimization/83501
16347 PR tree-optimization/81703
16349 * tree-ssa-strlen.c (get_string_cst): Rename...
16350 (get_string_len): ...to this. Handle global constants.
16351 (handle_char_store): Adjust.
16353 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16354 Jim Wilson <jimw@sifive.com>
16356 * config/riscv/riscv-protos.h (riscv_output_return): New.
16357 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16358 (riscv_attribute_table, riscv_output_return),
16359 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16360 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16361 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16363 (riscv_expand_prologue): Add early return for naked function.
16364 (riscv_expand_epilogue): Likewise.
16365 (riscv_function_ok_for_sibcall): Return false for naked function.
16366 (riscv_set_current_function): New.
16367 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16368 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16369 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16370 * doc/extend.texi (RISC-V Function Attributes): New.
16372 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16374 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16375 check for 128-bit long double before checking TCmode.
16376 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16377 128-bit long doubles before checking TFmode or TCmode.
16378 (FLOAT128_IBM_P): Likewise.
16380 2018-01-10 Martin Sebor <msebor@redhat.com>
16382 PR tree-optimization/83671
16383 * builtins.c (c_strlen): Unconditionally return zero for the empty
16385 Use -Warray-bounds for warnings.
16386 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16387 for non-constant array indices with COMPONENT_REF, arrays of
16388 arrays, and pointers to arrays.
16389 (gimple_fold_builtin_strlen): Determine and set length range for
16390 non-constant character arrays.
16392 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16394 PR middle-end/81897
16395 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16398 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16400 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16402 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16405 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16406 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16407 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16408 indexed_or_indirect_operand predicate.
16409 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16410 (*vsx_le_perm_load_v8hi): Likewise.
16411 (*vsx_le_perm_load_v16qi): Likewise.
16412 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16413 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16414 (*vsx_le_perm_store_v8hi): Likewise.
16415 (*vsx_le_perm_store_v16qi): Likewise.
16416 (eight unnamed splitters): Likewise.
16418 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16420 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16421 * config/rs6000/emmintrin.h: Likewise.
16422 * config/rs6000/mmintrin.h: Likewise.
16423 * config/rs6000/xmmintrin.h: Likewise.
16425 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16428 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16430 * tree.c (tree_nop_conversion): Return true for location wrapper
16432 (maybe_wrap_with_location): New function.
16433 (selftest::check_strip_nops): New function.
16434 (selftest::test_location_wrappers): New function.
16435 (selftest::tree_c_tests): Call it.
16436 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16437 (maybe_wrap_with_location): New decl.
16438 (EXPR_LOCATION_WRAPPER_P): New macro.
16439 (location_wrapper_p): New inline function.
16440 (tree_strip_any_location_wrapper): New inline function.
16442 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16445 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16446 stack_realign_offset for the largest alignment of stack slot
16448 (ix86_find_max_used_stack_alignment): New function.
16449 (ix86_finalize_stack_frame_flags): Use it. Set
16450 max_used_stack_alignment if we don't realign stack.
16451 * config/i386/i386.h (machine_function): Add
16452 max_used_stack_alignment.
16454 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16456 * config/arm/arm.opt (-mbranch-cost): New option.
16457 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16460 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16463 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16464 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16466 2018-01-10 Richard Biener <rguenther@suse.de>
16469 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16470 early out so it also covers the case where we have a non-NULL
16473 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16475 PR tree-optimization/83753
16476 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16477 for non-strided grouped accesses if the number of elements is 1.
16479 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16482 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16483 * i386.h (TARGET_USE_GATHER): Define.
16484 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16486 2018-01-10 Martin Liska <mliska@suse.cz>
16489 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16490 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16492 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16493 CLEANUP_NO_PARTITIONING is not set.
16495 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16497 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16498 for vectors, as a partial revert of r254296.
16499 * rtl.h (const_vec_p): Delete.
16500 (const_vec_duplicate_p): Don't test for vector CONSTs.
16501 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16502 * expmed.c (make_tree): Likewise.
16505 * common.md (E, F): Use CONSTANT_P instead of checking for
16507 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16508 checking for CONST_VECTOR.
16510 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16512 PR middle-end/83575
16513 * predict.c (force_edge_cold): Handle in more sane way edges
16514 with no prediction.
16516 2018-01-09 Carl Love <cel@us.ibm.com>
16518 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16520 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16521 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16522 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16523 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16524 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16525 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16526 * config/rs6000/rs6000-protos.h: Add extern defition for
16527 rs6000_generate_float2_double_code.
16528 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16530 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16531 (float2_v2df): Add define_expand.
16533 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16536 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16537 op_mode in the force_to_mode call.
16539 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16541 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16542 instead of checking each element individually.
16543 (aarch64_evpc_uzp): Likewise.
16544 (aarch64_evpc_zip): Likewise.
16545 (aarch64_evpc_ext): Likewise.
16546 (aarch64_evpc_rev): Likewise.
16547 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16548 instead of checking each element individually. Return true without
16550 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16551 whether all selected elements come from the same input, instead of
16552 checking each element individually. Remove calls to gen_rtx_REG,
16553 start_sequence and end_sequence and instead assert that no rtl is
16556 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16558 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16559 order of HIGH and CONST checks.
16561 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16563 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16564 if the destination isn't an SSA_NAME.
16566 2018-01-09 Richard Biener <rguenther@suse.de>
16568 PR tree-optimization/83668
16569 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16571 (canonicalize_loop_form): ... here, renamed from ...
16572 (canonicalize_loop_closed_ssa_form): ... this and amended to
16573 swap successor edges for loop exit blocks to make us use
16574 the RPO order we need for initial schedule generation.
16576 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16578 PR tree-optimization/64811
16579 * match.pd: When optimizing comparisons with Inf, avoid
16580 introducing or losing exceptions from comparisons with NaN.
16582 2018-01-09 Martin Liska <mliska@suse.cz>
16585 * asan.c (shadow_mem_size): Add gcc_assert.
16587 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16589 Don't save registers in main().
16592 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16593 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16594 * config/avr/avr.c (avr_set_current_function): Don't error if
16595 naked, OS_task or OS_main are specified at the same time.
16596 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16598 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16600 * common/config/avr/avr-common.c (avr_option_optimization_table):
16601 Switch on -mmain-is-OS_task for optimizing compilations.
16603 2018-01-09 Richard Biener <rguenther@suse.de>
16605 PR tree-optimization/83572
16606 * graphite.c: Include cfganal.h.
16607 (graphite_transform_loops): Connect infinite loops to exit
16608 and remove fake edges at the end.
16610 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16612 * ipa-inline.c (edge_badness): Revert accidental checkin.
16614 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16617 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16618 symbols; not inline clones.
16620 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16623 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16624 hard registers. Formatting fixes.
16626 PR preprocessor/83722
16627 * gcc.c (try_generate_repro): Pass
16628 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16629 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16632 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16633 Kito Cheng <kito.cheng@gmail.com>
16635 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16636 (riscv_leaf_function_p): Delete.
16637 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16639 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16641 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16643 (do_ifelse): New function.
16644 (do_isel): New function.
16645 (do_sub3): New function.
16646 (do_add3): New function.
16647 (do_load_mask_compare): New function.
16648 (do_overlap_load_compare): New function.
16649 (expand_compare_loop): New function.
16650 (expand_block_compare): Call expand_compare_loop() when appropriate.
16651 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16652 option description.
16653 (-mblock-compare-inline-loop-limit): New option.
16655 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16658 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16659 Reverse order of second and third operands in first alternative.
16660 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16661 of first and second elements in UNSPEC_VPERMR vector.
16662 (altivec_expand_vec_perm_le): Likewise.
16664 2018-01-08 Jeff Law <law@redhat.com>
16666 PR rtl-optimizatin/81308
16667 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16668 (process_switch): If group_case_labels makes a change, then set
16670 (pass_convert_switch::execute): If a switch is converted, then
16671 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16673 PR rtl-optimization/81308
16674 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16677 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16679 PR target/83663 - Revert r255946
16680 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16681 generation for cases where splatting a value is not useful.
16682 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16683 across a vec_duplicate and a paradoxical subreg forming a vector
16684 mode to a vec_concat.
16686 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16688 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16689 -march=armv8.3-a variants.
16690 * config/arm/t-multilib: Likewise.
16691 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16693 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16695 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16697 (cceq_ior_compare_complement): Give it a name so I can use it, and
16698 change boolean_or_operator predicate to boolean_operator so it can
16699 be used to generate a crand.
16700 (eqne): New code iterator.
16701 (bd/bd_neg): New code_attrs.
16702 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16703 a single define_insn.
16704 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16705 decrement (bdnzt/bdnzf/bdzt/bdzf).
16706 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16707 with the new names of the branch decrement patterns, and added the
16708 names of the branch decrement conditional patterns.
16710 2018-01-08 Richard Biener <rguenther@suse.de>
16712 PR tree-optimization/83563
16713 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16716 2018-01-08 Richard Biener <rguenther@suse.de>
16718 PR middle-end/83713
16719 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16721 2018-01-08 Richard Biener <rguenther@suse.de>
16723 PR tree-optimization/83685
16724 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16725 references to abnormals.
16727 2018-01-08 Richard Biener <rguenther@suse.de>
16730 * dwarf2out.c (output_indirect_strings): Handle empty
16731 skeleton_debug_str_hash.
16732 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16734 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16736 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16737 (emit_store_direct): Likewise.
16738 (arc_trampoline_adjust_address): Likewise.
16739 (arc_asm_trampoline_template): New function.
16740 (arc_initialize_trampoline): Use asm_trampoline_template.
16741 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16742 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16743 * config/arc/arc.md (flush_icache): Delete pattern.
16745 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16747 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16748 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16751 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16754 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16755 by not USED_FOR_TARGET.
16756 (make_pass_resolve_sw_modes): Likewise.
16758 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16760 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16763 2018-01-08 Richard Biener <rguenther@suse.de>
16765 PR middle-end/83580
16766 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16768 2018-01-08 Richard Biener <rguenther@suse.de>
16770 PR middle-end/83517
16771 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16773 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16775 PR middle-end/81897
16776 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16777 basic blocks with a small number of successors.
16778 (convert_control_dep_chain_into_preds): Improve handling of
16780 (dump_predicates): Split apart into...
16781 (dump_pred_chain): ...here...
16782 (dump_pred_info): ...and here.
16783 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16784 (can_chain_union_be_invalidated_p): Improve check for invalidation
16786 (uninit_uses_cannot_happen): Avoid unnecessary if
16787 convert_control_dep_chain_into_preds yielded nothing.
16789 2018-01-06 Martin Sebor <msebor@redhat.com>
16791 PR tree-optimization/83640
16792 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16793 subtracting negative offset from size.
16794 (builtin_access::overlap): Adjust offset bounds of the access to fall
16795 within the size of the object if possible.
16797 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16799 PR rtl-optimization/83699
16800 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16801 extract_bit_field_as_subreg to cases in which the extracted
16802 value is also a vector.
16804 * lra-constraints.c (process_alt_operands): Test for the equivalence
16805 substitutions when detecting a possible reload cycle.
16807 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16810 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16811 by default if flag_selective_schedling{,2}. Formatting fixes.
16813 PR rtl-optimization/83682
16814 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16815 if it has non-VECTOR_MODE element mode.
16816 (vec_duplicate_p): Likewise.
16818 PR middle-end/83694
16819 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16820 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16822 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16825 * config/i386/i386-builtin.def
16826 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16827 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16828 Require also OPTION_MASK_ISA_AVX512F in addition to
16829 OPTION_MASK_ISA_GFNI.
16830 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16831 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16832 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16833 to OPTION_MASK_ISA_GFNI.
16834 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16835 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16836 OPTION_MASK_ISA_AVX512BW.
16837 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16838 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16839 addition to OPTION_MASK_ISA_GFNI.
16840 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16841 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16842 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16843 to OPTION_MASK_ISA_GFNI.
16844 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16845 a requirement for all ISAs rather than any of them with a few
16847 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16849 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16850 bitmasks to be enabled with 3 exceptions, instead of requiring any
16851 enabled ISA with lots of exceptions.
16852 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16853 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16854 Change avx512bw in isa attribute to avx512f.
16855 * config/i386/sgxintrin.h: Add license boilerplate.
16856 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16857 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16858 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16859 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16861 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16862 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16863 temporarily sse2 rather than sse if not enabled already.
16866 * config/i386/sse.md (VI248_VLBW): Rename to ...
16867 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16868 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16869 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16870 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16871 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16872 mode iterator instead of VI248_VLBW.
16874 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16876 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16877 (record_modified): Skip clobbers; add debug output.
16878 (param_change_prob): Use sreal frequencies.
16880 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16882 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16883 punt for user-aligned variables.
16885 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16887 * tree-chrec.c (chrec_contains_symbols): Return true for
16890 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16893 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16894 of (x|y) == x for BICS pattern.
16896 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16898 PR tree-optimization/83605
16899 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16900 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16903 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16905 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16906 * config/epiphany/rtems.h: New file.
16908 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16909 Uros Bizjak <ubizjak@gmail.com>
16912 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16913 QIreg_operand instead of register_operand predicate.
16914 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16915 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16916 comments instead of -fmitigate[-_]rop.
16918 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16921 * cgraphunit.c (symbol_table::compile): Switch to text_section
16922 before calling assembly_start debug hook.
16923 * run-rtl-passes.c (run_rtl_passes): Likewise.
16926 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16928 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16929 range_int_cst_p rather than !symbolic_range_p before calling
16930 extract_range_from_multiplicative_op_1.
16932 2018-01-04 Jeff Law <law@redhat.com>
16934 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16935 redundant test in assertion.
16937 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16939 * doc/rtl.texi: Document machine_mode wrapper classes.
16941 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16943 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16944 using tree_to_uhwi.
16946 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16948 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16949 the VEC_PERM_EXPR fold to fail.
16951 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16954 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16955 to switched_sections.
16957 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16960 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16961 test for d.testing.
16963 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16966 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16967 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16969 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16972 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16973 is BLKmode and bitpos not zero or mode change is needed.
16975 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16978 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16981 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16984 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16985 instead of MULT rtx. Update all corresponding splitters.
16986 (*saddl_se): Ditto.
16987 (*ssub<modesuffix>): Ditto.
16988 (*ssubl_se): Ditto.
16989 (*cmp_sadd_di): Update split patterns.
16990 (*cmp_sadd_si): Ditto.
16991 (*cmp_sadd_sidi): Ditto.
16992 (*cmp_ssub_di): Ditto.
16993 (*cmp_ssub_si): Ditto.
16994 (*cmp_ssub_sidi): Ditto.
16995 * config/alpha/predicates.md (const23_operand): New predicate.
16996 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16997 Look for ASHIFT, not MULT inner operand.
16998 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17000 2018-01-04 Martin Liska <mliska@suse.cz>
17002 PR gcov-profile/83669
17003 * gcov.c (output_intermediate_file): Add version to intermediate
17005 * doc/gcov.texi: Document new field 'version' in intermediate
17006 file format. Fix location of '-k' option of gcov command.
17008 2018-01-04 Martin Liska <mliska@suse.cz>
17011 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17013 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17015 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17017 2018-01-03 Martin Sebor <msebor@redhat.com>
17019 PR tree-optimization/83655
17020 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17021 checking calls with invalid arguments.
17023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17025 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17026 (vectorizable_mask_load_store): Delete.
17027 (vectorizable_call): Return false for masked loads and stores.
17028 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17029 instead of gimple_assign_rhs1.
17030 (vectorizable_load): Handle IFN_MASK_LOAD.
17031 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17035 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17037 (vectorizable_mask_load_store): ...here.
17038 (vectorizable_load): ...and here.
17040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17042 * tree-vect-stmts.c (vect_build_all_ones_mask)
17043 (vect_build_zero_merge_argument): New functions, split out from...
17044 (vectorizable_load): ...here.
17046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17048 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17050 (vectorizable_mask_load_store): ...here.
17051 (vectorizable_store): ...and here.
17053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17055 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17057 (vectorizable_mask_load_store): ...here.
17059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17061 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17062 (vect_model_store_cost): Take a vec_load_store_type instead of a
17064 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17065 (vect_model_store_cost): Take a vec_load_store_type instead of a
17067 (vectorizable_mask_load_store): Update accordingly.
17068 (vectorizable_store): Likewise.
17069 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17073 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17074 IFN_MASK_LOAD calls here rather than...
17075 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17078 Alan Hayward <alan.hayward@arm.com>
17079 David Sherwood <david.sherwood@arm.com>
17081 * expmed.c (extract_bit_field_1): For vector extracts,
17082 fall back to extract_bit_field_as_subreg if vec_extract
17085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17086 Alan Hayward <alan.hayward@arm.com>
17087 David Sherwood <david.sherwood@arm.com>
17089 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17090 they are variable or constant sized.
17091 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17092 slots for constant-sized data.
17094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17095 Alan Hayward <alan.hayward@arm.com>
17096 David Sherwood <david.sherwood@arm.com>
17098 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17099 handling COND_EXPRs with boolean comparisons, try to find a better
17100 basis for the mask type than the boolean itself.
17102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17104 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17105 is calculated and how it can be overridden.
17106 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17107 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17109 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17113 Alan Hayward <alan.hayward@arm.com>
17114 David Sherwood <david.sherwood@arm.com>
17116 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17117 Remove the mode argument.
17118 (aarch64_simd_valid_immediate): Remove the mode and inverse
17120 * config/aarch64/iterators.md (bitsize): New iterator.
17121 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17122 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17123 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17124 aarch64_simd_valid_immediate.
17125 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17126 (aarch64_reg_or_bic_imm): Likewise.
17127 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17128 with an insn_type enum and msl with a modifier_type enum.
17129 Replace element_width with a scalar_mode. Change the shift
17130 to unsigned int. Add constructors for scalar_float_mode and
17131 scalar_int_mode elements.
17132 (aarch64_vect_float_const_representable_p): Delete.
17133 (aarch64_can_const_movi_rtx_p)
17134 (aarch64_simd_scalar_immediate_valid_for_move)
17135 (aarch64_simd_make_constant): Update call to
17136 aarch64_simd_valid_immediate.
17137 (aarch64_advsimd_valid_immediate_hs): New function.
17138 (aarch64_advsimd_valid_immediate): Likewise.
17139 (aarch64_simd_valid_immediate): Remove mode and inverse
17140 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17141 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17142 and aarch64_float_const_representable_p on the result.
17143 (aarch64_output_simd_mov_immediate): Remove mode argument.
17144 Update call to aarch64_simd_valid_immediate and use of
17145 simd_immediate_info.
17146 (aarch64_output_scalar_simd_mov_immediate): Update call
17149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17150 Alan Hayward <alan.hayward@arm.com>
17151 David Sherwood <david.sherwood@arm.com>
17153 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17154 (mode_nunits): Likewise CONST_MODE_NUNITS.
17155 * machmode.def (ADJUST_NUNITS): Document.
17156 * genmodes.c (mode_data::need_nunits_adj): New field.
17157 (blank_mode): Update accordingly.
17158 (adj_nunits): New variable.
17159 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17161 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17162 listed in adj_nunits.
17163 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17164 listed in adj_nunits. Don't emit case statements for such modes.
17165 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17166 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17167 nothing if adj_nunits is nonnull.
17168 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17169 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17170 (emit_mode_fbit): Update use of print_maybe_const_decl.
17171 (emit_move_size): Likewise. Treat the array as non-const
17173 (emit_mode_adjustments): Handle adj_nunits.
17175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17177 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17178 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17179 (VECTOR_MODES): Use it.
17180 (make_vector_modes): Take the prefix as an argument.
17182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17183 Alan Hayward <alan.hayward@arm.com>
17184 David Sherwood <david.sherwood@arm.com>
17186 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17187 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17188 for MODE_VECTOR_BOOL.
17189 * machmode.def (VECTOR_BOOL_MODE): Document.
17190 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17191 (make_vector_bool_mode): New function.
17192 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17194 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17195 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17197 * stor-layout.c (int_mode_for_mode): Likewise.
17198 * tree.c (build_vector_type_for_mode): Likewise.
17199 * varasm.c (output_constant_pool_2): Likewise.
17200 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17201 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17202 for MODE_VECTOR_BOOL.
17203 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17204 of mode class checks.
17205 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17206 instead of a list of mode class checks.
17207 (expand_vector_scalar_condition): Likewise.
17208 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17211 Alan Hayward <alan.hayward@arm.com>
17212 David Sherwood <david.sherwood@arm.com>
17214 * machmode.h (mode_size): Change from unsigned short to
17216 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17217 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17218 or if measurement_type is not polynomial.
17219 (fixed_size_mode::includes_p): Check for constant-sized modes.
17220 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17221 return a poly_uint16 rather than an unsigned short.
17222 (emit_mode_size): Change the type of mode_size from unsigned short
17223 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17224 (emit_mode_adjustments): Cope with polynomial vector sizes.
17225 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17227 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17229 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17230 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17231 * caller-save.c (setup_save_areas): Likewise.
17232 (replace_reg_with_saved_mem): Likewise.
17233 * calls.c (emit_library_call_value_1): Likewise.
17234 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17235 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17236 (gen_lowpart_for_combine): Likewise.
17237 * convert.c (convert_to_integer_1): Likewise.
17238 * cse.c (equiv_constant, cse_insn): Likewise.
17239 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17240 (cselib_subst_to_values): Likewise.
17241 * dce.c (word_dce_process_block): Likewise.
17242 * df-problems.c (df_word_lr_mark_ref): Likewise.
17243 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17244 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17245 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17246 (rtl_for_decl_location): Likewise.
17247 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17248 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17249 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17250 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17251 (expand_expr_real_1): Likewise.
17252 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17253 (pad_below): Likewise.
17254 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17255 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17256 * ira.c (get_subreg_tracking_sizes): Likewise.
17257 * ira-build.c (ira_create_allocno_objects): Likewise.
17258 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17259 (ira_sort_regnos_for_alter_reg): Likewise.
17260 * ira-costs.c (record_operand_costs): Likewise.
17261 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17262 (resolve_simple_move): Likewise.
17263 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17264 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17265 (lra_constraints): Likewise.
17266 (CONST_POOL_OK_P): Reject variable-sized modes.
17267 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17268 (add_pseudo_to_slot, lra_spill): Likewise.
17269 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17270 * optabs-query.c (get_best_extraction_insn): Likewise.
17271 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17272 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17273 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17274 * recog.c (offsettable_address_addr_space_p): Likewise.
17275 * regcprop.c (maybe_mode_change): Likewise.
17276 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17277 * regrename.c (build_def_use): Likewise.
17278 * regstat.c (dump_reg_info): Likewise.
17279 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17280 (find_reloads, find_reloads_subreg_address): Likewise.
17281 * reload1.c (eliminate_regs_1): Likewise.
17282 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17283 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17284 (simplify_binary_operation_1, simplify_subreg): Likewise.
17285 * targhooks.c (default_function_arg_padding): Likewise.
17286 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17287 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17288 (verify_gimple_assign_ternary): Likewise.
17289 * tree-inline.c (estimate_move_cost): Likewise.
17290 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17291 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17292 (get_address_cost_ainc): Likewise.
17293 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17294 (vect_supportable_dr_alignment): Likewise.
17295 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17296 (vectorizable_reduction): Likewise.
17297 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17298 (vectorizable_operation, vectorizable_load): Likewise.
17299 * tree.c (build_same_sized_truth_vector_type): Likewise.
17300 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17301 * var-tracking.c (emit_note_insn_var_location): Likewise.
17302 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17303 (ADDR_VEC_ALIGN): Likewise.
17305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17306 Alan Hayward <alan.hayward@arm.com>
17307 David Sherwood <david.sherwood@arm.com>
17309 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17311 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17312 or if measurement_type is polynomial.
17313 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17314 * combine.c (make_extraction): Likewise.
17315 * dse.c (find_shift_sequence): Likewise.
17316 * dwarf2out.c (mem_loc_descriptor): Likewise.
17317 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17318 (extract_bit_field, extract_low_bits): Likewise.
17319 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17320 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17321 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17322 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17323 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17324 * reload.c (find_reloads): Likewise.
17325 * reload1.c (alter_reg): Likewise.
17326 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17327 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17328 * tree-if-conv.c (predicate_mem_writes): Likewise.
17329 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17330 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17331 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17332 * valtrack.c (dead_debug_insert_temp): Likewise.
17333 * varasm.c (mergeable_constant_section): Likewise.
17334 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17337 Alan Hayward <alan.hayward@arm.com>
17338 David Sherwood <david.sherwood@arm.com>
17340 * expr.c (expand_assignment): Cope with polynomial mode sizes
17341 when assigning to a CONCAT.
17343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17344 Alan Hayward <alan.hayward@arm.com>
17345 David Sherwood <david.sherwood@arm.com>
17347 * machmode.h (mode_precision): Change from unsigned short to
17349 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17351 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17352 or if measurement_type is not polynomial.
17353 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17354 in which the mode is already known to be a scalar_int_mode.
17355 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17356 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17358 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17359 for GET_MODE_PRECISION.
17360 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17361 for GET_MODE_PRECISION.
17362 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17364 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17365 (expand_field_assignment, make_extraction): Likewise.
17366 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17367 (get_last_value): Likewise.
17368 * convert.c (convert_to_integer_1): Likewise.
17369 * cse.c (cse_insn): Likewise.
17370 * expr.c (expand_expr_real_1): Likewise.
17371 * lra-constraints.c (simplify_operand_subreg): Likewise.
17372 * optabs-query.c (can_atomic_load_p): Likewise.
17373 * optabs.c (expand_atomic_load): Likewise.
17374 (expand_atomic_store): Likewise.
17375 * ree.c (combine_reaching_defs): Likewise.
17376 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17377 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17378 * tree.h (type_has_mode_precision_p): Likewise.
17379 * ubsan.c (instrument_si_overflow): Likewise.
17381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17382 Alan Hayward <alan.hayward@arm.com>
17383 David Sherwood <david.sherwood@arm.com>
17385 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17386 polynomial numbers of units.
17387 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17388 (valid_vector_subparts_p): New function.
17389 (build_vector_type): Remove temporary shim and take the number
17390 of units as a poly_uint64 rather than an int.
17391 (build_opaque_vector_type): Take the number of units as a
17392 poly_uint64 rather than an int.
17393 * tree.c (build_vector_from_ctor): Handle polynomial
17394 TYPE_VECTOR_SUBPARTS.
17395 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17396 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17397 (build_vector_from_val): If the number of units is variable,
17398 use build_vec_duplicate_cst for constant operands and
17399 VEC_DUPLICATE_EXPR otherwise.
17400 (make_vector_type): Remove temporary is_constant ().
17401 (build_vector_type, build_opaque_vector_type): Take the number of
17402 units as a poly_uint64 rather than an int.
17403 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17405 * cfgexpand.c (expand_debug_expr): Likewise.
17406 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17407 (store_constructor, expand_expr_real_1): Likewise.
17408 (const_scalar_mask_from_tree): Likewise.
17409 * fold-const-call.c (fold_const_reduction): Likewise.
17410 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17411 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17412 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17413 (fold_relational_const): Likewise.
17414 (native_interpret_vector): Likewise. Change the size from an
17415 int to an unsigned int.
17416 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17417 TYPE_VECTOR_SUBPARTS.
17418 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17419 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17420 duplicating a non-constant operand into a variable-length vector.
17421 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17422 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17423 * ipa-icf.c (sem_variable::equals): Likewise.
17424 * match.pd: Likewise.
17425 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17426 * print-tree.c (print_node): Likewise.
17427 * stor-layout.c (layout_type): Likewise.
17428 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17429 * tree-cfg.c (verify_gimple_comparison): Likewise.
17430 (verify_gimple_assign_binary): Likewise.
17431 (verify_gimple_assign_ternary): Likewise.
17432 (verify_gimple_assign_single): Likewise.
17433 * tree-pretty-print.c (dump_generic_node): Likewise.
17434 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17435 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17436 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17437 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17438 (vect_shift_permute_load_chain): Likewise.
17439 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17440 (expand_vector_condition, optimize_vector_constructor): Likewise.
17441 (lower_vec_perm, get_compute_type): Likewise.
17442 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17443 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17444 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17445 (vect_recog_mask_conversion_pattern): Likewise.
17446 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17447 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17448 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17449 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17450 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17451 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17452 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17453 (supportable_widening_operation): Likewise.
17454 (supportable_narrowing_operation): Likewise.
17455 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17457 * varasm.c (output_constant): Likewise.
17459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17460 Alan Hayward <alan.hayward@arm.com>
17461 David Sherwood <david.sherwood@arm.com>
17463 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17464 so that both the length == 3 and length != 3 cases set up their
17465 own permute vectors. Add comments explaining why we know the
17466 number of elements is constant.
17467 (vect_permute_load_chain): Likewise.
17469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17470 Alan Hayward <alan.hayward@arm.com>
17471 David Sherwood <david.sherwood@arm.com>
17473 * machmode.h (mode_nunits): Change from unsigned char to
17475 (ONLY_FIXED_SIZE_MODES): New macro.
17476 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17477 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17478 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17480 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17481 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17482 or if measurement_type is not polynomial.
17483 * genmodes.c (ZERO_COEFFS): New macro.
17484 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17486 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17487 Use ZERO_COEFFS when emitting initializers.
17488 * data-streamer.h (bp_pack_poly_value): New function.
17489 (bp_unpack_poly_value): Likewise.
17490 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17491 for GET_MODE_NUNITS.
17492 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17493 for GET_MODE_NUNITS.
17494 * tree.c (make_vector_type): Remove temporary shim and make
17495 the real function take the number of units as a poly_uint64
17496 rather than an int.
17497 (build_vector_type_for_mode): Handle polynomial nunits.
17498 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17499 * emit-rtl.c (const_vec_series_p_1): Likewise.
17500 (gen_rtx_CONST_VECTOR): Likewise.
17501 * fold-const.c (test_vec_duplicate_folding): Likewise.
17502 * genrecog.c (validate_pattern): Likewise.
17503 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17504 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17505 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17506 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17507 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17508 * rtlanal.c (subreg_get_info): Likewise.
17509 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17510 (vect_grouped_load_supported): Likewise.
17511 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17512 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17513 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17514 (simplify_const_unary_operation, simplify_binary_operation_1)
17515 (simplify_const_binary_operation, simplify_ternary_operation)
17516 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17517 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17518 instead of CONST_VECTOR_NUNITS.
17519 * varasm.c (output_constant_pool_2): Likewise.
17520 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17521 explicit-encoded elements in the XVEC for variable-length vectors.
17523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17525 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17528 Alan Hayward <alan.hayward@arm.com>
17529 David Sherwood <david.sherwood@arm.com>
17531 * coretypes.h (fixed_size_mode): Declare.
17532 (fixed_size_mode_pod): New typedef.
17533 * builtins.h (target_builtins::x_apply_args_mode)
17534 (target_builtins::x_apply_result_mode): Change type to
17535 fixed_size_mode_pod.
17536 * builtins.c (apply_args_size, apply_result_size, result_vector)
17537 (expand_builtin_apply_args_1, expand_builtin_apply)
17538 (expand_builtin_return): Update accordingly.
17540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17542 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17543 * cselib.c (cselib_hash_rtx): Likewise.
17544 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17545 CONST_VECTOR encoding.
17547 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17548 Jeff Law <law@redhat.com>
17551 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17552 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17553 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17554 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17557 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17558 explicitly probe *sp in a noreturn function if there were any callee
17559 register saves or frame pointer is needed.
17561 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17564 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17565 BLKmode for ternary, binary or unary expressions.
17568 * var-tracking.c (delete_vta_debug_insn): New inline function.
17569 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17570 insns from get_insns () to NULL instead of each bb separately.
17571 Use delete_vta_debug_insn. No longer static.
17572 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17573 delete_vta_debug_insns callers.
17574 * rtl.h (delete_vta_debug_insns): Declare.
17575 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17576 instead of variable_tracking_main.
17578 2018-01-03 Martin Sebor <msebor@redhat.com>
17580 PR tree-optimization/83603
17581 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17582 arguments past the endof the argument list in functions declared
17583 without a prototype.
17584 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17585 Avoid checking when arguments are null.
17587 2018-01-03 Martin Sebor <msebor@redhat.com>
17590 * doc/extend.texi (attribute const): Fix a typo.
17591 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17592 issuing -Wsuggest-attribute for void functions.
17594 2018-01-03 Martin Sebor <msebor@redhat.com>
17596 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17597 offset_int::from instead of wide_int::to_shwi.
17598 (maybe_diag_overlap): Remove assertion.
17599 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17600 * gimple-ssa-sprintf.c (format_directive): Same.
17601 (parse_directive): Same.
17602 (sprintf_dom_walker::compute_format_length): Same.
17603 (try_substitute_return_value): Same.
17605 2018-01-03 Jeff Law <law@redhat.com>
17607 PR middle-end/83654
17608 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17609 non-constant residual for zero at runtime and avoid probing in
17610 that case. Reorganize code for trailing problem to mirror handling
17613 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17615 PR tree-optimization/83501
17616 * tree-ssa-strlen.c (get_string_cst): New.
17617 (handle_char_store): Call get_string_cst.
17619 2018-01-03 Martin Liska <mliska@suse.cz>
17621 PR tree-optimization/83593
17622 * tree-ssa-strlen.c: Include tree-cfg.h.
17623 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17624 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17625 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17627 (strlen_dom_walker::before_dom_children): Call
17628 gimple_purge_dead_eh_edges. Dump tranformation with details
17630 (strlen_dom_walker::before_dom_children): Update call by adding
17631 new argument cleanup_eh.
17632 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17634 2018-01-03 Martin Liska <mliska@suse.cz>
17637 * cif-code.def (VARIADIC_THUNK): New enum value.
17638 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17641 2018-01-03 Jan Beulich <jbeulich@suse.com>
17643 * sse.md (mov<mode>_internal): Tighten condition for when to use
17644 vmovdqu<ssescalarsize> for TI and OI modes.
17646 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17648 Update copyright years.
17650 2018-01-03 Martin Liska <mliska@suse.cz>
17653 * ipa-visibility.c (function_and_variable_visibility): Skip
17654 functions with noipa attribure.
17656 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17658 * gcc.c (process_command): Update copyright notice dates.
17659 * gcov-dump.c (print_version): Ditto.
17660 * gcov.c (print_version): Ditto.
17661 * gcov-tool.c (print_version): Ditto.
17662 * gengtype.c (create_file): Ditto.
17663 * doc/cpp.texi: Bump @copying's copyright year.
17664 * doc/cppinternals.texi: Ditto.
17665 * doc/gcc.texi: Ditto.
17666 * doc/gccint.texi: Ditto.
17667 * doc/gcov.texi: Ditto.
17668 * doc/install.texi: Ditto.
17669 * doc/invoke.texi: Ditto.
17671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17673 * vector-builder.h (vector_builder::m_full_nelts): Change from
17674 unsigned int to poly_uint64.
17675 (vector_builder::full_nelts): Update prototype accordingly.
17676 (vector_builder::new_vector): Likewise.
17677 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17678 (vector_builder::operator ==): Likewise.
17679 (vector_builder::finalize): Likewise.
17680 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17681 Take the number of elements as a poly_uint64 rather than an
17683 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17684 from unsigned int to poly_uint64.
17685 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17686 (vec_perm_indices::new_vector): Likewise.
17687 (vec_perm_indices::length): Likewise.
17688 (vec_perm_indices::nelts_per_input): Likewise.
17689 (vec_perm_indices::input_nelts): Likewise.
17690 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17691 number of elements per input as a poly_uint64 rather than an
17692 unsigned int. Use the original encoding for variable-length
17693 vectors, rather than clamping each individual element.
17694 For the second and subsequent elements in each pattern,
17695 clamp the step and base before clamping their sum.
17696 (vec_perm_indices::series_p): Handle polynomial element counts.
17697 (vec_perm_indices::all_in_range_p): Likewise.
17698 (vec_perm_indices_to_tree): Likewise.
17699 (vec_perm_indices_to_rtx): Likewise.
17700 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17701 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17702 (tree_vector_builder::new_binary_operation): Handle polynomial
17703 element counts. Return false if we need to know the number
17704 of elements at compile time.
17705 * fold-const.c (fold_vec_perm): Punt if the number of elements
17706 isn't known at compile time.
17708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17710 * vec-perm-indices.h (vec_perm_builder): Change element type
17711 from HOST_WIDE_INT to poly_int64.
17712 (vec_perm_indices::element_type): Update accordingly.
17713 (vec_perm_indices::clamp): Handle polynomial element_types.
17714 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17715 (vec_perm_indices::all_in_range_p): Likewise.
17716 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17718 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17719 polynomial vec_perm_indices element types.
17720 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17721 * fold-const.c (fold_vec_perm): Likewise.
17722 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17723 * tree-vect-generic.c (lower_vec_perm): Likewise.
17724 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17725 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17726 element type to HOST_WIDE_INT.
17728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17729 Alan Hayward <alan.hayward@arm.com>
17730 David Sherwood <david.sherwood@arm.com>
17732 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17733 rather than an int. Use plus_constant.
17734 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17735 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17738 Alan Hayward <alan.hayward@arm.com>
17739 David Sherwood <david.sherwood@arm.com>
17741 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17742 a HOST_WIDE_INT to a poly_int64.
17744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17745 Alan Hayward <alan.hayward@arm.com>
17746 David Sherwood <david.sherwood@arm.com>
17748 * calls.c (load_register_parameters): Cope with polynomial
17749 mode sizes. Require a constant size for BLKmode parameters
17750 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17751 forces a parameter to be padded at the lsb end in order to
17752 fill a complete number of words, require the parameter size
17753 to be ordered wrt UNITS_PER_WORD.
17755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17756 Alan Hayward <alan.hayward@arm.com>
17757 David Sherwood <david.sherwood@arm.com>
17759 * reload1.c (spill_stack_slot_width): Change element type
17760 from unsigned int to poly_uint64_pod.
17761 (alter_reg): Treat mode sizes as polynomial.
17763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17764 Alan Hayward <alan.hayward@arm.com>
17765 David Sherwood <david.sherwood@arm.com>
17767 * reload.c (complex_word_subreg_p): New function.
17768 (reload_inner_reg_of_subreg, push_reload): Use it.
17770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17771 Alan Hayward <alan.hayward@arm.com>
17772 David Sherwood <david.sherwood@arm.com>
17774 * lra-constraints.c (process_alt_operands): Reject matched
17775 operands whose sizes aren't ordered.
17776 (match_reload): Refer to this check here.
17778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17779 Alan Hayward <alan.hayward@arm.com>
17780 David Sherwood <david.sherwood@arm.com>
17782 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17783 that the mode size is in the set {1, 2, 4, 8, 16}.
17785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17786 Alan Hayward <alan.hayward@arm.com>
17787 David Sherwood <david.sherwood@arm.com>
17789 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17790 Use plus_constant instead of gen_rtx_PLUS.
17792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17793 Alan Hayward <alan.hayward@arm.com>
17794 David Sherwood <david.sherwood@arm.com>
17796 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17797 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17798 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17799 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17800 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17801 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17802 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17803 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17804 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17805 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17807 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17808 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17809 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17810 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17811 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17812 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17813 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17814 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17815 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17816 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17818 * expr.c (emit_move_resolve_push): Treat the input and result
17819 of PUSH_ROUNDING as a poly_int64.
17820 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17821 (emit_push_insn): Likewise.
17822 * lra-eliminations.c (mark_not_eliminable): Likewise.
17823 * recog.c (push_operand): Likewise.
17824 * reload1.c (elimination_effects): Likewise.
17825 * rtlanal.c (nonzero_bits1): Likewise.
17826 * calls.c (store_one_arg): Likewise. Require the padding to be
17827 known at compile time.
17829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17830 Alan Hayward <alan.hayward@arm.com>
17831 David Sherwood <david.sherwood@arm.com>
17833 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17834 Use plus_constant instead of gen_rtx_PLUS.
17836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17837 Alan Hayward <alan.hayward@arm.com>
17838 David Sherwood <david.sherwood@arm.com>
17840 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17841 rather than an int.
17843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17844 Alan Hayward <alan.hayward@arm.com>
17845 David Sherwood <david.sherwood@arm.com>
17847 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17848 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17849 via stack temporaries. Treat the mode size as polynomial too.
17851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17852 Alan Hayward <alan.hayward@arm.com>
17853 David Sherwood <david.sherwood@arm.com>
17855 * expr.c (expand_expr_real_2): When handling conversions involving
17856 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17857 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17858 as a poly_uint64 too.
17860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17861 Alan Hayward <alan.hayward@arm.com>
17862 David Sherwood <david.sherwood@arm.com>
17864 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17867 Alan Hayward <alan.hayward@arm.com>
17868 David Sherwood <david.sherwood@arm.com>
17870 * combine.c (can_change_dest_mode): Handle polynomial
17871 REGMODE_NATURAL_SIZE.
17872 * expmed.c (store_bit_field_1): Likewise.
17873 * expr.c (store_constructor): Likewise.
17874 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17875 and polynomial REGMODE_NATURAL_SIZE.
17876 (gen_lowpart_common): Likewise.
17877 * reginfo.c (record_subregs_of_mode): Likewise.
17878 * rtlanal.c (read_modify_subreg_p): Likewise.
17880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17881 Alan Hayward <alan.hayward@arm.com>
17882 David Sherwood <david.sherwood@arm.com>
17884 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17885 numbers of elements.
17887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17888 Alan Hayward <alan.hayward@arm.com>
17889 David Sherwood <david.sherwood@arm.com>
17891 * match.pd: Cope with polynomial numbers of vector elements.
17893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17894 Alan Hayward <alan.hayward@arm.com>
17895 David Sherwood <david.sherwood@arm.com>
17897 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17898 in a POINTER_PLUS_EXPR.
17900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17901 Alan Hayward <alan.hayward@arm.com>
17902 David Sherwood <david.sherwood@arm.com>
17904 * omp-simd-clone.c (simd_clone_subparts): New function.
17905 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17906 (ipa_simd_modify_function_body): Likewise.
17908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17909 Alan Hayward <alan.hayward@arm.com>
17910 David Sherwood <david.sherwood@arm.com>
17912 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17913 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17914 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17915 (expand_vector_condition, vector_element): Likewise.
17916 (subparts_gt): New function.
17917 (get_compute_type): Use subparts_gt.
17918 (count_type_subparts): Delete.
17919 (expand_vector_operations_1): Use subparts_gt instead of
17920 count_type_subparts.
17922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17923 Alan Hayward <alan.hayward@arm.com>
17924 David Sherwood <david.sherwood@arm.com>
17926 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17927 (vect_compile_time_alias): ...this new function. Do the calculation
17928 on poly_ints rather than trees.
17929 (vect_prune_runtime_alias_test_list): Update call accordingly.
17931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17932 Alan Hayward <alan.hayward@arm.com>
17933 David Sherwood <david.sherwood@arm.com>
17935 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17937 (vect_schedule_slp_instance): Likewise.
17939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17940 Alan Hayward <alan.hayward@arm.com>
17941 David Sherwood <david.sherwood@arm.com>
17943 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17944 constant and extern definitions for variable-length vectors.
17945 (vect_get_constant_vectors): Note that the number of units
17946 is known to be constant.
17948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17949 Alan Hayward <alan.hayward@arm.com>
17950 David Sherwood <david.sherwood@arm.com>
17952 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17953 of units as polynomial. Choose between WIDE and NARROW based
17956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17957 Alan Hayward <alan.hayward@arm.com>
17958 David Sherwood <david.sherwood@arm.com>
17960 * tree-vect-stmts.c (simd_clone_subparts): New function.
17961 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17964 Alan Hayward <alan.hayward@arm.com>
17965 David Sherwood <david.sherwood@arm.com>
17967 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17968 vectors as polynomial. Use build_index_vector for
17969 IFN_GOMP_SIMD_LANE.
17971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17972 Alan Hayward <alan.hayward@arm.com>
17973 David Sherwood <david.sherwood@arm.com>
17975 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17976 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17977 for variable-length vectors.
17978 (vectorizable_mask_load_store): Treat the number of units as
17979 polynomial, asserting that it is constant if the condition has
17980 already been enforced.
17981 (vectorizable_store, vectorizable_load): Likewise.
17983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17984 Alan Hayward <alan.hayward@arm.com>
17985 David Sherwood <david.sherwood@arm.com>
17987 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17988 of units as polynomial. Punt if we can't tell at compile time
17989 which vector contains the final result.
17991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17992 Alan Hayward <alan.hayward@arm.com>
17993 David Sherwood <david.sherwood@arm.com>
17995 * tree-vect-loop.c (vectorizable_induction): Treat the number
17996 of units as polynomial. Punt on SLP inductions. Use an integer
17997 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17998 cast of such a series for variable-length floating-point
18001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18002 Alan Hayward <alan.hayward@arm.com>
18003 David Sherwood <david.sherwood@arm.com>
18005 * tree.h (build_index_vector): Declare.
18006 * tree.c (build_index_vector): New function.
18007 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18008 of units as polynomial, forcibly converting it to a constant if
18009 vectorizable_reduction has already enforced the condition.
18010 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18011 to create a {1,2,3,...} vector.
18012 (vectorizable_reduction): Treat the number of units as polynomial.
18013 Choose vectype_in based on the largest scalar element size rather
18014 than the smallest number of units. Enforce the restrictions
18017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18018 Alan Hayward <alan.hayward@arm.com>
18019 David Sherwood <david.sherwood@arm.com>
18021 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18022 number of units as polynomial.
18024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18025 Alan Hayward <alan.hayward@arm.com>
18026 David Sherwood <david.sherwood@arm.com>
18028 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18029 * target.def (autovectorize_vector_sizes): Return the vector sizes
18030 by pointer, using vector_sizes rather than a bitmask.
18031 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18032 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18033 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18035 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18036 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18037 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18038 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18039 * omp-general.c (omp_max_vf): Likewise.
18040 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18041 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18042 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18043 * tree-vect-slp.c (vect_slp_bb): Likewise.
18044 * doc/tm.texi: Regenerate.
18045 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18047 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18048 the vector size as a poly_uint64 rather than an unsigned int.
18049 (current_vector_size): Change from an unsigned int to a poly_uint64.
18050 (get_vectype_for_scalar_type): Update accordingly.
18051 * tree.h (build_truth_vector_type): Take the size and number of
18052 units as a poly_uint64 rather than an unsigned int.
18053 (build_vector_type): Add a temporary overload that takes
18054 the number of units as a poly_uint64 rather than an unsigned int.
18055 * tree.c (make_vector_type): Likewise.
18056 (build_truth_vector_type): Take the number of units as a poly_uint64
18057 rather than an unsigned int.
18059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18060 Alan Hayward <alan.hayward@arm.com>
18061 David Sherwood <david.sherwood@arm.com>
18063 * target.def (get_mask_mode): Take the number of units and length
18064 as poly_uint64s rather than unsigned ints.
18065 * targhooks.h (default_get_mask_mode): Update accordingly.
18066 * targhooks.c (default_get_mask_mode): Likewise.
18067 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18068 * doc/tm.texi: Regenerate.
18070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18071 Alan Hayward <alan.hayward@arm.com>
18072 David Sherwood <david.sherwood@arm.com>
18074 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18075 * omp-general.c (omp_max_vf): Likewise.
18076 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18077 (expand_omp_simd): Handle polynomial safelen.
18078 * omp-low.c (omplow_simd_context): Add a default constructor.
18079 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18080 (lower_rec_simd_input_clauses): Update accordingly.
18081 (lower_rec_input_clauses): Likewise.
18083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18084 Alan Hayward <alan.hayward@arm.com>
18085 David Sherwood <david.sherwood@arm.com>
18087 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18088 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18089 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18090 (vect_analyze_slp_cost): Likewise.
18091 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18092 (vect_model_load_cost): Likewise.
18094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18095 Alan Hayward <alan.hayward@arm.com>
18096 David Sherwood <david.sherwood@arm.com>
18098 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18099 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18100 from an unsigned int * to a poly_uint64_pod *.
18101 (calculate_unrolling_factor): New function.
18102 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18105 Alan Hayward <alan.hayward@arm.com>
18106 David Sherwood <david.sherwood@arm.com>
18108 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18109 from an unsigned int to a poly_uint64.
18110 (_loop_vec_info::slp_unrolling_factor): Likewise.
18111 (_loop_vec_info::vectorization_factor): Change from an int
18113 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18114 (vect_get_num_vectors): New function.
18115 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18116 (vect_get_num_copies): Use vect_get_num_vectors.
18117 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18118 to an unsigned int *.
18119 (vect_analyze_data_refs): Change min_vf from an int * to a
18121 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18122 than an unsigned HOST_WIDE_INT.
18123 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18124 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18125 to an unsigned int *.
18126 (vect_analyze_data_ref_dependences): Likewise.
18127 (vect_compute_data_ref_alignment): Handle polynomial vf.
18128 (vect_enhance_data_refs_alignment): Likewise.
18129 (vect_prune_runtime_alias_test_list): Likewise.
18130 (vect_shift_permute_load_chain): Likewise.
18131 (vect_supportable_dr_alignment): Likewise.
18132 (dependence_distance_ge_vf): Take the vectorization factor as a
18133 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18134 (vect_analyze_data_refs): Change min_vf from an int * to a
18136 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18137 vfm1 as a poly_uint64 rather than an int. Make the same change
18138 for the returned bound_scalar.
18139 (vect_gen_vector_loop_niters): Handle polynomial vf.
18140 (vect_do_peeling): Likewise. Update call to
18141 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18142 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18144 * tree-vect-loop.c (vect_determine_vectorization_factor)
18145 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18146 (vect_get_known_peeling_cost): Likewise.
18147 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18148 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18149 (vect_transform_loop): Likewise. Use the lowest possible VF when
18150 updating the upper bounds of the loop.
18151 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18152 rather than an int.
18153 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18154 polynomial unroll factors.
18155 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18156 (vect_make_slp_decision): Likewise.
18157 (vect_supported_load_permutation_p): Likewise, and polynomial
18159 (vect_analyze_slp_cost): Handle polynomial vf.
18160 (vect_slp_analyze_node_operations): Likewise.
18161 (vect_slp_analyze_bb_1): Likewise.
18162 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18163 than an unsigned HOST_WIDE_INT.
18164 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18165 (vectorizable_load): Handle polynomial vf.
18166 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18168 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18171 Alan Hayward <alan.hayward@arm.com>
18172 David Sherwood <david.sherwood@arm.com>
18174 * match.pd: Handle bit operations involving three constants
18175 and try to fold one pair.
18177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18179 * tree-vect-loop-manip.c: Include gimple-fold.h.
18180 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18181 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18182 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18183 Add a path that uses a step of VF instead of 1, but disable it
18185 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18186 and niters_no_overflow parameters. Update calls to
18187 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18188 Create a new SSA name if the latter choses to use a ste other
18189 than zero, and return it via niters_vector_mult_vf_var.
18190 * tree-vect-loop.c (vect_transform_loop): Update calls to
18191 vect_do_peeling, vect_gen_vector_loop_niters and
18192 slpeel_make_loop_iterate_ntimes.
18193 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18194 (vect_gen_vector_loop_niters): Update declarations after above changes.
18196 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18198 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18199 128-bit round to integer instructions.
18200 (ceil<mode>2): Likewise.
18201 (btrunc<mode>2): Likewise.
18202 (round<mode>2): Likewise.
18204 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18206 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18207 unaligned VSX load/store on P8/P9.
18208 (expand_block_clear): Allow the use of unaligned VSX
18209 load/store on P8/P9.
18211 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18213 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18215 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18216 swap associated with both a load and a store.
18218 2018-01-02 Andrew Waterman <andrew@sifive.com>
18220 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18221 * config/riscv/riscv.md (clear_cache): Use it.
18223 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18225 * web.c: Remove out-of-date comment.
18227 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18229 * expr.c (fixup_args_size_notes): Check that any existing
18230 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18231 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18232 (emit_single_push_insn): ...here.
18234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18236 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18237 (const_vector_encoded_nelts): New function.
18238 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18239 (const_vector_int_elt, const_vector_elt): Declare.
18240 * emit-rtl.c (const_vector_int_elt_1): New function.
18241 (const_vector_elt): Likewise.
18242 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18243 of CONST_VECTOR_ELT.
18245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18247 * expr.c: Include rtx-vector-builder.h.
18248 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18249 directly on the tree encoding.
18250 (const_vector_from_tree): Likewise.
18251 * optabs.c: Include rtx-vector-builder.h.
18252 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18253 sequence of "u" values.
18254 * vec-perm-indices.c: Include rtx-vector-builder.h.
18255 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18256 directly on the vec_perm_indices encoding.
18258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18260 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18261 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18262 * rtx-vector-builder.h: New file.
18263 * rtx-vector-builder.c: Likewise.
18264 * rtl.h (rtx_def::u2): Add a const_vector field.
18265 (CONST_VECTOR_NPATTERNS): New macro.
18266 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18267 (CONST_VECTOR_DUPLICATE_P): Likewise.
18268 (CONST_VECTOR_STEPPED_P): Likewise.
18269 (CONST_VECTOR_ENCODED_ELT): Likewise.
18270 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18271 (unwrap_const_vec_duplicate): Likewise.
18272 (const_vec_series_p): Check for a non-duplicated vector encoding.
18273 Say that the function only returns true for integer vectors.
18274 * emit-rtl.c: Include rtx-vector-builder.h.
18275 (gen_const_vec_duplicate_1): Delete.
18276 (gen_const_vector): Call gen_const_vec_duplicate instead of
18277 gen_const_vec_duplicate_1.
18278 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18279 (gen_const_vec_duplicate): Use rtx_vector_builder.
18280 (gen_const_vec_series): Likewise.
18281 (gen_rtx_CONST_VECTOR): Likewise.
18282 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18283 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18284 Build a new vector rather than modifying a CONST_VECTOR in-place.
18285 (handle_special_swappables): Update call accordingly.
18286 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18287 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18288 Build a new vector rather than modifying a CONST_VECTOR in-place.
18289 (handle_special_swappables): Update call accordingly.
18291 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18293 * simplify-rtx.c (simplify_const_binary_operation): Use
18294 CONST_VECTOR_ELT instead of XVECEXP.
18296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18298 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18299 the selector elements to be different from the data elements
18300 if the selector is a VECTOR_CST.
18301 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18302 ssizetype for the selector.
18304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18306 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18307 before testing each element individually.
18308 * tree-vect-generic.c (lower_vec_perm): Likewise.
18310 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18312 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18313 * selftest-run-tests.c (selftest::run_tests): Call it.
18314 * vector-builder.h (vector_builder::operator ==): New function.
18315 (vector_builder::operator !=): Likewise.
18316 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18317 (vec_perm_indices::all_from_input_p): New function.
18318 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18319 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18320 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18321 instead of reading the VECTOR_CST directly. Detect whether both
18322 vector inputs are the same before constructing the vec_perm_indices,
18323 and update the number of inputs argument accordingly. Use the
18324 utility functions added above. Only construct sel2 if we need to.
18326 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18328 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18329 the broadcast of the low byte.
18330 (expand_mult_highpart): Use an explicit encoding for the permutes.
18331 * optabs-query.c (can_mult_highpart_p): Likewise.
18332 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18333 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18334 (vectorizable_bswap): Likewise.
18335 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18336 explicit encoding for the power-of-2 permutes.
18337 (vect_permute_store_chain): Likewise.
18338 (vect_grouped_load_supported): Likewise.
18339 (vect_permute_load_chain): Likewise.
18341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18343 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18344 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18345 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18346 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18347 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18348 (vect_gen_perm_mask_any): Likewise.
18350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18352 * int-vector-builder.h: New file.
18353 * vec-perm-indices.h: Include int-vector-builder.h.
18354 (vec_perm_indices): Redefine as an int_vector_builder.
18355 (auto_vec_perm_indices): Delete.
18356 (vec_perm_builder): Redefine as a stand-alone class.
18357 (vec_perm_indices::vec_perm_indices): New function.
18358 (vec_perm_indices::clamp): Likewise.
18359 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18360 (vec_perm_indices::new_vector): New function.
18361 (vec_perm_indices::new_expanded_vector): Update for new
18362 vec_perm_indices class.
18363 (vec_perm_indices::rotate_inputs): New function.
18364 (vec_perm_indices::all_in_range_p): Operate directly on the
18365 encoded form, without computing elided elements.
18366 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18367 encoding. Update for new vec_perm_indices class.
18368 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18369 the given vec_perm_builder.
18370 (expand_vec_perm_var): Update vec_perm_builder constructor.
18371 (expand_mult_highpart): Use vec_perm_builder instead of
18372 auto_vec_perm_indices.
18373 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18374 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18375 or double series encoding as appropriate.
18376 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18377 vec_perm_indices instead of auto_vec_perm_indices.
18378 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18379 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18380 (vect_permute_store_chain): Likewise.
18381 (vect_grouped_load_supported): Likewise.
18382 (vect_permute_load_chain): Likewise.
18383 (vect_shift_permute_load_chain): Likewise.
18384 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18385 (vect_transform_slp_perm_load): Likewise.
18386 (vect_schedule_slp_instance): Likewise.
18387 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18388 (vectorizable_mask_load_store): Likewise.
18389 (vectorizable_bswap): Likewise.
18390 (vectorizable_store): Likewise.
18391 (vectorizable_load): Likewise.
18392 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18393 vec_perm_indices instead of auto_vec_perm_indices. Use
18394 tree_to_vec_perm_builder to read the vector from a tree.
18395 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18396 vec_perm_builder instead of a vec_perm_indices.
18397 (have_whole_vector_shift): Use vec_perm_builder and
18398 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18399 truncation to calc_vec_perm_mask_for_shift.
18400 (vect_create_epilog_for_reduction): Likewise.
18401 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18402 from auto_vec_perm_indices to vec_perm_indices.
18403 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18404 instead of changing individual elements.
18405 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18406 the vector in d.perm.
18407 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18408 from auto_vec_perm_indices to vec_perm_indices.
18409 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18410 instead of changing individual elements.
18411 (arm_vectorize_vec_perm_const): Use new_vector to install
18412 the vector in d.perm.
18413 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18414 Update vec_perm_builder constructor.
18415 (rs6000_expand_interleave): Likewise.
18416 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18417 (rs6000_expand_interleave): Likewise.
18419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18421 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18422 to qimode could truncate the indices.
18423 * optabs.c (expand_vec_perm_var): Likewise.
18425 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18427 * Makefile.in (OBJS): Add vec-perm-indices.o.
18428 * vec-perm-indices.h: New file.
18429 * vec-perm-indices.c: Likewise.
18430 * target.h (vec_perm_indices): Replace with a forward class
18432 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18433 * optabs.h: Include vec-perm-indices.h.
18434 (expand_vec_perm): Delete.
18435 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18436 (expand_vec_perm_const): Declare.
18437 * target.def (vec_perm_const_ok): Replace with...
18438 (vec_perm_const): ...this new hook.
18439 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18440 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18441 * doc/tm.texi: Regenerate.
18442 * optabs.def (vec_perm_const): Delete.
18443 * doc/md.texi (vec_perm_const): Likewise.
18444 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18445 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18446 expand_vec_perm for constant permutation vectors. Assert that
18447 the mode of variable permutation vectors is the integer equivalent
18448 of the mode that is being permuted.
18449 * optabs-query.h (selector_fits_mode_p): Declare.
18450 * optabs-query.c: Include vec-perm-indices.h.
18451 (selector_fits_mode_p): New function.
18452 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18453 is defined, instead of checking whether the vec_perm_const_optab
18454 exists. Use targetm.vectorize.vec_perm_const instead of
18455 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18456 fit in the vector mode before using a variable permute.
18457 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18458 vec_perm_indices instead of an rtx.
18459 (expand_vec_perm): Replace with...
18460 (expand_vec_perm_const): ...this new function. Take the selector
18461 as a vec_perm_indices rather than an rtx. Also take the mode of
18462 the selector. Update call to shift_amt_for_vec_perm_mask.
18463 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18464 Use vec_perm_indices::new_expanded_vector to expand the original
18465 selector into bytes. Check whether the indices fit in the vector
18466 mode before using a variable permute.
18467 (expand_vec_perm_var): Make global.
18468 (expand_mult_highpart): Use expand_vec_perm_const.
18469 * fold-const.c: Includes vec-perm-indices.h.
18470 * tree-ssa-forwprop.c: Likewise.
18471 * tree-vect-data-refs.c: Likewise.
18472 * tree-vect-generic.c: Likewise.
18473 * tree-vect-loop.c: Likewise.
18474 * tree-vect-slp.c: Likewise.
18475 * tree-vect-stmts.c: Likewise.
18476 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18478 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18479 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18480 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18481 (aarch64_vectorize_vec_perm_const): ...this new function.
18482 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18483 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18484 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18485 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18486 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18487 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18488 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18490 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18491 check for NEON modes.
18492 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18493 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18494 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18495 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18497 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18498 the old VEC_PERM_CONST conditions.
18499 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18500 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18501 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18502 (ia64_vectorize_vec_perm_const_ok): Merge into...
18503 (ia64_vectorize_vec_perm_const): ...this new function.
18504 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18505 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18506 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18507 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18508 * config/mips/mips.c (mips_expand_vec_perm_const)
18509 (mips_vectorize_vec_perm_const_ok): Merge into...
18510 (mips_vectorize_vec_perm_const): ...this new function.
18511 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18512 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18513 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18514 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18515 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18516 (rs6000_expand_vec_perm_const): Delete.
18517 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18519 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18520 (altivec_expand_vec_perm_const_le): Take each operand individually.
18521 Operate on constant selectors rather than rtxes.
18522 (altivec_expand_vec_perm_const): Likewise. Update call to
18523 altivec_expand_vec_perm_const_le.
18524 (rs6000_expand_vec_perm_const): Delete.
18525 (rs6000_vectorize_vec_perm_const_ok): Delete.
18526 (rs6000_vectorize_vec_perm_const): New function.
18527 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18528 an element count and rtx array.
18529 (rs6000_expand_extract_even): Update call accordingly.
18530 (rs6000_expand_interleave): Likewise.
18531 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18532 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18533 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18534 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18535 (rs6000_expand_vec_perm_const): Delete.
18536 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18537 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18538 (altivec_expand_vec_perm_const_le): Take each operand individually.
18539 Operate on constant selectors rather than rtxes.
18540 (altivec_expand_vec_perm_const): Likewise. Update call to
18541 altivec_expand_vec_perm_const_le.
18542 (rs6000_expand_vec_perm_const): Delete.
18543 (rs6000_vectorize_vec_perm_const_ok): Delete.
18544 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18545 reference to the SPE evmerge intructions.
18546 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18547 an element count and rtx array.
18548 (rs6000_expand_extract_even): Update call accordingly.
18549 (rs6000_expand_interleave): Likewise.
18550 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18551 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18553 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18557 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18558 vector mode and that that mode matches the mode of the data
18560 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18561 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18562 directly using expand_vec_perm_1 when forcing selectors into
18564 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18568 * optabs-query.h (can_vec_perm_p): Delete.
18569 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18570 * optabs-query.c (can_vec_perm_p): Split into...
18571 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18572 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18573 particular selector is valid.
18574 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18575 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18576 (vect_grouped_load_supported): Likewise.
18577 (vect_shift_permute_load_chain): Likewise.
18578 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18579 (vect_transform_slp_perm_load): Likewise.
18580 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18581 (vectorizable_bswap): Likewise.
18582 (vect_gen_perm_mask_checked): Likewise.
18583 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18584 implementations of variable permutation vectors into account
18585 when deciding which selector to use.
18586 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18587 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18588 with a false third argument.
18589 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18590 to test whether the constant selector is valid and can_vec_perm_var_p
18591 to test whether a variable selector is valid.
18593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18595 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18596 * optabs-query.c (can_vec_perm_p): Likewise.
18597 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18598 instead of vec_perm_indices.
18599 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18600 (vect_gen_perm_mask_checked): Likewise,
18601 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18602 (vect_gen_perm_mask_checked): Likewise,
18604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18606 * optabs-query.h (qimode_for_vec_perm): Declare.
18607 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18608 (qimode_for_vec_perm): ...this new function.
18609 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18611 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18613 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18614 does not have a conditional at the top.
18616 2018-01-02 Richard Biener <rguenther@suse.de>
18618 * ipa-inline.c (big_speedup_p): Fix expression.
18620 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18623 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18626 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18630 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18631 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18632 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18633 cond_taken_branch_cost 3->4.
18635 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18637 PR tree-optimization/83581
18638 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18639 TODO_cleanup_cfg if any changes have been made.
18641 PR middle-end/83608
18642 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18643 convert_modes if target mode has the right side, but different mode
18646 PR middle-end/83609
18647 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18648 last argument when extracting from CONCAT. If either from_real or
18649 from_imag is NULL, use expansion through memory. If result is not
18650 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18651 the parts directly to inner mode, if even that fails, use expansion
18654 PR middle-end/83623
18655 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18656 check for bswap in mode rather than HImode and use that in expand_unop
18659 Copyright (C) 2018 Free Software Foundation, Inc.
18661 Copying and distribution of this file, with or without modification,
18662 are permitted in any medium without royalty provided the copyright
18663 notice and this notice are preserved.