1 2018-07-02 Martin Liska <mliska@suse.cz>
4 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5 (propagate_nothrow): Likewise.
7 2018-07-02 Martin Liska <mliska@suse.cz>
10 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
12 2018-07-02 David Malcolm <dmalcolm@redhat.com>
14 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
15 function in r262149, changing "loc" param from source_location to
16 const dump_location_t &.
17 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
18 declaration, as above.
20 2018-07-01 Paul Koning <ni1d@arrl.net>
22 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
23 -munit-asm, -mgnu-asm, -mdec-asm.
24 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
25 (pdp11_output_labelref): New.
26 (pdp11_output_def): New.
27 (pdp11_output_addr_vec_elt): New.
28 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
29 %# and %@ format codes.
30 (pdp11_option_override): New.
31 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
32 (pdp11_output_ident): New.
33 (pdp11_asm_named_section): New.
34 (pdp11_asm_init_sections): New.
35 (pdp11_file_start): New.
36 (pdp11_file_end): New.
37 (output_ascii): Use .ascii/.asciz for -mdec-asm.
38 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
39 %o, like %c but octal.
40 (pdp11_option_override): New.
41 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
43 (DATA_SECTION_ASM_OP): Ditto.
44 (READONLY_DATA_SECTION_ASM_OP): New.
45 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
46 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
47 (ASM_OUTPUT_LABELREF): Ditto.
48 (ASM_OUTPUT_DEF): Ditto.
49 (ASM_OUTPUT_EXTERNAL): New.
50 (ASM_OUTPUT_SOURCE_FILENAME): New.
51 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
52 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
53 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
54 %# and %@ format codes.
55 * config/pdp11/pdp11.opt (mgnu-asm): New.
56 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
57 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
58 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
60 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
62 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
63 dereferencing path[] beyond its length.
65 (debug_all_paths): New.
66 (rewire_first_differing_edge): New.
67 (adjust_paths_after_duplication): New.
68 (duplicate_thread_path): Call adjust_paths_after_duplication.
70 (thread_through_all_blocks): Add new argument to
71 duplicate_thread_path.
73 2018-06-30 Jim Wilson <jimw@sifive.com>
75 * config/riscv/predicates.md (p2m1_shift_operand): New.
76 (high_mask_shift_operand): New.
77 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
78 pattern using p2m1_shift_operand.
79 (lshsi3_zero_extend_3+2): New combiner pattern using
80 high_mask_shift_operand.
82 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
84 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
86 (vect_recog_rotate_pattern): ...here.
87 (vect_convert_input): Try to insert casts of invariants in the
89 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
90 preheader to be empty.
92 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
94 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
95 vector type. If given, install it in the new statement's
97 (vect_element_precision): New function.
98 (vect_unpromoted_value): New struct.
99 (vect_unpromoted_value::vect_unpromoted_value): New function.
100 (vect_unpromoted_value::set_op): Likewise.
101 (vect_look_through_possible_promotion): Likewise.
102 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
103 (vect_widened_op_tree, vect_convert_input): Likewise.
104 (vect_convert_inputs, vect_convert_output): Likewise.
105 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
106 to handle the optional cast of the multiplication result and
107 vect_widened_op_tree to detect the widened multiplication itself.
108 Do not require the input and output of promotion casts to have
109 the same sign, but base the signedness of the operation on the
110 input rather than the result. If the pattern includes two
111 promotions, check that those promotions have the same sign.
112 Do not restrict the MULT_EXPR handling to a double-width result;
113 handle quadruple-width results and wider. Use vect_convert_inputs
114 to convert the inputs to the common type.
115 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
116 to handle the optional cast of the ABS result. Also allow a sign
117 change or a sign extension between the ABS and MINUS.
118 Use vect_widened_op_tree to detect the widened subtraction and use
119 vect_convert_inputs to convert the inputs to the common type.
120 (vect_handle_widen_op_by_const): Delete.
121 (vect_recog_widen_op_pattern): New function.
122 (vect_recog_widen_mult_pattern): Use it.
123 (vect_recog_widen_shift_pattern): Likewise.
124 (vect_recog_widen_sum_pattern): Use
125 vect_look_through_possible_promotion to handle the promoted
128 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
130 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
131 the containing gimple_seq *.
132 * gimple-iterator.h (gsi_for_stmt): Declare it.
133 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
134 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
135 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
136 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
137 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
139 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
141 (vect_mark_pattern_stmts): ...here. Handle cases in which the
142 statement being replaced is part of an existing pattern
143 definition sequence, inserting the new pattern statements before
145 (vect_pattern_recog_1): Don't return a bool. If the statement
146 is already part of a pattern, instead apply pattern matching
147 to the pattern definition statements. Don't clear the
148 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
149 (vect_pattern_recog): Don't break after the first match;
150 continue processing the pattern definition statements instead.
151 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
153 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
155 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
156 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
157 (vect_recog_widen_sum_pattern): Use it.
159 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
161 * tree-vect-loop.c (vectorizable_reduction): Assert that the
162 phi is not a pattern statement and has not been replaced by
164 * tree-vect-patterns.c (type_conversion_p): Don't check
165 STMT_VINFO_IN_PATTERN_P.
166 (vect_recog_vector_vector_shift_pattern): Likewise.
167 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
168 the pattern statement rather than the original statement; check
169 directly for a WIDEN_MULT_EXPR here.
170 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
171 vect_is_simple_use to return the pattern statement rather
172 than the original statement; use is_pattern_stmt_p to check
173 for such a pattern statement.
174 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
175 to return the pattern statement rather than the original statement;
176 don't do the same transformation here.
177 (vect_is_simple_use): If the defining statement has been replaced
178 by a pattern statement, return the pattern statement instead.
179 Remove the corresponding (local) transformation from the vectype
182 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
184 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
185 end and default to null.
186 * tree-vect-loop.c (vect_create_epilog_for_reduction)
187 (vectorizable_reduction): Update calls accordingly, dropping the
188 gimple ** argument if the passed-back statement isn't needed.
189 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
190 (vect_recog_rotate_pattern): Likewise.
191 (vect_recog_mask_conversion_pattern): Likewise.
192 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
193 (vect_mask_constant_operand_p): Likewise.
194 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
195 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
196 (get_group_load_store_type, get_load_store_type): Likewise.
197 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
198 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
199 (vectorizable_conversion, vectorizable_assignment): Likewise.
200 (vectorizable_shift, vectorizable_operation): Likewise.
201 (vectorizable_store, vect_is_simple_cond): Likewise.
202 (vectorizable_condition, vectorizable_comparison): Likewise.
203 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
204 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
205 and move it to the end. Cope with null def_stmt_outs.
207 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
209 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
211 2018-06-29 Jeff Law <law@redhat.com>
213 * config/v850/v850.c (v850_legitimate_address_p): Handle large
214 displacements for TARGET_V850E2V3 and newer.
215 (TARGET_LRA_P): Remove. Defaults to LRA now.
216 * config/v850/v850.md (sign23byte_load): Remove.
217 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
218 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
220 2018-06-29 Martin Liska <mliska@suse.cz>
223 * coverage.c (coverage_init): Mangle full path name.
224 * doc/invoke.texi: Document the change.
225 * gcov-io.c (mangle_path): New.
226 * gcov-io.h (mangle_path): Likewise.
227 * gcov.c (mangle_name): Use mangle_path for path mangling.
229 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
231 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
232 if starting source register is not even.
234 2018-06-29 Martin Liska <mliska@suse.cz>
236 PR tree-optimization/86263
237 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
238 Make edge redirection.
240 2018-06-29 David Malcolm <dmalcolm@redhat.com>
242 * dumpfile.c (dump_loc): Add indentation based on scope depth.
243 (dump_scope_depth): New variable.
244 (get_dump_scope_depth): New function.
245 (dump_begin_scope): New function.
246 (dump_end_scope): New function.
247 * dumpfile.h (get_dump_scope_depth): New declaration.
248 (dump_begin_scope): New declaration.
249 (dump_end_scope): New declaration.
250 (class auto_dump_scope): New class.
251 (AUTO_DUMP_SCOPE): New macro.
252 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
255 2018-06-29 Richard Biener <rguenther@suse.de>
257 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
258 compute_all_dependences succeeds.
259 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
260 exceed --param loop-max-datarefs-for-datadeps.
262 2018-06-29 Jakub Jelinek <jakub@redhat.com>
264 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
266 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
269 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
270 alternative 0 in preferred_for_speed attribute.
272 2018-06-28 Paul Koning <ni1d@arrl.net>
274 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
275 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
276 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
277 * config/pdp11/pdp11.md: Correct "length" attribute calculation
278 for shift insn patterns.
280 2018-06-28 David Malcolm <dmalcolm@redhat.com>
282 * cgraph.c (cgraph_node::get_body): Replace assignments to
283 "dump_file" with calls to set_dump_file.
284 * dumpfile.c (alt_dump_file): Make static, and group with...
285 (alt_flags): ...this definition.
286 (dumps_are_enabled): New variable.
287 (refresh_dumps_are_enabled): New function.
288 (set_dump_file): New function.
289 (set_alt_dump_file): New function.
290 (gcc::dump_manager::dump_start): Replace assignments to
291 "dump_file" and "alt_dump_file" with calls to set_dump_file and
293 (gcc::dump_manager::dump_finish): Likewise.
294 * dumpfile.h (alt_dump_file): Delete decl.
295 (dumps_are_enabled): New variable decl.
296 (set_dump_file): New function decl.
297 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
299 * tree-nested.c (lower_nested_functions): Replace assignments to
300 "dump_file" with calls to set_dump_file.
302 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
304 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
305 goto_locus of each outgoing edge of each basic block.
307 2018-06-28 Richard Biener <rguenther@suse.de>
309 * dwarf2out.c (decl_scope_table): Remove.
310 (push_decl_scope): Likewise.
311 (pop_decl_scope): Likewise.
312 (gen_type_die_for_member): Do not call push/pop_decl_scope.
313 (gen_struct_or_union_type_die): Likewise.
314 (gen_tagged_type_die): Likewise.
315 (dwarf2out_init): Do not initialize decl_scope_table.
316 (dwarf2out_c_finalize): Do not free it.
318 2018-06-28 Richard Biener <rguenther@suse.de>
320 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
321 deciding whether to not re-use a DIE.
323 2018-06-28 Richard Biener <rguenther@suse.de>
325 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
326 DW_AT_abstract_origin attribute.
328 2018-06-28 Martin Liska <mliska@suse.cz>
330 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
331 Use newly introduced constants.
332 * tree-switch-conversion.h (struct jump_table_cluster):
333 Define max_ratio_for_size and max_ratio_for_speed.
335 2018-06-28 Martin Liska <mliska@suse.cz>
337 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
338 Add new checking assert to catch invalid state.
339 (jump_table_cluster::can_be_handled): Handle single case
341 (jump_table_cluster::is_beneficial): Bail out for such case.
342 (bit_test_cluster::find_bit_tests):
343 Add new checking assert to catch invalid state.
344 (bit_test_cluster::can_be_handled): Handle single case
346 (bit_test_cluster::is_beneficial): Bail out for such case.
347 (switch_decision_tree::analyze_switch_statement):
350 2018-06-28 Martin Liska <mliska@suse.cz>
352 * common.opt: Introduce -completion option.
353 * gcc.c (driver_handle_option): Handle it.
354 (driver::main): Print completions if completion
356 * opt-suggestions.c (option_proposer::get_completions):
358 (option_proposer::suggest_completion): Likewise.
359 (option_proposer::find_param_completions): Likewise.
360 (verify_autocompletions): Likewise.
361 (test_completion_valid_options): Likewise.
362 (test_completion_valid_params): Likewise.
363 (in_completion_p): Likewise.
364 (empty_completion_p): Likewise.
365 (test_completion_partial_match): Likewise.
366 (test_completion_garbage): Likewise.
367 (opt_proposer_c_tests): Likewise.
368 * opt-suggestions.h: Declare new functions.
369 * opts.c (common_handle_option): Handle OPT__completion_.
370 * selftest-run-tests.c (selftest::run_tests): Add
371 opt_proposer_c_tests.
372 * selftest.c (assert_str_startswith): New.
373 * selftest.h (assert_str_startswith): Likewise.
374 (opt_proposer_c_tests): New.
375 (ASSERT_STR_STARTSWITH): Likewise.
377 2018-06-28 Martin Liska <mliska@suse.cz>
379 * Makefile.in: Add opt-suggestions.o.
380 * gcc-main.c: Include opt-suggestions.h.
381 * gcc.c (driver::driver): Likewise.
382 (driver::~driver): Remove m_option_suggestions.
383 (driver::build_option_suggestions): Moved to option_proposer.
384 (driver::suggest_option): Likewise.
385 (driver::handle_unrecognized_options): Use option_proposer.
386 * gcc.h (class driver): Add new memver m_option_proposer.
387 * opt-suggestions.c: New file.
388 * opt-suggestions.h: New file.
390 2018-06-28 Martin Liska <mliska@suse.cz>
392 * vec.h (class auto_string_vec): New (moved from auto_argvec).
393 (auto_string_vec::~auto_string_vec): Likewise.
395 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
397 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
398 statement if id->reset_location is true.
399 (copy_edges_for_bb): Do not set goto_locus on the new edges if
400 id->reset_location is true.
401 (copy_phis_for_bb): Force input_location on the arguments if
402 id->reset_location is true.
403 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
404 is set on the function to be inlined.
405 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
406 prevent_decl_creation_for_types fields up and add reset_location field.
408 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
410 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
412 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
414 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
416 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
417 hard registers for the clobbered pseudo.
419 2018-06-27 Paul Koning <ni1d@arrl.net>
421 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
422 mutually exclusive options.
423 * config/pdp11/constraints.md (h): New constraint.
424 (O): Update definition to match shift code generation.
426 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
428 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
430 (output_jump): Change arguments.
431 (pdp11_fixed_cc_regs): New function.
432 (pdp11_cc_mode): Ditto.
433 (pdp11_expand_shift): Ditto.
434 (pdp11_assemble_shift): Ditto.
435 (pdp11_small_shift): Ditto.
436 (pdp11_branch_cost): Remove.
437 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
439 (pdp11_register_move_cost): Update for CC registers.
440 (pdp11_rtx_costs): Add case for LSHIFTRT.
441 (pdp11_output_jump): Add CCNZ mode conditional branches.
442 (notice_update_cc_on_set): Remove.
443 (pdp11_cc_mode): New function.
444 (simple_memory_operand): Correct pre/post decrement case.
445 (no_side_effect_operand): New function.
446 (pdp11_regno_reg_class): Add CC_REGS class.
447 (pdp11_fixed_cc_regs): New function.
448 (pdp11_small_shift): New function.
449 (pdp11_expand_shift): New function to expand shift insns.
450 (pdp11_assemble_shift): New function to output shifts.
451 (pdp11_branch_cost): Remove.
452 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
453 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
455 (PTRDIFF_TYPE): Ditto.
456 (ADJUST_INSN_LENGTH): New macro.
457 (FIXED_REGISTERS): Add CC registers.
458 (CALL_USED_REGISTERS): Ditto.
460 (REG_CLASS_NAMES): Ditto.
461 (REG_CLASS_CONTENTS): Ditto.
462 (SELECT_CC_MODE): Use new function.
463 (TARGET_FLAGS_REGNUM): New macro.
464 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
465 (cc0_reg_rtx): Remove.
466 (CC_STATUS_MDEP): Remove.
467 (CC_STATUS_MDEFP_INIT): Remove.
469 (NOTICE_UPDATE_CC): Remove.
470 (REGISTER_NAMES): Add CC registers.
471 (BRANCH_COST): Change to constant 1.
472 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
474 * config/pdp11/pdp11.opt (mbcopy): Remove.
475 (mbcopy-builtin): Remove.
476 (mbranch-cheap): Remove.
477 (mbranch-expensive): Remove.
478 * config/pdp11/predicates.md (expand_shift_operand): Update to
479 match shift code generation.
480 (ccnz_operator): New predicate.
481 * doc/invoke.texi (PDP-11 Options): Remove deleted options
482 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
483 Remove non-existent option -mabshi, -mno-abshi. Document mutually
485 * doc/md.texi (PDP-11): Document new D and h constraints. Update
486 description of O constraint.
488 2018-06-27 Jeff Law <law@redhat.com>
489 Austin Law <austinklaw@gmail.com>
491 * config/v850/v850.md (addsi3_set_flags): New pattern.
492 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
493 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
494 (zero_extendhisi2_v850_set_flags): Likewise.
495 (zero_extendqisi2_v850_set_flags): Likewise.
496 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
497 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
498 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
500 * config/v850/v850-protos.h (notice_update_cc): Remove.
501 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
502 (v850_print_operand): Handle 'D' and "d".
503 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
504 Add handling of arithmetic/logical operations compared against zero.
505 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
506 Do not look at v850_compare_op, instead get mode from last argument.
507 (v850_gen_compare): Remove
508 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
509 after reload for prologue insns.
510 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
512 (construct_save_jarl): Likewise.
513 (TARGET_FLAGS_REGNUM): Define.
514 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
515 (NOTICE_UPDATE_CC): Remove.
516 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
517 than cc0. Conditionalize on reload_completed.
518 (cmpsi_insn, setfcc_insn): Likewise.
519 (tst1 splitter): Turn into define_and_split which sets the flags
521 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
522 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
523 (cstoresf4, cstoredf4): Clobber the flags.
524 (cmpsi, cmpsf, cmpdf): Remove expanders.
525 (setf_insn): Remove pattern.
526 (addsi3): Turn into define_and_split which clobbers the flags after
527 reload and a suitable pattern (addsi3_clobber_flags) for use after
529 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
530 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
531 (ashrsi3, ashrsi3_v850e2): Likewise.
532 (bins): Clobber the flags.
533 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
534 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
535 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
536 (call_value_internal_short, call_value_internal_long): Likewise.
537 (callt_save_interrupt, callt_return_interrupt): Likewise.
538 (save_interrupt, return_interrupt): Likewise.
539 (callt_save_all_interrupt, save_all_interrupt): Likewise.
540 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
541 (restore_all_interrupt, _restore_all_interrupt): Likewise.
542 (All FP comparisons): Only allow after reload has completed.
544 (divh, divhu): Tweak output template.
545 (branch_z_normal, branch_z_invert): Remove
546 (branch_nz_normal, branch_nz_invert): Likewise.
547 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
549 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
550 * config/v850/v850.c (notice_update_cc): Remove.
551 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
552 (CC_NO_CARRY): Likewise.
553 (NOTICE_UPDATE_CC): Define to nothing.
554 * config/v850/v850.md: Remove block comment on cc0 handling
555 Remove "cc" attribute from all patterns. Remove cc_status handling
556 from all patterns. Minor formatting fixes.
558 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
560 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
561 (cortex-a76.cortex-a55): Likewise.
562 * config/aarch64/aarch64-tune.md: Regenerate.
563 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
564 cortex-a76.cortex-a55.
566 2018-06-27 Jeff Law <law@redhat.com>
568 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
569 (MULTILIB_DIRNAMES): Similarly.
571 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
573 * gimple.h (gimple_return_retbnd): Delete.
574 (gimple_return_set_retbnd): Likewise.
575 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
576 gimple_return_set_retbnd.
577 * gimple-pretty-print.c (dump_gimple_return): Remove call to
578 gimple_return_retbnd and adjust.
579 * tree-inline.h (struct copy_body_data): Remove retbnd field.
580 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
581 Explicitly return NULL in a couple more cases. Move assertion
582 on debug statements and remove unreachable code.
583 (reset_debug_binding): Do not test id->retbnd.
584 (expand_call_inline): Do not set it.
586 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
588 * configure.ac: Add --disable-gcov option.
589 * configure: Regenerate.
590 * Makefile.in: Honour @enable_gcov@.
591 * doc/install.texi: Document --disable-gcov.
593 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
595 * config/arm/arm-cpus.in (cortex-a76): New entry.
596 (cortex-a76.cortex-a55): Likewise.
597 * config/arm/arm-tables.opt: Regenerate.
598 * config/arm/arm-tune.md: Likewise.
599 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
600 * doc/invoke.texi (ARM Options): Document cortex-a76 and
601 cortex-a76.cortex-a55.
603 2018-06-27 Tamar Christina <tamar.christina@arm.com>
606 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
608 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
610 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
612 (EPILOGUE_USES): Likewise.
614 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
616 * tree-inline.c (remap_location): New function extracted from...
617 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
618 (copy_phis_for_bb): ...here. Call remap_location.
619 (copy_cfg_body): Adjust call to copy_edges_for_bb.
621 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
623 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
624 unaligned vsx for 16B memset.
626 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
629 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
630 ieee128_float_type_node to long_double_type_node unless
631 TARGET_LONG_DOUBLE_128 is set.
633 2018-06-26 David Malcolm <dmalcolm@redhat.com>
635 * cfgloop.c (get_loop_location): Convert return type from
636 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
637 by implicit construction from rtx_insn *, and using
638 dump_user_location_t::from_function_decl for the fallback case.
639 * cfgloop.h (get_loop_location): Convert return type from
640 location_t to dump_user_location_t.
641 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
642 dump_printf_loc to pass in a dump_location_t rather than a
643 location_t, via the gimple stmt.
644 * coverage.c (get_coverage_counts): Update calls to
645 dump_printf_loc to pass in dump_location_t rather than a
647 * doc/optinfo.texi (Dump types): Convert example of
648 dump_printf_loc from taking "locus" to taking "insn". Update
649 description of the "_loc" calls to cover dump_location_t.
650 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
652 (dump_user_location_t::dump_user_location_t): New constructors,
653 from gimple *stmt and rtx_insn *.
654 (dump_user_location_t::from_function_decl): New function.
655 (dump_loc): Make static.
656 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
657 const dump_location_t &.
658 (dump_generic_expr_loc): Delete.
659 (dump_printf_loc): Convert param "loc" from location_t to
660 const dump_location_t &.
661 (selftest::test_impl_location): New function.
662 (selftest::dumpfile_c_tests): New function.
663 * dumpfile.h: Include "profile-count.h".
664 (class dump_user_location_t): New class.
665 (struct dump_impl_location_t): New struct.
666 (class dump_location_t): New class.
667 (dump_printf_loc): Convert 2nd param from source_location to
668 const dump_location_t &.
669 (dump_generic_expr_loc): Delete.
670 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
671 const dump_location_t &.
672 * gimple-fold.c (fold_gimple_assign): Update call to
673 dump_printf_loc to pass in a dump_location_t rather than a
674 location_t, via the gimple stmt.
675 (gimple_fold_call): Likewise.
676 * gimple-loop-interchange.cc
677 (loop_cand::analyze_iloop_reduction_var): Update for change to
678 check_reduction_path.
679 (tree_loop_interchange::interchange): Update for change to
681 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
682 change in return-type of find_loop_location.
683 (graphite_regenerate_ast_isl): Likewise.
684 * graphite-optimize-isl.c (optimize_isl): Likewise.
685 * graphite.c (graphite_transform_loops): Likewise.
686 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
687 pass in a dump_location_t rather than a location_t, via the
689 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
690 * ipa.c (walk_polymorphic_call_targets): Likewise.
691 * loop-unroll.c (report_unroll): Convert "locus" param from
692 location_t to dump_location_t.
693 (decide_unrolling): Update for change to get_loop_location's
695 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
696 location_t to dump_user_location_t.
697 (grid_find_single_omp_among_assignments_1): Updates calls to
698 dump_printf_loc to pass in a dump_location_t rather than a
699 location_t, via the gimple stmt.
700 (grid_parallel_clauses_gridifiable): Convert "tloc" from
701 location_t to dump_location_t. Updates calls to dump_printf_loc
702 to pass in a dump_location_t rather than a location_t, via the
704 (grid_inner_loop_gridifiable_p): Likewise.
705 (grid_dist_follows_simple_pattern): Likewise.
706 (grid_gfor_follows_tiling_pattern): Likewise.
707 (grid_target_follows_gridifiable_pattern): Likewise.
708 (grid_attempt_target_gridification): Convert initialization
709 of local "grid" from memset to zero-initialization; FIXME: does
710 this require C++11? Update call to dump_printf_loc to pass in a
711 optinfo_location rather than a location_t, via the gimple stmt.
712 * profile.c (read_profile_edge_counts): Updates call to
713 dump_printf_loc to pass in a dump_location_t rather than a
715 (compute_branch_probabilities): Likewise.
716 * selftest-run-tests.c (selftest::run_tests): Call
718 * selftest.h (dumpfile_c_tests): New decl.
719 * tree-loop-distribution.c (pass_loop_distribution::execute):
720 Update for change in return type of find_loop_location.
721 * tree-parloops.c (parallelize_loops): Likewise.
722 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
723 "locus" from location_t to dump_user_location_t.
724 (canonicalize_loop_induction_variables): Likewise.
725 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
726 for change in return type of find_loop_location.
727 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
728 to dump_printf_loc to pass in a dump_location_t rather than a
729 location_t, via the stmt.
730 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
732 * tree-vect-loop-manip.c (find_loop_location): Convert return
733 type from source_location to dump_user_location_t.
734 (vect_do_peeling): Update for above change.
735 (vect_loop_versioning): Update for change in type of
737 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
738 from location_t to dump_user_location_t.
739 (vect_estimate_min_profitable_iters): Update for change in type
741 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
742 location_t to dump_location_t.
743 (vect_slp_bb): Update for change in type of vect_location.
744 * tree-vectorizer.c (vect_location): Convert from source_location
745 to dump_user_location_t.
746 (try_vectorize_loop_1): Update for change in vect_location's type.
747 (vectorize_loops): Likewise.
748 (increase_alignment): Likewise.
749 * tree-vectorizer.h (vect_location): Convert from source_location
750 to dump_user_location_t.
751 (find_loop_location): Convert return type from source_location to
752 dump_user_location_t.
753 (check_reduction_path): Convert 1st param from location_t to
754 dump_user_location_t.
755 * value-prof.c (check_counter): Update call to dump_printf_loc to
756 pass in a dump_user_location_t rather than a location_t; update
757 call to error_at for change in type of "locus".
758 (check_ic_target): Update call to dump_printf_loc to
759 pass in a dump_user_location_t rather than a location_t, via the
762 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
764 * config/s390/s390.h (enum processor_flags): Do not use
765 default tune parameter when -march was specified.
767 2018-06-26 Jakub Jelinek <jakub@redhat.com>
770 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
771 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
774 2018-06-26 Richard Biener <rguenther@suse.de>
776 PR tree-optimization/86287
778 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
779 (vect_analyze_loop): Initialize n_stmts.
781 2018-06-26 Richard Biener <rguenther@suse.de>
784 * fold-const.c (fold_convertible_p): Pointer extension
787 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
790 * dwarf2out.c (loc_list_has_views): Adjust comments.
791 (dw_loc_list): Split single cross-partition range with
794 2018-06-25 Jeff Law <law@redhat.com>
796 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
797 on -mbig-switch by default.
799 * config/v850/predicates.md (const_float_1_operand): Fix match_code
801 (const_float_0_operand): Remove unused predicate.
802 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
803 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
804 (recipsf2): New expander. Original pattern now called
806 (recipdf2, recipdf2_insn): Similarly.
807 (rsqrtsf2, rsqrtsf2_insn): Similarly
808 (rsqrtdf2, rsqrtdf2_insn): Similarly
810 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
812 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
813 Simplify logic for FreeBSD (twice).
815 2018-06-25 Martin Sebor <msebor@redhat.com>
817 PR tree-optimization/86204
818 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
819 a strnlen result if it's less than the length of the string.
821 2018-06-25 Martin Sebor <msebor@redhat.com>
823 PR tree-optimization/85700
824 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
825 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
826 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
828 2018-06-25 Martin Sebor <msebor@redhat.com>
830 * doc/extend.texi (Zero-length arrays): Update and clarify.
832 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
834 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
835 added IEEE/IBM long double multilib support on PowerPC little
836 endian Linux systems.
837 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
838 (MULTILIB_DEFAULTS): Likewise.
839 * config/rs6000/rs6000.c (rs6000_option_override_internal):
841 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
842 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
843 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
845 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
848 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
849 (REORDER_45): Likewise.
851 2018-06-25 Jeff Law <law@redhat.com>
853 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
854 dividend to 32 bits. Adjust length.
855 (udivmodhi4): Cleanup output template. Fix length.
857 2018-06-25 Carl Love <cel@us.ibm.com>
859 * config/rs6000/vsx.md: Change word selector to prefered location.
861 2018-06-25 Richard Biener <rguenther@suse.de>
863 PR tree-optimization/86304
864 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
865 epilogue-if-converted loops as well.
867 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
869 * lto-section-out.c (lto_begin_section): Do not print section
870 name for noaddr and unnumbered dumps.
872 2018-06-25 Richard Biener <rguenther@suse.de>
874 * tree-vectorizer.h (struct vec_info_shared): New structure
875 with parts split out from struct vec_info and loop_nest from
876 struct _loop_vec_info.
877 (struct vec_info): Adjust accordingly.
878 (struct _loop_vec_info): Likewise.
879 (LOOP_VINFO_LOOP_NEST): Adjust.
880 (LOOP_VINFO_DATAREFS): Likewise.
881 (LOOP_VINFO_DDRS): Likewise.
882 (struct _bb_vec_info): Likewise.
883 (BB_VINFO_DATAREFS): Likewise.
884 (BB_VINFO_DDRS): Likewise.
885 (struct _stmt_vec_info): Add dr_aux member.
886 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
887 (DR_MISALIGNMENT_UNINITIALIZED): New.
888 (set_dr_misalignment): Adjust.
889 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
890 (vect_analyze_loop): Adjust prototype.
891 (vect_analyze_loop_form): Likewise.
892 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
893 Compute dependences lazily.
894 (vect_record_base_alignments): Use shared datarefs/ddrs.
895 (vect_verify_datarefs_alignment): Likewise.
896 (vect_analyze_data_refs_alignment): Likewise.
897 (vect_analyze_data_ref_accesses): Likewise.
898 (vect_analyze_data_refs): Likewise.
899 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
900 constructor parameter for shared part.
901 (vect_analyze_loop_form): Pass in shared part and adjust.
902 (vect_analyze_loop_2): Pass in storage for the number of
903 stmts. Move loop nest finding to the caller. Compute
905 (vect_analyze_loop): Pass in shared part.
906 (vect_transform_loop): Verify shared datarefs are unchanged.
907 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
908 constructor parameter for shared part.
909 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
910 (vect_slp_bb): Verify shared datarefs are unchanged before
912 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
914 (new_stmt_vec_info): Initialize DR_AUX misalignment to
915 DR_MISALIGNMENT_UNINITIALIZED.
916 * tree-vectorizer.c (vec_info::vec_info): Add constructor
917 parameter for shared part.
918 (vec_info::~vec_info): Adjust.
919 (vec_info_shared::vec_info_shared): New.
920 (vec_info_shared::~vec_info_shared): Likewise.
921 (vec_info_shared::save_datarefs): Likewise.
922 (vec_info_shared::check_datarefs): Likewise.
923 (try_vectorize_loop_1): Construct shared part live for analyses
924 of a single loop for multiple vector sizes.
925 * tree-parloops.c (gather_scalar_reductions): Adjust.
927 2018-06-25 Richard Biener <rguenther@suse.de>
929 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
930 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
931 (vect_analyze_data_refs): Remove similar code from here and
932 simplify accordingly.
934 2018-06-25 Richard Biener <rguenther@suse.de>
936 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
937 for reverse storage order accesses rather than asserting
938 they cannot happen here.
940 2018-06-25 Tom de Vries <tdevries@suse.de>
943 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
944 Use data16 instead of .byte for insn prefix.
946 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
949 * parser.c (make_char_string_pack): Pass this literal chars
950 through cpp_interpret_string.
951 (cp_parser_userdef_numeric_literal): Check the result of
952 make_char_string_pack.
954 2018-06-24 Maya Rashish <coypu@sdf.org>
956 * ginclude/stddef.h: Simplify conditions around avoiding
957 re-definition of __size_t.
959 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
961 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
962 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
964 2018-06-22 Maya Rashish <coypu@sdf.org>
966 * doc/invoke.texi (mno-fancy-math-387): Update for changes
967 made to OpenBSD and NetBSD through the years.
969 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
971 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
972 behavior of vec_pack (vector double, vector double) to match
973 behavior of vec_float2 (vector double, vector double).
975 2018-06-22 Olivier Hainque <hainque@adacore.com>
977 * gimplify.c (gimplify_function_tree): Prevent creation
978 of a trampoline for the address of the current function
979 passed to entry/exit instrumentation hooks.
981 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
984 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
987 2018-06-22 Martin Liska <mliska@suse.cz>
989 PR tree-optimization/86263
990 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
991 Bail out if is_enabled is false.
992 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
994 (jump_table_cluster::is_enabled): New function.
996 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
998 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
999 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1000 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1001 (lto_input_ts_binfo_tree_pointers): Likewise.
1002 * tree-streamer-out.c (streamer_write_tree_bitfields,
1003 write_ts_binfo_tree_pointers): Likewise.
1004 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1006 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1008 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1010 2018-06-22 Martin Liska <mliska@suse.cz>
1012 * symbol-summary.h (get): Make it pure and inline move
1013 functionality from ::get function.
1014 (get): Remove and inline into ::get and ::get_create.
1015 (get_create): Move code from ::get function.
1017 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1020 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1021 -x assembler-with-cpp.
1023 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1025 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1026 _FILE_OFFSET_BITS=64 for C++.
1028 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1030 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1031 conversion insn that shows up when pr85657-3.c is compiled using
1032 IEEE 128-bit long double.
1033 (neg<mode>2_internal): Use the correct mode to check whether the
1034 mode is IBM extended.
1035 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1036 multiply and divide external functions from being created more
1039 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1041 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1043 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1044 the edge can be forwarded.
1045 (cfg_layout_merge_blocks): Likewise.
1047 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1049 * except.c (finish_eh_generation): Commit edge insertions only after
1050 the EH edges have been redirected from post-landing to landing pads.
1052 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1054 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1055 create_tmp_var_for to create the FRAME decl.
1056 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1058 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1060 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1061 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1062 debug statement when resetting its value.
1063 (expand_call_inline): Copy the locus of the call onto the assignment
1064 of the return value, if any. Use local variable in more cases.
1066 2018-06-21 Martin Liska <mliska@suse.cz>
1068 * ipa-pure-const.c (propagate_nothrow): Use
1069 funct_state_summaries->get.
1070 (dump_malloc_lattice): Likewise.
1071 (propagate_malloc): Likewise.
1073 2018-06-21 Richard Biener <rguenther@suse.de>
1075 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1076 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1077 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1079 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1080 BLOCK_ABSTRACT_ORIGIN unconditionally.
1082 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1084 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1086 * ipa-reference.c (ipa_reference_c_finalize): Delete
1087 ipa_ref_opt_sum_summaries and set it to NULL.
1089 2018-06-21 Tom de Vries <tdevries@suse.de>
1091 PR tree-optimization/85859
1092 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1093 test with comment from bb_no_side_effects_p.
1095 2018-06-21 Richard Biener <rguenther@suse.de>
1097 PR tree-optimization/86232
1098 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1099 max for constant niter.
1101 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1103 * config/aarch64/aarch64-simd.md
1104 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1106 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1108 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1109 Make opernads of the unspec commutative.
1111 2018-06-21 Richard Biener <rguenther@suse.de>
1113 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1114 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1115 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1116 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1118 (vect_record_base_alignments): Do not record base alignment
1119 for gathers or scatters.
1120 (vect_compute_data_ref_alignment): Drop return value that is always
1121 true. Bail out early for gathers or scatters.
1122 (vect_enhance_data_refs_alignment): Bail out early for gathers
1124 (vect_find_same_alignment_drs): Likewise.
1125 (vect_analyze_data_refs_alignment): Remove dead code.
1126 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1127 (vect_analyze_data_refs): For possible gathers or scatters do
1128 not create an alternate DR, just check their possible validity
1129 and mark them. Adjust DECL_NONALIASED handling to not rely
1131 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1132 update inits of gathers or scatters.
1133 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1134 Also copy gather/scatter flag to pattern vinfo.
1136 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1138 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1139 behavior of vec_packsu (vector unsigned long long, vector unsigned
1140 long long) to match behavior of vec_packs with same signature.
1142 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1143 Thomas Schwinge <thomas@codesourcery.com>
1144 Cesar Philippidis <cesar@codesourcery.com>
1146 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1147 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1148 (gimplify_adjust_omp_clauses): Likewise.
1149 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1150 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1151 (gimplify_omp_target_update): Update handling of acc update and
1153 * omp-low.c (install_var_field): Remove unused parameter
1154 base_pointers_restrict.
1155 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1156 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1158 (omp_target_base_pointers_restrict_p): Delete.
1159 (scan_omp_target): Update call to scan_sharing_clauses.
1160 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1162 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1163 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1164 (convert_local_omp_clauses): Likewise.
1165 * tree-pretty-print.c (dump_omp_clause): Likewise.
1166 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1168 (omp_clause_code_name): Likewise.
1170 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1173 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1176 PR tree-optimization/86231
1177 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1178 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1180 2018-06-20 Tom de Vries <tdevries@suse.de>
1182 PR tree-optimization/86097
1183 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1184 iv type if signedness of iv type is not the same as that of *nit.
1186 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1188 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1189 EDGE_EH edges, verify they are all EDGE_EH.
1191 2018-06-20 Maya Rashish <coypu@sdf.org>
1193 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1195 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1197 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1198 * config/aarch64/aarch64.c (xgene1_tunings): Add
1199 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1200 (aarch64_mode_valid_for_sched_fusion_p):
1201 Allow 16-byte modes.
1202 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1203 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1205 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1207 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1208 * config/aarch64/iterators.md (VQ2): New mode iterator.
1210 2018-06-20 Martin Liska <mliska@suse.cz>
1212 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1213 Change default ratio from 10 to 8.
1215 2018-06-20 Martin Liska <mliska@suse.cz>
1217 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1219 (bit_test_cluster::find_bit_tests): Likewise.
1220 (switch_decision_tree::analyze_switch_statement): Find clusters.
1221 * tree-switch-conversion.h (struct jump_table_cluster): Document
1224 2018-06-20 Martin Liska <mliska@suse.cz>
1226 * tree-switch-conversion.c (switch_conversion::collect):
1227 Record m_uniq property.
1228 (switch_conversion::expand): Bail out for special conditions.
1229 (group_cluster::~group_cluster): New.
1230 (group_cluster::group_cluster): Likewise.
1231 (group_cluster::dump): Likewise.
1232 (jump_table_cluster::emit): New.
1233 (switch_decision_tree::fix_phi_operands_for_edges): New.
1234 (struct case_node): Remove struct.
1235 (jump_table_cluster::can_be_handled): New.
1236 (case_values_threshold): Moved to header.
1237 (reset_out_edges_aux): Likewise.
1238 (jump_table_cluster::is_beneficial): New.
1239 (bit_test_cluster::can_be_handled): Likewise.
1240 (add_case_node): Remove.
1241 (bit_test_cluster::is_beneficial): New.
1242 (case_bit_test::cmp): New.
1243 (bit_test_cluster::emit): New.
1244 (expand_switch_as_decision_tree_p): Remove.
1245 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1246 (fix_phi_operands_for_edge): Likewise.
1247 (switch_decision_tree::analyze_switch_statement): New.
1248 (compute_cases_per_edge): Move ...
1249 (switch_decision_tree::compute_cases_per_edge): ... here.
1250 (try_switch_expansion): Likewise.
1251 (switch_decision_tree::try_switch_expansion): Likewise.
1252 (record_phi_operand_mapping): Likewise.
1253 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1254 (emit_case_decision_tree): Likewise.
1255 (switch_decision_tree::emit): Likewise.
1256 (balance_case_nodes): Likewise.
1257 (switch_decision_tree::balance_case_nodes): Likewise.
1258 (dump_case_nodes): Likewise.
1259 (switch_decision_tree::dump_case_nodes): Likewise.
1260 (emit_jump): Likewise.
1261 (switch_decision_tree::emit_jump): Likewise.
1262 (emit_cmp_and_jump_insns): Likewise.
1263 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1264 (emit_case_nodes): Likewise.
1265 (switch_decision_tree::emit_case_nodes): Likewise.
1266 (conditional_probability): Remove.
1267 * tree-switch-conversion.h (enum cluster_type): New.
1269 (struct cluster): Likewise.
1270 (cluster::cluster): Likewise.
1271 (struct simple_cluster): Likewise.
1272 (simple_cluster::simple_cluster): Likewise.
1273 (struct group_cluster): Likewise.
1274 (struct jump_table_cluster): Likewise.
1275 (struct bit_test_cluster): Likewise.
1276 (struct min_cluster_item): Likewise.
1277 (struct case_tree_node): Likewise.
1278 (case_tree_node::case_tree_node): Likewise.
1279 (jump_table_cluster::case_values_threshold): Likewise.
1280 (struct case_bit_test): Likewise.
1281 (struct switch_decision_tree): Likewise.
1282 (struct switch_conversion): Likewise.
1283 (switch_decision_tree::reset_out_edges_aux): Likewise.
1285 2018-06-20 Martin Liska <mliska@suse.cz>
1287 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1288 (hoist_edge_and_branch_if_true): Likewise.
1289 (expand_switch_using_bit_tests_p): Likewise.
1290 (struct case_bit_test): Likewise.
1291 (case_bit_test_cmp): Likewise.
1292 (emit_case_bit_tests): Likewise.
1293 (switch_conversion::switch_conversion): New class.
1294 (struct switch_conv_info): Remove old struct.
1295 (collect_switch_conv_info): More to ...
1296 (switch_conversion::collect): ... this.
1297 (check_range): Likewise.
1298 (switch_conversion::check_range): Likewise.
1299 (check_all_empty_except_final): Likewise.
1300 (switch_conversion::check_all_empty_except_final): Likewise.
1301 (check_final_bb): Likewise.
1302 (switch_conversion::check_final_bb): Likewise.
1303 (create_temp_arrays): Likewise.
1304 (switch_conversion::create_temp_arrays): Likewise.
1305 (free_temp_arrays): Likewise.
1306 (gather_default_values): Likewise.
1307 (switch_conversion::gather_default_values): Likewise.
1308 (build_constructors): Likewise.
1309 (switch_conversion::build_constructors): Likewise.
1310 (constructor_contains_same_values_p): Likewise.
1311 (switch_conversion::contains_same_values_p): Likewise.
1312 (array_value_type): Likewise.
1313 (switch_conversion::array_value_type): Likewise.
1314 (build_one_array): Likewise.
1315 (switch_conversion::build_one_array): Likewise.
1316 (build_arrays): Likewise.
1317 (switch_conversion::build_arrays): Likewise.
1318 (gen_def_assigns): Likewise.
1319 (switch_conversion::gen_def_assigns): Likewise.
1320 (prune_bbs): Likewise.
1321 (switch_conversion::prune_bbs): Likewise.
1322 (fix_phi_nodes): Likewise.
1323 (switch_conversion::fix_phi_nodes): Likewise.
1324 (gen_inbound_check): Likewise.
1325 (switch_conversion::gen_inbound_check): Likewise.
1326 (process_switch): Use the newly created class.
1327 (switch_conversion::expand): New.
1328 (switch_conversion::~switch_conversion): New.
1329 * tree-switch-conversion.h: New file.
1331 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1333 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1334 tree-vect-patterns.c.
1335 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1336 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1337 (vect_recog_sad_pattern): Likewise.
1338 (vect_recog_widen_sum_pattern): Likewise.
1339 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1340 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1341 (vect_recog_rotate_pattern): Likewise.
1342 (vect_recog_mult_pattern): Likewise.
1343 (vect_recog_vector_vector_shift_pattern): Likewise.
1344 (vect_recog_divmod_pattern): Likewise.
1345 (vect_recog_mixed_size_cond_pattern): Likewise.
1346 (vect_recog_bool_pattern): Likewise.
1347 (vect_recog_mask_conversion_pattern): Likewise.
1348 (vect_try_gather_scatter_pattern): Likewise.
1349 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1350 (vect_recog_over_widening_pattern): Likewise.
1351 (vect_recog_gather_scatter_pattern): Likewise.
1352 (vect_recog_func_ptr): Move from tree-vectorizer.h
1353 (vect_vect_recog_func_ptrs): Move further down the file.
1354 (vect_recog_func): Likewise. Remove the third argument.
1355 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1356 (vect_pattern_recog_1): Expect the pattern function to do any
1357 necessary target tests. Also expect it to provide a vector type.
1358 Remove the type_in handling.
1360 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1362 * tree-vect-patterns.c (vect_pattern_detected): New function.
1363 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1364 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1365 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1366 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1367 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1368 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1369 (vect_recog_mask_conversion_pattern)
1370 (vect_try_gather_scatter_pattern): Likewise.
1372 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1374 * tree-vect-patterns.c (vect_get_internal_def): New function.
1375 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1376 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1377 (search_type_for_mask_1): Use it.
1379 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1381 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1382 redundant WIDEN_SUM_EXPR handling.
1383 (vect_recog_sad_pattern): Likewise.
1385 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1387 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1388 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1389 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1390 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1391 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1393 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1395 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1396 use the stmt_vec_info of the original bb statement for the
1397 new zero assignment, even if the call is part of a pattern.
1399 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1401 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1402 that the sequence is attached to the original statement rather
1403 than the pattern statement.
1404 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1405 PATTERN_DEF_SEQ from the original statement rather than
1406 the main pattern statement.
1407 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1408 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1409 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1411 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1413 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1414 definition statements before the early exit for statements that aren't
1416 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1418 (vect_transform_loop): ...here. Process pattern definition
1419 statements without first checking whether the main pattern
1420 statement is live or relevant.
1422 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1424 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1425 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1427 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1429 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1430 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1431 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1432 (expand_block_compare): Change select_block_compare_mode call.
1433 (expand_strncmp_align_check): Use new functions, fix comment.
1434 (emit_final_str_compare_gpr): New function.
1435 (expand_strn_compare): Refactor and clean up code.
1436 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1438 2018-06-19 Tony Reix <tony.reix@atos.com>
1439 Damien Bergamini <damien.bergamini@atos.com>
1440 David Edelsohn <dje.gcc@gmail.com>
1442 * collect2.c (static_obj): New variable.
1443 (static_libs): New variable.
1444 (is_in_list): Uncomment declaration.
1445 (main): Track AIX libraries linked statically.
1446 (is_in_list): Uncomment definition.
1447 (scan_prog_file): Don't add AIX shared libraries initializer
1448 to constructor list if linking statically.
1450 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1452 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1454 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1456 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1458 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1461 2018-06-19 Martin Liska <mliska@suse.cz>
1463 * config/i386/i386.c (ix86_can_inline_p): Do not use
1464 ipa_fn_summaries::get_create.
1465 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1467 (devirtualization_time_bonus): Likewise.
1468 (ipcp_propagate_stage): Likewise.
1469 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1470 (edge_set_predicate): Likewise.
1471 (evaluate_conditions_for_known_args): Likewise.
1472 (evaluate_properties_for_edge): Likewise.
1473 (ipa_call_summary::reset): Tranform to ...
1474 (ipa_call_summary::~ipa_call_summary): ... this.
1475 (ipa_fn_summary::reset): Transform to ...
1476 (ipa_fn_summary::~ipa_fn_summary): ... this.
1477 (ipa_fn_summary_t::remove): Rename to ...
1478 (ipa_fn_summary_t::remove_callees): ... this.
1479 (ipa_fn_summary_t::duplicate): Use placement new
1480 instead of memory copy.
1481 (ipa_call_summary_t::duplicate): Likewise.
1482 (ipa_call_summary_t::remove): Remove.
1483 (dump_ipa_call_summary): Change get_create to get.
1484 (ipa_dump_fn_summary): Dump only when summary exists.
1485 (analyze_function_body): Use symbol_summary::get instead
1487 (compute_fn_summary): Likewise.
1488 (estimate_edge_devirt_benefit): Likewise.
1489 (estimate_edge_size_and_time): Likewise.
1490 (inline_update_callee_summaries): Likewise.
1491 (remap_edge_change_prob): Likewise.
1492 (remap_edge_summaries): Likewise.
1493 (ipa_merge_fn_summary_after_inlining): Likewise.
1494 (write_ipa_call_summary): Likewise.
1495 (ipa_fn_summary_write): Likewise.
1496 (ipa_free_fn_summary): Likewise.
1497 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1498 (struct ipa_call_summary): Likewise.
1499 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1501 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1502 (estimate_size_after_inlining): Likewise.
1503 (estimate_growth): Likewise.
1504 (growth_likely_positive): Likewise.
1505 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1506 (inline_call): Likewise.
1507 * ipa-inline.c (caller_growth_limits): Likewise.
1508 (can_inline_edge_p): Likewise.
1509 (can_inline_edge_by_limits_p): Likewise.
1510 (compute_uninlined_call_time): Likewise.
1511 (compute_inlined_call_time): Likewise.
1512 (want_inline_small_function_p): Likewise.
1513 (edge_badness): Likewise.
1514 (update_caller_keys): Likewise.
1515 (update_callee_keys): Likewise.
1516 (inline_small_functions): Likewise.
1517 (inline_to_all_callers_1): Likewise.
1518 (dump_overall_stats): Likewise.
1519 (early_inline_small_functions): Likewise.
1520 (early_inliner): Likewise.
1521 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1522 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1523 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1524 * ipa-split.c (execute_split_functions): Likewise.
1525 * symbol-summary.h: Likewise.
1526 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1528 2018-06-19 Richard Biener <rguenther@suse.de>
1530 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1531 (vectorize_loops): ... here. Fix dbgcnt handling.
1532 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1534 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1537 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1538 ieee128 argument takes up only one (vector) register, not two (floating
1541 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1543 * gimplify.c (gimplify_init_constructor): Really never clear for an
1544 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1546 2018-06-19 Richard Biener <rguenther@suse.de>
1548 PR tree-optimization/86179
1549 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1550 after failed recognition.
1552 2018-06-18 Martin Sebor <msebor@redhat.com>
1555 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1556 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1557 Handle integer subtraction.
1558 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1559 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1561 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1563 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1564 param from rtx to rtx_insn *.
1565 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1567 (frv_ifcvt_modify_insn): Likwise.
1568 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1569 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1570 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1571 as_a <rtx_insn *> cast to local "unprotected_region" once
1572 it's been established that it's not NULL or pc_rtx.
1573 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1574 param "sethi" from rtx to rtx_insn *.
1575 (nds32_group_float_insns): Likewise for param "insn".
1576 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1578 (vax_output_int_subtract): Likewise.
1579 * config/vax/vax.c (vax_output_int_add): Likewise for param
1581 (vax_output_int_subtract): Likewise.
1582 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1583 (emit_pattern_after): Likewise for param "after".
1584 (emit_insn_after): Likewise.
1585 (emit_jump_insn_after): Likewise.
1586 (emit_call_insn_after): Likewise.
1587 (emit_debug_insn_after): Likewise.
1588 (emit_pattern_before): Likewise for param "before".
1589 (emit_insn_before): Likewise.
1590 (emit_jump_insn_before): Likewise.
1591 * final.c (get_insn_template): Likewise for param "insn", removing
1593 * output.h (get_insn_template): Likewise for 2nd param.
1594 * rtl.h (emit_insn_before): Likewise.
1595 (emit_jump_insn_before): Likewise.
1596 (emit_debug_insn_before_noloc): Likewise.
1597 (emit_insn_after): Likewise.
1598 (emit_jump_insn_after): Likewise.
1599 (emit_call_insn_after): Likewise.
1600 (emit_debug_insn_after): Likewise.
1601 (set_insn_deleted): Likewise for param.
1603 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1606 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1607 floating point modes, so that IFmode is numerically greater than
1608 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1609 to declare the ordering. This prevents IFmode from being
1610 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1611 machine. Include rs6000-modes.h to share the fractional values
1612 between genmodes* and the rest of the compiler.
1616 * config/rs6000/rs6000-modes.h: New file.
1617 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1618 meaning of rs6000_long_double_size so that 126..128 selects an
1619 appropriate 128-bit floating point type.
1620 (rs6000_option_override_internal): Likewise.
1621 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1622 (TARGET_LONG_DOUBLE_128): Change the meaning of
1623 rs6000_long_double_size so that 126..128 selects an appropriate
1624 128-bit floating point type.
1625 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1626 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1627 source and destination to match the standard usage.
1628 (truncifkf2): Likewise.
1629 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1630 ISA 2.07 to use an explicit clobber, instead of passing in a
1632 (copysign<mode>3_soft): Likewise.
1634 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1636 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1637 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1638 (vect_slp_analyze_instance_dependence): Likewise.
1639 (vect_enhance_data_refs_alignment): Likewise.
1640 (vect_analyze_data_refs_alignment): Likewise.
1641 (vect_slp_analyze_and_verify_instance_alignment
1642 (vect_analyze_data_ref_accesses): Likewise.
1643 (vect_prune_runtime_alias_test_list): Likewise.
1644 (vect_analyze_data_refs): Likewise.
1645 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1646 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1647 (vect_analyze_scalar_cycles_1): Likewise.
1648 (vect_get_loop_niters): Likewise.
1649 (vect_analyze_loop_form_1): Likewise.
1650 (vect_update_vf_for_slp): Likewise.
1651 (vect_analyze_loop_operations): Likewise.
1652 (vect_analyze_loop): Likewise.
1653 (vectorizable_induction): Likewise.
1654 (vect_transform_loop): Likewise.
1655 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1656 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1657 (vect_make_slp_decision): Likewise.
1658 (vect_detect_hybrid_slp): Likewise.
1659 (vect_slp_analyze_operations): Likewise.
1660 (vect_slp_bb): Likewise.
1661 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1662 (vectorizable_bswap): Likewise.
1663 (vectorizable_call): Likewise.
1664 (vectorizable_simd_clone_call): Likewise.
1665 (vectorizable_conversion): Likewise.
1666 (vectorizable_assignment): Likewise.
1667 (vectorizable_shift): Likewise.
1668 (vectorizable_operation): Likewise.
1669 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1671 2018-06-18 Martin Sebor <msebor@redhat.com>
1673 PR tree-optimization/81384
1674 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1675 * builtins.c (expand_builtin_strnlen): New function.
1676 (expand_builtin): Call it.
1677 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1678 * builtins.def (BUILT_IN_STRNLEN): New.
1679 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1680 Warn for bounds in excess of maximum object size.
1681 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1682 single-value ranges. Handle strnlen.
1683 (handle_builtin_strlen): Handle strnlen.
1684 (strlen_check_and_optimize_stmt): Same.
1685 * doc/extend.texi (Other Builtins): Document strnlen.
1687 2018-06-18 Maya Rashish <coypu@sdf.org>
1689 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1690 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1691 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1693 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1695 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1697 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1699 * tree.c (escaped_string::escape): Replace cast to char * by
1700 const_cast<char *> (unescaped).
1702 2018-06-18 Nick Clifton <nickc@redhat.com>
1705 * tree.c (escaped_string): New class. Converts an unescaped
1706 string into its escaped equivalent.
1707 (warn_deprecated_use): Use the new class to convert the
1708 deprecation message, if present.
1709 (test_escaped_strings): New self test.
1710 (test_c_tests): Add test_escaped_strings.
1711 * doc/extend.texi (deprecated): Add a note that the
1712 deprecation message is affected by the -fmessage-length
1713 option, and that control characters will be escaped.
1714 (#pragma GCC error): Document this pragma.
1715 (#pragma GCC warning): Likewise.
1716 * doc/invoke.texi (-fmessage-length): Document this option's
1717 effect on the #warning and #error preprocessor directives and
1718 the deprecated attribute.
1720 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1722 * tree.c (decl_value_expr_lookup): Revert latest change.
1723 (decl_value_expr_insert): Likewise.
1725 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1727 * gimplify.c (nonlocal_vlas): Delete.
1728 (nonlocal_vla_vars): Likewise.
1729 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1731 (gimplify_body): Do not create and destroy nonlocal_vlas.
1732 * tree-nested.c: Include diagnostic.h.
1733 (use_pointer_in_frame): Tweak.
1734 (lookup_field_for_decl): Add assertion and declare the transformation.
1735 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1736 internal error when the reference is in a wrong context. Do not
1737 create a debug decl by default.
1738 (note_nonlocal_block_vlas): Delete.
1739 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1740 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1741 create a debug decl by default.
1742 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1743 call is in a wrong context.
1744 (fixup_vla_decls): New function.
1745 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1746 debug variables were created.
1747 * tree.c (decl_value_expr_lookup): Add checking assertion.
1748 (decl_value_expr_insert): Likewise.
1750 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1753 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1754 * tree-scalar-evolution.c (interpret_expr): Likewise.
1755 (expression_expensive_p): Likewise.
1756 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1757 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1758 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1759 (ssa_defined_by_minus_one_stmt_p): New.
1761 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1764 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1765 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1766 * dojump.c (do_jump): Likewise.
1767 * expr.c (expand_expr_real_2): Check operand type's sign.
1768 * fold-const.c (const_unop): Handle ABSU_EXPR.
1769 (fold_abs_const): Likewise.
1770 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1771 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1772 (strip_sign_op_1): Likesise.
1773 * match.pd: Add new pattern to generate ABSU_EXPR.
1774 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1775 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1776 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1777 * tree-inline.c (estimate_operator_cost): Likewise.
1778 * tree-pretty-print.c (dump_generic_node): Likewise.
1779 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1780 * tree.def (ABSU_EXPR): New.
1782 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1785 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1786 documented as preserved for backward compatibility only.
1787 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1789 PR rtl-optimization/86108
1790 * bb-reorder.c (create_forwarder_block): Renamed to ...
1791 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1792 jump from new landing pad to the second part.
1793 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1796 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1799 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1800 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1801 Only call store_expr for halves if the mode is the same.
1804 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1805 Fix up comment formatting.
1807 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1809 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1810 (typed_splay_tree::closure,
1811 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1812 (typed_splay_tree::typed_splay_tree,
1813 typed_splay_tree::operator =): Declared private.
1814 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1815 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1816 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1817 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1818 typed_splay_tree::splay_tree_splay,
1819 typed_splay_tree::splay_tree_foreach_helper,
1820 typed_splay_tree::splay_tree_insert,
1821 typed_splay_tree::splay_tree_remove,
1822 typed_splay_tree::splay_tree_lookup,
1823 typed_splay_tree::splay_tree_predecessor,
1824 typed_splay_tree::splay_tree_successor,
1825 typed_splay_tree::splay_tree_min,
1826 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1827 (typed_splay_tree::root, typed_splay_tree::comp,
1828 typed_splay_tree::delete_key,
1829 typed_splay_tree::delete_value): New data members.
1830 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1831 typed_splay_tree::remove.
1833 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1835 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1836 -mginv and -mno-ginv to the assembler.
1837 * config/mips/mips.opt (-mcrc): New option.
1839 * doc/invoke.text (-mcrc): Document.
1842 2018-06-15 Nick Clifton <nickc@redhat.com>
1845 * tree.c (escaped_string): New class. Converts an unescaped
1846 string into its escaped equivalent.
1847 (warn_deprecated_use): Use the new class to convert the
1848 deprecation message, if present.
1849 (test_escaped_strings): New self test.
1850 (test_c_tests): Add test_escaped_strings.
1851 * doc/extend.texi (deprecated): Add a note that the
1852 deprecation message is affected by the -fmessage-length
1853 option, and that control characters will be escaped.
1854 (#pragma GCC error): Document this pragma.
1855 (#pragma GCC warning): Likewise.
1856 * doc/invoke.texi (-fmessage-length): Document this option's
1857 effect on the #warning and #error preprocessor directives and
1858 the deprecated attribute.
1860 2018-06-15 Richard Biener <rguenther@suse.de>
1862 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1863 here, also noting vector size used.
1864 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1865 size used in MSG_OPTIMIZED_LOCATIONS dump.
1866 (pass_slp_vectorize::execute): Adjust.
1868 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1871 * config/arc/arc.c (arc_return_address_register): Fix
1874 2018-06-15 Richard Biener <rguenther@suse.de>
1877 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1878 leave useless conversion stripping to force_gimple_operand_gsi.
1879 (gimplify_build2): Likewise.
1880 (gimplify_build1): Likewise.
1882 2018-06-15 Richard Biener <rguenther@suse.de>
1885 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1886 before adjusting their block.
1888 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1890 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1891 multilibs for *-*-rtems*.
1892 * config/riscv/t-rtems: New file.
1894 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1897 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1898 unsigned_type_for returns NULL.
1901 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1902 subregs of multi-word pseudos unless the float mode has word size.
1904 2018-06-14 Richard Biener <rguenther@suse.de>
1907 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1908 duplicate and harmful type_hash_canon.
1909 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1911 2018-06-14 Richard Biener <rguenther@suse.de>
1914 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1917 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1919 * config/rtems.h (STDINT_LONG32): Define.
1921 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1922 Prachi Godbole <prachi.godbole@imgtec.com>
1924 * config/mips/mips-cpus.def: Define P6600.
1925 * config/mips/mips-tables.opt: Regenerate.
1926 * config/mips/mips.c (mips_ucbranch_type): New enum.
1927 (mips_rtx_cost_data): Add support for P6600.
1928 (mips_issue_rate): Likewise.
1929 (mips_multipass_dfa_lookahead): Likewise.
1930 (mips_avoid_hazard): Likewise.
1931 (mips_reorg_process_insns): Likewise.
1932 (mips_classify_branch_p6600): New function.
1933 * config/mips/mips.h (TUNE_P6600): New define.
1934 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1935 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1936 * config/mips/mips.md: Include p6600.md.
1937 (processor): Add p6600.
1938 * config/mips/p6600.md: New file.
1939 * doc/invoke.texi: Add p6600 to supported architectures.
1941 2018-06-13 Martin Sebor <msebor@redhat.com>
1943 PR tree-optimization/86114
1944 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1946 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1948 2018-06-13 Richard Biener <rguenther@suse.de>
1950 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1951 Properly set vector type of the intermediate stmt.
1952 * tree-vect-stmts.c (vectorizable_operation): The destination
1953 var always has vectype_out type.
1955 2018-06-13 Jeff Law <law@redhat.com>
1957 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1958 integer 0 for argument to print_rtl_with_bb.
1959 (rl78_reorg): Likewise.
1961 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1963 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1964 from rtx to rtx_insn *.
1965 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1967 (add_sched_insns_for_speculation): Likewise for local "target",
1968 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1969 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1970 from rtx_insn ** to rtx_code_label **.
1971 (reorg_emit_nops): Likewise.
1972 (c6x_reorg): Likewise for local "call_labels".
1973 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1975 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1976 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1977 the loops over LABEL_REFS.
1978 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1980 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1981 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1982 (split_branches): Strengthen local "olabel" from rtx to
1983 rtx_insn *, adding a safe_as_a cast.
1984 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1986 (add_insn_after): Likewise for first two params.
1987 (add_insn_before): Likewise.
1988 (remove_insn): Likewise for param.
1989 (emit_pattern_before_noloc): Likewise for second and third params.
1990 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1991 (emit_call_insn_before_noloc): Likewise.
1992 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1994 (emit_barrier_before): Likewise.
1995 (emit_label_before): Strengthen "label" param from "rtx" to
1996 "rtx_code_label *". Strengthen "before" param from "rtx" to
1998 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2000 (emit_pattern_after_noloc): Likewise.
2001 (emit_insn_after_noloc): Likewise.
2002 (emit_jump_insn_after_noloc): Likewise.
2003 (emit_call_insn_after_noloc): Likewise.
2004 (emit_debug_insn_after_noloc): Likewise.
2005 (emit_barrier_after): Likewise.
2006 (emit_label_after): Likewise for both params.
2007 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2008 "loc" param from "int" to "location_t".
2009 (emit_insn_after_setloc): Likewise.
2010 (emit_jump_insn_after_setloc): Likewise.
2011 (emit_call_insn_after_setloc): Likewise.
2012 (emit_debug_insn_after_setloc): Likewise.
2013 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2014 "loc" param from "int" to "location_t".
2015 (emit_pattern_before): Convert NULL_RTX to NULL.
2016 (emit_insn_before_setloc): Convert "loc" param from "int" to
2018 (emit_jump_insn_before_setloc): Likewise.
2019 (emit_call_insn_before_setloc): Likewise.
2020 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2021 rtx_insn *. Convert "loc" param from "int" to "location_t".
2022 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2023 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2024 Convert 3rd param from "int" to "location_t".
2025 (emit_barrier_before, emit_barrier_after, next_real_insn):
2026 Strengthen param from rtx to rtx_insn *.
2027 (emit_label_before): Strengthen 1st param from "rtx" to
2028 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2030 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2031 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2032 Strengthen 2nd param from "rtx" to "rtx_insn *".
2033 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2034 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2035 Likewise. Convert 3rd param from "int" to "location_t".
2036 (emit_label_after): Strengthen 1st param from "rtx" to
2038 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2040 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2041 from "rtx" to "rtx_insn *".
2043 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2045 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2046 bodies streamed in with -Q.
2047 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2048 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2049 * gimple-streamer-out.c: Include gimple-pretty-print.h
2050 (output_bb): Dump stmts streamed.
2051 * lto-section-out.c: Include print-tree.h
2052 (lto_begin_section): Dump sections created.
2053 (lto_output_decl_index): Dump decl encoded.
2054 * lto-streamer-out.c: Include print-tree.h
2055 (create_output_block): Dump output block created.
2056 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2057 (output_function): Dump function output.
2058 (output_constructor): Dump constructor streamed.
2059 (write_global_stream): Output indexes encoded.
2060 (produce_asm_for_decls): Dump streams encoded.
2061 * lto-streamer.c (streamer_dump_file): New global var.
2062 * lto-streamer.h (streamer_dump_file): Declare.
2063 * passes.c (ipa_write_summaries): Initialize streamer dump.
2064 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2067 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2070 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2071 offsets for register save directives. Emit a second batch of save
2072 directives, if need be, when the function accesses prior frames.
2074 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2076 * config/arc/fpu.md (fmasf4): Force operand to register.
2077 (fnmasf4): Likewise.
2079 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2081 * config/arc/arc-protos.h (arc_pad_return): Remove.
2082 * config/arc/arc.c (machine_function): Remove force_short_suffix
2084 (arc_print_operand): Adjust printing of '&'.
2085 (arc_verify_short): Remove conditional printing of short suffix.
2086 (arc_final_prescan_insn): Remove reference to size_reason.
2087 (pad_return): New function.
2088 (arc_reorg): Call pad_return.
2089 (arc_pad_return): Remove.
2090 (arc_init_machine_status): Remove reference to force_short_suffix.
2091 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2092 (attr length): When attribute iscompact is true force to 2
2093 regardless; in the case of maybe check if we want to force the
2094 instruction to have 4 bytes length.
2095 (nopv): Change it to generate 4 byte long nop as well.
2096 (blockage): New pattern.
2097 (simple_return): Remove call to arc_pad_return.
2098 (p_return_i): Likewise.
2100 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2102 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2104 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2106 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2109 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2111 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2112 for ARC700 and ARCv2.
2114 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2117 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2118 operands[2] instead of operands[1].
2121 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2123 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2124 case, check whether the outer register overlaps an unallocatable
2125 register, not just whether it fits the required class.
2127 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2129 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2131 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2132 (memrefs_conflict_p): Likewise.
2133 (init_alias_analysis): Likewise.
2134 * cfgexpand.c (expand_debug_expr): Likewise.
2135 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2136 * cse.c (fold_rtx): Likewise.
2137 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2138 * expr.c (emit_block_move_hints): Likewise.
2139 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2140 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2141 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2142 (emit_group_store): Likewise.
2143 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2144 to read the PRE/POST_MODIFY increment.
2145 * calls.c (store_one_arg): Use strip_offset.
2146 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2148 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2150 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2151 (simplify_binary_operation_1): Extend CONST_INT handling to
2153 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2154 than a HOST_WIDE_INT.
2155 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2157 (adjust_mems, add_stores): Update accodingly.
2158 (vt_canonicalize_addr): Track polynomial offsets.
2159 (emit_note_insn_var_location): Likewise.
2160 (vt_add_function_parameter): Likewise.
2161 (vt_initialize): Likewise.
2163 2018-06-12 Jeff Law <law@redhat.com>
2165 * config.gcc (alpha*-*-freebsd*): Remove.
2166 * config/alpha/freebsd.h: Remove.
2168 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2171 * spellcheck-tree.c (levenshtein_distance): Rename to...
2172 (get_edit_distance): ...this, and update for underlying renaming.
2173 * spellcheck-tree.h (levenshtein_distance): Rename to...
2174 (get_edit_distance): ...this.
2175 * spellcheck.c (levenshtein_distance): Rename to...
2176 (get_edit_distance): ...this. Convert from Levenshtein distance
2177 to Damerau-Levenshtein distance by supporting transpositions of
2178 adjacent characters. Rename "v1" to "v_next" and "v0" to
2180 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2181 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2182 update for underlying renaming.
2183 (selftest::levenshtein_distance_unit_test): Rename to...
2184 (selftest::test_get_edit_distance_unit): ...this, and update for
2185 underlying renaming.
2186 (selftest::test_find_closest_string): Add example from PR 69968
2187 where transposition helps
2188 (selftest::test_metric_conditions): Update for renaming.
2189 (selftest::test_metric_conditions): Likewise.
2190 (selftest::spellcheck_c_tests): Likewise.
2191 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2192 (get_edit_distance): ...this.
2193 (best_match::consider): Update for renaming.
2195 2018-06-12 Martin Sebor <msebor@redhat.com>
2197 PR tree-optimization/85259
2198 * builtins.c (compute_objsize): Handle constant offsets.
2199 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2200 true iff a warning has been issued.
2201 * gimple.h (gimple_nonartificial_location): New function.
2202 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2203 gimple_nonartificial_location and handle -Wno-system-headers.
2204 (handle_builtin_stxncpy): Same.
2206 2018-06-12 Martin Sebor <msebor@redhat.com>
2209 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2211 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2213 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2214 BUILTIN_VEC_XST entries for pointer to double and long long.
2216 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2219 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2221 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2224 2018-06-12 Martin Liska <mliska@suse.cz>
2226 * doc/options.texi: Document IntegerRange.
2228 2018-06-12 Martin Liska <mliska@suse.cz>
2230 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2231 * opt-functions.awk: Handle Deprecated flag.
2232 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2234 (read_cmdline_option): Report warning for a deprecated option.
2235 * opts.h (struct cl_option): Add new field cl_deprecated.
2236 (CL_ERR_DEPRECATED): New.
2238 2018-06-12 Martin Liska <mliska@suse.cz>
2240 * doc/options.texi: Document Deprecated option flag.
2242 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2244 * config/arc/arc-arch.h (arc_extras): New enum.
2245 (arc_cpu_t):Add field extra.
2246 (arc_cpu_types): Consider the extras.
2247 * config/arc/arc-cpus.def: Add extras info.
2248 * config/arc/arc-opts.h (processor_type): Consider extra field.
2249 * config/arc/arc.c (arc_override_options): Handle extra field.
2251 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2253 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2254 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2256 * config/arc/arc.c (arc_init): Update pic warning.
2257 (irq_range): Update irq range parsing warnings.
2258 (arc_override_options): Update various warning messages.
2259 (arc_handle_aux_attribute): Likewise.
2261 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2263 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2265 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2267 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2268 compiler messages directives.
2270 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2272 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2273 * config/mips/mips-tables.opt: Regenerate.
2274 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2276 * doc/invoke.texi: Document -march=i6500.
2278 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2280 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2281 (i6400_gpmul): Add cpu_unit.
2282 (i6400_gpdiv): Likewise.
2283 (i6400_msa_add_d): Update reservations.
2284 (i6400_msa_int_add) Likewise.
2285 (i6400_msa_short_logic3) Likewise.
2286 (i6400_msa_short_logic2) Likewise.
2287 (i6400_msa_short_logic) Likewise.
2288 (i6400_msa_move) Likewise.
2289 (i6400_msa_cmp) Likewise.
2290 (i6400_msa_short_float2) Likewise.
2291 (i6400_msa_div_d) Likewise.
2292 (i6400_msa_long_logic1) Likewise.
2293 (i6400_msa_long_logic2) Likewise.
2294 (i6400_msa_mult) Likewise.
2295 (i6400_msa_long_float2) Likewise.
2296 (i6400_msa_long_float4) Likewise.
2297 (i6400_msa_long_float5) Likewise.
2298 (i6400_msa_long_float8) Likewise.
2299 (i6400_fpu_fadd): Include frint type.
2300 (i6400_fpu_store): New define_insn_reservation.
2301 (i6400_fpu_load): Likewise.
2302 (i6400_fpu_move): Likewise.
2303 (i6400_fpu_fcmp): Likewise.
2304 (i6400_fpu_fmadd): Likewise.
2305 (i6400_int_mult): Include imul3nc type and update reservation.
2306 (i6400_int_div): Include idiv3 type and update reservation.
2307 (i6400_int_load): Update to check type not move_type.
2308 (i6400_int_store): Likewise.
2309 (i6400_int_prefetch): Set zero latency.
2311 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2313 * gcc.c: Document new %@{...} sequence.
2314 (LINK_COMMAND_SPEC): Use it for the -L switches.
2315 (cpp_unique_options): Use it for the -I switches.
2316 (at_file_argbuf): New global variable.
2317 (in_at_file): Likewise.
2318 (alloc_args): Create at_file_argbuf.
2319 (clear_args): Truncate at_file_argbuf.
2320 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2321 (open_at_file): New function.
2322 (close_at_file): Likewise.
2323 (create_at_file): Delete.
2324 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2327 (validate_switches_from_spec): Deal with %@{...} sequence.
2328 (validate_switches): Likewise.
2329 (driver::finalize): Call clear_args.
2331 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2333 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2335 2018-06-11 Martin Sebor <msebor@redhat.com>
2337 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2338 (Wno-class-memaccess): Add @opindex.
2339 (Wno-templates, Wno-multiple-inheritance): Same.
2340 (Wno-virtual-inheritance, Wno-namespaces): Same.
2341 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2342 (Wno-format-overflow, Wno-format-truncation): Same.
2343 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2344 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2345 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2346 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2347 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2348 (Wno-misspelled-isr): Same.
2350 2018-06-11 Martin Sebor <msebor@redhat.com>
2352 * PR tree-optimization/86083
2353 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2355 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2357 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2359 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2362 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2363 on the correct operand.
2364 (*movdi_internal64): Ditto.
2366 2018-06-11 Martin Liska <mliska@suse.cz>
2368 PR tree-optimization/86089
2369 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2371 2018-06-11 Julia Koval <julia.koval@intel.com>
2373 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2374 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2375 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2377 2018-06-11 Olivier Hainque <hainque@adacore.com>
2379 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2380 for Ada with strict dwarf2.
2382 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2385 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2388 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2390 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2392 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2394 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2397 2018-06-08 Martin Liska <mliska@suse.cz>
2399 * tree-cfg.h (debug_function): Fix argument type to match
2402 2018-06-08 Martin Liska <mliska@suse.cz>
2404 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2405 Remove usage of MPX-related (and removed) fields.
2406 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2408 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2410 * cfg.c (debug): Use TDF_NONE rather than 0.
2411 * cfghooks.c (debug): Likewise.
2412 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2413 (struct dump_option_value_info): Convert to...
2414 (struct kv_pair): ...this template type.
2415 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2417 (optinfo_verbosity_options): Likewise.
2418 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2420 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2421 than int for "optgroup_flags" param.
2422 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2424 (dump_dec): Likewise.
2425 (dump_finish): Use TDF_NONE rather than 0.
2426 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2427 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2428 than 0. Update for change to option_ptr.
2429 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2430 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2431 0. Update for changes to optinfo_verbosity_options and
2433 (opt_info_switch_p): Convert optgroup_flags from int to
2435 (dump_basic_block): Use dump_flags_t rather than int
2436 for "dump_kind" param.
2437 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2438 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2439 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2440 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2441 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2442 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2443 TDF_NONE): Convert from macros to...
2444 (enum dump_flag): ...this new enum.
2445 (dump_flags_t): Update to use enum.
2446 (operator|, operator&, operator~, operator|=, operator&=):
2447 Implement for dump_flags_t.
2448 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2449 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2450 Convert from macros to...
2451 (enum optgroup_flag): ...this new enum.
2452 (optgroup_flags_t): New typedef.
2453 (operator|, operator|=): Implement for optgroup_flags_t.
2454 (struct dump_file_info): Convert field "alt_flags" to
2455 dump_flags_t. Convert field "optgroup_flags" to
2457 (dump_basic_block): Use dump_flags_t rather than int for param.
2458 (dump_generic_expr_loc): Likewise.
2459 (dump_dec): Likewise.
2460 (dump_register): Convert param "optgroup_flags" to
2462 (opt_info_enable_passes): Likewise.
2463 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2465 * gimple-pretty-print.c (debug): Likewise.
2466 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2467 (merged_store_group::apply_stores): Likewise.
2468 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2469 * gimple.c (verify_gimple_pp): Likewise.
2470 * graphite-poly.c (print_pbb_body): Likewise.
2471 * passes.c (pass_manager::register_one_dump_file): Convert
2472 local "optgroup_flags" to optgroup_flags_t.
2473 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2475 (debug_body): Likewise.
2476 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2477 to optgroup_flags_t.
2478 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2480 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2481 (convert_mult_to_fma): Likewise.
2482 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2483 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2484 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2485 "dump_kind" to dump_flags_t.
2487 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2489 * config/rs6000/rs6000.c (min, max): Delete.
2491 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2493 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2494 -mabi=spe and -mabi=no-spe.
2496 2018-06-08 Martin Liska <mliska@suse.cz>
2498 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2499 where we expect an existing summary.
2501 2018-06-08 Martin Liska <mliska@suse.cz>
2503 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2504 * ipa-inline.h (estimate_edge_growth): Likewise.
2506 2018-06-08 Martin Liska <mliska@suse.cz>
2508 * cgraph.c (function_version_hasher::hash): Use
2509 cgraph_node::get_uid ().
2510 (function_version_hasher::equal):
2511 * cgraph.h (cgraph_node::get_uid): New method.
2512 * ipa-inline.c (update_caller_keys): Use
2513 cgraph_node::get_uid ().
2514 (update_callee_keys): Likewise.
2515 * ipa-utils.c (searchc): Likewise.
2516 (ipa_reduced_postorder): Likewise.
2517 * lto-cgraph.c (input_node): Likewise.
2518 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2519 * symbol-summary.h (symtab_insertion): Likewise.
2520 (symtab_removal): Likewise.
2521 (symtab_duplication): Likewise.
2522 * tree-pretty-print.c (dump_function_header): Likewise.
2523 * tree-sra.c (convert_callers_for_node): Likewise.
2525 2018-06-08 Martin Liska <mliska@suse.cz>
2527 * cgraph.c (symbol_table::create_edge): Always assign a new
2529 (symbol_table::free_edge): Do not recycle numbers.
2530 * cgraph.h (cgraph_edge::get): New method.
2531 * symbol-summary.h (symtab_removal): Use it.
2532 (symtab_duplication): Likewise.
2533 (call_summary::hashable_uid): Remove.
2535 2018-06-08 Martin Liska <mliska@suse.cz>
2537 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2538 (initialize_growth_caches): Remove.
2539 (free_growth_caches): Likewise.
2540 (do_estimate_edge_time): Use edge_growth_cache.
2541 (do_estimate_edge_size): Likewise.
2542 (do_estimate_edge_hints): Likewise.
2543 * ipa-inline.c (reset_edge_caches): Likewise.
2544 (recursive_inlining): Likewise.
2545 (inline_small_functions): Likewise.
2546 * ipa-inline.h (initialize_growth_caches): Remove.
2547 (estimate_edge_size): Likewise.
2548 (estimate_edge_time): Likewise.
2549 (estimate_edge_hints): Likewise.
2550 (reset_edge_growth_cache): Likewise.
2551 * symbol-summary.h (call_summary::remove): New method.
2553 2018-06-08 Martin Liska <mliska@suse.cz>
2555 * ipa-cp.c (class edge_clone_summary): New summary.
2556 (grow_edge_clone_vectors): Remove.
2557 (ipcp_edge_duplication_hook): Remove.
2558 (class edge_clone_summary_t): New call_summary class.
2559 (ipcp_edge_removal_hook): Remove.
2560 (edge_clone_summary_t::duplicate): New function.
2561 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2562 (create_specialized_node): Likewise.
2563 (ipcp_driver): Initialize edge_clone_summaries and do not
2566 2018-06-08 Martin Liska <mliska@suse.cz>
2568 * symbol-summary.h (get): New function.
2569 (call_summary::m_initialize_when_cloning): New class member.
2571 2018-06-08 Martin Liska <mliska@suse.cz>
2573 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2574 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2575 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2576 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2577 (remove_cgraph_node_from_order): Use the removed_nodes set.
2578 (do_per_function_toporder): Likwise.
2579 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2580 instead of summary_uid.
2581 (symtab_removal): Likewise.
2582 (symtab_duplication): Likewise.
2584 2018-06-08 Martin Liska <mliska@suse.cz>
2586 * ipa-cp.c (ipcp_store_bits_results): Use
2587 ipcp_transformation_sum.
2588 (ipcp_store_vr_results): Likewise.
2589 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2591 (ipcp_transformation_initialize): ... this.
2592 (ipa_set_node_agg_value_chain):
2593 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2594 (write_ipcp_transformation_info): Likewise.
2595 (read_ipcp_transformation_info): Likewise.
2596 (ipcp_update_bits): Likewise.
2597 (ipcp_update_vr): Likewise.
2598 (ipcp_transform_function): Likewise.
2599 * ipa-prop.h: Rename ipcp_transformation_summary to
2600 ipcp_transformation.
2601 (class ipcp_transformation_t): New function summary.
2602 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2603 (ipa_get_agg_replacements_for_node): Likewise.
2605 2018-06-08 Martin Liska <mliska@suse.cz>
2607 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2609 (class funct_state_summary_t): New function_summary class.
2610 (has_function_state): Remove.
2611 (get_function_state): Likewise.
2612 (set_function_state): Likewise.
2613 (add_new_function): Likewise.
2614 (funct_state_summary_t::insert): New function.
2615 (duplicate_node_data): Remove.
2616 (remove_node_data): Remove.
2617 (funct_state_summary_t::duplicate): New function.
2618 (register_hooks): Create new funct_state_summaries.
2619 (pure_const_generate_summary): Use it.
2620 (pure_const_write_summary): Likewise.
2621 (pure_const_read_summary): Likewise.
2622 (propagate_pure_const): Likewise.
2623 (propagate_nothrow): Likewise.
2624 (dump_malloc_lattice): Likewise.
2625 (propagate_malloc): Likewise.
2626 (execute): Do not register hooks, just remove summary
2628 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2631 2018-06-08 Martin Liska <mliska@suse.cz>
2633 * ipa-reference.c (remove_node_data): Remove.
2634 (duplicate_node_data): Likewise.
2635 (class ipa_ref_var_info_summary_t): New class.
2636 (class ipa_ref_opt_summary_t): Likewise.
2637 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2638 (get_reference_optimization_summary): Use
2639 ipa_ref_opt_sum_summaries.
2640 (set_reference_vars_info): Remove.
2641 (set_reference_optimization_summary): Likewise.
2642 (ipa_init): Create summaries.
2643 (init_function_info): Use function summary.
2644 (ipa_ref_opt_summary_t::duplicate): New function.
2645 (ipa_ref_opt_summary_t::remove): New function.
2646 (get_read_write_all_from_node): Fix GNU coding style.
2647 (propagate): Use function summary.
2648 (write_node_summary_p): Fix GNU coding style.
2649 (stream_out_bitmap): Likewise.
2650 (ipa_reference_read_optimization_summary): Use function summary.
2651 (ipa_reference_c_finalize): Do not release hooks.
2653 2018-06-08 Martin Liska <mliska@suse.cz>
2655 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2656 (analyze_function_body): Extract multiple calls of get_create.
2657 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2658 * ipa-inline.c (recursive_inlining): Use ::get method.
2659 * ipa-inline.h (estimate_edge_growth): Likewise.
2661 2018-06-08 Martin Liska <mliska@suse.cz>
2663 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2665 (hsa_function_summary::hsa_function_summary): Use the new enum
2667 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2668 * hsa-gen.c (hsa_get_host_function): Likewise.
2669 (get_brig_function_name): Likewise.
2670 * ipa-hsa.c (process_hsa_functions): Likewise.
2671 (ipa_hsa_write_summary): Likewise.
2672 * symbol-summary.h (symtab_duplication): Use ::get function/
2673 (get): New function.
2675 2018-06-08 Martin Liska <mliska@suse.cz>
2677 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2679 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2680 (hsa_register_kernel): Likewise.
2681 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2682 * hsa-gen.c (hsa_get_host_function): Likewise.
2683 (get_brig_function_name): Likewise.
2684 (generate_hsa): Likewise.
2685 (pass_gen_hsail::execute): Likewise.
2686 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2687 (devirtualization_time_bonus): Likewise.
2688 (ipcp_propagate_stage): Likewise.
2689 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2690 (edge_set_predicate): Likewise.
2691 (evaluate_conditions_for_known_args): Likewise.
2692 (evaluate_properties_for_edge): Likewise.
2693 (ipa_fn_summary::reset): Likewise.
2694 (ipa_fn_summary_t::duplicate): Likewise.
2695 (dump_ipa_call_summary): Likewise.
2696 (ipa_dump_fn_summary): Likewise.
2697 (analyze_function_body): Likewise.
2698 (compute_fn_summary): Likewise.
2699 (estimate_edge_devirt_benefit): Likewise.
2700 (estimate_edge_size_and_time): Likewise.
2701 (estimate_calls_size_and_time): Likewise.
2702 (estimate_node_size_and_time): Likewise.
2703 (inline_update_callee_summaries): Likewise.
2704 (remap_edge_change_prob): Likewise.
2705 (remap_edge_summaries): Likewise.
2706 (ipa_merge_fn_summary_after_inlining): Likewise.
2707 (ipa_update_overall_fn_summary): Likewise.
2708 (read_ipa_call_summary): Likewise.
2709 (inline_read_section): Likewise.
2710 (write_ipa_call_summary): Likewise.
2711 (ipa_fn_summary_write): Likewise.
2712 (ipa_free_fn_summary): Likewise.
2713 * ipa-hsa.c (process_hsa_functions): Likewise.
2714 (ipa_hsa_write_summary): Likewise.
2715 (ipa_hsa_read_section): Likewise.
2716 * ipa-icf.c (sem_function::merge): Likewise.
2717 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2718 (do_estimate_edge_time): Likewise.
2719 (estimate_size_after_inlining): Likewise.
2720 (estimate_growth): Likewise.
2721 (growth_likely_positive): Likewise.
2722 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2723 (inline_call): Likewise.
2724 * ipa-inline.c (caller_growth_limits): Likewise.
2725 (can_inline_edge_p): Likewise.
2726 (can_inline_edge_by_limits_p): Likewise.
2727 (compute_uninlined_call_time): Likewise.
2728 (compute_inlined_call_time): Likewise.
2729 (want_inline_small_function_p): Likewise.
2730 (edge_badness): Likewise.
2731 (update_caller_keys): Likewise.
2732 (update_callee_keys): Likewise.
2733 (recursive_inlining): Likewise.
2734 (inline_small_functions): Likewise.
2735 (inline_to_all_callers_1): Likewise.
2736 (dump_overall_stats): Likewise.
2737 (early_inline_small_functions): Likewise.
2738 (early_inliner): Likewise.
2739 * ipa-inline.h (estimate_edge_growth): Likewise.
2740 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2741 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2742 * ipa-prop.h (IPA_NODE_REF): Likewise.
2743 (IPA_EDGE_REF): Likewise.
2744 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2745 (propagate_malloc): Likewise.
2746 * ipa-split.c (execute_split_functions): Likewise.
2747 * symbol-summary.h: Rename get to get_create.
2749 (get_create): Likewise.
2750 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2752 2018-06-08 Martin Liska <mliska@suse.cz>
2754 * symbol-summary.h (release): Move definition out of class
2756 (symtab_removal): Likewise.
2757 (symtab_duplication): Likewise.
2759 2018-06-08 Martin Liska <mliska@suse.cz>
2761 * symbol-summary.h (function_summary): Move constructor
2762 implementation out of class declaration.
2763 (release): Likewise.
2764 (symtab_insertion): Likewise.
2765 (symtab_removal): Likewise.
2766 (symtab_duplication): Likewise.
2769 2018-06-08 Martin Liska <mliska@suse.cz>
2771 * Makefile.in: Remove support for MPX (macros, related functions,
2772 fields in cgraph_node, ...).
2773 * builtin-types.def (BT_BND): Likewise.
2774 (BT_FN_BND_CONST_PTR): Likewise.
2775 (BT_FN_CONST_PTR_BND): Likewise.
2776 (BT_FN_VOID_PTR_BND): Likewise.
2777 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2778 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2779 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2780 (expand_builtin_mempcpy_with_bounds): Likewise.
2781 (expand_builtin_memset_with_bounds): Likewise.
2782 (expand_builtin_memset_args): Likewise.
2783 (std_expand_builtin_va_start): Likewise.
2784 (expand_builtin): Likewise.
2785 (expand_builtin_with_bounds): Likewise.
2786 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2787 (DEF_LIB_BUILTIN_CHKP): Likewise.
2788 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2789 (DEF_CHKP_BUILTIN): Likewise.
2790 (BUILT_IN_MEMCPY): Likewise.
2791 (BUILT_IN_MEMMOVE): Likewise.
2792 (BUILT_IN_MEMPCPY): Likewise.
2793 (BUILT_IN_MEMSET): Likewise.
2794 (BUILT_IN_STPCPY): Likewise.
2795 (BUILT_IN_STRCAT): Likewise.
2796 (BUILT_IN_STRCHR): Likewise.
2797 (BUILT_IN_STRCPY): Likewise.
2798 (BUILT_IN_STRLEN): Likewise.
2799 (BUILT_IN_MEMCPY_CHK): Likewise.
2800 (BUILT_IN_MEMMOVE_CHK): Likewise.
2801 (BUILT_IN_MEMPCPY_CHK): Likewise.
2802 (BUILT_IN_MEMSET_CHK): Likewise.
2803 (BUILT_IN_STPCPY_CHK): Likewise.
2804 (BUILT_IN_STRCAT_CHK): Likewise.
2805 (BUILT_IN_STRCPY_CHK): Likewise.
2806 * calls.c (store_bounds): Likewise.
2807 (emit_call_1): Likewise.
2808 (special_function_p): Likewise.
2809 (maybe_warn_nonstring_arg): Likewise.
2810 (initialize_argument_information): Likewise.
2811 (finalize_must_preallocate): Likewise.
2812 (compute_argument_addresses): Likewise.
2813 (expand_call): Likewise.
2814 * cfgexpand.c (expand_call_stmt): Likewise.
2815 (expand_return): Likewise.
2816 (expand_gimple_stmt_1): Likewise.
2817 (pass_expand::execute): Likewise.
2818 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2819 (cgraph_node::remove): Likewise.
2820 (cgraph_node::dump): Likewise.
2821 (cgraph_node::verify_node): Likewise.
2822 * cgraph.h (chkp_function_instrumented_p): Likewise.
2823 (symtab_node::get_alias_target): Likewise.
2824 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2825 (cgraph_local_p): Likewise.
2826 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2827 (cgraph_edge::rebuild_references): Likewise.
2828 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2829 (walk_polymorphic_call_targets): Likewise.
2830 (cgraph_node::expand_thunk): Likewise.
2831 (symbol_table::output_weakrefs): Likewise.
2832 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2833 (ix86_handle_option): Likewise.
2834 * config/i386/constraints.md: Likewise.
2835 * config/i386/i386-builtin-types.def (BND): Likewise.
2839 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2840 (BDESC_FIRST): Likewise.
2842 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2843 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2844 * config/i386/i386.c (enum reg_class): Likewise.
2845 (ix86_target_string): Likewise.
2846 (ix86_option_override_internal): Likewise.
2847 (ix86_conditional_register_usage): Likewise.
2848 (ix86_valid_target_attribute_inner_p): Likewise.
2849 (ix86_set_indirect_branch_type): Likewise.
2850 (ix86_set_current_function): Likewise.
2851 (ix86_function_arg_regno_p): Likewise.
2852 (init_cumulative_args): Likewise.
2853 (ix86_function_arg_advance): Likewise.
2854 (ix86_function_arg): Likewise.
2855 (ix86_pass_by_reference): Likewise.
2856 (ix86_function_value_regno_p): Likewise.
2857 (ix86_function_value_1): Likewise.
2858 (ix86_function_value_bounds): Likewise.
2859 (ix86_return_in_memory): Likewise.
2860 (ix86_setup_incoming_vararg_bounds): Likewise.
2861 (ix86_va_start): Likewise.
2862 (indirect_thunk_need_prefix): Likewise.
2863 (print_reg): Likewise.
2864 (ix86_print_operand): Likewise.
2865 (ix86_expand_call): Likewise.
2866 (ix86_output_function_return): Likewise.
2867 (reg_encoded_number): Likewise.
2868 (BDESC_VERIFYS): Likewise.
2869 (ix86_init_mpx_builtins): Likewise.
2870 (ix86_init_builtins): Likewise.
2871 (ix86_emit_cmove): Likewise.
2872 (ix86_emit_move_max): Likewise.
2873 (ix86_expand_builtin): Likewise.
2874 (ix86_builtin_mpx_function): Likewise.
2875 (ix86_get_arg_address_for_bt): Likewise.
2876 (ix86_load_bounds): Likewise.
2877 (ix86_store_bounds): Likewise.
2878 (ix86_load_returned_bounds): Likewise.
2879 (ix86_store_returned_bounds): Likewise.
2880 (ix86_class_likely_spilled_p): Likewise.
2881 (ix86_hard_regno_mode_ok): Likewise.
2882 (x86_order_regs_for_local_alloc): Likewise.
2883 (ix86_mitigate_rop): Likewise.
2884 (ix86_bnd_prefixed_insn_p): Likewise.
2885 (ix86_mpx_bound_mode): Likewise.
2886 (ix86_make_bounds_constant): Likewise.
2887 (ix86_initialize_bounds): Likewise.
2888 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2889 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2890 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2891 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2892 (TARGET_CHKP_BOUND_MODE): Likewise.
2893 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2894 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2895 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2896 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2897 * config/i386/i386.h (TARGET_MPX): Likewise.
2898 (TARGET_MPX_P): Likewise.
2899 (VALID_BND_REG_MODE): Likewise.
2900 (FIRST_BND_REG): Likewise.
2901 (LAST_BND_REG): Likewise.
2902 (enum reg_class): Likewise.
2903 (BND_REG_P): Likewise.
2904 (BND_REGNO_P): Likewise.
2905 (BNDmode): Likewise.
2906 (ADJUST_INSN_LENGTH): Likewise.
2907 * config/i386/i386.md: Likewise.
2908 * config/i386/i386.opt: Likewise.
2909 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2910 (defined): Likewise.
2911 (LINK_MPX): Likewise.
2912 (MPX_SPEC): Likewise.
2913 (LIBMPX_SPEC): Likewise.
2914 (LIBMPXWRAPPERS_SPEC): Likewise.
2915 (CHKP_SPEC): Likewise.
2916 * config/i386/predicates.md: Likewise.
2917 * dbxout.c (dbxout_type): Likewise.
2918 * doc/extend.texi: Likewise.
2919 * doc/invoke.texi: Likewise.
2920 * doc/md.texi: Likewise.
2921 * doc/tm.texi: Likewise.
2922 * doc/tm.texi.in: Likewise.
2923 * dwarf2out.c (is_base_type): Likewise.
2924 (gen_formal_types_die): Likewise.
2925 (gen_subprogram_die): Likewise.
2926 (gen_type_die_with_usage): Likewise.
2927 (gen_decl_die): Likewise.
2928 (dwarf2out_late_global_decl): Likewise.
2929 * expr.c (expand_assignment): Likewise.
2930 (emit_storent_insn): Likewise.
2931 (store_expr_with_bounds): Likewise.
2932 (store_expr): Likewise.
2933 (expand_expr_real_1): Likewise.
2934 * expr.h (store_expr_with_bounds): Likewise.
2935 * function.c (use_register_for_decl): Likewise.
2936 (struct bounds_parm_data): Likewise.
2937 (assign_parms_augmented_arg_list): Likewise.
2938 (assign_parm_find_entry_rtl): Likewise.
2939 (assign_parm_is_stack_parm): Likewise.
2940 (assign_parm_load_bounds): Likewise.
2941 (assign_bounds): Likewise.
2942 (assign_parms): Likewise.
2943 (expand_function_start): Likewise.
2944 * gcc.c (CHKP_SPEC): Likewise.
2945 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2946 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2947 (wrestrict_dom_walker::check_call): Likewise.
2948 * gimple.c (gimple_build_call_from_tree): Likewise.
2949 * gimple.h (enum gf_mask): Likewise.
2950 (gimple_call_with_bounds_p): Likewise.
2951 (gimple_call_set_with_bounds): Likewise.
2952 * gimplify.c (gimplify_init_constructor): Likewise.
2953 * ipa-cp.c (initialize_node_lattices): Likewise.
2954 (propagate_constants_across_call): Likewise.
2955 (find_more_scalar_values_for_callers_subset): Likewise.
2956 * ipa-hsa.c (process_hsa_functions): Likewise.
2957 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2958 * ipa-icf.c (sem_function::merge): Likewise.
2959 * ipa-inline.c (early_inliner): Likewise.
2960 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2961 (warn_function_cold): Likewise.
2962 (propagate_pure_const): Likewise.
2963 * ipa-ref.h (enum GTY): Likewise.
2964 * ipa-split.c (find_retbnd): Likewise.
2965 (consider_split): Likewise.
2966 (split_function): Likewise.
2967 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2968 * ipa.c (walk_polymorphic_call_targets): Likewise.
2969 (symbol_table::remove_unreachable_nodes): Likewise.
2970 (process_references): Likewise.
2971 (cgraph_build_static_cdtor_1): Likewise.
2972 * lto-cgraph.c (lto_output_node): Likewise.
2973 (output_refs): Likewise.
2974 (compute_ltrans_boundary): Likewise.
2975 (input_overwrite_node): Likewise.
2976 (input_node): Likewise.
2977 (input_cgraph_1): Likewise.
2978 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2979 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2980 (class pass_chkp_instrumentation_passes): Likewise.
2981 (make_pass_chkp_instrumentation_passes): Likewise.
2982 * passes.def: Likewise.
2983 * rtl.h (struct GTY): Likewise.
2984 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2985 * stor-layout.c (layout_type): Likewise.
2986 * symtab.c: Likewise.
2987 * target.def: Likewise.
2988 * targhooks.c (default_chkp_bound_type): Likewise.
2989 (default_chkp_bound_mode): Likewise.
2990 (default_builtin_chkp_function): Likewise.
2991 (default_chkp_function_value_bounds): Likewise.
2992 (default_chkp_make_bounds_constant): Likewise.
2993 (default_chkp_initialize_bounds): Likewise.
2994 * targhooks.h (default_chkp_bound_type): Likewise.
2995 (default_chkp_bound_mode): Likewise.
2996 (default_builtin_chkp_function): Likewise.
2997 (default_chkp_function_value_bounds): Likewise.
2998 (default_chkp_make_bounds_constant): Likewise.
2999 (default_chkp_initialize_bounds): Likewise.
3000 * toplev.c (compile_file): Likewise.
3001 (process_options): Likewise.
3002 * tree-core.h (DEF_BUILTIN): Likewise.
3003 (DEF_BUILTIN_CHKP): Likewise.
3004 * tree-inline.c (declare_return_variable): Likewise.
3005 (remap_gimple_stmt): Likewise.
3006 (copy_bb): Likewise.
3007 (initialize_inlined_parameters): Likewise.
3008 (expand_call_inline): Likewise.
3009 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3010 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3011 (make_pass_ipa_chkp_produce_thunks): Likewise.
3012 (make_pass_chkp): Likewise.
3013 (make_pass_chkp_opt): Likewise.
3014 (make_pass_chkp_instrumentation_passes): Likewise.
3015 * tree-pretty-print.c (dump_generic_node): Likewise.
3016 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3017 * tree-ssa-dce.c (propagate_necessity): Likewise.
3018 (eliminate_unnecessary_stmts): Likewise.
3019 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3020 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3021 * tree-ssa-sccvn.h: Likewise.
3022 * tree-ssa-strlen.c (get_string_length): Likewise.
3023 (valid_builtin_call): Likewise.
3024 (adjust_last_stmt): Likewise.
3025 (handle_builtin_strchr): Likewise.
3026 (handle_builtin_strcpy): Likewise.
3027 (handle_builtin_stxncpy): Likewise.
3028 (handle_builtin_memcpy): Likewise.
3029 (handle_builtin_strcat): Likewise.
3030 (strlen_check_and_optimize_stmt): Likewise.
3031 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3032 * tree-streamer-in.c: Likewise.
3033 * tree-streamer.c (record_common_node): Likewise.
3034 * tree.c (tree_code_size): Likewise.
3035 (wide_int_to_tree_1): Likewise.
3036 (type_contains_placeholder_1): Likewise.
3037 (build_common_tree_nodes): Likewise.
3038 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3039 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3040 (POINTER_BOUNDS_P): Likewise.
3041 (BOUNDED_TYPE_P): Likewise.
3042 (BOUNDED_P): Likewise.
3043 (CALL_WITH_BOUNDS_P): Likewise.
3044 (pointer_bounds_type_node): Likewise.
3045 * value-prof.c (gimple_ic): Likewise.
3046 * var-tracking.c (vt_add_function_parameters): Likewise.
3047 * varasm.c (make_decl_rtl): Likewise.
3048 (assemble_start_function): Likewise.
3049 (output_constant): Likewise.
3050 (maybe_assemble_visibility): Likewise.
3051 * varpool.c (ctor_for_folding): Likewise.
3052 * chkp-builtins.def: Remove.
3053 * ipa-chkp.c: Remove.
3054 * ipa-chkp.h: Remove.
3055 * rtl-chkp.c: Remove.
3056 * rtl-chkp.h: Remove.
3057 * tree-chkp-opt.c: Remove.
3058 * tree-chkp.c: Remove.
3059 * tree-chkp.h: Remove.
3061 2018-06-07 Carl Love <cel@us.ibm.com>
3063 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3064 vextract_fp_from_shortl): Add BE support.
3066 2018-06-07 Paul Koning <ni1d@arrl.net>
3068 * compare-elim.c (try_merge_compare): Don't merge compare if
3069 address contains a side effect.
3070 (try_eliminate_compare): Likewise.
3072 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3074 * config.gcc: Support "tremont".
3075 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3076 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3078 * config/i386/i386.c (m_TREMONT): Define.
3079 (processor_target_table): Add "tremont".
3080 (PTA_TREMONT): Define.
3081 (ix86_lea_outperforms): Add TARGET_TREMONT.
3082 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3083 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3084 and M_INTEL_GOLDMONT_PLUS.
3085 (fold_builtin_cpu): Add "tremont".
3086 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3087 (ix86_option_override_internal): Add "tremont".
3088 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3089 (processor_type): Add PROCESSOR_TREMONT.
3090 * config/i386/x86-tune.def: Add m_TREMONT.
3091 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3093 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3095 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3096 symbol defined for msp430i* devices to be lower case.
3098 gcc/testsuite/gcc.target/msp430/
3099 * msp430i-device-symbol.c: New test.
3100 * msp430f-device-symbol.c: New test.
3101 * msp430.h: New test header file.
3103 2018-06-07 Richard Biener <rguenther@suse.de>
3105 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3106 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3107 Properly wrap signed arithmetic if overflow wraps.
3109 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3111 PR tree-optimization/69615
3112 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3113 of a cast from a same precision integral SSA_NAME in a bb dominated
3114 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3115 cast to utype if rhs2 has already a compatible type.
3117 2018-06-07 Richard Biener <rguenther@suse.de>
3119 PR tree-optimization/85935
3120 * graphite-scop-detection.c (find_params_in_bb): Analyze
3121 condition operands with respect to the correct loop. Assert
3122 the analysis doesn't fail.
3124 2018-06-04 Carl Love <cel@us.ibm.com>
3126 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3127 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3128 as it is slightly cheaper.
3129 (first_match_or_eos_index_<mode>):
3130 Calculate index using natural element order.
3131 (first_match_index_<mode>):
3132 Calculate index using natural element order.
3133 (first_match_or_eos_index_<mode>):
3134 Calculate index using natural order.
3135 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3136 for BE and LE modes.
3137 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3138 P9V_BUILTIN_VCLZLSBB_V16QI.
3139 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3142 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3144 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3145 indentation and line wrap for many prototypes. Add missing
3146 @smallexample directives around block of prototypes for vec_xl and
3149 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3151 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3152 track if we pass or return IEEE 128-bit floating point.
3153 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3154 C++ mangling that is compatible with GCC 8.1.
3155 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3156 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3157 floating point types.
3158 (rs6000_function_arg_advance_1): Likewise.
3159 (rs6000_mangle_type): Optionally generate mangled names that match
3160 what GCC 8.1 generated for IEEE 128-bit floating point types.
3161 (rs6000_globalize_decl_name): If we have an external function that
3162 passes or returns IEEE 128-bit types, generate a weak reference
3163 from the mangled name used in GCC 8.1 to the current mangled
3165 (rs6000_init_builtins): Make __ibm128 use the long double type if
3166 long double is IBM extended double. Make __float128 use the long
3167 double type if long double is IEEE 128-bit.
3170 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3171 macro for __ibm128 built-in functions.
3172 (PACK_IF): Add __ibm128 pack/unpack functions.
3173 (UNPACK_IF): Likewise.
3174 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3175 enable long double built-in functions if long double is IEEE
3176 128-bit floating point.
3177 (rs6000_invalid_builtin): Update long double built-in function
3179 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3180 functions, adjust the built-in function to use the long double
3181 built-in function if __ibm128 and long double are the same type.
3182 * doc/extend.texi (PowerPC builtins): Update documention for
3183 __builtin_{,un}pack_longdouble. Add documentation for
3184 __builtin_{,un}pack_ibm128.
3186 2018-06-06 Jim Wilson <jimw@sifive.com>
3188 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3189 (struct machine_function): New field interrupt_mode.
3190 (riscv_handle_type_attribute): New function. Add forward declaration.
3191 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3192 (riscv_expand_epilogue): Check interrupt_mode field.
3193 (riscv_set_current_function): Check interrupt attribute args and
3194 set interrupt_mode field.
3195 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3196 (riscv_sret, riscv_uret): New.
3197 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3198 new arguments to interrupt attribute.
3200 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3203 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3204 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3206 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3208 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3209 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3210 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3211 several redundant entries.
3213 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3215 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3216 type from "rtx" to "rtx_insn *".
3217 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3218 for local "call_insn", removing cast.
3219 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3221 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3223 PR tree-optimization/86066
3224 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3225 for BIT_INSERT_EXPR stores.
3227 2018-06-06 Richard Biener <rguenther@suse.de>
3229 PR tree-optimization/86062
3230 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3231 component refs ontop
3232 of to be offsetted base.
3234 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3236 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3237 to be static and remove check on interrupt attribute name.
3239 gcc/testsuite/gcc.target/msp430/
3240 * function-attributes-4.c: New test.
3241 * static-interrupts.c: New test.
3243 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3245 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3246 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3248 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3251 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3253 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3254 Remove second argument, change how error is called.
3255 (aarch64_layout_arg): Remove second argument from
3256 aarch64_err_no_fpadvsimd call.
3257 (aarch64_init_cumulative_args): Ditto.
3258 (aarch64_gimplify_va_arg_expr): Ditto.
3259 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3261 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3263 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3264 (*simple_return_indirect_internal<mode>): Rename from
3265 simple_return_indirect_internal. Use W mode iterator.
3266 (rstorssp): New expander.
3267 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3268 (clrssbsy): New expander.
3269 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3271 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3273 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3275 (cmse_check_pointed_object): Likewise.
3277 2018-06-05 Martin Liska <mliska@suse.cz>
3279 PR gcov-profile/47618
3280 * doc/invoke.texi: Document how -fprofile-dir format
3283 2018-06-05 Richard Biener <rguenther@suse.de>
3285 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3286 removal pretend DOM info isn't available so we do not update
3287 it and only remove edges, not dominated blocks. Actually free
3288 DOM info in case we removed something. Remove unreachable blocks.
3289 (mfb_keep_latches): Work with either DOM info or marked backedges.
3290 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3291 first. Mark backedges if DOM info isn't available.
3292 (Re-)compute DOM info after cleanup_control_flow_pre.
3294 2018-06-05 Richard Biener <rguenther@suse.de>
3296 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3297 (locus_discrim_hasher::hash): Adjust.
3298 (locus_discrim_hasher::equal): Likewise.
3299 (next_discriminator_for_locus): Work on line directly.
3300 (same_line_p): Pass in expanded locus1 as well.
3301 (assign_discriminators): Avoid redundant location expansions.
3303 2018-06-05 Richard Biener <rguenther@suse.de>
3305 PR tree-optimization/86046
3306 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3307 if required after clearing TREE_ADDRESSABLE.
3309 2018-06-05 Richard Biener <rguenther@suse.de>
3311 PR tree-optimization/86047
3312 * tree-ssa-loop.c (for_each_index): Glob handling of all
3313 decls and constants and really handle all of them.
3315 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3318 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3319 qualifier_void_pointer and qualifier_const_void_pointer.
3320 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3321 (arm_init_builtins): Handle the above.
3322 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3323 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3324 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3327 2018-06-05 Martin Liska <mliska@suse.cz>
3329 * auto-profile.c (read_autofdo_file): Do not use
3330 gcov_ctr_summary struct.
3331 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3332 * coverage.c (struct counts_entry): Likewise.
3333 (read_counts_file): Read just single summary entry.
3334 (get_coverage_counts): Use gcov_summary struct.
3335 * coverage.h (get_coverage_counts): Likewise.
3336 * gcov-dump.c (dump_working_sets): Likewise.
3337 (tag_summary): Dump just single summary.
3338 * gcov-io.c (gcov_write_summary): Write just histogram
3340 (gcov_read_summary): Read just single summary.
3341 (compute_working_sets): Use gcov_summary struct.
3342 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3343 of GCOV_COUNTERS_SUMMABLE.
3344 (GCOV_COUNTERS_SUMMABLE): Remove.
3345 (GCOV_FIRST_VALUE_COUNTER): Replace with
3346 GCOV_COUNTER_V_INTERVAL.
3347 (struct gcov_ctr_summary): Remove.
3348 (struct gcov_summary): Directly use fields of former
3350 (compute_working_sets): Use gcov_summary struct.
3351 * gcov.c (read_count_file): Do not use ctrs fields.
3352 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3354 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3356 * profile.c: Likewise.
3357 * profile.h: Likewise.
3359 2018-06-05 Martin Liska <mliska@suse.cz>
3361 PR gcov-profile/84846
3362 * gcov.c (output_lines): Print working directory only
3363 in intermediate format.
3365 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3367 * config/s390/s390-builtin-types.def: Add void function type.
3368 * config/s390/s390-builtins.def: Use the function type for the
3371 2018-06-04 Jim Wilson <jimw@sifive.com>
3373 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3375 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3376 and maybe_eh_return. Change regno to unsigned int. Use new args to
3377 handle EH_RETURN_DATA_REGNO registers properly.
3378 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3379 (riscv_expand_epilogue): Update comment. Change argument name and
3380 type. Update code to use new name and type. Pass new args to
3381 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3383 * config/riscv/riscv.md (NORMAL_RETURN): New.
3384 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3385 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3386 (eh_return): Call gen_eh_return_internal and emit barrier.
3387 (eh_return_internal): Call riscv_expand_epilogue.
3389 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3391 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3392 bit_insertion field and declare can_be_merged_into method.
3393 (merged_store_group::can_be_merged_into): New method.
3394 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3395 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3396 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3398 2018-06-04 Richard Biener <rguenther@suse.de>
3400 PR tree-optimization/85955
3401 * builtins.c (fold_builtin_sincos): Convert pointers to
3402 destination to appropriate type before dereferencing.
3404 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3406 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3408 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3410 * expr.c (expand_expr_real_1): Force the operand into memory if
3411 its TYPE_MODE is BLKmode and if there is no integer mode for
3412 the number of bits being extracted.
3414 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3418 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3419 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3421 2018-06-04 Richard Biener <rguenther@suse.de>
3423 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3424 (cleanup_tree_cfg_noloop): ... single caller. Do
3425 start_recording_case_labels later.
3427 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3429 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3430 to _IMMINTRIN_H_INCLUDED.
3431 * config/i386/pconfigintrin.h: Ditto.
3432 * config/i386/waitpkgintrin.h: Ditto.
3433 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3434 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3435 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3436 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3437 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3438 waitpkgintrin.h and cldemoteintrin.h.
3440 2018-06-04 Richard Biener <rguenther@suse.de>
3442 PR tree-optimization/86038
3443 * tracer.c (find_best_successor): Check probability for
3444 being initialized, bail out if not.
3446 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3449 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3450 of bits to ignore when comparing architectures.
3452 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3454 PR tree-optimization/69615
3455 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3456 maximum or minimum of the type, try to merge it also as if
3457 range1 is + [-, x - 1] or + [x + 1, -].
3460 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3462 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3464 PR tree-optimization/86034
3465 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3466 the unsigned bitfield type in a bit insertion sequence if it does not
3467 have a larger precision than the bitfield size.
3468 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3470 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3472 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3474 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3476 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3477 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3478 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3479 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3481 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3483 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3484 Disable -fdelete-null-pointer-checks for ELF toolchain.
3486 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3487 Kito Cheng <kito.cheng@gmail.com>
3489 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3490 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3491 (nds32*-*-*): Add glibc and uclibc conditions.
3492 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3493 (TARGET_EXCEPT_UNWIND_INFO): Define.
3494 * config/nds32/elf.h: New file.
3495 * config/nds32/linux.h: New file.
3496 * config/nds32/nds32-elf.opt: New file.
3497 * config/nds32/nds32-linux.opt: New file.
3498 * config/nds32/nds32-fp-as-gp.c
3499 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3500 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3502 (nds32_asm_file_end): Ditto.
3503 (nds32_print_operand): Ditto.
3504 (nds32_insert_attributes): Ditto.
3505 (nds32_init_libfuncs): New function.
3506 (TARGET_HAVE_TLS): Define.
3507 (TARGET_INIT_LIBFUNCS): Define.
3508 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3510 (TARGET_ELF): Apply different mcmodel setting.
3511 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3512 been migrated into elf.h and linux.h files.
3513 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3514 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3515 (mcmodel): The content has been migrated into nds32-elf.opt and
3516 nds32-linux.opt files.
3517 * config/nds32/t-elf: New file.
3518 * config/nds32/t-linux: New file.
3520 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3521 Shiva Chen <shiva0217@gmail.com>
3523 * config/nds32/constants.md (unspec_volatile_element): Add
3524 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3525 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3527 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3528 (make_pass_nds32_fp_as_gp): Declare.
3529 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3531 (nds32_asm_function_end_prologue): Remove unused asm output.
3532 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3533 (nds32_asm_file_start): Output necessary fp_as_gp information.
3534 (nds32_option_override): Adjust register usage.
3535 (nds32_expand_prologue): Consider fp_as_gp situation.
3536 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3537 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3540 (simple_return): Ditto.
3541 (omit_fp_begin): Output special directive for fp_as_gp.
3542 (omit_fp_end): Output special directive for fp_as_gp.
3543 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3544 mforbid-fp-as-gp): New options.
3546 2018-06-01 Mark Wielaard <mark@klomp.org>
3548 * dwarf2out.c (dwarf2out_finish): Remove generation of
3549 DW_AT_loclists_base.
3551 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3553 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3554 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3555 (struct merged_store_group): Add bit_insertion field.
3556 (dump_char_array): Use standard hexadecimal format.
3557 (merged_store_group::merged_store_group): Set bit_insertion to false.
3558 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3559 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3560 also print the mask in the dump file.
3561 (pass_store_merging::gate): Minor tweak.
3562 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3563 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3564 stores with INTEGER_CST stores.
3565 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3566 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3567 and use it throughout. Generate bit insertion sequences if need be.
3568 (pass_store_merging::process_store): Remove redundant condition.
3569 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3571 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3573 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3574 the 128-bit floating point types. Fix function comment.
3576 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3578 * config/aarch64/aarch64-simd.md
3579 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3581 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3584 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3586 PR tree-optimization/85989
3587 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3589 (backprop::intersect_uses): Check it when deciding whether this
3590 is a backedge reference.
3591 (backprop::process_block): Add each phi to m_visited_phis
3592 after visiting it, then clear it at the end.
3594 2018-06-01 Richard Biener <rguenther@suse.de>
3596 * tree-vectorizer.h (vect_dr_stmt): New function.
3597 (vect_get_load_cost): Adjust.
3598 (vect_get_store_cost): Likewise.
3599 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3600 Use vect_dr_stmt instead of DR_SMTT.
3601 (vect_record_base_alignments): Likewise.
3602 (vect_calculate_target_alignment): Likewise.
3603 (vect_compute_data_ref_alignment): Likewise and make static.
3604 (vect_update_misalignment_for_peel): Likewise.
3605 (vect_verify_datarefs_alignment): Likewise.
3606 (vector_alignment_reachable_p): Likewise.
3607 (vect_get_data_access_cost): Likewise. Pass down
3608 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3609 (vect_get_peeling_costs_all_drs): Likewise.
3610 (vect_peeling_hash_get_lowest_cost): Likewise.
3611 (vect_enhance_data_refs_alignment): Likewise.
3612 (vect_find_same_alignment_drs): Likewise.
3613 (vect_analyze_data_refs_alignment): Likewise.
3614 (vect_analyze_group_access_1): Likewise.
3615 (vect_analyze_group_access): Likewise.
3616 (vect_analyze_data_ref_access): Likewise.
3617 (vect_analyze_data_ref_accesses): Likewise.
3618 (vect_vfa_segment_size): Likewise.
3619 (vect_small_gap_p): Likewise.
3620 (vectorizable_with_step_bound_p): Likewise.
3621 (vect_prune_runtime_alias_test_list): Likewise.
3622 (vect_analyze_data_refs): Likewise.
3623 (vect_supportable_dr_alignment): Likewise.
3624 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3625 (vect_gen_prolog_loop_niters): Likewise.
3626 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3627 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3629 (vect_recog_mask_conversion_pattern): Likewise.
3630 (vect_try_gather_scatter_pattern): Likewise.
3631 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3632 to vect_get_store_cost.
3633 (vect_get_store_cost): Get stmt_info instead of DR.
3634 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3635 (vect_get_load_cost): Get stmt_info instead of DR.
3637 2018-06-01 Richard Biener <rguenther@suse.de>
3640 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3641 vars wrapped in MEM_REFs.
3643 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3645 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3646 Fix subreg tests so that we only return a choice between
3647 GENERAL_REGS and FP_REGS if the original classes included both.
3649 2018-06-01 Richard Biener <rguenther@suse.de>
3652 * tree-ssa-structalias.c (get_function_part_constraint):
3653 Handle NULL fi->decl.
3654 (find_func_aliases_for_call): Properly handle indirect
3655 fi from direct call.
3656 (find_func_clobbers): Likewise.
3657 (ipa_pta_execute): Likewise.
3658 (create_variable_info_for): For functions that are ifunc_resolver
3659 resolve to a varinfo that contains the result of the resolver call.
3660 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3663 2018-05-31 Michael Collison <michael.collison@arm.com>
3665 * config/aarch64/aarch64.md:
3666 (*fix_to_zero_extenddfdi2): New pattern.
3667 * gcc.target/aarch64/fix_extend1.c: New testcase.
3669 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3673 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3674 and BUILT_IN_STRNCMP_EQ.
3675 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3676 BUILT_IN_STRNCMP_EQ.
3677 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3678 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3679 (gimple_fold_builtin): Likewise.
3680 * tree-ssa-strlen.c (compute_string_length): New function.
3681 (determine_min_obsize): New function.
3682 (handle_builtin_string_cmp): New function to handle calls to
3683 string compare functions.
3684 (strlen_optimize_stmt): Add handling to builtin string compare
3686 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3687 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3688 * tree.c (build_common_builtin_nodes): Add new defines of
3689 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3691 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3694 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3695 functions with naked attribute.
3697 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3699 * config/i386/sse.md (avx_vec_concat<mode>):
3700 Substitute concat_tg_mode mode attribute with xtg_mode.
3701 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3702 (concat_tg_mode): Remove mode attribute.
3704 2018-05-31 Martin Sebor <msebor@redhat.com>
3707 * calls.c (alloc_max_size): Correct a logic error/typo.
3708 Treat excessive arguments as infinite. Warn for invalid arguments.
3709 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3711 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3714 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3715 and movx for Haswell.
3717 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3718 Cesar Philippidis <cesar@codesourcery.com>
3721 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3722 when emitting error on private/firstprivate reductions.
3723 * omp-low.c (lower_omp_target): Avoid reference-type processing
3724 on pointers for firstprivate clause.
3726 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3728 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3731 * config/aarch64/aarch64-simd.md
3732 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3733 (aarch64_ld1_x3_<mode>): Likewise
3734 (aarch64_st1x2<VALLDIF:mode>): Likewise
3735 (aarch64_st1_x2_<mode>): Likewise
3736 (aarch64_st1x3<VALLDIF:mode>): Likewise
3737 (aarch64_st1_x3_<mode>): Likewise
3738 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3739 (vld1_s8_x3): Likewise.
3740 (vld1_u16_x3): Likewise.
3741 (vld1_s16_x3): Likewise.
3742 (vld1_u32_x3): Likewise.
3743 (vld1_s32_x3): Likewise.
3744 (vld1_u64_x3): Likewise.
3745 (vld1_s64_x3): Likewise.
3746 (vld1_f16_x3): Likewise.
3747 (vld1_f32_x3): Likewise.
3748 (vld1_f64_x3): Likewise.
3749 (vld1_p8_x3): Likewise.
3750 (vld1_p16_x3): Likewise.
3751 (vld1_p64_x3): Likewise.
3752 (vld1q_u8_x3): Likewise.
3753 (vld1q_s8_x3): Likewise.
3754 (vld1q_u16_x3): Likewise.
3755 (vld1q_s16_x3): Likewise.
3756 (vld1q_u32_x3): Likewise.
3757 (vld1q_s32_x3): Likewise.
3758 (vld1q_u64_x3): Likewise.
3759 (vld1q_s64_x3): Likewise.
3760 (vld1q_f16_x3): Likewise.
3761 (vld1q_f32_x3): Likewise.
3762 (vld1q_f64_x3): Likewise.
3763 (vld1q_p8_x3): Likewise.
3764 (vld1q_p16_x3): Likewise.
3765 (vld1q_p64_x3): Likewise.
3766 (vst1_s64_x2): Likewise.
3767 (vst1_u64_x2): Likewise.
3768 (vst1_f64_x2): Likewise.
3769 (vst1_s8_x2): Likewise.
3770 (vst1_p8_x2): Likewise.
3771 (vst1_s16_x2): Likewise.
3772 (vst1_p16_x2): Likewise.
3773 (vst1_s32_x2): Likewise.
3774 (vst1_u8_x2): Likewise.
3775 (vst1_u16_x2): Likewise.
3776 (vst1_u32_x2): Likewise.
3777 (vst1_f16_x2): Likewise.
3778 (vst1_f32_x2): Likewise.
3779 (vst1_p64_x2): Likewise.
3780 (vst1q_s8_x2): Likewise.
3781 (vst1q_p8_x2): Likewise.
3782 (vst1q_s16_x2): Likewise.
3783 (vst1q_p16_x2): Likewise.
3784 (vst1q_s32_x2): Likewise.
3785 (vst1q_s64_x2): Likewise.
3786 (vst1q_u8_x2): Likewise.
3787 (vst1q_u16_x2): Likewise.
3788 (vst1q_u32_x2): Likewise.
3789 (vst1q_u64_x2): Likewise.
3790 (vst1q_f16_x2): Likewise.
3791 (vst1q_f32_x2): Likewise.
3792 (vst1q_f64_x2): Likewise.
3793 (vst1q_p64_x2): Likewise.
3794 (vst1_s64_x3): Likewise.
3795 (vst1_u64_x3): Likewise.
3796 (vst1_f64_x3): Likewise.
3797 (vst1_s8_x3): Likewise.
3798 (vst1_p8_x3): Likewise.
3799 (vst1_s16_x3): Likewise.
3800 (vst1_p16_x3): Likewise.
3801 (vst1_s32_x3): Likewise.
3802 (vst1_u8_x3): Likewise.
3803 (vst1_u16_x3): Likewise.
3804 (vst1_u32_x3): Likewise.
3805 (vst1_f16_x3): Likewise.
3806 (vst1_f32_x3): Likewise.
3807 (vst1_p64_x3): Likewise.
3808 (vst1q_s8_x3): Likewise.
3809 (vst1q_p8_x3): Likewise.
3810 (vst1q_s16_x3): Likewise.
3811 (vst1q_p16_x3): Likewise.
3812 (vst1q_s32_x3): Likewise.
3813 (vst1q_s64_x3): Likewise.
3814 (vst1q_u8_x3): Likewise.
3815 (vst1q_u16_x3): Likewise.
3816 (vst1q_u32_x3): Likewise.
3817 (vst1q_u64_x3): Likewise.
3818 (vst1q_f16_x3): Likewise.
3819 (vst1q_f32_x3): Likewise.
3820 (vst1q_f64_x3): Likewise.
3821 (vst1q_p64_x3): Likewise.
3823 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3825 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3826 user_label_prefix to name.
3828 * tree-core.h: Update comment about the format of NAME string
3829 passed to handler in attribute_spec.
3831 * config/msp430/msp430.md: Remove erroneous subreg expression from
3832 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3833 zero_extend{q,h}isi2.
3835 2018-05-30 Borislav Petkov <bp@suse.de>
3837 * doc/extend.texi: Document some architecture specific
3838 constraints and sort entries.
3840 2018-05-30 Martin Sebor <msebor@redhat.com>
3843 * builtins.c (expand_builtin_stpcpy_1): New function.
3844 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3845 only if the former succeeds.
3847 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3849 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3852 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3854 * doc/invoke.texi (-flinker-output): Document
3856 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3858 * passes.c (ipa_write_summaries): Only modify statements if body
3860 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3861 incrementally linking.
3862 (ipa_passes): Likewise.
3863 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3864 pass down resolution info.
3865 * common.opt (flag_incremental_link): Update info.
3866 * gcc.c (plugin specs): Turn flinker-output=* to
3867 -plugin-opt=-linker-output-known
3868 * toplev.c (compile_file): Also cut compilation when doing incremental
3870 * flag-types. (enum lto_partition_model): Add
3871 LTO_LINKER_OUTPUT_NOLTOREL.
3872 (invoke.texi): Add -flinker-output docs.
3873 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3874 link same way as WPA; do not stream in dead initializers.
3876 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3877 darf2out_register_external_decl): Support incremental link.
3879 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3881 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3883 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3885 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3886 it down to simple_object_copy_lto_debug_sections.
3887 (run_gcc): Determine incremental LTO link time and configure
3888 lto1 into non-wpa mode, disable renaming of debug sections.
3890 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3892 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3893 descriptions of various incorrectly documented functions.
3895 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3898 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3899 address check not strict.
3901 2018-05-30 Richard Biener <rguenther@suse.de>
3903 PR tree-optimization/85964
3904 * tracer.c (better_p): Drop initialized count check, we only
3905 call the function with initialized counts now.
3906 (find_best_successor): Do find a best edge if one
3907 has uninitialized count.
3908 (find_best_predecessor): Likewise. Do BB frequency check only
3909 if count is initialized.
3911 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3913 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3914 (aarch64_ldrstr_offset_compare): New.
3915 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3916 load/store orderings.
3917 (aarch64_gen_adjusted_ldpstp): Likewise.
3919 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3921 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3922 Check for subset of GENERAL_REGS and FP_REGS.
3923 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3926 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3928 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3929 and wi::to_poly_offset. Add the current offset and then check
3930 whether the sum fits, rather than using an unchecked addition of
3931 a checked term. Check for a shwi rather than a uhwi.
3932 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3933 (store_constructor): Use poly_int_tree_p.
3934 (expand_expr_real_1): Likewise.
3935 * function.c (assign_temp): Likewise.
3936 * fold-const.c (const_binop): Use poly_int_tree_p and
3938 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3940 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3941 to_poly_offset for MEM offsets.
3942 * ipa-icf.c (sem_variable::equals): Likewise.
3943 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3944 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3945 wi::to_poly_offset for BIT_FIELD_REF offsets.
3946 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3948 * var-tracking.c (emit_note_insn_var_location): Use
3949 tree_to_poly_uint64.
3951 2018-05-29 Jim Wilson <jimw@sifive.com>
3953 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3955 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3958 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3959 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3961 (sse4_1_round<mode>2): Use nonimmediate_operand
3962 for operand 1 predicate.
3964 2018-05-29 Martin Sebor <msebor@redhat.com>
3965 Richard Biener <rguenther@suse.de>
3968 * calls.c (get_size_range): Call determine_value_range instead
3969 of get_value_range..
3970 * tree-vrp.h (determine_value_range): Declared new function.
3971 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3973 2018-05-29 Richard Biener <rguenther@suse.de>
3975 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3976 sure to use non-pattern stmts for get_earlier_stmt arguments.
3977 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3978 called on pattern stmts.
3979 (get_later_stmt): Likewise.
3981 2018-05-29 Martin Liska <mliska@suse.cz>
3983 PR gcov-profile/85759
3984 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3987 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3989 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3991 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3995 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3996 VEC_PACK_FLOAT_EXPR): New tree codes.
3997 * tree-pretty-print.c (op_code_prio): Handle
3998 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3999 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4000 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4001 * tree-inline.c (estimate_operator_cost): Likewise.
4002 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4003 * fold-const.c (const_binop): Likewise.
4004 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4005 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4006 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4007 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4008 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4009 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4010 * expr.c (expand_expr_real_2): Likewise.
4011 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4012 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4013 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4015 * optabs.c (expand_widen_pattern_expr): For
4016 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4017 sign from result type rather than operand's type.
4018 (expand_binop_directly): For vec_packu_float_optab and
4019 vec_packs_float_optab allow result type to be different from operand's
4021 * optabs-tree.c (optab_for_tree_code): Handle
4022 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4023 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4024 * tree-vect-generic.c (expand_vector_operations_1): Handle
4025 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4026 VEC_PACK_FLOAT_EXPR.
4027 * tree-vect-stmts.c (supportable_widening_operation): Handle
4029 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4030 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4031 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4032 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4033 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4035 (vec_pack<floatprefix>_float_<mode>): New expander.
4036 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4038 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4039 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4040 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4041 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4042 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4044 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4045 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4046 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4047 VEC_PACK_FLOAT_EXPR): Document.
4049 2018-05-29 Richard Biener <rguenther@suse.de>
4051 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4053 (stmt_vec_info_vec): Make pointer.
4054 (init_stmt_vec_info_vec): Remove.
4055 (free_stmt_vec_info_vec): Likewise.
4056 (set_stmt_vec_info_vec): New function.
4057 (free_stmt_vec_infos): Likewise.
4058 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4059 (set_vinfo_for_stmt): Likewise.
4060 (get_earlier_stmt): Likewise.
4061 (get_later_stmt): Likewise.
4062 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4063 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4064 (vec_info::~vec_info): Free stmt_vec_infos.
4065 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4066 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4067 (pass_slp_vectorize::execute): Likewise.
4068 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4069 (free_stmt_vec_info_vec): Likewise.
4070 (set_stmt_vec_info_vec): New function.
4071 (free_stmt_vec_infos): Likewise.
4072 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4073 the global stmt_vec_info_vec.
4074 * tree-parloops.c (gather_scalar_reductions): Use
4075 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4078 2018-05-29 Richard Biener <rguenther@suse.de>
4080 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4082 2018-05-29 Martin Liska <mliska@suse.cz>
4083 David Malcolm <dmalcolm@redhat.com>
4085 * vec.c (test_reverse): New.
4086 (vec_c_tests): Add new test.
4087 * vec.h (vl_ptr>::reverse): New function.
4089 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4091 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4093 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4096 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4098 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4100 2018-05-28 Richard Biener <rguenther@suse.de>
4102 PR tree-optimization/85933
4103 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4104 look at stmts marked as vectorizable.
4106 2018-05-28 Richard Biener <rguenther@suse.de>
4108 PR tree-optimization/85934
4109 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4110 vector boolean check before scalar optimization.
4112 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4114 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4117 2018-05-28 Mark Wielaard <mark@klomp.org>
4119 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4120 if it is an expression containing a minus sign.
4122 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4124 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4126 2018-05-27 Paul Koning <ni1d@arrl.net>
4128 * config/pdp11/pdp11.md (truncsihi2): Remove.
4130 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4131 Chung-Ju Wu <jasonwucj@gmail.com>
4133 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4135 (unaligned_store_dw): Ditto.
4136 * config/nds32/nds32-memory-manipulation.c
4137 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4138 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4139 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4140 (emit_setmem_word_loop): Rename to ...
4141 (emit_setmem_doubleword_loop): ... this.
4142 (nds32_gen_dup_4_byte_to_word_value): New function.
4143 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4144 (nds32_expand_setmem_loop): Refine implementation.
4145 (nds32_expand_setmem_loop_v3m): Ditto.
4146 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4149 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4151 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4153 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4155 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4156 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4157 machine->attr_no_prologue_p.
4158 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4159 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4160 (nds32_expand_epilogue): Consider attr_naked_p.
4161 (nds32_expand_epilogue_v3pop): Likewise.
4162 (nds32_can_use_return_insn): Likewise.
4163 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4164 attr_no_prologue_p fields.
4165 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4167 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4170 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4172 * config/i386/sse.md
4173 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4175 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4177 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4179 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4181 (*<floatsuffix>floatv2div2sf2): Rename to ...
4182 (*float<floatunssuffix>v2div2sf2): ... this.
4183 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4184 (float<floatunssuffix>v2div2sf2_mask): ... this.
4185 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4186 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4187 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4189 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4191 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4193 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4195 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4197 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4199 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4200 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4201 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4202 gen_ufix_truncv8dfv8si2.
4203 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4204 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4205 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4206 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4207 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4208 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4209 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4210 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4212 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4216 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4218 2018-05-25 Jim Wilson <jimw@sifive.com>
4220 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4221 * config/riscv/riscv.c (struct machine_function): Add
4222 interrupt_handler_p and attribute_checked_p fields.
4223 (riscv_attribute_table): Add interrupt.
4224 (riscv_interrupt_type_p): New.
4225 (riscv_save_reg_p): Save extra regs for interrupt handler.
4226 (riscv_use_save_libcall): Return false for interrupt handler.
4227 (riscv_first_stack_step): Add forward declaration.
4228 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4229 for interrupt handler with large frame. Use it for saved reg list.
4230 (riscv_expand_prologue): Move flag_stack_usage_info support to
4231 eliminate duplication.
4232 (riscv_expand_epilogue): Generate mret for interrupt handler.
4233 (riscv_epilogue_uses): New.
4234 (riscv_can_use_return_insn): Return false for interrupt handler.
4235 (riscv_function_ok_for_sibcall): Likewise.
4236 (riscv_set_current_function): Add interrupt handler support.
4237 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4238 * config/riscv/riscv.md (UNSPECV_MRET): New.
4240 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4242 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4244 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4246 PR tree-optimization/85712
4247 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4248 this candidate has already been replaced in-situ by a copy.
4250 2018-05-25 Jason Merrill <jason@redhat.com>
4252 PR c++/80485 - inline function non-zero address.
4253 * symtab.c (nonzero_address): Check DECL_COMDAT.
4255 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4258 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4259 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4260 extension of SImode operation. Use const123_operand predicate.
4261 (*saddsi_1): Remove.
4262 (*saddl_se_1): Ditto.
4264 (*ssubl_se_1): Ditto.
4265 * config/alpha/predicates.md (const123_operand): New predicate.
4266 * config/alpha/constraints.md (P): Use IN_RANGE.
4268 2018-05-25 Richard Biener <rguenther@suse.de>
4270 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4272 (ref_maybe_used_by_stmt_p): Likewise.
4273 (stmt_may_clobber_ref_p): Likewise.
4274 (stmt_may_clobber_ref_p_1): Likewise.
4275 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4277 (ref_maybe_used_by_stmt_p): Likewise.
4278 (stmt_may_clobber_ref_p): Likewise.
4279 (stmt_may_clobber_ref_p_1): Likewise.
4280 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4281 the alias oracle to disambiguate DRs with stmts DR analysis
4283 (vect_analyze_data_refs): Do not give up on not analyzable
4284 DRs for BB vectorization. Remove code truncating the dataref
4287 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4290 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4291 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4292 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4294 2018-05-25 Richard Biener <rguenther@suse.de>
4296 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4297 function, combining stmt data ref gathering and fatal analysis
4299 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4300 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4301 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4302 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4303 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4305 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4307 PR tree-optimization/85720
4308 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4309 SCC if all partitions are builtins.
4310 (version_loop_by_alias_check): New parameter. Generate cancelable
4311 runtime alias check if all partitions are builtins.
4312 (distribute_loop): Update call to above function.
4314 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4316 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4317 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4318 (parm_default_def_partition_arg): Ditto.
4319 (set_parm_default_def_partition): Ditto.
4320 (get_parm_default_def_partitions): Ditto and make it static.
4321 (get_undefined_value_partitions): Ditto and make it static.
4322 (remove_ssa_form): Refactor call to init_var_map here.
4323 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4324 computation for loop region.
4325 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4326 (register_default_def): Delete.
4327 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4328 (parm_default_def_partition_arg): Ditto.
4329 (set_parm_default_def_partition): Ditto.
4330 (get_parm_default_def_partitions): Ditto and make it static.
4331 (get_undefined_value_partitions): Ditto and make it static.
4332 (coalesce_with_default, coalesce_with_default): Update comment.
4333 (create_coalesce_list_for_region): New func factored out from
4334 create_outofssa_var_map.
4335 (populate_coalesce_list_for_outofssa): New func factored out from
4336 create_outofssa_var_map and coalesce_ssa_name.
4337 (create_outofssa_var_map): Delete.
4338 (coalesce_ssa_name): Refactor to support live range computation.
4339 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4340 (get_parm_default_def_partitions): Delete.
4341 (get_undefined_value_partitions): Ditto.
4342 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4343 computation for loop region.
4344 (new_tree_live_info, loe_visit_block): Ditto.
4345 (live_worklist, set_var_live_on_entry): Ditto.
4346 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4347 * tree-ssa-live.h (struct _var_map): New fields.
4348 (init_var_map): Change decl.
4349 (region_contains_p): New.
4351 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4353 * tree-ssa-live.h (live_merge_and_clear): Delete.
4355 2018-05-25 Richard Biener <rguenther@suse.de>
4358 * tree-dump.c (dequeue_and_dump): Remove access to removed
4359 operand 2 of a SWITCH_EXPR.
4361 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4363 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4364 multiplication and division.
4365 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4366 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4367 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4368 (cond_udiv_optab, cond_umod_optab): New optabs.
4369 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4370 (IFN_COND_RDIV): New internal functions.
4371 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4372 TRUNC_MOD_EXPR and RDIV_EXPR.
4373 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4374 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4376 (SVE_INT_BINARY): Include mult.
4377 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4378 (optab, sve_int_op): Handle mult.
4379 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4381 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4382 for SVE_INT_BINARY_SD.
4384 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4386 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4387 (optab, sve_int_op): Handle div and udiv.
4388 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4389 for SVE_INT_BINARY_SD.
4390 (*<optab><mode>3): New insn for the same.
4392 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4394 * tree-vect-patterns.c: Include predict.h.
4395 (vect_recog_divmod_pattern): Restrict check for division support
4396 to when optimizing for size.
4398 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4400 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4401 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4402 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4403 (gimple_match_op::set_op): Likewise.
4404 (gimple_resimplify4): Declare.
4405 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4406 (expr::gen_transform): Likewise.
4407 (decision_tree::gen): Generate a simplification routine for 4 operands.
4408 * gimple-match-head.c (gimple_simplify): Add an overload for
4409 4 operands. In the top-level function, handle up to 4 call
4410 arguments and call gimple_resimplify4.
4411 (gimple_resimplify4): New function.
4412 (build_call_internal): Pass a fourth operand.
4413 (maybe_push_to_seq): Likewise.
4414 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4415 Fold VEC_COND_EXPRs of an operation and a default value into
4416 an IFN_COND_* function if possible.
4417 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4419 (SVE_COND_FP_BINARY): Include them.
4420 (optab, sve_fp_op): Handle them.
4421 (SVE_INT_BINARY_REV): New code iterator.
4422 (SVE_COND_FP_BINARY_REV): New int iterator.
4423 (commutative): New int attribute.
4424 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4426 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4428 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4429 (*cond_<optab><mode>): New patterns for reversed operands.
4431 2018-05-25 Richard Biener <rguenther@suse.de>
4433 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4434 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4435 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4436 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4437 * tree-vect-data-refs.c (everywhere): Adjust users.
4438 * tree-vect-loop.c (everywhere): Likewise.
4439 * tree-vect-slp.c (everywhere): Likewise.
4440 * tree-vect-stmts.c (everywhere): Likewise.
4441 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4443 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4445 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4447 (gcc_cv_as_section_exclude): ... this.
4448 Try Solaris as #exclude syntax.
4449 * configure: Regenerate.
4450 * config.in: Regenerate.
4451 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4453 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4454 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4456 * varasm.c (default_elf_asm_named_section): Don't check if
4457 HAVE_GAS_SECTION_EXCLUDE is defined.
4459 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4461 * doc/md.texi: Update the documentation of the cond_* optabs
4462 to mention the new final operand. Fix GET_MODE_NUNITS call.
4463 Describe the scalar case too.
4464 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4465 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4467 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4468 (get_conditional_internal_fn): Update comment.
4469 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4470 accumulator value as a final argument to conditional functions.
4471 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4472 a define_expand and add an "else" operand. Assert for now that
4473 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4474 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4475 (*cond_<optab><mode>): New patterns.
4476 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4477 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4478 (UNSPEC_COND_EOR): Delete.
4479 (optab): Remove associated mappings.
4480 (SVE_INT_BINARY): New code iterator.
4481 (sve_int_op): Remove int attribute and add "minus" to the code
4483 (SVE_COND_INT_OP): Delete.
4484 (SVE_COND_FP_OP): Rename to...
4485 (SVE_COND_FP_BINARY): ...this.
4487 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4489 * optabs.c (can_reuse_operands_p): New function.
4490 (maybe_legitimize_operands): Try to reuse the results for
4493 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4495 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4496 Add {q} suffix to insn mnemonic.
4498 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4500 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4501 (msp430_warn_func_return): New.
4503 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4505 * fold-const.c (tree_nonzero_bits): New function.
4506 * fold-const.h (tree_nonzero_bits): Likewise.
4507 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4508 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4510 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4514 * varasm.c (assemble_alias): Check ifunc_resolver only on
4517 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4520 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4521 when memory input operand is handled.
4523 2018-05-24 Luis Machado <luis.machado@linaro.org>
4525 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4527 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4529 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4531 * match.pd: Delay FMA folds until after vectorization.
4533 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4536 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4537 address check not strict.
4539 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4541 * gimple-match.h (gimple_match_op): New class.
4542 (mprts_hook): Replace parameters with a gimple_match_op *.
4543 (maybe_build_generic_op): Likewise.
4544 (gimple_simplified_result_is_gimple_val): Replace parameters with
4545 a const gimple_match_op *.
4546 (gimple_simplify): Replace code_helper * and tree * parameters with
4547 a gimple_match_op * parameter.
4548 (gimple_resimplify1): Replace code_helper *, tree and tree *
4549 parameters with a gimple_match_op * parameter.
4550 (gimple_resimplify2): Likewise.
4551 (gimple_resimplify3): Likewise.
4552 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4553 parameters with a gimple_match_op * parameter.
4554 * gimple-match-head.c (gimple_simplify): Change prototypes of
4555 auto-generated functions to take a gimple_match_op * instead of
4556 separate code_helper * and tree * parameters. Make the same
4557 change in the top-level overload and update calls to the
4558 gimple_resimplify routines. Update calls to the auto-generated
4559 functions and to maybe_push_res_to_seq in the publicly-facing
4560 operation-specific gimple_simplify overloads.
4561 (gimple_match_op::MAX_NUM_OPS): Define.
4562 (gimple_resimplify1): Replace rcode and ops with a single res_op
4563 parameter. Update call to gimple_simplify.
4564 (gimple_resimplify2): Likewise.
4565 (gimple_resimplify3): Likewise.
4566 (mprts_hook): Replace parameters with a gimple_match_op *.
4567 (maybe_build_generic_op): Likewise.
4568 (build_call_internal): Replace type, nargs and ops with
4569 a gimple_match_op *.
4570 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4571 with a single gimple_match_op *. Update calls to mprts_hook,
4572 build_call_internal and gimple_simplified_result_is_gimple_val.
4573 Factor out code that is common to the tree_code and combined_fn cases.
4574 * genmatch.c (expr::gen_transform): Replace tem_code and
4575 tem_ops with a gimple_match_op called tem_op. Update calls
4576 to the gimple_resimplify functions and maybe_push_res_to_seq.
4577 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4578 res_ops. Update call to the gimple_resimplify functions.
4579 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4580 (decision_tree::gen): Make the functions take a gimple_match_op *
4581 called res_op instead of separate res_code and res_ops parameters.
4582 Update call accordingly.
4583 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4584 and ops with a single res_op parameter. Update calls to
4585 maybe_build_generic_op and maybe_push_res_to_seq.
4586 (fold_stmt_1): Update calls to gimple_simplify and
4587 replace_stmt_with_simplification.
4588 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4589 and gimple_simplified_result_is_gimple_val.
4590 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4592 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4593 with a gimple_match_op *.
4594 (vn_nary_build_or_lookup): Likewise. Update call to
4595 vn_nary_build_or_lookup_1.
4596 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4597 gimple_match_op *. Update calls to the gimple_resimplify routines
4598 and to gimple_simplified_result_is_gimple_val.
4599 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4600 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4601 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4602 (visit_nary_op): Likewise.
4603 (visit_reference_op_load): Likewise.
4605 2018-05-23 Luis Machado <luis.machado@linaro.org>
4607 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4608 modifier for printing the step amount.
4610 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4613 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4616 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4618 * doc/sourcebuild.texi (Endianness): New subsubsection.
4620 2018-05-23 Luis Machado <luis.machado@linaro.org>
4622 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4623 <prefetch_dynamic_strides>: New const bool field.
4624 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4625 prefetch_dynamic_strides.
4626 (exynosm1_prefetch_tune): Likewise.
4627 (thunderxt88_prefetch_tune): Likewise.
4628 (thunderx_prefetch_tune): Likewise.
4629 (thunderx2t99_prefetch_tune): Likewise.
4630 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4632 (aarch64_override_options_internal): Update to set
4633 PARAM_PREFETCH_DYNAMIC_STRIDES.
4634 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4635 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4636 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4637 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4638 prefetch-dynamic-strides setting.
4640 2018-05-23 Luis Machado <luis.machado@linaro.org>
4642 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4643 <minimum_stride>: New const int field.
4644 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4645 minimum_stride field defaulting to -1.
4646 (exynosm1_prefetch_tune): Likewise.
4647 (thunderxt88_prefetch_tune): Likewise.
4648 (thunderx_prefetch_tune): Likewise.
4649 (thunderx2t99_prefetch_tune): Likewise.
4650 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4651 <default_opt_level>: Set to 3.
4652 (aarch64_override_options_internal): Update to set
4653 PARAM_PREFETCH_MINIMUM_STRIDE.
4654 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4655 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4656 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4657 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4658 stride is constant and is below the minimum stride threshold.
4660 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4662 * config/arm/arm-cpus.in (mode26): Delete.
4663 (armv4): Delete mode26 reference.
4664 * config/arm/arm.c (arm_configure_build_target): Delete use of
4667 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4669 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4671 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4672 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4674 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4675 (floatunsdidf2): Ditto.
4677 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4679 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4680 (fixuns_trunc<mode>si2_avx512f): Ditto.
4681 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4682 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4683 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4685 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4687 PR rtl-optimization/79985
4688 * df-scan.c (df_insn_refs_collect): Remove special case for
4689 global registers and asm statements.
4691 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4693 * extend.texi (Global Register Variables): Rewrite the bullet list.
4694 Note that the register is available for allocation. Note that access
4695 via inline asm must use constraints. Add note about async-signal
4696 handlers. Remove paragraph about automagic register selection.
4698 2018-05-23 Richard Biener <rguenther@suse.de>
4700 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4701 of fixed offset from memset VN.
4703 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4705 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4707 (alloc_cand_and_find_basis): Initialize first_interp field.
4708 (slsr_process_mul): Modify first_interp field.
4709 (slsr_process_add): Likewise.
4710 (slsr_process_cast): Modify first_interp field for each new
4712 (slsr_process_copy): Likewise.
4713 (dump_candidate): Dump first_interp field.
4714 (replace_mult_candidate): Process all interpretations, not just
4716 (replace_rhs_if_not_dup): Likewise.
4717 (replace_one_candidate): Likewise.
4719 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4721 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4723 (aarch64_needs_frame_chain): New function.
4724 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4726 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4729 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4730 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4731 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4732 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4733 as true for strict-align.
4734 (aarch64_can_inline_p): Perform checks even when callee has no
4735 attributes to check for strict alignment.
4736 * doc/extend.texi (AArch64 Function Attributes): Document
4738 * doc/invoke.texi: (AArch64 Options): Likewise.
4740 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4742 PR tree-optimization/85853
4743 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4744 the handling of the root of the node to...
4745 (vect_slp_analyze_node_operations_1): ...this new function,
4746 and run the whole thing with the child nodes' def types
4747 set according to their SLP node's def type.
4749 2018-05-23 Richard Biener <rguenther@suse.de>
4752 * tree-data-ref.c (create_runtime_alias_checks): Defer
4753 and ignore overflow warnings.
4755 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4757 PR tree-optimization/85822
4758 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4761 2018-05-23 Richard Biener <rguenther@suse.de>
4763 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4764 memset constants via native_interpret_expr.
4766 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4769 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4771 (cgraph_node::create_alias): Likewise.
4772 (cgraph_node::get_availability): Check ifunc_resolver instead
4773 of looking up ifunc attribute.
4774 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4775 * varasm.c (do_assemble_alias): Likewise.
4776 (assemble_alias): Likewise.
4777 (default_binds_local_p_3): Likewise.
4778 * cgraph.h (cgraph_node): Add ifunc_resolver.
4779 (cgraph_node::only_called_directly_or_aliased_p): Return false
4781 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4783 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4784 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4785 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4786 instead of looking up ifunc attribute.
4788 2018-05-22 Luis Machado <luis.machado@linaro.org>
4790 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4792 2018-05-22 Martin Sebor <msebor@redhat.com>
4795 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4796 only when expasion succeeds.
4797 (expand_builtin_strcmp): Same.
4798 (expand_builtin_strncmp): Same.
4800 2018-05-22 Martin Sebor <msebor@redhat.com>
4802 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4804 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4805 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4807 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4808 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4810 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4811 Add check for legitimate_address.
4812 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4813 (aarch64_swap_ldrstr_operands): New.
4814 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4817 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4818 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4820 * config/aarch64/aarch64.md: New patterns to generate stp
4822 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4823 single words and double words.
4824 (load_pair_sw, load_pair_dw): Likewise.
4825 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4827 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4829 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4830 for different mode ldpstp and add peephole for merged zero stores.
4832 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4834 (aarch64_gen_store_pair): Rename calls to match new patterns.
4835 (aarch64_gen_load_pair): Rename calls to match new patterns.
4836 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4837 (load_pair<DREG:mode><DREG2:mode>): ... This.
4838 (store_pair<mode>): Rename to...
4839 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4840 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4842 (V_INT_EQUIV): Handle SImode.
4843 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4846 2018-05-22 Martin Sebor <msebor@redhat.com>
4849 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4850 or ajust the presumed bound on an operation to avoid unnecessary
4853 2018-05-22 Martin Sebor <msebor@redhat.com>
4855 PR tree-optimization/85826
4856 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4857 assuming that a DECL necesarily has a constant size.
4859 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4862 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4864 2018-05-22 Richard Biener <rguenther@suse.de>
4866 PR tree-optimization/85834
4867 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4868 non-constant and non-zero memset arguments.
4870 2018-05-22 Martin Liska <mliska@suse.cz>
4873 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4875 2018-05-22 Richard Biener <rguenther@suse.de>
4877 PR tree-optimization/85863
4878 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4879 comparisons when vectype is specified.
4880 (vectorizable_condition): Do not specify vectype for
4881 vect_is_simple_cond when SLP vectorizing.
4883 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4886 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4887 define __ibm128 as long double.
4888 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4889 as a distinct type when IEEE 128-bit support is enabled.
4890 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4891 128-bit types to use the correct functions.
4892 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4893 convert between 128-bit floating point types that have different
4894 modes but the same representation, instead of using gen_lowpart to
4896 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4898 (IFKF_reg): New attributes to give the register constraints for
4900 (extend<mode>tf2_internal): New insns to mark an explicit
4901 conversion between 128-bit floating point types that have a
4902 different mode but share the same representation.
4904 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4906 PR tree-optimization/85814
4907 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4908 a null return from get_strinfo when unsharing the next
4909 strinfo in the chain.
4911 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4914 * varasm.c (weak_finish): Clean up weak_decls.
4916 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4918 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4919 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4920 UNSPEC_UADALP values.
4921 * config/aarch64/iterators.md (ABAL): New int iterator.
4924 (sur): Add mappings for the above.
4925 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4927 (aarch64_<sur>abal<mode>_4): Likewise.
4928 (aarch64_<sur>adalp<mode>_3): Likewise.
4929 (<sur>sadv16qi): New define_expand.
4931 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4933 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4934 (*movdf_internal): Ditto.
4935 (*rcpsf2_sse): Ditto.
4936 (*rsqrtsf2_sse): Ditto.
4937 (*sqrt<mode>2_sse): Ditto.
4939 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4941 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4943 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4944 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4945 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4946 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4948 * config/aarch64/arm_neon.h: Likewise.
4949 * config/aarch64/iterators.md (VQ_I): New.
4951 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4953 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4955 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4958 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4960 * config/nds32/constraints.md (S): New constraint.
4961 * config/nds32/nds32.md (call_internal): Use constraint S.
4962 (call_value_internal): Likewise.
4963 (sibcall_internal): Likewise.
4964 (sibcall_value_internal): Likewise.
4966 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4967 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4972 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4973 Chung-Ju Wu <jasonwucj@gmail.com>
4975 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4976 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4977 (nds32_rtx_costs_impl): Simplify.
4978 (nds32_address_cost_impl): Simplify.
4979 (nds32_init_rtx_costs): New function.
4980 (nds32_rtx_costs_speed_prefer): Likewise.
4981 (nds32_rtx_costs_size_prefer): Likewise.
4982 (nds32_address_cost_speed_prefer): Likewise.
4983 (nds32_address_cost_speed_fwprop): Likewise.
4984 (nds32_address_cost_size_prefer): Likewise.
4985 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4986 * config/nds32/nds32.c (nds32_option_override): Use
4987 nds32_init_rtx_costs function.
4989 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4991 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4992 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4993 (TARGET_PIPELINE_N8): Likewise.
4994 (TARGET_PIPELINE_N10): Likewise.
4995 (TARGET_PIPELINE_N13): Likewise.
4996 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4998 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5000 * config/nds32/nds32-fpu.md: Update copyright year.
5002 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5004 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5006 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5008 * config/nds32/nds32.c
5009 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5010 * config/nds32/nds32.opt (minline-asm-r15): New option.
5012 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5014 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5016 * config/nds32/nds32.md (abssi2): New pattern.
5018 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5020 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5021 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5022 Merge insn pattern from sse_cvtsi2ss<round_name> and
5023 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5024 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5025 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5026 using SWI48 mode iterator.
5027 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5028 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5029 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5030 pattern from sse_cvttss2si<round_saeonly_name>
5031 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5032 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5033 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5034 using SWI48 mode iterator.
5035 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5036 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5037 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5038 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5039 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5040 using SWI48 mode iterator.
5041 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5042 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5043 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5044 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5045 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5046 SWI48 mode iterator.
5047 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5048 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5049 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5050 pattern from sse_cvttsd2si<round_saeonly_name>
5051 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5053 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5055 * config/nds32/nds32-md-auxiliary.c
5056 (nds32_valid_smw_lwm_base_p): Refine.
5057 (nds32_output_smw_single_word): Refine.
5058 (nds32_output_smw_double_word): New.
5059 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5061 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5063 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5064 (nds32_output_stack_pop): Refine.
5065 (nds32_expand_unaligned_load): Refine.
5066 (nds32_expand_unaligned_store): Refine.
5068 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5069 Chung-Ju Wu <jasonwucj@gmail.com>
5071 * config/nds32/constants.md: Add TP_REGNUM constant.
5072 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5073 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5075 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5076 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5077 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5078 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5079 and PIC code generation.
5080 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5082 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5084 * config/nds32/nds32.md: Support TLS and PIC.
5085 * config/nds32/nds32.c: Support TLS and PIC.
5086 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5087 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5090 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5092 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5093 mode with E_ prefix.
5095 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5096 Chung-Ju Wu <jasonwucj@gmail.com>
5098 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5099 * config/nds32/nds32-md-auxiliary.c
5100 (symbolic_reference_mentioned_p): New.
5101 (nds32_legitimize_ict_address): New.
5102 (nds32_expand_ict_move): New.
5103 (nds32_indirect_call_referenced_p): New.
5104 (nds32_symbol_binds_local_p): Delete.
5105 (nds32_long_call_p): Modify.
5106 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5107 * config/nds32/nds32-protos.h
5108 (symbolic_reference_mentioned_p): Declare.
5109 (nds32_legitimize_ict_address): Declare.
5110 (nds32_expand_ict_move): Declare.
5111 (nds32_indirect_call_referenced_p): Declare.
5112 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5113 (nds32_relax_group): Use nds32_ict_const_p as condition.
5114 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5115 (nds32_asm_file_start): Output ict_model directive in asm code.
5116 (nds32_legitimate_address_p): Consider indirect call.
5117 (nds32_print_operand): Consider indirect call.
5118 (nds32_print_operand_address): Consider indirect call.
5119 (nds32_insert_attributes): Handle "indirect_call" attribute.
5120 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5121 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5122 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5123 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5124 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5125 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5126 (TARGET_ICT_MODEL_SMALL): Define.
5127 (TARGET_ICT_MODEL_LARGE): Define.
5128 * config/nds32/nds32.md (movsi): Consider ict model.
5129 (call, call_value): Consider ict model.
5130 (sibcall, sibcall_value): Consider ict model.
5131 * config/nds32/nds32.opt (mict-model): New option.
5132 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5135 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5136 Monk Chiang <sh.chiang04@gmail.com>
5137 Jim Wilson <jimw@sifive.com>
5139 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5140 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5141 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5142 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5143 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5144 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5145 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5146 compute save_libcall_adjustment properly.
5147 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5148 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5149 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5150 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5151 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5152 (ABI_SPEC): Handle mabi=ilp32e.
5153 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5154 (RVE): Add RVE mask.
5155 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5156 <-march>: Add rv32e as an example.
5158 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5161 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5162 (intra_create_variable_infos): Handle C++ constructors.
5164 2018-05-18 Martin Liska <mliska@suse.cz>
5166 * passes.def: Remove a redundant pass.
5168 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5171 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5173 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5175 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5177 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5179 (armv2, armv2a, armv3, armv3m): Delete architectures.
5180 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5181 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5182 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5184 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5185 (*mulsidi3adddi): Likewise.
5186 (mulsidi3): Likewise.
5187 (*mulsidi3_nov6): Likewise.
5188 (umulsidi3): Likewise.
5189 (umulsidi3_nov6): Likewise.
5190 (umaddsidi4): Likewise.
5191 (*umulsidi3adddi): Likewise.
5192 (smulsi3_highpart): Likewise.
5193 (*smulsi3_highpart_nov6): Likewise.
5194 (umulsi3_highpart): Likewise.
5195 (*umulsi3_highpart_nov6): Likewise.
5196 * config/arm/arm.h (arm_arch3m): Delete.
5197 * config/arm/arm.c (arm_arch3m): Delete.
5198 (arm_option_override_internal): Update armv3-related comment.
5199 (arm_configure_build_target): Delete use of isa_bit_mode32.
5200 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5201 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5202 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5204 (mulusa3): Likewise.
5205 * config/arm/arm-protos.h (arm_arch3m): Delete.
5206 * config/arm/arm-tables.opt: Regenerate.
5207 * config/arm/arm-tune.md: Likewise.
5208 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5209 deleted architectures.
5211 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5213 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5214 (armv5t, armv5te): New features.
5215 (ARMv5, ARMv5e): Delete fgroups.
5216 (ARMv5t, ARMv5te): Adjust for above changes.
5218 (armv5, armv5e): Delete arches.
5219 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5221 (*call_reg_arm): Likewise.
5222 (*call_value_reg_armv5): Likewise.
5223 (*call_value_reg_arm): Likewise.
5224 (*call_symbol): Likewise.
5225 (*call_value_symbol): Likewise.
5226 (*sibcall_insn): Likewise.
5227 (*sibcall_value_insn): Likewise.
5229 (prefetch): Likewise.
5230 (define_split and define_peephole2 dependent on arm_arch5):
5232 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5234 (TARGET_ARM_QBIT): Likewise.
5235 (TARGET_DSP_MULTIPLY): Likewise.
5236 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5237 (arm_arch5, arm_arch5e): Delete.
5238 (arm_arch5t, arm_arch5te): Declare.
5239 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5240 (arm_arch5t): Declare.
5241 (arm_option_reconfigure_globals): Update for the above.
5242 (arm_options_perform_arch_sanity_checks): Update comment, replace
5243 use of arm_arch5 with arm_arch5t.
5244 (use_return_insn): Likewise.
5245 (arm_emit_call_insn): Likewise.
5246 (output_return_instruction): Likewise.
5247 (arm_final_prescan_insn): Likewise.
5248 (arm_coproc_builtin_available): Likewise.
5249 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5250 arm_arch5e with arm_arch5t and arm_arch5te.
5251 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5252 (arm_arch5t, arm_arch5te): Declare.
5253 * config/arm/arm-tables.opt: Regenerate.
5254 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5255 * config/arm/t-multilib: Likewise.
5256 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5257 instead of arm_arch5.
5258 (*call_reg_thumb1): Likewise.
5259 (*call_value_reg_thumb1_v5): Likewise.
5260 (*call_value_reg_thumb1): Likewise.
5261 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5263 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5265 2018-05-18 Martin Liska <mliska@suse.cz>
5267 PR gcov-profile/84846
5268 * doc/gcov.texi: Document -t option of gcov tool.
5270 2018-05-18 Martin Liska <mliska@suse.cz>
5272 PR gcov-profile/84846
5273 * gcov.c (print_usage): Add new -t option.
5274 (process_args): Handle the option.
5275 (generate_results): Use stdout as output when requested by
5278 2018-05-18 Martin Liska <mliska@suse.cz>
5280 PR gcov-profile/84846
5281 * coverage.c (coverage_init): Write PWD to .gcno file.
5282 * doc/gcov.texi: Document how working directory is printed.
5283 * gcov-dump.c (dump_gcov_file): Print PWD.
5284 * gcov.c (output_intermediate_file): Likewise.
5285 (read_graph_file): Read PWD string.
5286 (output_lines): Print PWD.
5288 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5291 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5292 for retval and return false if all args to phi are zero.
5294 2018-05-18 Richard Biener <rguenther@suse.de>
5296 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5298 (evrp_dom_walker::before_dom_children): Call it.
5300 2018-05-18 Richard Biener <rguenther@suse.de>
5302 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5303 results when processing array refs with variable index.
5305 2018-05-18 Toon Moene <toon@moene.org>
5307 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5308 directly after that of -floop-interchange. Indicate that both
5309 options are enabled by default when specifying -O3.
5311 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5313 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5314 iterator. Delete separate integer-mode vec_set<mode> expander.
5315 (aarch64_simd_vec_setv2di): Delete.
5316 (vec_setv2di): Delete.
5317 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5318 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5319 the "w, r" alternative.
5321 2018-05-18 Martin Liska <mliska@suse.cz>
5323 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5324 * tree-pass.h (make_pass_lower_switch_O0): New function.
5325 * tree-switch-conversion.c (node_has_low_bound): Remove.
5326 (node_has_high_bound): Likewise.
5327 (node_is_bounded): Likewise.
5328 (class pass_lower_switch): Make it a template type and create
5330 (pass_lower_switch::execute): Add template argument.
5331 (make_pass_lower_switch): New function.
5332 (make_pass_lower_switch_O0): New function.
5333 (do_jump_if_equal): Remove.
5334 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5335 all the hard work to tree optimization passes.
5337 2018-05-18 Martin Liska <mliska@suse.cz>
5339 * dbgcnt.c (limit_low): Renamed from limit.
5340 (limit_high): New variable.
5341 (dbg_cnt_is_enabled): Check for upper limit.
5342 (dbg_cnt): Adjust dumping.
5343 (dbg_cnt_set_limit_by_index): Add new argument for high
5345 (dbg_cnt_set_limit_by_name): Likewise.
5346 (dbg_cnt_process_single_pair): Parse new format.
5347 (dbg_cnt_process_opt): Use strtok.
5348 (dbg_cnt_list_all_counters): Remove 'value' and add
5350 * doc/invoke.texi: Document changes.
5352 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5354 * doc/sourcebuild.texi (scalar_all_fma): Document.
5355 * tree.def (FMA_EXPR): Delete.
5356 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5357 * internal-fn.c (ternary_direct): New macro.
5358 (expand_ternary_optab_fn): Likewise.
5359 (direct_ternary_optab_supported_p): Likewise.
5360 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5361 * builtins.c (fold_builtin_fma): Delete.
5362 (fold_builtin_3): Don't call it.
5363 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5364 * expr.c (expand_expr_real_2): Likewise.
5365 * fold-const.c (operand_equal_p): Likewise.
5366 (fold_ternary_loc): Likewise.
5367 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5368 * gimple.c (DEFTREECODE): Likewise.
5369 * gimplify.c (gimplify_expr): Likewise.
5370 * optabs-tree.c (optab_for_tree_code): Likewise.
5371 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5372 * tree-eh.c (operation_could_trap_p): Likewise.
5373 (stmt_could_throw_1_p): Likewise.
5374 * tree-inline.c (estimate_operator_cost): Likewise.
5375 * tree-pretty-print.c (dump_generic_node): Likewise.
5376 (op_code_prio): Likewise.
5377 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5378 * tree-ssa-operands.c (get_expr_operands): Likewise.
5379 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5380 * fold-const-call.h (fold_fma): Delete.
5381 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5382 CFN_FNMA and CFN_FNMS.
5384 * genmatch.c (combined_fn): New enum.
5385 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5386 (commutative_op): New function.
5387 (commutate): Use it. Handle more than 2 operands.
5388 (dt_operand::gen_gimple_expr): Use commutative_op.
5389 (parser::parse_expr): Allow :c to be used with non-binary
5390 operators if the commutative operand is known.
5391 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5392 CFN_FMS, CFN_FNMA and CFN_FNMS.
5393 (backprop::process_assign_use): Remove FMA_EXPR handling.
5394 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5395 (gen_hsa_fma): New function.
5396 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5397 IFN_FNMA and IFN_FNMS.
5398 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5399 * gimple-fold.h (follow_all_ssa_edges): Declare.
5400 * gimple-fold.c (follow_all_ssa_edges): New function.
5401 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5402 gimple_build interface and use follow_all_ssa_edges to fold the result.
5403 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5404 instead of checking for optabs directly.
5405 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5406 rather than FMA_EXPRs.
5407 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5408 call to IFN_FMA instead of an FMA_EXPR.
5410 2018-05-17 Jim Wilson <jimw@sifive.com>
5412 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5413 sign extended promoted subreg, and the range does not have the sign bit
5414 set, then do a sign extend.
5416 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5417 test, check for sign extended subreg and/or constant operands, and
5418 do a sign extend in that case.
5420 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5422 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5423 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5425 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5426 Change logics_shift_reg to logics_shift_imm.
5427 (thunderx2t99_fp_loadpair_basic): Delete.
5428 (thunderx2t99_fp_storepair_basic): Delete.
5429 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5430 (thunderx2t99_asimd_polynomial): Delete.
5431 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5432 and neon_fp_mul_d_scalar_q.
5433 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5434 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5435 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5436 (thunderx2t99_asimd_lut): Add missing tbl types.
5437 (thunderx2t99_asimd_ext): Delete.
5438 (thunderx2t99_asimd_load1_1_mult): Delete.
5439 (thunderx2t99_asimd_load1_2_mult): Delete.
5440 (thunderx2t99_asimd_load1_ldp): New.
5441 (thunderx2t99_asimd_load1): New.
5442 (thunderx2t99_asimd_load2): Add missing *load2* types.
5443 (thunderx2t99_asimd_load3): New.
5444 (thunderx2t99_asimd_load4): New.
5445 (thunderx2t99_asimd_store1_1_mult): Delete.
5446 (thunderx2t99_asimd_store1_2_mult): Delete.
5447 (thunderx2t99_asimd_store2_mult): Delete.
5448 (thunderx2t99_asimd_store2_onelane): Delete.
5449 (thunderx2t99_asimd_store_stp): New.
5450 (thunderx2t99_asimd_store1): New.
5451 (thunderx2t99_asimd_store2): New.
5452 (thunderx2t99_asimd_store3): New.
5453 (thunderx2t99_asimd_store4): New.
5455 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5457 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5458 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5460 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5461 Segher Boessenkool <segher@kernel.crashing.org>
5464 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5467 2018-05-17 Richard Biener <rguenther@suse.de>
5469 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5470 for pruning loop and prune defs feeding only already visited PHIs.
5472 2018-05-17 Richard Biener <rguenther@suse.de>
5474 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5476 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5477 Richard Biener <rguenther@suse.de>
5479 PR tree-optimization/85793
5480 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5481 for VMAT_ELEMENTWISE.
5483 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5485 * internal-fn.h (lookup_internal_fn): Declare
5486 * internal-fn.c (lookup_internal_fn): New function.
5487 * gimple.c (gimple_build_call_from_tree): Handle calls to
5489 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5490 internal function names.
5491 * tree-pretty-print.c (dump_generic_node): Likewise.
5492 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5494 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5496 * gimple-fold.h (gimple_build): Make the function forms take
5497 combined_fn rather than built_in_function.
5498 (gimple_simplify): Likewise.
5499 * gimple-match-head.c (gimple_simplify): Likewise.
5500 * gimple-fold.c (gimple_build): Likewise.
5501 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5502 rather than gimple_build_call_internal.
5503 (get_initial_defs_for_reduction): Likewise.
5504 (vect_create_epilog_for_reduction): Likewise.
5505 (vectorizable_live_operation): Likewise.
5507 2018-05-17 Martin Liska <mliska@suse.cz>
5509 * gimple-ssa-sprintf.c (format_directive): Do not use
5510 space in between 'G_' and '('.
5512 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5515 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5516 even if the mask is not all ones.
5519 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5521 (ix86_gimple_fold_builtin): Likewise.
5524 * config/i386/i386.c: Include tree-vector-builder.h.
5525 (ix86_vector_shift_count): New function.
5526 (ix86_fold_builtin): Fold shift builtins by scalar count.
5527 (ix86_gimple_fold_builtin): Likewise.
5529 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5530 _mm512_setzero): New intrinsics.
5532 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5533 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5535 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5536 code generation for cases where splatting a value is not useful.
5537 * simplify-rtx.c (simplify_ternary_operation): Simplify
5538 vec_merge across a vec_duplicate and a paradoxical subreg forming
5539 a vector mode to a vec_concat.
5541 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5543 * config.gcc: Support "goldmont-plus".
5544 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5546 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5547 PROCESSOR_GOLDMONT_PLUS.
5548 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5549 (processor_target_table): Add "goldmont-plus".
5550 (PTA_GOLDMONT_PLUS): Define.
5551 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5552 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5553 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5554 (fold_builtin_cpu): Add "goldmont-plus".
5555 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5556 (ix86_option_override_internal): Add "goldmont-plus".
5557 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5558 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5559 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5560 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5562 2018-05-17 Richard Biener <rguenther@suse.de>
5564 PR tree-optimization/85757
5565 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5566 remove defs that only feed that PHI from further processing.
5568 2018-05-16 Jim Wilson <jimw@sifive.com>
5570 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5572 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5574 2018-05-16 Mark Wielaard <mark@klomp.org>
5576 * dwarf2out.c (count_index_strings): New function.
5577 (output_indirect_strings): Call count_index_strings and generate
5578 header for dwarf_version >= 5.
5580 2018-05-16 Mark Wielaard <mark@klomp.org>
5582 * dwarf2out.c (dwarf_FORM): New function.
5583 (set_indirect_string): Use dwarf_FORM.
5584 (reset_indirect_string): Likewise.
5585 (size_of_die): Likewise.
5586 (value_format): Likewise.
5587 (output_die): Likewise.
5588 (add_skeleton_AT_string): Likewise.
5589 (output_macinfo_op): Likewise.
5590 (index_string): Likewise.
5591 (output_index_string_offset): Likewise.
5592 (output_index_string): Likewise.
5593 (count_index_strings): Likewise.
5595 2018-05-16 Carl Love <cel@us.ibm.com>
5597 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5598 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5600 2018-05-16 Martin Jambor <mjambor@suse.cz>
5602 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5603 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5605 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5607 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5608 (fnma<mode>4): Likewise.
5609 (fms<mode>4): Likewise.
5610 (fnms<mode>4): Likewise.
5611 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5612 (aarch64_fnma<mode>4): Likewise.
5613 (aarch64_fms<mode>4): Likewise.
5614 (aarch64_fnms<mode>4): Likewise.
5615 (aarch64_fnmadd<mode>4): Likewise.
5617 2018-05-16 Jason Merrill <jason@redhat.com>
5619 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5621 2018-05-16 Richard Biener <rguenther@suse.de>
5623 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5624 (dump_stmt_cost): Declare.
5625 (add_stmt_cost): Dump cost we add.
5626 (add_stmt_costs): New function.
5627 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5629 (vect_analyze_stmt): Adjust prototype.
5630 (vectorizable_condition): Likewise.
5631 (vectorizable_live_operation): Likewise.
5632 (vectorizable_reduction): Likewise.
5633 (vectorizable_induction): Likewise.
5634 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5635 cost vector to pass to vectorizable_ and record afterwards.
5636 (vect_model_reduction_cost): Take cost vector argument and adjust.
5637 (vect_model_induction_cost): Likewise.
5638 (vectorizable_reduction): Likewise.
5639 (vectorizable_induction): Likewise.
5640 (vectorizable_live_operation): Likewise.
5641 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5642 SLP_TREE_NUMBER_OF_VEC_STMTS.
5643 (vect_analyze_slp_cost_1): Remove.
5644 (vect_analyze_slp_cost): Likewise.
5645 (vect_slp_analyze_node_operations): Take visited args and
5646 a target cost vector. Avoid processing already visited stmt sets.
5647 (vect_slp_analyze_operations): Use a local cost vector to gather
5648 costs and register those of non-discarded instances.
5649 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5650 (vect_schedule_slp_instance): Remove copying of
5651 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5653 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5654 adding cost. Record cost entry location.
5655 (vect_prologue_cost_for_slp_op): Function to compute cost of
5656 a constant or invariant generated for SLP vect in the prologue,
5657 split out from vect_analyze_slp_cost_1.
5658 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5659 (vect_model_promotion_demotion_cost): Likewise.
5660 (vect_model_store_cost): Likewise, make static.
5661 (vect_model_load_cost): Likewise.
5662 (vectorizable_bswap): Add cost vector arg and adjust.
5663 (vectorizable_call): Likewise.
5664 (vectorizable_simd_clone_call): Likewise.
5665 (vectorizable_conversion): Likewise.
5666 (vectorizable_assignment): Likewise.
5667 (vectorizable_shift): Likewise.
5668 (vectorizable_operation): Likewise.
5669 (vectorizable_store): Likewise.
5670 (vectorizable_load): Likewise.
5671 (vectorizable_condition): Likewise.
5672 (vectorizable_comparison): Likewise.
5673 (can_vectorize_live_stmts): Likewise.
5674 (vect_analyze_stmt): Likewise.
5675 (vect_transform_stmt): Adjust calls to vectorizable_*.
5676 * tree-vectorizer.c: Include gimple-pretty-print.h.
5677 (dump_stmt_cost): New function.
5679 2018-05-16 Richard Biener <rguenther@suse.de>
5681 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5682 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5683 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5684 (check_name): New callback.
5685 (dse_classify_store): Track cycles via a visited bitmap of PHI
5686 defs and simplify handling of in-loop and across loop dead stores
5687 and properly fail for loop-variant refs. Handle byte-tracking with
5688 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5691 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5693 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5694 (vect_get_mask_type_for_stmt): Likewise.
5695 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5697 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5698 to determine the statement's vector type and the vector type that
5699 should be used for calculating nunits. Deal with cases in which
5700 the type has to be deferred.
5701 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5702 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5703 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5704 (vect_determine_vf_for_stmt): New functions, split out from...
5705 (vect_determine_vectorization_factor): ...here.
5706 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5707 (vect_get_mask_type_for_stmt): New functions, split out from
5708 vect_determine_vectorization_factor.
5710 2018-05-16 Richard Biener <rguenther@suse.de>
5712 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5713 verify the [VEC_]COND_EXPR embedded comparison.
5715 2018-05-15 Martin Sebor <msebor@redhat.com>
5717 PR tree-optimization/85753
5718 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5719 RECORD_TYPE in addition to ARRAY_TYPE.
5721 2018-05-15 Martin Sebor <msebor@redhat.com>
5724 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5726 2018-05-15 Richard Biener <rguenther@suse.de>
5728 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5729 add by_clobber_p one. Change algorithm to collect all defs
5730 representing uses we need to walk and try reducing them to
5731 a single one before failing.
5732 (dse_dom_walker::dse_optimize_stmt): Adjust.
5734 2018-05-13 Mark Wielaard <mark@klomp.org>
5736 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5737 (size_of_loc_descr): Likewise.
5738 (output_loc_operands): Likewise.
5739 (output_loc_operands_raw): Likewise.
5740 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5741 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5742 (hash_loc_operands): Likewise.
5743 (compare_loc_operands): Likewise.
5745 2018-05-14 Mark Wielaard <mark@klomp.org>
5747 * dwarf2out.c (count_index_addrs): New function.
5748 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5750 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5752 PR tree-optimization/83648
5753 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5754 return value as malloc candidate.
5756 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5759 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5760 param as true in call to suggest_attribute.
5762 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5764 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5765 -mreadonly-in-sdata.
5767 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5769 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5771 (aarch64_crypto_aesd_fused): Likewise.
5773 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5775 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5776 (movsi_aarch64): Likewise.
5777 (load_pairsi): Likewise.
5778 (load_pairdi): Likewise.
5779 (store_pairsi): Likewise.
5780 (store_pairdi): Likewise.
5781 (load_pairsf): Likewise.
5782 (load_pairdf): Likewise.
5783 (store_pairsf): Likewise.
5784 (store_pairdf): Likewise.
5785 (zero_extend): Likewise.
5786 (trunc): Swap alternatives.
5787 (fcvt_target): Add '?' to prefer w over r.
5789 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5792 * config/i386/i386.md: Disallow non-commutative arithmetics in
5793 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5794 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5795 in the peephole2 before it.
5797 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5799 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5800 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5801 (ix86_handle_option): Handle -mcldemote.
5802 * config.gcc: New header.
5803 * config/i386/cldemoteintrin.h: New file.
5804 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5805 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5807 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5808 OPTION_MASK_ISA_CLDEMOTE.
5809 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5810 (ix86_valid_target_attribute_inner_p): Ditto.
5811 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5812 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5813 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5814 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5815 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5817 * config/i386/i386.opt: Add -mcldemote.
5818 * config/i386/x86intrin.h: New header.
5819 * doc/invoke.texi: Add -mcldemote.
5821 2018-05-14 Richard Biener <rguenther@suse.de>
5823 * doc/match-and-simplify.texi: Adjust :s documentation.
5825 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5827 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5828 intended memcpy size.
5829 (REORDER_45): Likewise.
5831 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5833 * sort.cc: New file.
5834 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5835 * vec.c (qsort_chk): Use gcc_qsort.
5836 * Makefile.in (OBJS-libcommon): Add sort.o.
5837 (build/sort.o): New target. Use it...
5838 (BUILD_RTL): ... here, and...
5839 (build/gencfn-macros): ... here, and...
5840 (build/genmatch): ... here.
5842 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5843 Chung-Ju Wu <jasonwucj@gmail.com>
5845 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5846 * config/nds32/nds32-graywolf.md: New file.
5847 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5848 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5850 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5851 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5852 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5853 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5854 * config/nds32/pipelines.md: Include n15 settings.
5856 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5857 Chung-Ju Wu <jasonwucj@gmail.com>
5859 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5860 * config/nds32/nds32-n13.md: New file.
5861 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5862 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5864 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5865 * config/nds32/nds32.md (pipeline_model): Add n13.
5866 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5867 * config/nds32/pipelines.md: Include n13 settings.
5869 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5870 Chung-Ju Wu <jasonwucj@gmail.com>
5872 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5873 * config/nds32/nds32-n10.md: New file.
5874 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5875 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5877 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5878 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5879 * config/nds32/nds32.md (pipeline_model): Add n10.
5880 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5881 * config/nds32/pipelines.md: Include n10 settings.
5883 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5884 Kito Cheng <kito.cheng@gmail.com>
5885 Chung-Ju Wu <jasonwucj@gmail.com>
5887 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5888 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5889 Add enum values for DSP extension instructions.
5890 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5892 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5893 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5895 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5896 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5897 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5898 * config/nds32/nds32-intrinsic.md: Likewise.
5899 * config/nds32/nds32_intrinsic.h: Likewise.
5900 * config/nds32/nds32-md-auxiliary.c: Likewise.
5901 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5902 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5903 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5904 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5905 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5906 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5907 TYPE_DMAC in switch statement.
5908 * config/nds32/nds32.c: New checking and implementation for DSP
5909 extension instructions.
5910 * config/nds32/nds32.h: Likewise.
5911 * config/nds32/nds32.md: Likewise.
5912 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5913 * config/nds32/predicates.md: Implement new predicates for DSP
5916 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5918 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5919 Reformat alternatives and attributes so it is easier to identify
5920 which constraints/attributes go with which instruction.
5921 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5922 (mov<mode>_softfloat32, FMOVE64): Likewise.
5923 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5924 (mov<mode>_softfloat64, FMOVE64): Likewise.
5926 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5928 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5930 (Basic PowerPC Built-in Functions): The new name of the
5931 subsection previously known as "PowerPC Built-in Functions".
5932 (Basic PowerPC Built-in Functions Available on all Configurations):
5934 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5935 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5936 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5937 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5939 2018-05-11 Martin Jambor <mjambor@suse.cz>
5942 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5945 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5948 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5950 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5952 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5953 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5954 (ix86_handle_option): Handle -mwaitpkg.
5955 * config.gcc: New header.
5956 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5957 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5958 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5960 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5961 OPTION_MASK_ISA_WAITPKG.
5962 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5963 (ix86_option_override_internal): Add PTA_WAITPKG.
5964 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5965 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5966 IX86_BUILTIN_TPAUSE.
5967 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5968 __builtin_ia32_umwait and __builtin_ia32_tpause.
5969 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5970 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5971 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5972 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5973 UNSPECV_TPAUSE): New.
5974 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5975 * config/i386/i386.opt: Add -mwaitpkg.
5976 * config/i386/waitpkgintrin.h: New file.
5977 * config/i386/x86intrin.h: New header.
5978 * doc/invoke.texi: Add -mwaitpkg.
5980 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5983 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5985 (cortex-m0): Use armv6s-m isa.
5986 (cortex-m0plus): Likewise.
5987 (cortex-m1): Likewise.
5988 (cortex-m0.small-multiply): Likewise.
5989 (cortex-m0plus.small-multiply): Likewise.
5990 (cortex-m1.small-multiply): Likewise.
5992 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5993 Jakub Jelinek <jakub@redhat.com>
5995 PR tree-optimization/85692
5996 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5997 source permute as well.
5999 2018-05-11 Martin Liska <mliska@suse.cz>
6002 * doc/extend.texi: Document LLVM style format for no_sanitize
6005 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6007 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6008 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6009 (mode_supports_vsx_dform_quad): Likewise.
6010 (mode_supports_vmx_dform): Move these functions to be next to the
6011 other mode_supports functions.
6012 (mode_supports_dq_form): Likewise.
6013 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6014 mode_supports_dq_form.
6015 (reg_offset_addressing_ok_p): Likewise.
6016 (offsettable_ok_by_alignment): Likewise.
6017 (rs6000_legitimate_offset_address_p): Likewise.
6018 (legitimate_lo_sum_address_p): Likewise.
6019 (rs6000_legitimize_address): Likewise.
6020 (rs6000_legitimize_reload_address): Likewise.
6021 (rs6000_secondary_reload_inner): Likewise.
6022 (rs6000_preferred_reload_class): Likewise.
6023 (rs6000_output_move_128bit): Likewise.
6025 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6027 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6028 Generate SImode target register for null target.
6029 <case IX86_BUILTIN_XGETBV>: Ditto.
6030 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6031 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6033 2018-05-10 Carl Love <cel@us.ibm.com>
6035 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6036 dcbtt and dcbtstt if operands[2] is 0.
6038 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6041 * config/i386/sse.md (usadv64qi): New expander.
6043 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6045 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6046 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6047 -maltivec=be support.
6048 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6049 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6050 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6051 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6052 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6053 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6054 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6055 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6056 altivec_vsumsws): Adjust.
6057 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6058 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6059 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6061 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6062 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6063 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6064 (altivec_lve<VI_char>x): Delete expand.
6065 (*altivec_lve<VI_char>x_internal): Rename to...
6066 (altivec_lve<VI_char>x): ... this.
6067 (altivec_lvxl_<mode>): Delete expand.
6068 (*altivec_lvxl_<mode>_internal): Rename to ...
6069 (altivec_lvxl_<mode>): ... this.
6070 (altivec_stvxl_<mode>): Delete expand.
6071 (*altivec_stvxl_<mode>_internal): Rename to ...
6072 (altivec_stvxl_<mode>): ... this.
6073 (altivec_stve<VI_char>x): Delete expand.
6074 (*altivec_stve<VI_char>x_internal): Rename to ...
6075 (altivec_stve<VI_char>x): ... this.
6076 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6077 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6078 reduc_plus_scal_<mode>): Adjust.
6079 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6081 (rs6000_cpu_cpp_builtins): Adjust.
6082 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6083 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6084 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6085 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6086 -maltivec=be support.
6087 (rs6000_split_vec_extract_var): Adjust.
6088 (rs6000_split_v4si_init): Adjust.
6089 (swap_selector_for_mode): Delete.
6090 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6091 altivec_expand_stvex_be): Delete.
6092 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6093 -maltivec=be support.
6094 (rs6000_gimple_fold_builtin): Ditto.
6095 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6097 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6098 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6099 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6100 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6101 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6102 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6103 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6104 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6105 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6106 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6107 anonymous split): Adjust.
6108 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6109 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6111 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6113 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6114 when --with-gxx-include-dir is also specified.
6115 * configure: Regenerate.
6117 2018-05-09 Jim Wilson <jimw@sifive.com>
6120 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6121 * config/riscv/t-withmultilib: New.
6122 * config/riscv/withmultilib.h: New.
6123 * doc/install.texi: Document RISC-V --with-multilib-list support.
6125 2018-05-09 Richard Biener <rguenther@suse.de>
6127 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6129 (vect_bb_vectorization_profitable_p): Adjust. Compute
6130 actual scalar cost using the cost vector and the add_stmt_cost
6133 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6135 PR rtl-optimization/85645
6136 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6137 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6139 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6141 PR rtl-optimization/85645
6142 * shrink-wrap.c (spread_components): Return a boolean saying if
6143 anything was changed.
6144 (try_shrink_wrapping_separate): Iterate spread_components until
6145 nothing changes anymore.
6147 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6149 PR rtl-optimization/85645
6150 * regrename.c (build_def_use): Also kill the chains that include the
6151 destination of a REG_CFA_REGISTER note.
6153 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6155 PR rtl-optimization/85645
6156 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6157 insn that has a REG_CFA_REGISTER note.
6159 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6161 * cfgexpand.c (expand_clobber): New function.
6162 (expand_gimple_stmt_1): Use it.
6163 * tree-vect-stmts.c (vect_clobber_variable): New function,
6165 (vectorizable_simd_clone_call): ...here.
6166 (vectorizable_store): Emit a clobber either side of an
6167 IFN_STORE_LANES sequence.
6168 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6170 2018-05-09 Tom de Vries <tom@codesourcery.com>
6173 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6174 (define_insn "trap_if_false"): Add exit after trap.
6176 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6178 PR rtl-optimization/85638
6179 * bb-reorder.c: Include common/common-target.h.
6180 (create_forwarder_block): New function extracted from...
6181 (fix_up_crossing_landing_pad): ...here. Rename into...
6182 (dw2_fix_up_crossing_landing_pad): ...this.
6183 (sjlj_fix_up_crossing_landing_pad): New function.
6184 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6185 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6186 from both partitions and exit the loop after one iteration.
6188 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6191 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6193 (Basic PowerPC Built-in Functions): The new name of the
6194 subsection previously known as "PowerPC Built-in Functions".
6195 (Basic PowerPC Built-in Functions Available on all Configurations):
6197 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6199 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6201 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6203 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6206 2018-05-08 Jim Wilson <jimw@sifive.com>
6208 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6209 (LD_EMUL_SUFFIX): New.
6210 (LINK_SPEC): Use it.
6212 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6214 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6216 (Basic PowerPC Built-in Functions): The new name of the
6217 subsection previously known as "PowerPC Built-in Functions".
6218 (Basic PowerPC Built-in Functions Available on all Configurations):
6220 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6222 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6224 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6226 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6229 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6232 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6233 after cmpelim optimization.
6235 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6237 * config.gcc: Support "goldmont".
6238 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6239 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6241 * config/i386/i386.c (m_GOLDMONT): Define.
6242 (processor_target_table): Add "goldmont".
6243 (PTA_GOLDMONT): Define.
6244 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6245 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6246 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6247 (fold_builtin_cpu): Add "goldmont".
6248 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6249 (ix86_option_override_internal): Add "goldmont".
6250 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6251 (processor_type): Add PROCESSOR_GOLDMONT.
6252 * config/i386/i386.md: Add CPU "glm".
6253 * config/i386/glm.md: New file.
6254 * config/i386/x86-tune.def: Add m_GOLDMONT.
6255 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6257 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6260 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6262 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6263 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6264 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6267 * config/i386/i386.c (ix86_fold_builtin): Handle
6268 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6271 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6272 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6274 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6277 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6278 (check_arch): Likewise.
6279 (check_fpu): Return the result rather than printing it.
6280 (end arch): Fix operator precedence.
6281 (end cpu): Likewise.
6282 (END): Print the result from check_fpu.
6284 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6285 Alan Hayward <alan.hayward@arm.com>
6286 David Sherwood <david.sherwood@arm.com>
6288 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6289 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6290 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6291 (*fcmuo<mode>_and): New patterns.
6293 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6295 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6296 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6297 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6298 (cmp_op, sve_imm_con): New code attributes.
6299 (SVE_COND_INT_CMP, imm_con): Delete.
6300 (cmp_op): Remove above unspecs from int attribute.
6301 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6303 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6304 comparison-specific unspecs.
6305 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6306 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6307 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6308 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6309 (*vec_fcm<cmp_op><mode>): Rename to...
6310 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6311 (*vec_fcmuo<mode>): Rename to...
6312 (*fcmuo<mode>): ...this and adjust likewise.
6313 (*pred_fcm<cmp_op><mode>): New pattern.
6314 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6315 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6317 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6319 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6320 (aarch64_emit_sve_predicated_cond): New function.
6321 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6322 (aarch64_emit_unspec_cond_or): Replace with...
6323 (aarch64_emit_sve_or_conds): ...this new function. Use
6324 aarch64_emit_sve_ptrue_op for the individual comparisons and
6325 aarch64_emit_binop to OR them together.
6326 (aarch64_emit_inverted_unspec_cond): Replace with...
6327 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6328 aarch64_emit_sve_ptrue_op for the comparison and
6329 aarch64_emit_unop to invert the result.
6330 (aarch64_expand_sve_vec_cmp_float): Update after the above
6331 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6333 2018-05-07 Nathan Sidwell <nathan@acm.org>
6335 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6336 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6337 (Backwards Compatibility): Likewise.
6339 2018-05-07 Luis Machado <luis.machado@linaro.org>
6343 2018-05-07 Luis Machado <luis.machado@linaro.org>
6345 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6346 <prefetch_dynamic_strides>: New const bool field.
6347 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6348 prefetch_dynamic_strides.
6349 (exynosm1_prefetch_tune): Likewise.
6350 (thunderxt88_prefetch_tune): Likewise.
6351 (thunderx_prefetch_tune): Likewise.
6352 (thunderx2t99_prefetch_tune): Likewise.
6353 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6355 (aarch64_override_options_internal): Update to set
6356 PARAM_PREFETCH_DYNAMIC_STRIDES.
6357 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6358 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6359 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6360 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6361 prefetch-dynamic-strides setting.
6363 2018-05-07 Luis Machado <luis.machado@linaro.org>
6365 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6366 <minimum_stride>: New const int field.
6367 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6368 minimum_stride field.
6369 (exynosm1_prefetch_tune): Likewise.
6370 (thunderxt88_prefetch_tune): Likewise.
6371 (thunderx_prefetch_tune): Likewise.
6372 (thunderx2t99_prefetch_tune): Likewise.
6373 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6374 (aarch64_override_options_internal): Update to set
6375 PARAM_PREFETCH_MINIMUM_STRIDE.
6376 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6377 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6378 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6379 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6380 stride is constant and is below the minimum stride threshold.
6382 2018-05-07 Luis Machado <luis.machado@linaro.org>
6384 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6387 2018-05-07 Luis Machado <luis.machado@linaro.org>
6389 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6390 <prefetch_dynamic_strides>: New const bool field.
6391 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6392 prefetch_dynamic_strides.
6393 (exynosm1_prefetch_tune): Likewise.
6394 (thunderxt88_prefetch_tune): Likewise.
6395 (thunderx_prefetch_tune): Likewise.
6396 (thunderx2t99_prefetch_tune): Likewise.
6397 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6399 (aarch64_override_options_internal): Update to set
6400 PARAM_PREFETCH_DYNAMIC_STRIDES.
6401 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6402 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6403 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6404 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6405 prefetch-dynamic-strides setting.
6407 2018-05-07 Luis Machado <luis.machado@linaro.org>
6409 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6410 <minimum_stride>: New const int field.
6411 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6412 minimum_stride field.
6413 (exynosm1_prefetch_tune): Likewise.
6414 (thunderxt88_prefetch_tune): Likewise.
6415 (thunderx_prefetch_tune): Likewise.
6416 (thunderx2t99_prefetch_tune): Likewise.
6417 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6418 (aarch64_override_options_internal): Update to set
6419 PARAM_PREFETCH_MINIMUM_STRIDE.
6420 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6421 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6422 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6423 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6424 stride is constant and is below the minimum stride threshold.
6426 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6429 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6430 the type is addressable. Don't force op into register if it has
6433 2018-05-05 Roland McGrath <mcgrathr@google.com>
6436 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6437 any section for which we don't know a specific type it should have,
6438 regardless of name. Previously this was done only for the exact
6439 names ".init_array", ".fini_array", and ".preinit_array".
6440 (default_elf_asm_named_section): Add comment about
6441 relationship with default_section_type_flags and SECTION_NOTYPE.
6442 (get_section): Don't consider it a type conflict if one side has
6443 SECTION_NOTYPE and the other doesn't, as long as neither has the
6444 SECTION_BSS et al used in the default_section_type_flags logic.
6446 2018-05-05 Tom de Vries <tom@codesourcery.com>
6449 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6450 (workaround_barsyncs): New function.
6451 (nvptx_reorg): Use workaround_barsyncs.
6452 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6453 (define_expand "nvptx_membar_cta"): New define_expand.
6454 (define_insn "*nvptx_membar_cta"): New insn.
6456 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6458 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6459 To improve optimization opportunities.
6460 * builtin-types.def: The new needed builtin types for the above.
6462 2018-05-04 Richard Biener <rguenther@suse.de>
6464 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6465 * gimple-ssa-store-merging.c
6466 (imm_store_chain_info::output_merged_store): Remove redundant create,
6467 release split_store vector contents on failure.
6468 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6469 scalar stmt vector on cache hit.
6471 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6473 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6475 * config.gcc (powerpc-xilinx-eabi*): Remove.
6476 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6478 (fusion_addis_mem_combo_load): Ditto.
6479 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6481 (rs6000_cpu_cpp_builtins): Ditto.
6482 * config/rs6000/rs6000-linux.c
6483 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6484 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6485 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6487 (rs6000_setup_reg_addr_masks): Ditto.
6488 (rs6000_init_hard_regno_mode_ok): Ditto.
6489 (rs6000_option_override_internal): Ditto.
6490 (legitimate_lo_sum_address_p): Ditto.
6491 (rs6000_legitimize_address): Ditto.
6492 (rs6000_legitimize_reload_address): Ditto.
6493 (rs6000_legitimate_address_p): Ditto.
6494 (abi_v4_pass_in_fpr): Ditto.
6495 (setup_incoming_varargs): Ditto.
6496 (rs6000_gimplify_va_arg): Ditto.
6497 (rs6000_split_multireg_move): Ditto.
6498 (rs6000_savres_strategy): Ditto.
6499 (rs6000_emit_prologue_components): Ditto.
6500 (rs6000_emit_epilogue_components): Ditto.
6501 (rs6000_emit_prologue): Ditto.
6502 (rs6000_emit_epilogue): Ditto.
6503 (rs6000_elf_file_end): Ditto.
6504 (rs6000_function_value): Ditto.
6505 (rs6000_libcall_value): Ditto.
6506 * config/rs6000/rs6000.h: Ditto.
6507 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6508 (TARGET_MINMAX): ... this. New.
6509 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6510 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6511 (*movsi_internal1_single): Delete.
6512 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6513 mfpu=, mxilinx-fpu): Delete.
6514 * config/rs6000/singlefp.h: Delete.
6515 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6516 * config/rs6000/t-rs6000: Ditto.
6517 * config/rs6000/t-xilinx: Delete.
6518 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6519 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6520 (VStype_simple): Delete.
6521 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6522 * config/rs6000/xfpu.h: Delete.
6523 * config/rs6000/xfpu.md: Delete.
6524 * config/rs6000/xilinx.h: Delete.
6525 * config/rs6000/xilinx.opt: Delete.
6526 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6527 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6529 2018-05-04 Tom de Vries <tom@codesourcery.com>
6532 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6535 2018-05-04 Richard Biener <rguenther@suse.de>
6538 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6539 (expand_complex_div_wide): Likewise.
6540 (expand_complex_operations_1): Likewise.
6541 (expand_complex_libcall): Preserve EH info of the original stmt.
6542 (tree_lower_complex): Handle removed blocks.
6543 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6544 on complex multiplication and division libcall builtins.
6546 2018-05-04 Richard Biener <rguenther@suse.de>
6549 * fold-const.c (negate_expr_p): Restrict negation of operand
6550 zero of a division to when we know that can happen without
6552 (fold_negate_expr_1): Likewise.
6554 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6557 * real.h (real_nextafter): Declare.
6558 * real.c (real_nextafter): New function.
6559 * fold-const-call.c (fold_const_nextafter): New function.
6560 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6561 CASE_CFN_NEXTTOWARD.
6562 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6563 even when arg1_mode is different from arg0_mode.
6565 2018-05-03 Nathan Sidwell <nathan@acm.org>
6567 * doc/extend.texi (Deprecated Features): Remove
6569 (Backwards Compatibility): Likewise.
6570 * doc/invoke.texi (C++ Language Options): Likewise.
6571 (C++ Dialect Options): Likewise.
6573 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6576 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6577 _mm512_mask_mullox_epi64): New intrinsics.
6579 2018-05-03 Tom de Vries <tom@codesourcery.com>
6582 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6583 dump files): Add offload-tree.
6585 2018-05-03 Richard Biener <rguenther@suse.de>
6587 PR tree-optimization/85615
6588 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6589 to loops not nested in BBs loop father to avoid creating multi-entry
6592 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6594 PR tree-optimization/70291
6595 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6596 arguments. Change return type to tree. Emit libcall as a new
6597 statement rather than replacing existing one when inplace_p is true.
6598 (expand_complex_multiplication_components): New function.
6599 (expand_complex_multiplication): Expand floating-point complex
6600 multiplication using the above.
6601 (expand_complex_division): Rename inner_type parameter to type.
6602 Update expand_complex_libcall call-site.
6603 (expand_complex_operations_1): Update expand_complex_multiplication
6604 and expand_complex_division call-sites.
6606 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6609 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6610 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6611 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6612 the highest significant bit of the shift count mask is clear. In
6613 check whether and[sq]i3 is needed verify that all significant bits
6614 of the shift count other than the highest are set.
6616 2018-05-02 Tom de Vries <tom@codesourcery.com>
6619 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6620 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6621 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6622 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6623 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6624 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6625 __builtin_goacc_parlevel_size.
6627 2018-05-02 Richard Biener <rguenther@suse.de>
6629 PR tree-optimization/85597
6630 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6631 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6634 2018-05-02 Tom de Vries <tom@codesourcery.com>
6637 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6638 dump files): Add ltrans-tree.
6640 2018-05-02 Tom de Vries <tom@codesourcery.com>
6643 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6644 dump files): Add wpa-ipa.
6646 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6648 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6649 powerpc*-*-linux*paired* target.
6650 * config/rs6000/750cl.h: Delete.
6651 * config/rs6000/paired.h: Delete.
6652 * config/rs6000/paired.md: Delete.
6653 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6655 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6656 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6657 comment. Remove paired float support.
6658 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6659 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6661 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6662 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6664 * config/rs6000/rs6000.c: Remove paired float support.
6665 (paired_expand_vector_init, paired_expand_vector_move,
6666 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6667 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6668 paired_expand_builtin, paired_expand_predicate_builtin,
6669 paired_init_builtins): Delete.
6670 * config/rs6000/rs6000.h: Remove paired float support.
6671 * config/rs6000/rs6000.md: Remove paired float support.
6672 (move_from_CR_ov_bit): Delete.
6673 * config/rs6000/rs6000.opt (mpaired): Delete.
6674 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6675 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6677 2018-05-02 Richard Biener <rguenther@suse.de>
6680 * gimplify.c (gimplify_save_expr): When in SSA form allow
6681 SAVE_EXPRs to compute to SSA vars.
6683 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6686 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6687 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6688 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6689 clobber operands[2], instead use a new pseudo. Formatting fixes.
6691 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6693 PR tree-optimization/85586
6694 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6695 exit early for statements in the same group if the accesses are
6698 2018-05-02 Tom de Vries <tom@codesourcery.com>
6701 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6704 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6706 PR tree-optimization/85143
6707 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6709 2018-05-01 Tom de Vries <tom@codesourcery.com>
6712 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6713 not found" error message.
6715 2018-05-01 Tom de Vries <tom@codesourcery.com>
6718 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6719 * vec.c (test_ordered_remove_if): New function.
6720 (vec_c_tests): Call test_ordered_remove_if.
6721 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6722 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6723 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6724 VEC_ORDERED_REMOVE_IF.
6726 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6728 PR tree-optimization/82665
6729 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6730 pointer subtraction where arguments come from a memchr call.
6732 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6734 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6735 --push-state --as-needed and --pop-state instead of --as-needed and
6736 --no-as-needed if ld supports it.
6737 * configure: Regenerated.
6740 * doc/install.texi2html: Replace _002d with - and _002a with * in
6741 generated html files using sed.
6743 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6746 * gcc-rich-location.c (blank_line_before_p): New function.
6747 (use_new_line): New function.
6748 (gcc_rich_location::add_fixit_insert_formatted): New function.
6749 * gcc-rich-location.h
6750 (gcc_rich_location::add_fixit_insert_formatted): New function.
6752 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6754 * selftest.c (assert_streq): Rename "expected" and "actual" to
6755 "val1" and "val2". Extend NULL-handling to cover both inputs
6756 symmetrically, while still requiring both to be non-NULL for a pass.
6757 * selftest.h (assert_streq): Rename "expected" and "actual" to
6759 (ASSERT_EQ): Likewise.
6760 (ASSERT_EQ_AT): Likewise.
6761 (ASSERT_KNOWN_EQ): Likewise.
6762 (ASSERT_KNOWN_EQ_AT): Likewise.
6763 (ASSERT_NE): Likewise.
6764 (ASSERT_MAYBE_NE): Likewise.
6765 (ASSERT_MAYBE_NE_AT): Likewise.
6766 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6767 the assertion to pass.
6768 (ASSERT_STREQ_AT): Likewise.
6770 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6772 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6773 interaction with -pie.
6775 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6777 * selftest.h: Fix alphabetization of per-source-file selftest
6780 2018-04-30 Jason Merrill <jason@redhat.com>
6782 PR c++/61982 - dead stores to destroyed objects.
6783 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6786 2018-04-30 Jason Merrill <jason@redhat.com>
6788 * tree.c (build_clobber): New.
6789 * tree.h: Declare it.
6790 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6792 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6794 * diagnostic-show-locus.c (layout::layout): Update for
6795 location_get_source_line returning a char_span.
6796 (struct char_span): Move to input.h.
6797 (struct correction): Update for fields in char_span becoming
6799 (struct source_line): Update for location_get_source_line
6800 returning a char_span.
6801 (layout::print_line): Likewise.
6802 * edit-context.c (edited_file::print_content): Likewise.
6803 (edited_file::print_diff_hunk): Likewise.
6804 (edited_file::print_run_of_changed_lines): Likewise.
6805 (edited_file::get_num_lines): Likewise.
6806 (edited_line::edited_line): Likewise.
6807 * final.c (asm_show_source): Likewise.
6808 * input.c (location_get_source_line): Convert return type
6809 from const char * to char_span, losing the final "line_len"
6811 (dump_location_info): Update for the above.
6812 (get_substring_ranges_for_loc): Likewise. Use a char_span
6813 when handling the literal within the line.
6814 (test_reading_source_line): Update for location_get_source_line
6815 returning a char_span.
6816 * input.h (class char_span): Move here from
6817 diagnostic-show-locus.c, converting from a struct to a class.
6818 Make data members private.
6819 (char_span::operator bool): New.
6820 (char_span::length): New.
6821 (char_span::get_buffer): New.
6822 (char_span::operator[]): New.
6823 (char_span::subspan): Make const.
6824 (char_span::xstrdup): New.
6825 (location_get_source_line): Convert return type from const char *
6826 to char_span, losing the final "line_size" param.
6828 2018-04-30 Jan Hubicka <jh@suse.cz>
6830 * lto-wrapper.c (ltrans_priorities): New static var.
6831 (cmp_priority): New.
6832 (run_gcc): Read priorities and if doing parallel build order
6833 the Makefile by them.
6835 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6837 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6839 2018-04-30 Richard Biener <rguenther@suse.de>
6841 * tree-cfg.c (verify_address): Remove base argument, add
6842 flag whether to check TREE_ADDRESSABLE and do that.
6843 (verify_expr): Remove.
6844 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6845 (verify_gimple_assign_single): Likewise.
6846 (verify_gimple_switch): Likewise.
6847 (verify_expr_location_1): Dereference tp once. Add (disabled)
6848 piece from verify_expr.
6849 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6851 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6853 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6855 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6857 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6858 (small_data_pattern): Likewise.
6859 (arc_rewrite_small_data): Likewise.
6860 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6861 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6862 (get_symbol_alignment): New function.
6863 (legitimate_small_data_address_p): Likewise.
6864 (legitimate_scaled_address): Update, call
6865 legitimate_small_data_address_p.
6866 (output_sdata): New static variable.
6867 (arc_print_operand): Update how we handle small data operands.
6868 (arc_print_operand_address): Likewise.
6869 (arc_legitimate_address_p): Update, use
6870 legitimate_small_data_address_p.
6871 (arc_rewrite_small_data_p): Remove.
6872 (arc_rewrite_small_data_1): Likewise.
6873 (arc_rewrite_small_data): Likewise.
6874 (small_data_pattern): Likewise.
6875 (compact_sda_memory_operand): Update to use
6876 legitimate_small_data_address_p and get_symbol_alignment.
6877 (prepare_move_operands): Don't rewite sdata pattern.
6878 (prepare_extend_operands): Remove.
6879 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6881 (zero_extendqisi2): Likewise.
6882 (zero_extendhisi2): Likewise.
6883 (extendqihi2): Likewise.
6884 (extendqisi2): Likewise.
6885 (extendhisi2): Likewise.
6889 * config/arc/constraints.md (Usd): Change it to memory constraint.
6891 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6893 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6894 as source of std instructions.
6895 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6896 allow 6-bit constants as source for store instructions.
6897 (movdi_insn): Update instruction pattern to allow 6-bit constants
6898 as source for store instructions.
6900 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6902 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6904 2018-04-30 Nathan Sidwell <nathan@acm.org>
6905 Sandra Loosemore <sandra@codesourcery.com>
6907 * dumpfile.c (dump_open): Allow '-' for stdout.
6908 * doc/invoke.texi (Developer Options): Document dump filename
6909 determination early. Document stdin/stdout selection.
6911 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6913 Microblaze Target: PIC data text relative
6915 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6916 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6918 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6919 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6920 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6921 New addressing mode for data-text relative position indepenedent code.
6922 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6923 'ADDRESS_SYMBOLIC_TXT_REL'.
6924 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6925 (microblaze_legitimate_pic_operand): Exclude function calls from
6926 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6927 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6929 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6930 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6931 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6932 for 'address + offset'.
6933 (microblaze_expand_prologue): Add new function prologue call for
6935 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6936 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6937 table in case of TARGET_PIC_DATA_TEXT_REL.
6938 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6939 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6940 Add new macros 'UNSPEC_TEXT',
6941 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6942 + exclude function calls from 'UNSPEC_PLT' in case of data text
6944 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6945 new target hook for generating address diff vector tables in case of
6947 * doc/tm.texi : Regenerate.
6948 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6949 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6950 of addr diff vector generation.
6951 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6952 target hook definition.
6953 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6954 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6955 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6956 option for data text relative.
6958 2018-04-30 Richard Biener <rguenther@suse.de>
6960 * tree-chrec.h (evolution_function_is_constant_p): Remove
6962 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6964 2018-04-30 Richard Biener <rguenther@suse.de>
6967 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6969 2018-04-30 Richard Biener <rguenther@suse.de>
6971 PR tree-optimization/28364
6972 PR tree-optimization/85275
6973 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6974 copying first exit test.
6976 2018-04-28 Mark Wielaard <mark@klomp.org>
6978 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6980 (dwarf_AT): Handle DW_AT_addr_base.
6981 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6983 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6986 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6987 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6988 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6989 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6991 2018-04-28 Richard Biener <rguenther@suse.de>
6993 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6994 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6995 to reflect use. Only add interesting stmts.
6997 2018-04-27 Martin Jambor <mjambor@suse.cz>
7000 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7001 the jump function allows for passing through aggregate values.
7003 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7005 * input.h (in_system_header_at): Convert from macro to inline
7007 (from_macro_expansion_at): Likewise.
7008 (from_macro_definition_at): Likewise.
7010 2018-04-27 Jeff Law <law@redhat.com>
7012 * config.gcc: Mark tile* targets as deprecated/obsolete.
7014 2018-04-27 Richard Biener <rguenther@suse.de>
7016 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7019 2018-04-27 Richard Biener <rguenther@suse.de>
7021 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7023 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7025 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7026 with Yd constraint. Set "preferred_for_speed" attribute from
7027 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7029 (*movdi_internal): Ditto.
7030 (movti_interunit splitters): Remove
7031 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7032 (movdi_interunit splitters): Ditto.
7033 * config/i386/constraints.md (Ye): Remove.
7034 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7036 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7039 * config/aarch64/constraints.md (Usg): Limit to 31.
7042 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7044 PR tree-optimization/85529
7045 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7046 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7047 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7048 zero extension or masking of the MSB bit.
7049 (optimize_range_tests): Add FIRST_BB argument, pass it through
7050 to optimize_range_tests_var_bound.
7051 (maybe_optimize_range_tests, reassociate_bb): Adjust
7052 optimize_range_tests callers.
7054 2018-04-26 Richard Biener <rguenther@suse.de>
7055 Jakub Jelinek <jakub@redhat.com>
7057 * cgraph.h (symbol_table): Just declare debug method here.
7058 * symtab.c (symbol_table::debug): Define.
7060 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7062 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7064 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7066 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7067 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7068 (*movdi_internal): Substitute Yi and Yj constraint with x
7069 and Ym and Yn constraint with y constraint. Update "isa"
7070 attribute and set "preferred_for_speed" attribute from
7071 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7072 (*movsi_internal): Ditto.
7073 (*movdf_internal): Ditto.
7074 (*movsf_internal): Ditto.
7075 (*zero_extendsidi2): Ditto.
7076 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7077 (sse2_loadld): Ditto.
7078 (*vec_extract<ssevecmodelower>_0): Ditto.
7079 (*vec_extractv4si_0_zext_sse4): Ditto.
7080 (vec_concatv2di): Ditto.
7081 (*vec_dup<mode>): Ditto.
7082 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7083 * config/i386/constraints.md (Yi): Remove.
7088 2018-04-26 Nathan Sidwell <nathan@acm.org>
7090 * dumpfile.c (dump_open): New.
7091 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7092 (dump_finish): Detect stdio/stderr by value not name.
7094 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7096 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7098 2018-04-26 Tom de Vries <tom@codesourcery.com>
7101 * config/nvptx/nvptx.c (verify_neutering_jumps)
7102 (verify_neutering_labels): New function
7103 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7105 2018-04-26 Tom de Vries <tom@codesourcery.com>
7108 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7109 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7112 2018-04-26 Richard Biener <rguenther@suse.de>
7113 Tom de Vries <tom@codesourcery.com>
7116 * lto-streamer-out.c (output_function): Fixup loops if required to match
7117 discovery done in the reader.
7119 2018-04-26 Richard Biener <rguenther@suse.de>
7121 PR tree-optimization/85116
7122 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7123 have a loop exit from the single latch predecessor. Remove
7124 case of header with just condition.
7125 (ch_base::copy_headers): Exclude infinite loops from any
7127 (pass_ch::execute): Record exits.
7129 2018-04-26 Richard Biener <rguenther@suse.de>
7131 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7132 prologue cost vector and pass it to vect_get_load_cost.
7133 (vect_get_peeling_costs_all_drs): Likewise.
7134 (vect_peeling_hash_get_lowest_cost): Likewise.
7135 (vect_enhance_data_refs_alignment): Likewise.
7137 2018-04-26 Richard Biener <rguenther@suse.de>
7140 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7141 checking of integer<->pointer conversions.
7142 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7143 sign-/zero-extending pointer types.
7144 (expand_omp_for_static_chunk): Likewise.
7146 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7147 Jean Lee <xiaoyur347@gmail.com>
7149 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7150 (TARGET_ASAN_SHADOW_OFFSET): Define.
7151 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7152 true for -fsanitize=address.
7154 2018-04-25 Mark Wielaard <mark@klomp.org>
7156 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7159 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7161 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7162 than "alu", remove explicit "memory" and "imm_disp" attributes.
7163 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7166 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7167 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7168 gen_lowpart_no_emit.
7170 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7173 * config/i386/i386.c (ix86_expand_builtin): Change memory
7174 operand to XI, extend p0 to Pmode.
7175 * config/i386/i386.md: Change unspec volatile and operand
7176 1 mode to XI, change operand 0 mode to P.
7178 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7180 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7181 GET_MODE_MASK before any checking.
7182 (nds32_can_use_bset_p): Likewise.
7183 (nds32_can_use_btgl_p): Likewise.
7185 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7187 * config/nds32/nds32-doubleword.md: New define_split pattern for
7188 illegal register number.
7190 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7192 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7194 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7196 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7198 2018-04-25 Richard Biener <rguenther@suse.de>
7200 * lto-streamer.h (LTO_major_version): Bump to 8.
7202 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7204 * BASE-VER: Set to 9.0.0.
7206 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7208 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7209 in __abskf2 and __powikf2.
7211 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7214 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7215 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7216 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7217 Use the above on operand 2. Reindent.
7218 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7220 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7223 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7225 * config/i386/i386.opt (mcet): Removed.
7226 * doc/install.texi: Remove -mcet documentation.
7227 * doc/invoke.texi: Likewise.
7229 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7232 * doc/install.texi: Remove -mcet from bootstrap-cet.
7234 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7237 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7238 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7242 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7243 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7244 containing a CONST_VECTOR.
7246 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7248 * doc/install.texi: Update newlib dependency for nvptx.
7250 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7253 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7254 instead of INTVAL when shifting x left.
7256 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7258 PR tree-optimization/85478
7259 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7260 vect_grouped_store_supported for single element vectors.
7262 2018-04-24 Richard Biener <rguenther@suse.de>
7265 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7266 load cost increase to the case of non-constant step.
7268 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7271 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7272 destination if any_malformed_asm.
7274 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7277 * expr.c (store_field): In the bitfield case, if the value comes from
7278 a function call and is returned in registers by means of a PARALLEL,
7279 do not change the mode of the temporary unless BLKmode and VOIDmode.
7281 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7283 PR rtl-optimization/85423
7284 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7285 dependencies to debug insns when the previous insn is non-debug.
7287 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7289 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7290 enums into a single definition.
7291 (fls): Fix predicates and printing.
7294 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7296 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7297 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7298 and short u6 immediate.
7299 (check_if_valid_sleep_operand): Remove.
7300 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7302 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7304 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7305 flag_always_save_lp condition.
7306 * config/nds32/nds32.opt (malways-save-lp): New option.
7308 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7310 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7311 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7312 * config/nds32/nds32.h
7313 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7314 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7316 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7318 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7319 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7321 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7322 Chung-Ju Wu <jasonwucj@gmail.com>
7324 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7326 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7327 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7329 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7331 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7333 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7335 * config/nds32/nds32-protos.h (nds32_data_alignment,
7336 nds32_local_alignment): Declare.
7337 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7338 nds32_local_alignment): New functions.
7339 (TARGET_CONSTANT_ALIGNMENT): Define.
7340 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7342 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7344 * config/nds32/nds32.c
7345 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7346 (TARGET_MODES_TIEABLE_P): Likewise.
7348 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7350 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7353 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7354 Chung-Ju Wu <jasonwucj@gmail.com>
7356 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7357 for unaligned access.
7358 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7360 * config/nds32/nds32-intrinsic.md: Likewise.
7361 * config/nds32/nds32_intrinsic.h: Likewise.
7362 * config/nds32/nds32.h (nds32_builtins): Likewise.
7363 * config/nds32/nds32.opt (munaligned-access): New option.
7364 * config/nds32/nds32.c (nds32_asm_file_start): Display
7365 flag_unaligned_access status.
7367 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7369 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7370 -mno-relax is present.
7371 * config/riscv/linux.h (LINK_SPEC): Ditto.
7373 2018-04-20 Martin Sebor <msebor@redhat.com>
7376 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7378 (gimple_fold_builtin_stxcpy_chk): Same.
7379 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7381 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7384 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7385 __powikf2 when long double is IEEE 128-bit.
7387 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7389 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7390 step to make sure stack always aligned.
7392 2018-04-20 Carl Love <cel@us.ibm.com>
7395 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7396 size check for arg0.
7398 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7399 Tom de Vries <tom@codesourcery.com>
7402 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7403 Emit insns for calls too.
7404 (nvptx_find_par): Always look for worker-level predecessor insn.
7405 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7407 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7408 (nvptx_process_pars): Propagate frames for calls.
7410 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7413 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7415 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7416 (ix86_handle_option): Don't handle OPT_mibt.
7417 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7419 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7421 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7422 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7423 (ix86_target_macros): Define __CET__ with flag_cf_protection
7424 for -fcf-protection.
7425 * config/i386/i386.c (isa2_opts): Remove -mibt.
7426 * config/i386/i386.h (TARGET_IBT): Removed.
7427 (TARGET_IBT_P): Likewise.
7428 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7429 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7430 * config/i386/i386.opt (mcet): Update help message.
7433 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7434 -mcet as an alias for -mshstk.
7436 2018-04-20 Richard Biener <rguenther@suse.de>
7439 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7440 complexity by forcing a single use of the multiply operand.
7442 2018-04-20 Martin Jambor <mjambor@suse.cz>
7445 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7446 recursion dependency to only apply to non-clones.
7448 2018-04-20 Martin Jambor <mjambor@suse.cz>
7451 * ipa-cp.c (create_specialized_node): Check that clones of
7452 self-recursive edges exist during IPA-CP.
7454 2018-04-19 Toon Moene <toon@moene.org>
7456 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7459 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7461 PR tree-optimization/85467
7462 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7463 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7464 VECTOR_CST element to type.
7466 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7469 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7470 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7471 (builtin_longjmp): Likewise.
7472 (save_stack_nonlocal): New pattern.
7473 (restore_stack_nonlocal): Likewise.
7475 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7478 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7479 Replace ASM_OUTPUT_LABEL with fprintf.
7481 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7484 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7485 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7486 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7487 define __IBT__ and __SHSTK__ for -fcf-protection.
7488 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7490 (ix86_trampoline_init): Likewise.
7491 (x86_output_mi_thunk): Likewise.
7492 (ix86_notrack_prefixed_insn_p): Likewise.
7493 (ix86_option_override_internal): Don't disallow -fcf-protection.
7494 * config/i386/i386.md (rdssp<mode>): Also enable for
7496 (incssp<mode>): Likewise.
7497 (nop_endbr): Likewise.
7498 * config/i386/i386.opt (mcet): Change help message to built-in
7502 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7503 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7504 enable CET built-in functions.
7506 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7508 * common/config/i386/i386-common.c
7509 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7510 OPTION_MASK_ISA_MOVDIRI_UNSET,
7511 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7512 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7513 * config.gcc (movdirintrin.h): New header.
7514 * config/i386/cpuid.h (bit_MOVDIRI,
7515 bit_MOVDIR64B): New bits.
7516 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7518 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7519 (VOID, PVOID, PCVOID)): New function types.
7520 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7521 __builtin_ia32_directstoreu_u64,
7522 __builtin_ia32_movdir64b): New builtins.
7523 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7524 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7526 (ix86_valid_target_attribute_inner_p): Ditto.
7527 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7528 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7529 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7530 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7531 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7532 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7533 (movdiri<mode>, movdir64b_<mode>): New.
7534 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7535 * config/i386/immintrin.h: Include movdirintrin.h.
7536 * config/i386/movdirintrin.h: New file.
7537 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7539 2018-04-19 Richard Biener <rguenther@suse.de>
7542 * cfg.c (clear_bb_flags): When loop state says we have
7543 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7545 2018-04-19 Richard Biener <rguenther@suse.de>
7547 PR tree-optimization/84737
7548 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7549 copying restrict info.
7550 (vect_setup_realignment): Use it.
7551 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7552 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7553 the first DR to all generated stores.
7554 (vectorizable_load): Likewise for loads.
7556 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7558 PR tree-optimization/85446
7559 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7560 the integral and pointer types to have the same precision.
7562 * doc/install.texi: Document --disable-cet being the default and
7565 2018-04-18 Martin Liska <mliska@suse.cz>
7567 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7570 2018-04-18 Martin Liska <mliska@suse.cz>
7573 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7576 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7577 arguments if they are comparable.
7579 2018-04-18 Martin Liska <mliska@suse.cz>
7582 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7585 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7588 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7591 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7592 ENDBR after calling __morestack.
7594 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7597 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7598 by using gcc_base_ver to generate a gcc_driver_version, and use
7599 it when generating GCC_DRIVER_NAME.
7600 * configure: Regenerate.
7602 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7605 * config.gcc: Obsolete powerpc*-*-*spe*.
7607 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7610 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7611 (stabstr_D): Change type of unum from unsigned int to
7612 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7615 2018-04-17 Jim Wilson <jimw@sifive.com>
7618 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7619 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7620 Set arg_pointer_offset after using pretend_args_size.
7622 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7624 PR rtl-optimization/85431
7625 * dse.c (record_store): Ignore zero width stores.
7628 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7629 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7630 __builtin_stack_restore rather than after it.
7631 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7632 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7633 argument instead of virtual_dynamic_stack_rtx.
7635 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7637 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7639 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7640 Add note to error message to explain internal mapping of overloaded
7641 built-in function name to non-overloaded built-in function name.
7642 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7645 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7648 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7649 where the inputs overlap with the output.
7651 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7653 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7654 (=v, v) alternative and explicit "memory" attribute.
7655 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7656 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7658 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7659 "sselog1" type instead of "sselog".
7660 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7661 "sselog". Remove explicit "memory" attribute.
7662 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7663 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7665 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7666 "sselog1" type instead of "sselog". Remove explicit "memory"
7668 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7669 use "sselog1" type instead of "sselog". Remove explicit "memory"
7671 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7672 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7674 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7675 "sselog1" type instead of "sselog". Remove explicit "memory"
7677 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7678 use "sselog1" type instead of "sselog". Remove explicit "memory"
7682 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7685 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7688 2018-04-17 Martin Jambor <mjambor@suse.cz>
7691 * ipa-cp.c (create_specialized_node): Call
7692 expand_all_artificial_thunks if necessary.
7694 2018-04-17 Martin Liska <mliska@suse.cz>
7697 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7698 in message, remote space in between '_G' and '('.
7700 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7703 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7704 avx512f_vmcmp<mode>3<round_saeonly_name>,
7705 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7706 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7707 avx512f_rndscale<mode><round_saeonly_name>,
7708 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7709 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7710 Use %<iptr>2 instead of %2 for -masm=intel.
7711 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7712 avx512f_vcvttss2usi<round_saeonly_name>,
7713 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7715 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7716 avx512f_vcvttsd2usi<round_saeonly_name>,
7717 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7718 Use %q1 instead of %1 for -masm=intel.
7719 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7720 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7721 of %3 for -masm=intel.
7722 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7724 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7726 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7728 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7729 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7731 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7733 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7734 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7735 %g1 and one with %0 and %1.
7736 (avx512er_vmrcp28<mode><round_saeonly_name>,
7737 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7739 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7740 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7741 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7742 of %0 and %{%4%} for -masm=intel.
7743 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7744 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7745 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7746 order of %0 and %{%5%}%{z%} for -masm=intel.
7748 2018-04-17 Jan Hubicka <jh@suse.cz>
7751 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7753 2018-04-17 Martin Liska <mliska@suse.cz>
7756 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7757 for target_clone error message. Make default implementation
7758 clone to be a local declaration.
7759 (separate_attrs): Add new argument and check for an empty
7761 (expand_target_clones): Handle it.
7762 (ipa_target_clone): Make redirection just for target_clones
7765 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7766 Tom de Vries <tom@codesourcery.com>
7769 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7770 tiled basic blocks without omp continue statements.
7772 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7775 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7776 vec_extract expression as having side effects to make sure it gets
7779 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7782 * config/i386/i386.c (get_builtin_code_for_version): Check
7785 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7788 * gcc/config.gcc: Support "skylake".
7789 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7791 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7792 (processor_target_table): Add "skylake".
7793 (ix86_option_override_internal): Add "skylake".
7794 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7795 PROCESSOR_CANNONLAKE.
7796 (get_builtin_code_for_version): Fix priority for
7797 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7798 PROCESSOR_SKYLAKE-AVX512.
7799 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7800 (processor_type): Add PROCESSOR_SKYLAKE.
7802 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7803 Jason Merrill <jason@redhat.com>
7806 * convert.c (convert_to_integer_1): Use direct recursion for
7807 enumeral types and types with a precision less than the number
7808 of bits in their mode.
7810 2018-04-16 Julia Koval <julia.koval@intel.com>
7813 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7814 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7816 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7819 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7820 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7821 and -mno-direct-move.
7823 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7826 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7827 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7828 Ensure negative shifts result in {0}.
7830 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7832 PR rtl-optimization/79916
7833 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7834 regs (if any) to define how to gnerate SD moves when LRA is in
7837 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7839 PR rtl-optimization/85393
7840 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7841 * except.c (expand_dw2_landing_pad_for_region): Make static.
7842 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7843 a label and unconditional jump to old_bb, rather than
7844 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7847 PR rtl-optimization/85376
7848 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7849 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7850 instead of a specific value.
7852 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7853 Bin Cheng <bin.cheng@arm.com>
7855 PR tree-optimization/82965
7856 PR tree-optimization/83991
7857 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7858 by_profile_only parameter.
7859 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7860 information if the loop was predicted to iterate too many times.
7861 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7863 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7866 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7869 2018-04-13 Martin Liska <mliska@suse.cz>
7870 Jakub Jelinek <jakub@redhat.com>
7873 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7874 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7875 * builtins.c (expand_builtin_memory_copy_args): Use
7876 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7877 handle dest_addr == pc_rtx.
7879 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7882 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7883 asked to not generate direct moves.
7884 (fix_trunc<mode>si2_stfiwx): Similar.
7885 (fix_trunc<mode>si2_internal): Similar.
7887 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7890 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7891 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7892 lookup if dest in some wider mode is known to be const0_rtx and
7893 if so, record permanent equivalence for it to be ZERO_EXTEND of
7894 the narrower mode destination.
7896 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7898 * lto-streamer-out.c (output_function): Revert 259346.
7899 * omp-expand.c (expand_oacc_for): Likewise.
7901 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7903 PR rtl-optimization/85354
7904 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7905 * sel-sched.c (sel_global_init): ... here.
7907 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7910 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7911 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7912 mode for PE-COFF targets.
7913 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7914 (i386_pe_asm_lto_end): Likewise.
7915 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7916 (TARGET_ASM_LTO_END): Likewise.
7917 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7918 (i386_pe_asm_lto_start): New function.
7919 (i386_pe_asm_lto_end): Likewise.
7921 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7922 Richard Biener <rguenther@suse.de>
7925 * lto-streamer-out.c (output_function): Fix CFG loop state before
7927 * omp-expand.c (expand_oacc_for): Handle calls to internal
7928 functions like regular functions.
7930 2018-04-12 Richard Biener <rguenther@suse.de>
7933 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7934 for the early LTO debug to properly generate references to it
7935 during DIE emission. Do not re-use that for the skeleton for
7937 (dwarf2out_early_finish): Likewise.
7939 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7942 * config/i386/sse.md
7943 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7944 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7945 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7946 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7947 and output is a reg, avoid creating invalid lowpart subreg, but
7948 instead split into a 512-bit move. Don't split if not AVX512VL,
7949 input is xmm16+ reg and output is a mem.
7950 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7951 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7952 xmm16+ reg and output is a mem.
7954 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7956 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7957 also for flag_dwarf2_cfi_asm.
7959 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7961 PR rtl-optimization/85342
7962 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7963 a bool scalar var inside of the loop instead. Don't try to update
7964 recog_data.operand after failed apply_change_group.
7966 2018-04-12 Tom de Vries <tom@codesourcery.com>
7969 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7970 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7971 array with flexible array member as array without given dimension.
7972 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7973 argument for undefined param to true.
7975 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7978 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7979 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7980 from PowerPC section.
7981 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7982 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7983 help text that is too long.
7984 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7985 help text that is too long.
7986 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7987 help text that is too long.
7989 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7991 * config/alpha/alpha.md (stack_probe_internal): Rename
7992 from "probe_stack". Update all callers.
7994 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7996 PR rtl-optimization/84566
7997 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7998 sched_macro_fuse_insns.
8000 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8003 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8004 (compute_block_dependences): ... from here.
8006 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8008 PR tree-optimization/85331
8009 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8010 from int to HOST_WIDE_INT.
8012 2018-04-11 Martin Jambor <mjambor@suse.cz>
8015 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8016 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8017 not the same as the source val.
8018 (cgraph_edge_brings_value_p): New parameter.
8019 (gather_edges_for_value): Pass destination value to
8020 cgraph_edge_brings_value_p.
8021 (perhaps_add_new_callers): Likewise.
8022 (get_info_about_necessary_edges): Likewise and exclude values brought
8023 only by self-recursive edges.
8024 (create_specialized_node): Redirect only clones of self-calling edges.
8025 (+self_recursive_pass_through_p): New function.
8026 (find_more_scalar_values_for_callers_subset): Use it.
8027 (find_aggregate_values_for_callers_subset): Likewise.
8028 (known_aggs_to_agg_replacement_list): Removed.
8029 (decide_whether_version_node): Re-calculate known constants for all
8030 remaining context clones.
8032 2018-04-11 Richard Biener <rguenther@suse.de>
8035 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8036 from early DWARF output.
8037 (dwarf2out_early_finish): Output line info unconditionally into
8038 early DWARF and add reference to it.
8040 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8043 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8044 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8045 other than V2DFmode using iptr mode attribute.
8046 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8048 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8050 PR rtl-optimization/84659
8051 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8053 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8056 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8058 (output_loc_list): Pass address of a dummy size variable even in the
8059 locview handling loop.
8060 (index_location_lists): Add comment on why skip_loc_list_entry can't
8063 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8066 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8069 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8072 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8073 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8074 and -mstring-compare-inline-limit.
8076 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8079 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8080 for stack clash protection in a register whenever we need it to be in
8083 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8085 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8086 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8088 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8091 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8093 (mlong-double-): Ditto.
8094 * config/rs6000/sysv4.opt (msdata=): Ditto.
8095 (mtls-size=): Ditto.
8097 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8099 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8100 erroneous entries for
8101 "vector int vec_ldl (int, long int *)", and
8102 "vector unsigned int vec_ldl (int, unsigned long int *)".
8103 Add comments and entries for
8104 "vector bool char vec_ldl (int, bool char *)",
8105 "vector bool short vec_ldl (int, bool short *)",
8106 "vector bool int vec_ldl (int, bool int *)",
8107 "vector bool long long vec_ldl (int, bool long long *)",
8108 "vector pixel vec_ldl (int, pixel *)",
8109 "vector long long vec_ldl (int, long long *)",
8110 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8111 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8112 type tree bool_long_long_type_node and correct definition of
8113 bool_V2DI_type_node to make reference to this new type tree.
8114 (rs6000_mangle_type): Replace erroneous reference to
8115 bool_long_type_node with bool_long_long_type_node.
8116 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8117 comments to emphasize sign distinctions for char and int types and
8118 replace RS6000_BTI_bool_long constant with
8119 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8120 use of RS6000_BTI_pixel.
8121 (bool_long_type_node): Remove this macro definition.
8122 (bool_long_long_type_node): New macro definition
8124 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8126 PR rtl-optimization/85300
8127 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8128 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8129 simplify_unary_operation fails.
8131 2018-04-10 Martin Liska <mliska@suse.cz>
8133 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8134 cgraph_edge and ipa_ref.
8136 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8140 * config/i386/sse.md
8141 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8142 computation of the VEC_MERGE selector from mask.
8143 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8144 Fix decoding of the VEC_MERGE selector into mask.
8146 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8148 PR tree-optimization/85286
8149 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8151 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8153 * final.c (final_1): Set insn_last_address as well as
8154 insn_current_address.
8156 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8159 * explow.c (emit_stack_probe): Call validize_mem on memory location
8160 before passing it to gen_probe_stack. Create address operand and
8161 legitimize it for the probe_stack_address case.
8163 2018-04-09 Jan Hubicka <jh@suse.cz>
8166 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8167 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8168 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8170 (free_lang_data): Rebuild type inheritance graph.
8172 2018-04-09 Martin Sebor <msebor@redhat.com>
8174 * invoke.texi (-finline-small-functions): Mention other optimization
8176 (-findirect-inlining, -fpartial-inlining): Same.
8177 (-finline-functions-called-once): Same.
8178 (-freorder-blocks-and-partition): Same.
8180 2018-04-09 Jan Hubicka <jh@suse.cz>
8183 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8184 jumps; choose last target that matches the criteria (i.e.
8185 no partition changes for non-crossing jumps).
8186 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8187 support for redirecting crossing jumps to non-crossing.
8189 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8191 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8192 also for naked functions.
8194 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8196 * config/arc/arc.md (add_shift): New pattern.
8197 (add_shift2): Likewise.
8198 (sub_shift): Likewise.
8199 (sub_shift_cmp0_noout): Likewise.
8200 (compare_si_ashiftsi): Likewise.
8201 (xbfu_cmp0_noout): New combine pattern.
8202 (xbfu_cmp0"): Likewise.
8203 (movsi_set_cc_insn): Place the predicable variant first.
8204 (commutative_binary_cmp0_noout): Remove clobber.
8205 (commutative_binary_cmp0): New pattern.
8206 (noncommutative_binary_cmp0): Likewise.
8207 (noncommutative_binary_cmp0_noout): Likewise.
8208 (noncommutative_binary_comparison_result_used): Removed.
8209 (rsub_cmp0): New pattern.
8210 (rsub_cmp0_noout): Likewise.
8211 (extzvsi): Changed, keep only meaningful variants.
8212 (SQH, SEZ): New iterators.
8213 (SQH_postfix): New mode attribute.
8214 (SEZ_prefix): New code attribute.
8215 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8216 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8217 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8219 (noncommutative_operator): Check the availability of barrel
8222 2018-04-09 Richard Biener <rguenther@suse.de>
8224 PR tree-optimization/85284
8225 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8226 Only use the niter constraining form of simple_iv when the exit
8229 2018-04-09 Tom de Vries <tom@codesourcery.com>
8232 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8233 (define_expand "*memory_barrier"): New define_expand.
8234 (define_insn "memory_barrier"): New insn.
8236 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8238 PR rtl-optimization/80463
8239 PR rtl-optimization/83972
8240 PR rtl-optimization/83480
8242 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8243 correct producer for the insn.
8244 (tidy_control_flow): Fixup seqnos in case of debug insns.
8246 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8248 PR rtl-optimization/83913
8250 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8251 different sched-times when merging exprs.
8253 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8255 PR rtl-optimization/83962
8257 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8258 tidy_fallthru_edge and tidy_control_flow.
8260 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8262 PR rtl-optimization/83530
8264 * sel-sched.c (force_next_insn): New global variable.
8265 (remove_insn_for_debug): When force_next_insn is true, also leave only
8266 next insn in the ready list.
8267 (sel_sched_region): When the region wasn't scheduled, make another pass
8268 over it with force_next_insn set to 1.
8270 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8272 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8274 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8275 for interrupt control.
8276 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8277 functions for interrupt control.
8278 * config/nds32/nds32-intrinsic.md: Likewise.
8279 * config/nds32/nds32_intrinsic.h: Likewise.
8280 * config/nds32/nds32.h (nds32_builtins): Likewise.
8282 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8284 * config/nds32/nds32.c (nds32_init_machine_status,
8285 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8286 strict_aligned_p field.
8287 (nds32_expand_to_rtl_hook): New function.
8288 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8289 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8291 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8292 Chung-Ju Wu <jasonwucj@gmail.com>
8294 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8295 * config/nds32/nds32-n7.md: New file.
8296 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8299 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8300 * config/nds32/nds32.md (pipeline_model): Add n7.
8301 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8302 * config/nds32/pipelines.md: Include n7 settings.
8304 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8305 Chung-Ju Wu <jasonwucj@gmail.com>
8307 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8308 * config/nds32/nds32-e8.md: New file.
8309 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8312 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8313 * config/nds32/nds32.md (pipeline_model): Add e8.
8314 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8315 * config/nds32/pipelines.md: Include e8 settings.
8317 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8318 Chung-Ju Wu <jasonwucj@gmail.com>
8320 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8321 * config/nds32/nds32-n8.md: New file.
8322 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8323 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8325 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8326 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8327 * config/nds32/nds32.md (pipeline_model): Add n8.
8328 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8329 * config/nds32/pipelines.md: Include n8 settings.
8331 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8332 Chung-Ju Wu <jasonwucj@gmail.com>
8334 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8335 * config/nds32/nds32-n9-2r1w.md: New file.
8336 * config/nds32/nds32-n9-3r2w.md: New file.
8337 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8338 nds32_register_ports): New or modify for cpu n9.
8339 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8341 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8342 * config/nds32/nds32-utils.c: New file.
8343 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8344 TARGET_MUL_SLOW): Define.
8345 * config/nds32/nds32.md (pipeline_model): New attribute.
8346 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8347 New options that support cpu n9.
8348 * config/nds32/pipelines.md: Include n9 settings.
8349 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8351 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8353 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8354 information if necessary.
8355 (output_cond_branch_compare_zero): Likewise.
8356 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8357 (nds32_target_alignment): Refine for alignment.
8358 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8359 (FUNCTION_BOUNDARY): Modify.
8360 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8362 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8364 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8366 * config/nds32/constants.md (unspec_volatile_element): Add values for
8367 TLB operation and data prefetch.
8368 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8369 functions for TLB operation and data prefetch.
8370 * config/nds32/nds32-intrinsic.md: Likewise.
8371 * config/nds32/nds32_intrinsic.h: Likewise.
8372 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8373 (nds32_print_operand): Likewise.
8374 * config/nds32/nds32.h (nds32_builtins): Likewise.
8376 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8377 Andrew Pinski <pinsika@gcc.gnu.org>
8380 * match.pd: Use constant_boolean_node of correct type instead of
8381 boolean_true_node or boolean_false_node for simplifying
8382 pointer comparisons to zero.
8384 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8386 PR tree-optimization/80021
8387 * tree.c (verify_type_variant): Make error call in verify_variant_match
8388 translatable and remove final full stop.
8390 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8392 * config/nds32/constants.md (unspec_volatile_element): Add
8393 UNSPEC_VOLATILE_EH_RETURN.
8394 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8395 nds32_output_stack_pop): Support dwarf exception handling process.
8396 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8397 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8398 exception handling process.
8399 (nds32_compute_stack_frame): Likewise.
8400 (nds32_return_addr_rtx): Likewise.
8401 (nds32_initial_elimination_offset): Likewise.
8402 (nds32_expand_prologue): Likewise.
8403 (nds32_expand_epilogue): Likewise.
8404 (nds32_dynamic_chain_address): New function.
8405 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8407 (DYNAMIC_CHAIN_ADDRESS): Define.
8408 (EH_RETURN_DATA_REGNO): Define.
8409 (EH_RETURN_STACKADJ_RTX): Define.
8410 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8411 patterns for dwarf exception handling.
8413 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8415 * config/nds32/nds32.h: Clean up obsolete macros.
8417 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8419 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8420 Add enum values for particular instructions.
8421 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8422 particular intrinsic functions.
8423 * config/nds32/nds32-intrinsic.md: Likewise.
8424 * config/nds32/nds32_intrinsic.h: Likewise.
8425 * config/nds32/nds32.h (nds32_builtins): Likewise.
8426 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8427 (btst, ave): New patterns for particular instructions.
8429 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8431 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8432 Add enum values for atomic load/store and memory sync.
8433 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8435 * config/nds32/nds32-intrinsic.md: Likewise.
8436 * config/nds32/nds32_intrinsic.h: Likewise.
8437 * config/nds32/nds32.h (nds32_builtins): Likewise.
8439 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8441 PR tree-optimization/85257
8442 * fold-const.c (native_encode_vector): If not all elts could fit
8443 and off is -1, return 0 rather than offset.
8444 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8445 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8446 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8447 adjust buffer in native_interpret_expr call.
8449 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8451 * config/nds32/constants.md (unspec_volatile_element): Add cache
8452 control enum values.
8453 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8454 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8455 * config/nds32/nds32.c (nds32_cctl_names): New.
8456 (nds32_print_operand): Handle cache control register names.
8457 * config/nds32/nds32.h (nds32_builtins): New enum values.
8458 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8460 * config/nds32/nds32.md (type): Add mmu.
8461 * config/nds32/pipelines.md (simple_insn): Add mmu.
8463 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8465 * config/nds32/nds32.md (type): Remove call.
8466 * config/nds32/pipelines.md (simple_insn): Likewise.
8468 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8470 * config/nds32/constants.md (unspec_volatile_element): Add
8471 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8472 UNSPEC_VOLATILE_FMFCFG.
8473 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8474 description for fmfcfg and fmfcsr.
8475 (bdesc_1arg): Add fmtcsr.
8476 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8477 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8478 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8479 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8480 unspec_fmfcfg): New patterns.
8481 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8482 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8483 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8484 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8485 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8486 __nds32__fmfcfg): Define.
8488 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8490 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8491 intrinsic register names.
8492 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8493 intrinsic register enum values and macros.
8495 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8497 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8498 for load/store addressing form.
8499 (nds32_print_operand_address): Likewise.
8501 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8504 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8505 based on LABEL_REF. Remove useless assertion.
8506 (pic_address_needs_scratch): Fix formatting.
8507 (sparc_legitimize_pic_address): Minor tweaks.
8508 (sparc_delegitimize_address): Adjust assertion accordingly.
8509 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8510 into symbolic_operand.
8511 (movsi_high_pic_label_ref): Likewise.
8512 (movsi_lo_sum_pic_label_ref): Likewise.
8513 (movdi_pic_label_ref): Likewise.
8514 (movdi_high_pic_label_ref): Likewise.
8515 (movdi_lo_sum_pic_label_ref): Likewise.
8517 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8519 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8520 custom LIB_SPEC setup.
8522 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8523 Kito Cheng <kito.cheng@gmail.com>
8525 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8526 * config/riscv/freebsd.h: New.
8528 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8530 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8531 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8534 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8535 Kito Cheng <kito.cheng@gmail.com>
8537 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8538 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8539 * config/nds32/nds32-protos.h (nds32_output_call,
8540 nds32_output_return): Declare.
8541 * config/nds32/nds32.md: Refine all the call and return patterns.
8543 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8546 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8547 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8549 PR rtl-optimization/84872
8550 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8551 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8554 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8556 * expr.c (copy_blkmode_to_reg): Revert 254862.
8557 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8559 2018-04-06 Richard Biener <rguenther@suse.de>
8562 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8563 after seeing a component reference with an adjacent field. Treat
8564 refs to arrays at struct end of external decls similar to
8565 refs to unconstrained commons.
8567 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8570 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8571 look through SAVE_EXPRs with non-side-effects argument. Adjust
8573 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8576 2018-04-06 Richard Biener <rguenther@suse.de>
8579 * alias.c (find_base_term): New wrapper around find_base_term
8580 unwinding CSELIB_VAL_PTR changes.
8581 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8584 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8586 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8588 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8589 constant definitions.
8590 ("nop"): lr 0,0 -> nopr r0
8591 ("nop_lr0", "nop_lr1"): New insn definitions.
8593 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8595 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8596 NDS32_V3PUSH_AVAILABLE_P macro.
8598 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8599 Chung-Ju Wu <jasonwucj@gmail.com>
8601 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8602 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8603 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8604 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8605 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8606 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8607 * config/nds32/constraints.md: New constraints and checking for hard
8608 float configuration.
8609 * config/nds32/iterators.md: New mode iterator and attribute for hard
8610 float configuration.
8611 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8613 * config/nds32/nds32-fpu.md: New file.
8614 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8615 deal with hard float code generation.
8616 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8618 (abi_type, float_reg_number): New enum type.
8619 * config/nds32/nds32-predicates.c: New predicates for hard float.
8620 * config/nds32/nds32-protos.h: Declare functions for hard float.
8621 * config/nds32/nds32.c: Implementation for hard float configuration.
8622 * config/nds32/nds32.h: Definitions for hard float configuration.
8623 * config/nds32/nds32.md: Include hard float machine description and
8624 modify patterns for hard float configuration.
8625 * config/nds32/nds32.opt: New options for hard float configuration.
8626 * config/nds32/predicates.md: New predicates for hard float
8629 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8631 * common/config/nds32/nds32-common.c
8632 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8634 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8637 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8638 CONSTRUCTOR_ELT (ctor, ...)->value.
8640 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8643 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8645 2018-04-05 Tom de Vries <tom@codesourcery.com>
8648 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8651 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8652 Kito Cheng <kito.cheng@gmail.com>
8654 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8655 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8656 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8657 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8659 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8660 Kito Cheng <kito.cheng@gmail.com>
8662 * config/nds32/constraints.md (Ufe): New memory constraint.
8663 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8664 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8665 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8667 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8668 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8670 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8672 * config/nds32/nds32.md: Use optimize_size in the condition for
8673 alu-shift instructions.
8675 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8677 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8679 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8681 * config/nds32/nds32.md (negsi2): Refine pattern.
8683 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8684 Chung-Ju Wu <jasonwucj@gmail.com>
8686 * config/nds32/iterators.md (shift_rotate): New code iterator.
8687 (shift): New code attribute.
8688 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8689 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8690 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8691 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8692 bit-wise operations.
8693 (andsi3, *andsi3): Ditto.
8694 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8695 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8696 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8697 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8698 nds32_ior_operand, nds32_xor_operand): New predicates.
8700 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8702 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8703 (addsi3, subsi3): ... this.
8705 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8707 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8709 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8711 * config/nds32/nds32.md: Adjust indention.
8713 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8715 * config/nds32/nds32.md (feature): New attribute.
8717 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8719 * config/nds32/nds32.md (subtype): New attribute.
8721 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8724 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8725 expansion to perform a bitwise AND of the argument followed by a
8726 boolean negation of the result.
8728 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8730 PR rtl-optimization/84878
8731 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8732 the basic block. Assert the use reference is not artificial and that
8733 it has an associated insn.
8735 2018-04-04 Michael Matz <matz@suse.de>
8737 * builtins.c (compute_objsize): Pass correct operand
8738 to array_at_struct_end_p.
8740 2018-04-04 Richard Biener <rguenther@suse.de>
8743 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8744 from contexts for DINFO_LEVEL_TERSE and below.
8746 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8748 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8749 resiter_operand condition.
8750 * config/nds32/nds32.md (*move<mode>): Ditto.
8752 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8753 Monk Chiang <sh.chiang04@gmail.com>
8755 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8757 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8759 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8761 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8762 Kito Cheng <kito.cheng@gmail.com>
8764 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8765 nds32_cond_code_str, output_cond_branch,
8766 output_cond_branch_compare_zero, nds32_expand_cbranch,
8767 nds32_expand_cstore, nds32_expand_movcc,
8768 nds32_output_cbranchsi4_equality_zero,
8769 nds32_output_cbranchsi4_equality_reg,
8770 nds32_output_cbranchsi4_equality_reg_or_const_int,
8771 nds32_output_cbranchsi4_greater_less_zero: New functions.
8772 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8773 nds32_expand_cstore, nds32_expand_movcc,
8774 nds32_output_cbranchsi4_equality_zero,
8775 nds32_output_cbranchsi4_equality_reg,
8776 nds32_output_cbranchsi4_equality_reg_or_const_int,
8777 nds32_output_cbranchsi4_greater_less_zero): Declare.
8778 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8779 nds32_rimm11s_operand): New predicates.
8780 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8781 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8784 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8786 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8787 * config/nds32/nds32.md: Ditto.
8788 * config/nds32/pipelines.md: Ditto.
8790 2018-04-04 Richard Biener <rguenther@suse.de>
8792 PR tree-optimization/85168
8793 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8794 propagating abnormals.
8796 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8798 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8800 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8801 Kito Cheng <kito.cheng@gmail.com>
8803 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8804 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8805 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8806 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8807 * config/nds32/nds32.md (sibcall_internal): New.
8808 (sibcall_register): Remove.
8809 (sibcall_immediate): Remove.
8810 (sibcall_value_internal): New.
8811 (sibcall_value_register): Remove.
8812 (sibcall_value_immediate): Remove.
8813 * config/nds32/predicates.md (nds32_general_register_operand): New.
8814 (nds32_call_address_operand): New.
8816 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8818 PR rtl-optimization/85167
8819 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8820 bb_defs if *split_p, instead preinitialize it to NULL.
8822 PR tree-optimization/85156
8823 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8824 evaluating the argument multiple times.
8826 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8828 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8830 (_mm_cvtpd_ps): Likewise.
8831 (_mm_cvttpd_epi32): Likewise.
8832 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8833 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8834 vector, pixel, and bool following altivec.h include.
8836 2018-04-03 Martin Sebor <msebor@redhat.com>
8838 * doc/extend.texi (Common Function Attributes): Clarify.
8839 (const attribute): Likewise.
8840 (pure attribute): Likewise.
8842 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8845 * config/i386/i386.c (ix86_expand_vector_set): Use
8846 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8848 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8850 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8851 instructions when changing rounding bits to preserve precision bits
8852 in the x87 control word.
8854 2018-04-03 Martin Liska <mliska@suse.cz>
8856 PR tree-optimization/82491
8857 * rtl.h (strip_offset_and_add): Replace += suboffset with
8858 poly_uint64 () + suboffset.
8860 2018-03-29 Martin Liska <mliska@suse.cz>
8861 Martin Jambor <mjambor@suse.cz>
8864 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8865 param_type is not an integral or pointer type.
8867 2018-04-03 Richard Biener <rguenther@suse.de>
8869 * sese.h (recompute_all_dominators): Remove.
8871 2018-04-02 Martin Sebor <msebor@redhat.com>
8873 * doc/invoke.texi (-Wrestrict): Fix typos.
8875 2018-04-02 Jim Wilson <jimw@sifive.com>
8877 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8878 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8879 (<optab>di3, <optab>si3_extend): Likewise.
8880 (<optab>si3_mask, <optab>si3_mask_1): New.
8881 (<optab>di3_mask, <optab>di3_mask_1): New.
8882 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8883 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8884 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8886 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8888 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8891 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8893 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8894 (nds32_canonicalize_comparison): New function.
8896 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8897 Kito Cheng <kito.cheng@gmail.com>
8898 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8900 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8901 * config/nds32/constants.md (unspec_volatile_element): Add
8902 UNSPEC_VOLATILE_RELAX_GROUP.
8903 * config/nds32/nds32-relax-opt.c: New file.
8904 * config/nds32/nds32-predicates.c
8905 (nds32_symbol_load_store_p): New function.
8906 * config/nds32/nds32-protos.h
8907 (nds32_symbol_load_store_p): Declare function.
8908 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8909 * config/nds32/nds32.c
8910 (nds32_register_pass): New function to register pass.
8911 (nds32_register_passes): New function to register passes.
8912 * config/nds32/nds32.md (relax_group): New pattern.
8913 * config/nds32/nds32.opt (mrelax-hint): New option.
8914 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8916 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8918 * config/nds32/t-nds32: Modify files dependency.
8920 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8922 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8923 (PROFILE_HOOK): Define its implementation.
8925 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8927 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8928 type and 32-bit size.
8930 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8933 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8934 (V_128_256): New mode iterator.
8935 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8936 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8937 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8939 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8940 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8942 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8945 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8946 NaN inputs correctly.
8948 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8951 * config/rs6000/vsx.md (??r): New mode attribute.
8952 (*vsx_mov<mode>_64bit): Use it.
8953 (*vsx_mov<mode>_32bit): Likewise.
8955 2018-03-30 Martin Sebor <msebor@redhat.com>
8957 PR tree-optimization/84818
8958 * builtins.c (check_access): Use warning_n.
8960 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8963 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8965 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8968 2018-03-30 Julia Koval <julia.koval@intel.com>
8971 * x86-tune.def (movx, partial_reg_dependency): Enable for
8974 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8977 * lra-constraints.c (process_alt_operands): Move setting
8978 this_alternative_matches below.
8980 2018-03-29 Martin Liska <mliska@suse.cz>
8983 * doc/invoke.texi: Document how LTO works with debug info.
8984 Describe auto-load support of binutils. Mention 'x86-64'
8985 as valid option value of -march option.
8987 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8989 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8992 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8993 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8994 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8997 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9000 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9001 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9002 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9003 for RS6000_BTM_POWERPC64.
9004 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9005 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9006 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9011 2018-03-28 Carl Love <cel@us.ibm.com>
9014 2017-09-27 Carl Love <cel@us.ibm.com>
9016 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9017 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9018 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9021 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9023 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9024 instead of __vector bool.
9025 (_mm_max_pu8): Likewise.
9026 (_mm_min_pi16): Likewise.
9028 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9031 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9032 (DIVWEUO): Likewise.
9034 (DIVDEUO): Likewise.
9035 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9036 DIVWEUO and DIVDEUO.
9037 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9038 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9039 (div_extend): Likewise.
9040 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9042 (__builtin_divweuo): Likewise.
9043 (__builtin_divdeo): Likewise.
9044 (__builtin_divdeuo): Likewise.
9046 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9049 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9050 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9052 PR tree-optimization/82004
9053 * gimple-match-head.c (optimize_pow_to_exp): New function.
9054 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9055 Don't fold to exp if optimize_pow_to_exp is false.
9057 2018-03-28 Martin Liska <mliska@suse.cz>
9060 * calls.c (initialize_argument_information): Fix trailing space.
9061 * common.opt: Fix typo and provide better explanation for
9062 -fsanitize-coverage option.
9063 * config/i386/i386.opt: Fix typo.
9065 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9066 Martin Liska <mliska@suse.cz>
9069 * gimplify.c (asan_poison_variable): Don't do the check for
9070 gimplify_omp_ctxp here.
9071 (gimplify_decl_expr): Do it here.
9072 (gimplify_target_expr): Likewise.
9074 2018-03-28 Martin Liska <mliska@suse.cz>
9077 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9078 chkp_type_bounds_count if MPX is not enabled.
9080 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9082 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9084 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9087 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9088 function to create the function decl for complex long double
9089 multiply and divide for -mabi=ieeelongdouble.
9090 (init_float128_ieee): Call it.
9092 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9095 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9096 -fcf-protection=branch -mibt.
9097 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9099 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9102 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9103 arm_word_relocations.
9105 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9108 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9109 extern array declarations.
9111 2018-03-27 Richard Biener <rguenther@suse.de>
9114 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9115 explicit single_use checks.
9117 2018-03-27 Richard Biener <rguenther@suse.de>
9119 PR tree-optimization/85082
9120 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9123 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9125 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9126 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9127 Turn on fasynchronous-unwind-tables and funwind-tables.
9129 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9132 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9133 (*bmi_blsr_<mode>_ccz): Ditto.
9135 2018-03-26 Tom de Vries <tom@codesourcery.com>
9137 PR tree-optimization/85063
9138 * omp-general.c (offloading_function_p): New function. Factor out
9140 * omp-offload.c (pass_omp_target_link::gate): ... here.
9141 * omp-general.h (offloading_function_p): Declare.
9142 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9143 with attribute omp declare target for offloading functions.
9145 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9147 PR tree-optimization/84005
9148 * tree-data-ref.h (get_base_for_alignment): Declare.
9149 * tree-data-ref.c (get_base_for_alignment_1): New function.
9150 (get_base_for_alignment): Likewise.
9151 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9152 get_base_for_alignment to find a suitable base object, instead
9153 of always using drb->base_address.
9155 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9158 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9159 known size by default.
9161 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9164 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9165 and non BLKmode operands.
9167 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9170 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9171 Clean up attributes.
9173 2018-03-23 Richard Biener <rguenther@suse.de>
9176 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9177 we are going to emit early debug for LTO.
9179 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9182 * function.c (match_asm_constraints_1): Don't optimize if input
9183 doesn't satisfy general_operand predicate for output's mode.
9186 * alias.c (write_dependence_p): Don't require for x_canonicalized
9187 non-VOIDmode if x has VOIDmode.
9190 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9191 just don't try to optimize it rather than assert it never happens.
9193 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9195 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9196 macro expansions for definition of ST_INTERNAL_<mode> and
9197 LD_INTERNAL_<mode> builtins.
9198 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9200 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9202 (altivec_expand_st_builtin): Likewise.
9203 (altivec_expand_builtin): Remove calls to deleted functions.
9204 (rs6000_address_for_altivec): Delete this function.
9205 * config/rs6000/vector.md: Remove expands for
9206 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9208 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9211 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9212 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9213 re-computing once computed.
9214 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9215 (arm_init_machine_status): Initialize
9216 machine->static_chain_stack_bytes.
9218 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9221 * doc/extend.texi: Add four new prototypes for vec_ld.
9222 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9223 definitions for more logical presentation.
9224 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9225 entries for V1TI variants of __builtin_altivec_ld builtin.
9226 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9227 handling of V1TI variant of LVX icode pattern.
9228 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9229 (rs6000_gimple_fold_builtin): Likewise.
9230 (altivec_init_builtins): Add code to define
9231 __builtin_altivec_lvx_v1ti function.
9233 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9236 * function.c (match_asm_constraints_1): Don't do the optimization
9237 if input isn't a REG, SUBREG, MEM or constant.
9239 2018-03-22 Tom de Vries <tom@codesourcery.com>
9241 PR tree-optimization/84956
9242 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9243 bb_has_abnormal_pred.
9245 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9248 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9249 DECL_INITIAL (decl) to decl at the end.
9250 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9253 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9255 * doc/extend.texi (__builtin_tgmath): Document when complex
9256 integer types are treated as _Complex _Float64.
9258 2018-03-21 Tom de Vries <tom@codesourcery.com>
9260 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9262 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9264 PR tree-optimization/84960
9265 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9266 if it is ENTRY block, move them into single succ of ENTRY in that case.
9268 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9270 PR tree-optimization/84811
9271 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9272 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9273 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9274 (known_subrange_p): Update accordingly. Cast each value involved
9275 in the size comparison, rather than casting the result of the
9278 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9280 PR tree-optimization/84982
9281 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9282 by flipping the least significant bit rather than all bits from
9283 bitpos to bitpos + bitsize - 1.
9285 2018-03-21 Nathan Sidwell <nathan@acm.org>
9287 * doc/extend.texi (Deprecated Features): Remove mention of
9288 long-deleted deprecations.
9290 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9293 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9294 * configure: Regenerate.
9296 2018-03-21 Tom de Vries <tom@codesourcery.com>
9298 PR tree-optimization/83126
9299 * tree-parloops.c (num_phis): New function.
9300 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9302 2018-03-21 Nathan Sidwell <nathan@acm.org>
9304 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9305 mention anon-struct/union members and trailing attributes.
9307 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9309 PR tree-optimization/84969
9310 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9311 builtin memset partitions if they set different rhs values.
9313 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9315 PR rtl-optimization/84989
9316 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9317 VEC_DUPLICATE with scalar result mode.
9319 2018-03-21 Martin Liska <mliska@suse.cz>
9322 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9323 not intended return statement.
9325 2018-03-21 Martin Liska <mliska@suse.cz>
9328 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9329 (chkp_find_bound_slots_1): Limit number of iterations.
9331 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9334 * Minor grammar fixes for x86 options.
9336 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9339 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9340 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9342 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9345 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9346 (altivec_lvx_<mode>_1op): Likewise.
9347 (altivec_stvx_<mode>_2op): Likewise.
9348 (altivec_stvx_<mode>_1op): Likewise.
9349 (altivec_lvx_<VM2:mode>): New define_expand.
9350 (altivec_stvx_<VM2:mode>): Likewise.
9351 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9352 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9353 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9354 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9355 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9356 (rs6000_gen_lvx): Likewise.
9357 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9358 (altivec_expand_stv_builtin): Likewise.
9359 (altivec_expand_builtin): Likewise.
9360 * config/rs6000/vector.md: Likewise.
9362 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9365 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9368 2018-03-20 Richard Biener <rguenther@suse.de>
9371 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9372 sign-conversions as zero, fall back to standard scalar_stmt
9375 2018-03-20 Martin Liska <mliska@suse.cz>
9378 * predict.c (rebuild_frequencies): Handle case when we have
9379 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9381 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9384 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9385 flag_section_anchors.
9386 * varasm.c (use_blocks_for_decl_p): Remove hack for
9387 dw2_force_const_mem.
9390 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9392 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9393 be created, use lowpart_subreg of operands[0] rather than operands[0]
9395 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9396 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9397 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9398 and n constraint instead of aarch64_shift_imm_di and Usd.
9399 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9400 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9402 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9405 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9406 to favor GPR over NEON registers.
9407 (<shift>di3_neon): Likewise.
9409 2018-03-20 Tom de Vries <tom@codesourcery.com>
9412 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9413 (nvptx_process_pars): Emit bar.sync asap and alap.
9415 2018-03-20 Tom de Vries <tom@codesourcery.com>
9418 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9419 seen_label if seen_label is already set.
9421 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9424 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9425 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9426 Use 1U instead of 1. Formatting fixes.
9429 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9430 instead of TREE_TYPE (s1) for the return value.
9432 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9434 PR tree-optimization/84946
9435 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9436 bitsize + bitsize in poly_uint64 rather than poly_int64.
9439 * dwarf2asm.c: Include fold-const.c.
9440 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9441 of decl rather than decl itself.
9443 PR rtl-optimization/84643
9444 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9446 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9449 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9450 calling assemble_variable.
9452 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9455 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9456 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9458 2018-03-19 Jim Wilson <jimw@sifive.com>
9461 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9462 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9463 (riscv_first_stack_step): Likewise.
9464 (riscv_option_override): Use STACK_BOUNDARY instead of
9466 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9468 (BIGGEST_ALIGNMENT): Set to 128.
9469 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9470 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9473 2018-03-19 Richard Biener <rguenther@suse.de>
9475 PR tree-optimization/84933
9476 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9477 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9479 2018-03-19 Richard Biener <rguenther@suse.de>
9481 PR tree-optimization/84859
9482 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9483 (cond_if_else_store_replacement): Perform sinking operation on
9484 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9485 Generalize what a BB with a single eligible store is.
9487 2018-03-19 Richard Biener <rguenther@suse.de>
9489 PR tree-optimization/84929
9490 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9491 chrec_is_positive against non-chrec arg.
9493 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9496 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9498 2018-03-18 Martin Liska <mliska@suse.cz>
9500 PR rtl-optimization/84635
9501 * regrename.c (build_def_use): Use matches_mode only when
9504 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9506 PR tree-optimization/84913
9507 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9508 vectorize chains of COND_EXPRs.
9510 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9512 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9514 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9516 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9518 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9520 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9522 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9523 Kito Cheng <kito.cheng@gmail.com>
9525 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9526 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9527 (nds32_adjust_reg_alloc_order): New function.
9528 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9530 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9532 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9533 nds32_print_operand, nds32_print_operand_address): Use
9534 HOST_WIDE_INT_PRINT_DEC instead.
9536 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9538 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9540 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9543 * config/i386/i386.c (initial_ix86_tune_features,
9544 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9546 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9547 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9548 rather than 1u << ix86_tune. Formatting fix.
9549 (ix86_option_override_internal): Change ix86_arch_mask from
9550 unsigned int to unsigned HOST_WIDE_INT, initialize to
9551 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9552 (ix86_function_specific_restore): Likewise.
9554 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9557 * postreload.c (reload_combine_recognize_pattern): Perform
9558 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9559 truncate_int_for_mode the result for the destination's mode.
9562 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9563 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9565 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9568 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9571 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9572 regno_allocno_class_array and sorted_pseudos.
9573 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9574 insns where regno is used.
9576 2018-03-16 Martin Liska <mliska@suse.cz>
9579 * multiple_target.c (create_dispatcher_calls): Redirect
9580 reference in the symbol table.
9582 2018-03-16 Martin Liska <mliska@suse.cz>
9585 * multiple_target.c (create_dispatcher_calls): Redirect also
9588 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9592 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9593 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9594 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9596 2018-03-16 Julia Koval <julia.koval@intel.com>
9598 * doc/invoke.texi (Skylake Server): Add CLWB.
9599 Cannonlake): Remove CLWB.
9601 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9603 PR tree-optimization/84841
9604 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9606 (FLOAT_ONE_CONST_TYPE): Define.
9607 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9608 (sort_by_operand_rank): Put entries with higher constant_type last
9609 rather than first to match comments.
9611 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9613 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9616 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9619 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9620 check and use address of target always.
9622 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9625 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9626 (indirect_thunk_bnd_needed): Likewise.
9627 (indirect_thunks_used): Likewise.
9628 (indirect_thunks_bnd_used): Likewise.
9629 (indirect_return_needed): New.
9630 (indirect_return_bnd_needed): Likewise.
9631 (output_indirect_thunk_function): Add a bool argument for
9633 (output_indirect_thunk_function): Don't generate alias for
9634 function return thunk.
9635 (ix86_code_end): Call output_indirect_thunk_function to generate
9636 function return thunks.
9637 (ix86_output_function_return): Set indirect_return_bnd_needed
9638 and indirect_return_needed instead of indirect_thunk_bnd_needed
9639 and indirect_thunk_needed.
9641 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9643 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9644 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9645 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9647 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9648 Paul Hua <paul.hua.gm@gmail.com>
9651 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9653 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9655 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9656 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9659 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9662 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9663 instead of GET_MODE_SIZE when comparing Units.
9665 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9668 * varasm.c (hash_section): Return an unchangeble hash value
9669 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9670 Return !aarch64_can_use_per_function_literal_pools_p ().
9672 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9675 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9676 rather than address of cmode as last argument to prepare_cmp_insn.
9678 2018-03-15 Julia Koval <julia.koval@intel.com>
9680 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9681 F_AVX512VNNI, F_AVX512BITALG): New.
9683 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9686 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9687 insn for floating-point loads and stores.
9689 2018-03-14 Carl Love <cel@us.ibm.com>
9691 * config/rs6000/rs6000-c.c: Add macro definitions for
9692 ALTIVEC_BUILTIN_VEC_PERMXOR.
9693 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9694 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9695 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9696 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9698 * config/doc/extend.texi: Add prototypes for vec_permxor.
9700 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9703 * diagnostic-show-locus.c (class layout_point): Convert m_line
9704 from int to linenum_type.
9705 (line_span::comparator): Use linenum "compare" function when
9706 comparing line numbers.
9707 (test_line_span): New function.
9708 (layout_range::contains_point): Convert param "row" from int to
9710 (layout_range::intersects_line_p): Likewise.
9711 (layout::will_show_line_p): Likewise.
9712 (layout::print_source_line): Likewise.
9713 (layout::should_print_annotation_line_p): Likewise.
9714 (layout::print_annotation_line): Likewise.
9715 (layout::print_leading_fixits): Likewise.
9716 (layout::annotation_line_showed_range_p): Likewise.
9717 (struct line_corrections): Likewise for field m_row.
9718 (line_corrections::line_corrections): Likewise for param "row".
9719 (layout::print_trailing_fixits): Likewise.
9720 (layout::get_state_at_point): Likewise.
9721 (layout::get_x_bound_for_row): Likewise.
9722 (layout::print_line): Likewise.
9723 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9724 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9725 * input.c (selftest::test_linenum_comparisons): New function.
9726 (selftest::input_c_tests): Call it.
9727 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9728 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9729 * selftest.h (ASSERT_GT): New macro.
9730 (ASSERT_GT_AT): New macro.
9731 (ASSERT_LT): New macro.
9732 (ASSERT_LT_AT): New macro.
9734 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9736 PR rtl-optimization/84780
9737 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9739 2018-03-14 Martin Liska <mliska@suse.cz>
9741 * tree.c (record_node_allocation_statistics): Use
9742 get_stats_node_kind.
9743 (get_stats_node_kind): New function extracted from
9744 record_node_allocation_statistics.
9745 (free_node): Use get_stats_node_kind.
9747 2018-03-14 Richard Biener <rguenther@suse.de>
9749 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9750 that the value-set of ANTIC_IN doesn't grow.
9753 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9755 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9756 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9758 2018-03-14 Julia Koval <julia.koval@intel.com>
9760 * config.gcc (icelake-client, icelake-server): New.
9762 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9763 (initial_ix86_arch_features): Ditto.
9764 (PTA_SKYLAKE): Add SGX.
9765 (PTA_ICELAKE): Remove.
9766 (PTA_ICELAKE_CLIENT): New.
9767 (PTA_ICELAKE_SERVER): New.
9768 (ix86_option_override_internal): Split up icelake on icelake client and
9770 (get_builtin_code_for_version): Ditto.
9771 (fold_builtin_cpu): Ditto.
9772 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9773 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9774 * config/i386/i386.h (processor_type): Ditto.
9775 * doc/invoke.texi: Ditto.
9777 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9780 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9781 INTEGER_CST offset, add it together with bitpos / 8 and
9782 sign extend based on POINTER_SIZE.
9786 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9789 * config/i386/constraints.md (Yc): New register constraint.
9790 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9791 Use Yc constraint for alternative 2 of operand 0. Remove
9792 preferred_for_speed attribute.
9794 2018-03-14 Richard Biener <rguenther@suse.de>
9796 PR tree-optimization/84830
9797 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9798 with the old one to avoid oscillations.
9800 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9803 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9805 (assign_by_spills): Return a flag of reload assignment failure.
9806 Do not process the reload assignment failures. Do not spill other
9807 reload pseudos if they has the same reg class. Update n if
9809 (lra_assign): Add a return arg. Set up from the result of
9810 assign_by_spills call.
9811 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9812 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9813 usage_insns if it is not NULL.
9814 (spill_hard_reg_in_range): New function.
9815 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9816 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9817 function prototypes.
9818 (lra_assign): Change prototype.
9819 * lra.c (lra): Add code to deal with fails by splitting hard reg
9822 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9824 * config/riscv/riscv.opt (mrelax): New option.
9825 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9826 "norelax" when riscv_mrelax is disabled.
9827 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9829 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9832 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9833 reassociation for int modes.
9835 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9837 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9838 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9840 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9841 * config/aarch64/aarch64-sve.md
9842 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9843 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9844 (*extend<mode><Vwide>2): Rename to...
9845 (aarch64_sve_extend<mode><Vwide>2): ...this.
9846 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9847 renaming the old pattern to...
9848 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9850 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9851 define_expand, renaming the old pattern to...
9852 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9853 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9854 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9855 account when deciding which SVE instruction the optab should use.
9856 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9858 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9860 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9861 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9862 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9863 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9864 (tlsdesc_small_<mode>): Turn a define_expand and use
9865 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9866 (tlsdesc_small_advsimd_<mode>): ...this.
9867 (tlsdesc_small_sve_<mode>): New pattern.
9869 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9871 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9872 (UNSPEC_UMUL_HIGHPART): New constants.
9873 (MUL_HIGHPART): New int iteraor.
9874 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9875 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9877 (*<su>mul<mode>3_highpart): New define_insn.
9879 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9882 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9885 2018-03-13 Martin Liska <mliska@suse.cz>
9888 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9890 (sem_item_optimizer::~sem_item_optimizer): Release it.
9891 (sem_item_optimizer::merge_classes): Register variable aliases.
9892 (sem_item_optimizer::fixup_pt_set): New function.
9893 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9894 * ipa-icf.h: Declare new variables and functions.
9896 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9899 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9900 integer_pow2p@2 and test integer_pow2p in condition.
9901 (A < 0 ? C : 0): Similarly for @1.
9904 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9905 characters starting at p contain '\0' character, don't look beyond
9909 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9910 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9913 * reg-stack.c (change_stack): Change update_end var from int to
9914 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9915 also call set_block_for_insn on the newly added insns and rescan.
9918 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9919 on the last operand.
9922 * tree.c (stabilize_reference_1): Return save_expr (e) for
9923 STATEMENT_LIST even if it doesn't have side-effects.
9925 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9927 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9929 2018-03-12 Renlin Li <renlin.li@arm.com>
9931 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9932 aarch64_output_scalar_simd_mov_immediate.
9934 2018-03-12 Martin Sebor <msebor@redhat.com>
9936 PR tree-optimization/83456
9937 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9938 for perfectly overlapping calls to memcpy.
9939 (gimple_fold_builtin_memory_chk): Same.
9940 (gimple_fold_builtin_strcpy): Handle no-warning.
9941 (gimple_fold_builtin_stxcpy_chk): Same.
9942 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9944 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9946 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9947 parameter. Use it for SFmode.
9948 (rs6000_function_arg_advance_1): Adjust.
9949 (rs6000_function_arg): Adjust.
9950 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9952 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9954 PR rtl-optimization/84169
9955 PR rtl-optimization/84780
9956 * combine.c (can_combine_p): Check for a 2-insn combination whether
9957 the destination register is used between the two insns, too.
9959 2018-03-12 Richard Biener <rguenther@suse.de>
9961 PR tree-optimization/84803
9962 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9963 for refs DR analysis didn't process.
9965 2018-03-12 Richard Biener <rguenther@suse.de>
9967 PR tree-optimization/84777
9968 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9969 force-vectorize loops ignore whether we are optimizing for size.
9971 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9973 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9974 (TARGET_MD_ASM_ADJUST): Define.
9976 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9977 Kito Cheng <kito.cheng@gmail.com>
9978 Chung-Ju Wu <jasonwucj@gmail.com>
9980 * config/nds32/nds32.c (nds32_compute_stack_frame,
9981 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9982 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9983 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9984 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9985 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9986 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9987 * config/nds32/nds32.md (prologue, epilogue): Use macro
9988 NDS32_V3PUSH_AVAILABLE_P to do checking.
9990 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9993 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9994 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9995 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9996 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9997 addition of most attributes on !orig_type_die or the attribute not
9998 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10000 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10001 Chung-Ju Wu <jasonwucj@gmail.com>
10003 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10004 __NDS32_VH__ macro.
10005 * config/nds32/nds32.opt (mvh): New option.
10007 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10008 Chung-Ju Wu <jasonwucj@gmail.com>
10010 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10012 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10013 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10016 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10017 Chung-Ju Wu <jasonwucj@gmail.com>
10019 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10021 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10022 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10024 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10025 Kito Cheng <kito.cheng@gmail.com>
10026 Chung-Ju Wu <jasonwucj@gmail.com>
10028 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10029 UNSPEC_FFMISM and UNSPEC_FLMISM.
10030 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10031 for ffb, ffmism and flmism.
10032 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10033 (unspec_ffmism): Ditto.
10034 (unspec_flmism): Ditto.
10035 (nds32_expand_builtin_impl): Check if string extension is available.
10036 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10037 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10039 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10042 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10045 * lra-assigns.c (assign_by_spills): Return a flag of reload
10046 assignment failure. Do not process the reload assignment
10047 failures. Do not spill other reload pseudos if they has the same
10049 (lra_assign): Add a return arg. Set up from the result of
10050 assign_by_spills call.
10051 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10052 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10053 usage_insns if it is not NULL.
10054 (spill_hard_reg_in_range): New function.
10055 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10056 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10057 function prototypes.
10058 (lra_assign): Change prototype.
10059 * lra.c (lra): Add code to deal with fails by splitting hard reg
10062 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10065 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10067 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10070 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10071 (dw_val_node): Add val_symbolic_view.
10072 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10073 (symview_upper_bound): New.
10074 (new_line_info_table): Initialize symviews_since_reset.
10075 (dwarf2out_source_line): Count symviews_since_reset and set
10076 symview_upper_bound.
10077 (dw_val_equal_p): Handle symview.
10078 (add_AT_symview): New.
10079 (print_dw_val): Handle symview.
10080 (attr_checksum, attr_checksum_ordered): Likewise.
10081 (same_dw_val_p, size_of_die): Likewise.
10082 (value_format, output_die): Likewise.
10083 (add_high_low_attributes): Use add_AT_symview for entry_view.
10084 (dwarf2out_finish): Reset symview_upper_bound, clear
10087 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10090 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10091 Add strict argument and use it.
10092 (rs6000_split_multireg_move): Update for new strict argument.
10093 (mem_operand_gpr): Disallow all non-offsettable addresses.
10094 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10096 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10099 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10100 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10101 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10104 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10105 decl, use remap_type if we want to use the type.
10107 2018-03-09 Martin Sebor <msebor@redhat.com>
10109 PR tree-optimization/84526
10110 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10112 (builtin_access::generic_overlap): Be prepared to handle non-array
10115 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10117 PR rtl-optimization/84682
10118 * lra-constraints.c (process_address_1): Check is_address flag
10119 for address constraints.
10120 (process_alt_operands): Likewise.
10121 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10122 preprocess_constraints.
10123 * recog.h (preprocess_constraints): Add oploc parameter.
10125 * recog.c (preprocess_constraints): Test address_operand for
10126 CT_ADDRESS constraints.
10128 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10131 * lra-assigns.c (assign_by_spills): Return a flag of reload
10132 assignment failure. Do not process the reload assignment
10133 failures. Do not spill other reload pseudos if they has the same
10135 (lra_assign): Add a return arg. Set up from the result of
10136 assign_by_spills call.
10137 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10138 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10139 usage_insns if it is not NULL.
10140 (spill_hard_reg_in_range): New function.
10141 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10142 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10143 function prototypes.
10144 (lra_assign): Change prototype.
10145 * lra.c (lra): Add code to deal with fails by splitting hard reg
10148 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10151 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10152 Accept complain bool parameter. Only emit errors if it is true.
10153 (arm_parse_cpu_option_name): Likewise.
10154 (arm_target_thumb_only): Adjust callers of the above.
10155 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10156 prototype to take a default true bool parameter.
10157 (arm_parse_arch_option_name): Likewise.
10159 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10160 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10164 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10165 * configure: Regenerate.
10166 * configure.ac ("linker --version-script option"): New.
10167 ("linker soname option"): New.
10169 2018-03-09 Richard Biener <rguenther@suse.de>
10171 PR tree-optimization/84775
10172 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10173 immediate uses of predicate stmts and mark them modified.
10176 PR tree-optimization/84178
10177 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10179 (version_loop_for_if_conversion): Delay update_ssa call.
10180 (tree_if_conversion): Delay update_ssa until after predicate
10183 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10186 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10187 when the function accesses prior frames.
10189 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10192 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10193 gen_llsym, otherwise call maybe_gen_llsym.
10195 PR inline-asm/84742
10196 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10197 has ',' character inside of it.
10199 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10202 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10203 as clobbering CC_REGNUM.
10205 2018-03-08 Richard Biener <rguenther@suse.de>
10207 PR middle-end/84552
10208 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10209 (follow_copies_to_constant): Do not follow SSA names registered
10212 2018-03-08 Richard Biener <rguenther@suse.de>
10214 PR tree-optimization/84178
10215 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10217 (version_loop_for_if_conversion): Delay update_ssa call.
10218 (tree_if_conversion): Delay update_ssa until after predicate
10221 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10223 PR tree-optimization/84178
10224 * tree-if-conv.c (release_bb_predicate): Remove the
10225 the assertion that the stmts have NULL use_ops.
10226 Discard the statements, asserting that they haven't
10227 yet been added to a BB.
10229 2018-03-08 Richard Biener <rguenther@suse.de>
10231 PR tree-optimization/84746
10232 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10233 (phi_translate): Pass in destination ANTIC_OUT set.
10234 (phi_translate_1): Likewise. For a simplified result lookup
10235 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10236 (phi_translate_set): Adjust.
10237 (do_pre_regular_insertion): Likewise.
10238 (do_pre_partial_partial_insertion): Likewise.
10240 2018-03-08 Martin Liska <mliska@suse.cz>
10242 PR gcov-profile/84735
10243 * doc/gcov.texi: Document usage of profile files.
10244 * gcov-io.h: Document changes in the format.
10246 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10250 * dwarf2out.c (struct dw_line_info_table): Update comments for
10252 (FORCE_RESET_NEXT_VIEW): New.
10253 (FORCE_RESETTING_VIEW_P): New.
10254 (RESETTING_VIEW_P): Check for -1 too.
10255 (ZERO_VIEW_P): Likewise.
10256 (new_line_info_table): Force-reset next view.
10257 (dwarf2out_begin_function): Likewise.
10258 (dwarf2out_source_line): Simplify zero_view_p initialization.
10259 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10260 view directly. Omit view when omitting .loc at line 0.
10262 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10264 PR tree-optimization/84740
10265 * tree-switch-conversion.c (process_switch): Call build_constructors
10266 only if info.phi_count is non-zero.
10268 PR tree-optimization/84739
10269 * tree-tailcall.c (find_tail_calls): Check call arguments against
10270 DECL_ARGUMENTS (current_function_decl) rather than
10271 DECL_ARGUMENTS (func) when checking for tail recursion.
10273 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10275 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10276 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10277 Volker Reichelt's entry and add entries for people that perform
10278 GCC fuzzy testing and report numerous bugs.
10280 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10283 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10284 readonly data in sdata, if that is disabled.
10285 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10286 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10287 -mreadonly-in-sdata option.
10289 2018-03-07 Martin Sebor <msebor@redhat.com>
10291 PR tree-optimization/84468
10292 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10293 basic block when looking for nul assignment.
10295 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10298 * except.h (output_function_exception_table): Adjust prototype.
10299 * except.c (output_function_exception_table): Remove FNNAME parameter
10300 and add SECTION parameter. Ouput one part of the table at a time.
10301 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10302 the first part of the exception table and emit unwind directives.
10303 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10304 (i386_pe_seh_cold_init): Likewise.
10305 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10306 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10307 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10308 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10309 * config/i386/winnt.c: Include except.h.
10310 (struct seh_frame_state): Add reg_offset, after_prologue and
10311 in_cold_section fields.
10312 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10313 (i386_pe_seh_cold_init): New function.
10314 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10315 to seh->in_cold_section.
10316 (seh_emit_push): Record the offset of the push.
10317 (seh_emit_save): Record the offet of the save.
10318 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10319 Test seh->after_prologue to disregard the epilogue.
10320 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10321 (i386_pe_end_cold_function): New function.
10323 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10326 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10327 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10330 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10332 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10333 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10335 PR middle-end/84723
10336 * multiple_target.c: Include tree-inline.h and intl.h.
10337 (expand_target_clones): Diagnose and fail if node->definition and
10338 !tree_versionable_function_p (node->decl).
10340 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10342 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10344 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10345 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10346 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10348 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10351 * combine.c (try_combine): Use reg_or_subregno instead of handling
10352 just paradoxical SUBREGs and REGs.
10354 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10356 * config/arc/arc.c (arc_finalize_pic): Remove function.
10357 (arc_must_save_register): We use single base PIC register, remove
10358 checks to save/restore the PIC register.
10359 (arc_expand_prologue): Likewise.
10360 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10362 (arc_verify_short): Likewise.
10363 (arc_attr_type): Likewise.
10364 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10365 (walk_stores): Likewise.
10366 (arc_address_cost): Make it static.
10367 (arc_verify_short): Likewise.
10368 (branch_dest): Likewise.
10369 (arc_attr_type): Likewise.
10370 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10371 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10372 (arc_final_prescan_insn): Remove inserting the nops due to
10373 hardware hazards. It is done in reorg step.
10374 (insn_length_variant_t): Remove.
10375 (insn_length_parameters_t): Likewise.
10376 (arc_insn_length_parameters): Likewise.
10377 (arc_get_insn_variants): Likewise.
10378 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10380 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10382 PR inline-asm/84683
10383 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10386 PR tree-optimization/84687
10387 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10389 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10391 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10393 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10394 Rename to ppc_speculation_barrier.
10395 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10396 __builtin_ppc_speculation_barrier.
10398 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10401 * combine.c (combine_simplify_rtx): Don't try to simplify if
10402 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10405 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10407 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10408 to 32 bytes when compiling for POWER9.
10410 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10413 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10414 regparm >= 3 with no arg reg available also for calls with
10415 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10418 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10420 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10422 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10425 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10427 2018-03-05 Richard Biener <rguenther@suse.de>
10429 PR tree-optimization/84486
10430 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10431 When inserting a __builtin_assume_aligned call set the LHS
10432 SSA name alignment info accordingly.
10434 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10436 PR tree-optimization/84114
10437 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10438 Avoid reassociation of FLOAT_MODE addition.
10440 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10442 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10443 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10444 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10445 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10446 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10447 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10448 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10450 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10451 __builtin_ia32_wbinvd): New builtins.
10452 (SPECIAL_ARGS2): New.
10453 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10454 (SPECIAL_ARGS2): New.
10455 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10456 (ix86_valid_target_attribute_inner_p): Ditto.
10457 (ix86_init_mmx_sse_builtins): Add special_args2.
10458 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10459 TARGET_WBNOINVD_P): New.
10460 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10461 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10462 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10463 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10464 * config/i386/pconfigintrin.h: New file.
10465 * config/i386/wbnoinvdintrin.h: Ditto.
10466 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10468 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10470 2018-03-05 Richard Biener <rguenther@suse.de>
10472 PR tree-optimization/84670
10473 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10475 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10476 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10477 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10478 doesn't grow if all successors (recursively) were visited at least
10481 2018-03-05 Richard Biener <rguenther@suse.de>
10483 PR tree-optimization/84650
10484 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10485 if executed in the loop pipeline.
10487 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10489 * doc/configfiles.texi (Configuration Files): Move info about
10490 conditionalizing $target-protos.h to...
10491 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10492 differs from $target-protos.h.
10494 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10495 Chung-Ju Wu <jasonwucj@gmail.com>
10497 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10498 * config/nds32/nds32-multiple.md (setmemsi): Define.
10499 * config/nds32/nds32-memory-manipulation.c
10500 (nds32_gen_dup_4_byte_to_word_value): New.
10501 (emit_setmem_word_loop): New.
10502 (emit_setmem_byte_loop): New.
10503 (nds32_expand_setmem_loop): New.
10504 (nds32_expand_setmem_loop_v3m): New.
10505 (nds32_expand_setmem_unroll): New.
10506 (nds32_expand_setmem): New.
10508 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10509 Chung-Ju Wu <jasonwucj@gmail.com>
10511 * config/nds32/nds32-memory-manipulation.c
10512 (nds32_emit_load_store): New.
10513 (nds32_emit_post_inc_load_store): New.
10514 (nds32_emit_mem_move): New.
10515 (nds32_emit_mem_move_block): New.
10516 (nds32_expand_movmemsi_loop_unknown_size): New.
10517 (nds32_expand_movmemsi_loop_known_size): New.
10518 (nds32_expand_movmemsi_loop): New.
10519 (nds32_expand_movmemsi_unroll): New.
10520 (nds32_expand_movmemqi): Rename ...
10521 (nds32_expand_movmemsi): ... to this.
10522 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10523 (movmemsi): ... to this.
10524 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10525 (nds32_expand_movmemsi): ... to this.
10527 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10528 Monk Chiang <sh.chiang04@gmail.com>
10529 Chung-Ju Wu <jasonwucj@gmail.com>
10531 * config/nds32/nds32-protos.h
10532 (nds32_expand_load_multiple): New arguments.
10533 (nds32_expand_store_multiple): Ditto.
10534 (nds32_valid_multiple_load_store): Rename ...
10535 (nds32_valid_multiple_load_store_p): ... to this.
10536 * config/nds32/nds32-memory-manipulation.c
10537 (nds32_expand_load_multiple): Refine implementation.
10538 (nds32_expand_store_multiple): Ditto.
10539 * config/nds32/nds32-multiple.md
10540 (load_multiple): Update nds32_expand_load_multiple interface.
10541 (store_multiple): Update nds32_expand_store_multiple interface.
10542 * config/nds32/nds32-predicates.c
10543 (nds32_valid_multiple_load_store): Rename ...
10544 (nds32_valid_multiple_load_store_p): ... to this and refine
10546 * config/nds32/predicates.md
10547 (nds32_load_multiple_and_update_address_operation): New predicate.
10548 (nds32_store_multiple_and_update_address_operation): New predicate.
10550 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10551 Chung-Ju Wu <jasonwucj@gmail.com>
10553 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10554 (combo): New attribute.
10555 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10557 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10559 * config/nds32/nds32.opt: Change -mcmodel= default value.
10561 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10562 Monk Chiang <sh.chiang04@gmail.com>
10563 Chung-Ju Wu <jasonwucj@gmail.com>
10565 * config/nds32/constants.md (unspec_element): New enum.
10566 * config/nds32/constraints.md (Umw): New constraint.
10567 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10568 * config/nds32/nds32-intrinsic.md: Likewise.
10569 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10570 (nds32_valid_smw_lwm_base_p): New.
10571 (nds32_output_smw_single_word): New.
10572 (nds32_output_lmw_single_word): New.
10573 (nds32_expand_unaligned_load): New.
10574 (nds32_expand_unaligned_store): New.
10575 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10576 (nds32_output_smw_single_word): Declare.
10577 (nds32_output_lmw_single_word): Declare.
10578 (nds32_expand_unaligned_load): Declare.
10579 (nds32_expand_unaligned_store): Declare.
10580 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10581 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10582 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10583 NDS32_BUILTIN_UASTORE_DW.
10584 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10587 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10588 Kito Cheng <kito.cheng@gmail.com>
10589 Chung-Ju Wu <jasonwucj@gmail.com>
10591 * config/nds32/nds32-intrinsic.c
10592 (nds32_expand_builtin_null_ftype_reg): Delete.
10593 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10594 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10595 (nds32_read_argument): New.
10596 (nds32_legitimize_target): Ditto.
10597 (nds32_legitimize_argument): Ditto.
10598 (nds32_check_constant_argument): Ditto.
10599 (nds32_expand_unop_builtin): Ditto.
10600 (nds32_expand_unopimm_builtin): Ditto.
10601 (nds32_expand_binop_builtin): Ditto.
10602 (nds32_builtin_decl_impl): Ditto.
10603 (builtin_description): Ditto.
10604 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10605 (nds32_init_builtins_impl): Ditto.
10606 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10607 (nds32_builtin_decl): New.
10608 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10609 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10611 2018-03-02 Jeff Law <law@redhat.com>
10613 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10614 (redundant_insn, fill_simple_delay_slots): Likewise.
10615 (fill_slots_from_thread): Likewise.
10616 * resource.c (mark_referenced_resources): Likewise.
10617 (mark_set_resources, find_dead_or_set_registers): Likewise.
10619 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10621 * substring-locations.h (format_warning_va): Formatting fix for
10622 ATTRIBUTE_GCC_DIAG.
10623 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10625 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10626 * substring-locations.c: Include intl.h.
10627 (format_warning_va): Turned into small wrapper around
10628 format_warning_n_va, renamed to ...
10629 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10630 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10632 (format_warning_at_substring_n): New function.
10633 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10634 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10635 format_warning_at_substring with just a shorter name instead of
10636 const function pointer.
10637 (fmtwarn_n): New function.
10638 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10639 appropriate, get rid of all the fmtstr temporaries, move conditionals
10640 with G_() wrapped string literals directly into fmtwarn arguments,
10641 cast dir.len to (int), formatting fixes.
10643 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10645 * doc/invoke.texi: Remove "Cilk Plus" references.
10647 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10648 Richard Biener <rguenther@suse.de>
10651 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10652 for error or warning attributes if CALL_FROM_THUNK_P is set.
10655 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10658 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10659 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10662 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10663 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10665 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10666 instead of -1U in last predictors element's probability member.
10668 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10671 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10672 arguments if they are comparable.
10674 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10676 PR tree-optimization/84634
10677 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10678 masks and masked_loop_p with a single loop_masks, making sure it's
10679 null for bb vectorization.
10681 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10683 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10684 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10685 loop->force_vectorize to check whether there is no alias.
10687 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10690 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10692 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10694 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10695 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10696 instead of a loop around prev_real_insn.
10697 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10700 PR inline-asm/84625
10701 * config/i386/i386.c (ix86_print_operand): Use conditional
10702 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10705 2018-03-02 Richard Biener <rguenther@suse.de>
10707 PR tree-optimization/84427
10708 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10709 (bitmap_set_subtract_values): Rewrite to handle multiple
10712 (prune_clobbered_mems): Likewise.
10713 (phi_translate): Take edge instead of pred/phiblock.
10714 (phi_translate_1): Likewise.
10715 (phi_translate_set): Likewise. Insert all translated
10716 exprs for a value into the set, keeping possibly multiple
10717 expressions per value.
10718 (compute_antic_aux): Adjust for phi_translate changes.
10719 When intersecting union the expressions and prune those
10720 not in the final value set, keeping possibly multiple
10721 expressions per value. Do not use value-insertion
10722 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10723 all expressions. Add verification that the value-sets
10724 only shrink during iteration.
10725 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10726 (do_pre_regular_insertion): Likewise.
10727 (do_pre_partial_partial_insertion): Likewise.
10729 2018-03-02 Richard Biener <rguenther@suse.de>
10732 * config/darwin.c (saved_debug_info_level): New static global.
10733 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10734 (darwin_asm_lto_end): Restore debug info generation settings.
10736 2018-03-01 Martin Liska <mliska@suse.cz>
10739 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10740 volatile arguments.
10742 2018-03-01 Richard Biener <rguenther@suse.de>
10745 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10746 type annotation with LTO when debug was disabled at compile-time.
10748 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10750 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10752 (mips_final_postscan_insn): Likewise.
10754 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10756 PR rtl-optimization/84528
10757 * alias.c (init_alias_target): Add commentary.
10758 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10759 a unique base value if the frame pointer is not eliminated
10760 to the stack pointer.
10762 2018-03-01 Tom de Vries <tom@codesourcery.com>
10764 PR rtl-optimization/83327
10765 * lra-int.h (hard_regs_spilled_into): Declare.
10766 * lra.c (hard_regs_spilled_into): Define.
10767 (init_reg_info): Init hard_regs_spilled_into.
10768 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10769 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10770 (process_bb_lives): Handle hard_regs_spilled_into.
10771 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10772 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10774 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10776 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10777 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10778 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10779 * config/rs6000/aix72.h: New file.
10781 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10783 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10784 instead of warning_at with conditional singular and plural messages
10788 * stor-layout.c (update_alignment_for_field): For
10789 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10790 && !DECL_PACKED (field), do the alignment update, just use
10791 only desired_align instead of MAX (type_align, desired_align)
10793 (place_field): Don't do known_align < desired_align handling
10794 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10795 is non-NULL, instead do it after rli->prev_field handling and
10796 only if not within a bitfield word. For DECL_PACKED (field)
10797 use type_align of BITS_PER_UNIT.
10799 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10801 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10802 superfluous parentheses and trailing spaces.
10804 2018-02-28 Richard Biener <rguenther@suse.de>
10806 PR tree-optimization/84584
10807 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10808 SCoPs with fake exit edge.
10810 2018-02-28 Martin Liska <mliska@suse.cz>
10813 * timevar.c (timer::print): Fix format to properly print 100%
10816 2018-02-28 Richard Biener <rguenther@suse.de>
10818 PR middle-end/84607
10819 * genmatch.c (capture_info::walk_match): Do not mark
10820 captured expressions without operands as expr_p given
10821 they act more like predicates and should be subject to
10822 "lost tail" side-effect preserving.
10824 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10826 PR rtl-optimization/81611
10827 * auto-inc-dec.c (attempt_change): Move dead note from
10828 mem_insn if it's the next use of regno
10829 (find_address): Take address use of reg holding
10830 non-incremented value. Add parm to limit search to the named
10832 (merge_in_block): Attempt to use a mem insn that is the next
10833 use of the original regno.
10835 2018-02-27 Martin Sebor <msebor@redhat.com>
10838 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10839 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10841 2018-02-27 Martin Sebor <msebor@redhat.com>
10843 PR translation/84207
10844 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10845 n argument to unsigned HOST_WIDE_INT.
10846 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10847 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10848 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10851 2018-02-27 Richard Biener <rguenther@suse.de>
10853 PR tree-optimization/84512
10854 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10855 Do not use the estimate returned from record_stmt_cost for
10856 the scalar iteration cost but sum properly using add_stmt_cost.
10858 2018-02-27 Richard Biener <rguenther@suse.de>
10860 PR tree-optimization/84466
10861 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10862 Adjust last change to less strictly validate use operands.
10864 2018-02-27 Martin Liska <mliska@suse.cz>
10866 PR gcov-profile/84548
10867 * gcov.c (process_file): Allow partial overlap and consider it
10868 also as group functions.
10869 (output_lines): Properly calculate range of lines for a group.
10871 2018-02-27 Martin Liska <mliska@suse.cz>
10873 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10874 'ggc' suffixes. Change first column width.
10875 (timer::print): Fix formatting of the column.
10877 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10879 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10880 preserve inline entry blocks for the sake of debug inline
10881 entry point markers alone.
10882 (remove_unused_locals): Suggest in comments a better place to
10883 force the preservation of inline entry blocks that are
10884 otherwise unused, but do not preserve them.
10886 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10888 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10890 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10893 * config/i386/constraints.md (Bs): Replace
10894 ix86_indirect_branch_register with
10895 TARGET_INDIRECT_BRANCH_REGISTER.
10897 * config/i386/i386.md (indirect_jump): Likewise.
10898 (tablejump): Likewise.
10899 (*sibcall_memory): Likewise.
10900 (*sibcall_value_memory): Likewise.
10901 Peepholes of indirect call and jump via memory: Likewise.
10902 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10903 (*sibcall_value_GOT_32): Likewise.
10904 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10905 (GOT_memory_operand): Likewise.
10906 (call_insn_operand): Likewise.
10907 (sibcall_insn_operand): Likewise.
10908 (GOT32_symbol_operand): Likewise.
10909 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10911 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10913 PR rtl-optimization/83496
10914 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10915 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10916 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10917 redundant insn, if any.
10918 (relax_delay_slots): Likewise.
10919 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10921 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10923 PR tree-optimization/83965
10924 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10925 that grouped statements are part of a reduction chain. Return
10926 true if the statement is not marked as a reduction itself but
10927 is part of a group.
10928 (vect_recog_dot_prod_pattern): Don't check whether the statement
10929 is part of a group here.
10930 (vect_recog_sad_pattern): Likewise.
10931 (vect_recog_widen_sum_pattern): Likewise.
10933 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10936 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10938 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10941 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10943 (ix86_output_indirect_function_return): New prototype.
10944 (ix86_split_simple_return_pop_internal): Likewise.
10945 * config/i386/i386.c (indirect_return_via_cx): New.
10946 (indirect_return_via_cx_bnd): Likewise.
10947 (indirect_thunk_name): Handle return va CX_REG.
10948 (output_indirect_thunk_function): Create alias for
10949 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10950 (ix86_output_indirect_jmp): Remove the bool argument.
10951 (ix86_output_indirect_function_return): New function.
10952 (ix86_split_simple_return_pop_internal): Likewise.
10953 * config/i386/i386.md (*indirect_jump): Don't pass false
10954 to ix86_output_indirect_jmp.
10955 (*tablejump_1): Likewise.
10956 (simple_return_pop_internal): Change it to define_insn_and_split.
10957 Call ix86_split_simple_return_pop_internal to split it for
10958 -mfunction-return=.
10959 (simple_return_indirect_internal): Call
10960 ix86_output_indirect_function_return instead of
10961 ix86_output_indirect_jmp.
10963 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10966 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10967 memset and value initialization afterwards.
10969 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10971 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10973 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10976 * common/config/aarch64/aarch64-common.c
10977 (aarch_option_optimization_table[]): Switch
10978 off fomit-frame-pointer
10980 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10981 Chung-Ju Wu <jasonwucj@gmail.com>
10983 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10985 (store_multiple): Ditto.
10987 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10989 * config.gcc: Add --with-cpu support for nds32 target.
10990 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10991 * config/nds32/nds32.opt: Add -mcpu= option.
10993 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10995 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10996 isel=yes): Warn for these deprecated options.
10998 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11000 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11001 ISA_2_5_MASKS_EMBEDDED.
11003 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11005 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11006 p->max as pointers rather than using iterative_hash_expr.
11008 2018-02-23 Carl Love <cel@us.ibm.com>
11010 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11011 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11013 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11014 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11015 P8V_BUILTIN_VEC_VUNSIGNED2.
11017 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11020 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11021 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11023 * lra-constraints.c (curr_insn_transform): Set up
11024 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11026 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11027 * lra-eliminations.c (spill_pseudos): Ditto.
11028 (process_insn_for_elimination): Ditto.
11029 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11030 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11031 LRA_NON_CLOBBERED_ALT.
11033 2018-02-22 Martin Sebor <msebor@redhat.com>
11035 PR tree-optimization/84480
11036 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11037 to maybe_diag_stxncpy_trunc. Call it.
11038 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11039 from gimple_fold_builtin_strcpy. Print inlining stack.
11040 (handle_builtin_stxncpy): Print inlining stack.
11041 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11043 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11046 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11047 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11048 and -fcheck-pointer-bounds are used together.
11049 (indirect_thunk_prefix): New enum.
11050 (indirect_thunk_need_prefix): New function.
11051 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11052 "_nt" instead of "_bnd" for NOTRACK prefix.
11053 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11054 (output_indirect_thunk_function): Likewise.
11056 (ix86_code_end): Update output_indirect_thunk_function calls.
11057 (ix86_output_indirect_branch_via_reg): Replace
11058 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11059 (ix86_output_indirect_branch_via_push): Likewise.
11060 (ix86_output_function_return): Likewise.
11061 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11062 incompatible with -fcf-protection=branch and
11063 -fcheck-pointer-bounds.
11065 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11068 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11069 Change gcc_assert call to output_operand_lossage.
11071 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11073 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11075 2018-02-22 DJ Delorie <dj@redhat.com>
11076 Sebastian Perta <sebastian.perta@renesas.com>
11077 Oleg Endo <olegendo@gcc.gnu.org>
11079 * config/rx/rx.c (rx_rtx_costs): New function.
11080 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11082 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11084 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11086 2018-02-22 Martin Liska <mliska@suse.cz>
11089 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11090 Add "native" as a possible value.
11092 2018-02-22 Martin Liska <mliska@suse.cz>
11095 * config/i386/i386.c (ix86_option_override_internal):
11096 Add "native" as a possible value for -march and -mtune.
11098 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11101 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11102 to all type variants.
11104 PR tree-optimization/84503
11105 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11106 width as info->bitpos + info->bitsize - start.
11107 (merged_store_group::merge_overlapping): Simplify width computation.
11108 (check_no_overlap): New function.
11109 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11110 start + width and last_order of the group, fail if check_no_overlap
11112 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11113 to group if check_no_overlap fails.
11115 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11117 * config/rs6000/altivec.md: Delete contraint arguments to
11118 define_expand, define_split, and define_peephole2, and in
11119 define_insn_and_split if always unused.
11120 * config/rs6000/darwin.md: Ditto.
11121 * config/rs6000/dfp.md: Ditto.
11122 * config/rs6000/rs6000.md: Ditto.
11123 * config/rs6000/sync.md: Ditto.
11124 * config/rs6000/vector.md: Ditto.
11125 * config/rs6000/vsx.md: Ditto.
11127 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11129 * config/rs6000/altivec.md: Write output control strings as braced
11130 blocks instead of double-quoted strings.
11131 * config/rs6000/darwin.md: Ditto.
11132 * config/rs6000/rs6000.md: Ditto.
11133 * config/rs6000/vector.md: Ditto.
11134 * config/rs6000/vsx.md: Ditto.
11136 2018-02-21 Jason Merrill <jason@redhat.com>
11138 PR c++/84314 - ICE with templates and fastcall attribute.
11139 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11141 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11143 * ipa-cp.c (determine_versionability): Fix comment typos.
11145 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11148 * ipa-cp.c (determine_versionability): Do not version functions caling
11151 2018-02-21 Martin Liska <mliska@suse.cz>
11154 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11155 Add "native" as a possible value.
11156 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11157 the macro when native cpu detection is available.
11159 2018-02-21 Martin Liska <mliska@suse.cz>
11162 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11163 Add "native" as a possible value.
11164 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11165 when native cpu detection is available.
11167 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11168 Martin Sebor <msebor@redhat.com>
11170 PR tree-optimization/84478
11171 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11173 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11174 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11175 support which is conservatively correct, for 2 only stay conservative
11176 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11177 argument to the 2 argument get_range_strlen, adjust 6 arg
11178 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11180 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11181 (gimple_fold_builtin_strlen): Pass true as last argument to
11184 2018-02-20 Martin Sebor <msebor@redhat.com>
11186 PR middle-end/84095
11187 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11188 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11189 (builtin_memref::builtin_memref): Factor out parts into
11190 set_base_and_offset and call it.
11192 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11194 PR middle-end/84406
11195 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11196 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11197 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11198 search at the associated MODE_INT.
11200 2018-02-20 Jeff Law <law@redhat.com>
11202 PR middle-end/82123
11203 PR tree-optimization/81592
11204 PR middle-end/79257
11205 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11206 for range data rather than using global data.
11207 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11208 range data rather than using global data.
11209 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11210 pass it to children as needed.
11211 (struct directive::fmtresult): Similarly.
11212 (struct directive::set_width): Similarly.
11213 (struct directive::set_precision): Similarly.
11214 (format_integer, format_directive, parse_directive): Similarly.
11215 (format_none): Accept unnamed vr_values parameter.
11216 (format_percent, format_floating, format_character): Similarly.
11217 (format_string, format_plain): Similarly.
11218 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11219 the EVRP range analyzer for range data rather than using global data.
11220 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11221 gimple-ssa-evrp-analyze.h
11222 (class sprintf_dom_walker): Add after_dom_children member function.
11223 Add evrp_range_analyzer member.
11224 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11225 range analyzer as needed.
11226 (sprintf_dom_walker::after_dom_children): New member function.
11227 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11229 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11230 (evrp_range_analyzer::pop_to_marker): Likewise.
11232 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11234 PR tree-optimization/84419
11235 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11236 with the required type if its current type is compatible but
11239 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11241 PR middle-end/82004
11242 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11243 after vectorization.
11245 2018-02-20 Martin Liska <mliska@suse.cz>
11248 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11249 possible values if we don't have a hint.
11251 2018-02-20 Martin Liska <mliska@suse.cz>
11255 * final.c (shorten_branches): Build align_tab array with one
11257 * opts.c (finish_options): Add alignment option limit check.
11258 (MAX_CODE_ALIGN): Likewise.
11259 (MAX_CODE_ALIGN_VALUE): Likewise.
11260 * doc/invoke.texi: Document maximum allowed option value for
11261 all -falign-* options.
11263 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11266 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11267 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11268 * var-tracking.c (emit_note_insn_var_location): Remove all references
11269 to NOTE_INSN_CALL_ARG_LOCATION.
11270 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11271 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11272 Use copy_rtx_if_shared.
11273 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11274 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11275 (dwarf2out_var_location): Remove handling of
11276 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11278 * final.c (final_scan_insn): Remove all references to
11279 NOTE_INSN_CALL_ARG_LOCATION.
11280 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11281 before dumping final insns.
11282 * except.c (emit_note_eh_region_end): Remove all references to
11283 NOTE_INSN_CALL_ARG_LOCATION.
11284 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11285 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11286 * config/arc/arc.c (hwloop_optimize): Likewise.
11287 * config/arm/arm.c (create_fix_barrier): Likewise.
11288 * config/s390/s390.c (s390_chunkify_start): Likewise.
11289 * config/sh/sh.c (find_barrier): Likewise.
11290 * config/i386/i386.c (rest_of_insert_endbranch,
11291 ix86_seh_fixup_eh_fallthru): Likewise.
11292 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11293 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11294 * config/frv/frv.c (frv_function_prologue): Likewise.
11295 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11297 (note_outside_basic_block_p): Remove all references to
11298 NOTE_INSN_CALL_ARG_LOCATION.
11299 * gengtype.c (adjust_field_rtx_def): Likewise.
11300 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11302 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11303 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11306 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11309 PR tree-optimization/84452
11310 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11311 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11314 2018-02-19 Martin Liska <mliska@suse.cz>
11317 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11319 2018-02-19 Martin Liska <mliska@suse.cz>
11320 Richard Sandiford <richard.sandiford@linaro.org>
11322 PR tree-optimization/82491
11323 * gimple-fold.c (get_base_constructor): Make earlier bail out
11326 2018-02-19 Carl Love <cel@us.ibm.com>
11328 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11329 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11331 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11332 P8V_BUILTIN_VEC_NEG.
11334 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11336 * config/rl78/rl78.md (movdf): New define expand.
11338 2018-02-19 Martin Liska <mliska@suse.cz>
11341 * doc/invoke.texi: Fix typo.
11342 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11344 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11346 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11347 handle rs6000_single_float and rs6000_double_float specially for
11350 2018-02-16 Jeff Law <law@redhat.com>
11352 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11353 (pop_regs): Use it.
11355 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11358 * ipa-inline.c (inline_small_functions): Fix a typo.
11360 2018-02-16 Nathan Sidwell <nathan@acm.org>
11362 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11364 2018-02-16 Carl Love <cel@us.ibm.com>
11366 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11367 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11368 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11369 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11370 expansion to P8V_BUILTIN_VEC_FLOAT2.
11372 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11374 PR rtl-optimization/70023
11375 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11376 src_regno into account.
11378 2018-02-16 Carl Love <cel@us.ibm.com>
11380 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11381 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11382 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11383 * config/rs6000/rs6000.c: Remove case statements for
11384 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11385 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11386 and P9V_BUILTIN_VEC_VINSERT4B.
11387 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11388 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11389 * config/rs6000/vsx.md:
11390 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11393 2018-02-16 Carl Love <cel@us.ibm.com>
11395 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11397 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11399 * config/rs6000/rs6000-c.c: Add the definitions for
11400 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11401 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11402 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11403 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11404 definition for insert4b and define insn *insert3b_internal.
11405 * doc/extend.texi: Add documentation for vec_extract4b.
11407 2018-02-16 Nathan Sidwell <nathan@acm.org>
11409 * doc/extend.texi (Backwards Compatibility): Mention friend
11410 injection. Note for-scope is deprecated.
11411 * doc/invoke.texi (-ffriend-injection): Deprecate.
11413 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11415 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11416 that moved to I2, also allow destinations that are a paradoxical
11417 subreg (instead of a normal reg).
11419 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11422 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11425 2018-02-16 Richard Biener <rguenther@suse.de>
11427 PR tree-optimization/84037
11428 PR tree-optimization/84016
11430 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11431 Adjust vec_construct for the fact we need additional higher latency
11432 128bit inserts for AVX256 and AVX512 vector builds.
11433 (ix86_add_stmt_cost): Scale vector construction cost for
11436 2018-02-16 Richard Biener <rguenther@suse.de>
11438 PR tree-optimization/84417
11439 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11440 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11441 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11443 2018-02-16 Martin Liska <mliska@suse.cz>
11446 * internal-fn.def (ASAN_CHECK): Set proper flags.
11447 (ASAN_MARK): Likewise.
11449 2018-02-16 Julia Koval <julia.koval@intel.com>
11451 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11452 from PTA_CANNONLAKE.
11454 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11457 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11458 Use ++iter rather than iter++ for std::list iterators.
11459 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11460 defer deleting them until all nodes in the forest are processed. Do
11461 free even leaf nodes. Change to_process into auto_vec.
11464 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11465 * vec.h (vec_default_construct): Use memset instead of placement new
11466 if BROKEN_VALUE_INITIALIZATION is defined.
11467 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11468 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11471 PR rtl-optimization/83723
11472 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11473 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11474 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11476 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11478 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11480 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11482 PR rtl-optimization/81443
11483 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11484 from inner REGs to paradoxical SUBREGs.
11486 2018-02-16 Richard Biener <rguenther@suse.de>
11488 PR tree-optimization/84399
11489 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11490 For operands we can analyze at their definition make sure we can
11491 analyze them at each use as well.
11493 2018-02-16 Richard Biener <rguenther@suse.de>
11495 PR tree-optimization/84190
11496 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11497 volatile accesses if the decl isn't volatile.
11499 2018-02-15 Jason Merrill <jason@redhat.com>
11501 PR c++/84314 - ICE with templates and fastcall attribute.
11502 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11503 TYPE_CANONICAL on an existing type.
11505 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11507 PR tree-optimization/84383
11508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11509 dstoff nor call operand_equal_p if dstbase is NULL.
11511 PR tree-optimization/84334
11512 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11513 also a CONSTANT_CLASS_P, punt.
11515 2018-02-14 Jim Wilson <jimw@sifive.com>
11517 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11518 first SMALL_OPERAND check. New local min_second_step. Move assert
11519 to where locals are set. Add TARGET_RVC support.
11520 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11522 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11524 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11526 2018-02-14 Martin Sebor <msebor@redhat.com>
11528 PR tree-optimization/83698
11529 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11530 arrays constrain the offset range to their bounds.
11531 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11532 (builtin_access::overlap): Avoid setting the size of overlap if it's
11534 (maybe_diag_overlap): Also consider arrays when deciding what values
11535 of offsets to include in diagnostics.
11537 2018-02-14 Martin Sebor <msebor@redhat.com>
11540 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11541 that correspond to the kind of a declaration.
11543 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11546 * config/pa/pa.md: Load address of PIC label using the linkage table
11547 if the label is nonlocal.
11549 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11551 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11552 warning message if user requests -maltivec=be.
11553 * doc/invoke.texi: Document deprecation of -maltivec=be.
11555 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11558 * config/rs6000/rs6000-c.c: Update definitions for
11559 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11560 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11562 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11565 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11566 add _get_ssp intrinsics. Remove argument from
11567 __builtin_ia32_rdssp[d|q].
11568 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11569 * config/i386/i386-builtin.def: Remove argument from
11570 __builtin_ia32_rdssp[d|q].
11571 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11572 ix86_expand_special_args_builtin for _rdssp[d|q].
11573 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11574 Clear register before usage.
11575 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11576 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11578 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11580 PR tree-optimization/84357
11581 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11582 operand 1 of an ARRAY_REF too.
11584 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11587 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11588 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11590 (set_of_reg): New struct.
11591 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11592 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11593 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11595 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11596 Split into bitclr, bitset, bitinvert patterns if appropriate.
11597 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11598 use rx_fuse_in_memory_bitop.
11599 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11600 to named insn, correct maximum insn length.
11602 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11605 * machmode.def: Define a complex mode for PARTIAL_INT.
11606 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11608 * doc/rtl.texi: Document CSPImode.
11609 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11611 (msp430_hard_regno_nregs_with_padding): Likewise.
11613 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11616 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11618 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11620 PR rtl-optimization/84169
11621 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11622 we generated a parallel as new i3 and we split that to new i2 and i3
11623 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11624 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11625 those to i2, not i1. Partially rewrite this scan code.
11627 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11630 * stor-layout.c (place_field): For variable length fields, adjust
11631 offset_align afterwards not just based on the field's alignment,
11632 but also on the size.
11634 PR middle-end/84309
11635 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11636 of exps and logs in the use_exp2 case.
11638 2018-02-13 Jeff Law <law@redhat.com>
11640 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11641 entry for "vector".
11643 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11646 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11650 * common.opt (gas-loc-support, gas-locview-support): New.
11651 (ginline-points, ginternal-reset-location-views): New.
11652 * doc/invoke.texi: Document them. Use @itemx where intended.
11653 (gvariable-location-views): Adjust.
11654 * target.def (reset_location_view): New.
11655 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11656 (TARGET_RESET_LOCATION_VIEW): New.
11657 * doc/tm.texi: Rebuilt.
11658 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11659 (dwarf2out_default_as_locview_support): New.
11660 (output_asm_line_debug_info): Use option variables.
11661 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11662 (output_loc_list): Likewise.
11663 (add_high_low_attributes): Check option variables.
11664 Don't output entry view attribute in strict mode.
11665 (gen_inlined_subroutine_die): Check option variables.
11666 (dwarf2out_inline_entry): Likewise.
11667 (init_sections_and_labels): Likewise.
11668 (dwarf2out_early_finish): Likewise.
11669 (maybe_reset_location_view): New, from...
11670 (dwarf2out_var_location): ... here. Call it.
11671 * debug.h (dwarf2out_default_as_loc_support): Declare.
11672 (dwarf2out_default_as_locview_support): Declare.
11673 * hooks.c (hook_int_rtx_insn_0): New.
11674 * hooks.h (hook_int_rtx_insn_0): Declare.
11675 * toplev.c (process_options): Take -gas-loc-support and
11676 -gas-locview-support from dwarf2out. Enable
11677 -gvariable-location-views by default only with locview
11678 assembler support. Enable -ginternal-reset-location-views by
11679 default only if the target defines the corresponding hook.
11680 Enable -ginline-points by default if location views are
11681 enabled; force it disabled if statement frontiers are
11683 * tree-inline.c (expand_call_inline): Check option variables.
11684 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11686 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11688 PR tree-optimization/84321
11689 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11690 handling. Also check whether the anti-range contains any values
11691 that satisfy the mask; switch to a VR_RANGE if not.
11693 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11696 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11698 2018-02-13 Martin Jambor <mjambor@suse.cz>
11701 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11702 of call statements, also set location of a load to a temporary.
11704 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11706 * config/rl78/rl78.c (add_vector_labels): New function.
11707 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11708 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11709 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11710 which checks that no arguments are passed.
11711 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11712 * doc/extend.texi: Documentation for the new attribute.
11714 2018-02-13 Andreas Schwab <schwab@suse.de>
11716 * config/riscv/linux.h (CPP_SPEC): Define.
11718 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11721 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11722 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11723 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11724 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11725 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11726 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11727 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11728 temporarily for AES and PCLMUL builtins.
11730 PR tree-optimization/84339
11731 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11732 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11735 PR middle-end/84309
11736 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11737 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11738 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11740 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11742 * omp-simd-clone.h: New file.
11743 * omp-simd-clone.c: Include omp-simd-clone.h.
11744 (expand_simd_clones): No longer static.
11745 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11746 cgraph.h and omp-simd-clone.h.
11747 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11748 (vect_recog_widen_shift_pattern): Formatting fix.
11749 (vect_pattern_recog_1): Don't check optab for calls.
11752 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11753 operands[2] into a REG before using gen_lowpart on it.
11755 2018-02-12 Jeff Law <law@redhat.com>
11758 * config/sh/sh.c (find_barrier): Consider a sibling call
11761 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11762 successfully back substituting a reg.
11764 2018-02-12 Richard Biener <rguenther@suse.de>
11766 PR tree-optimization/84037
11767 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11768 parameter, move visited init to caller.
11769 (vect_slp_analyze_operations): Separate cost from validity
11770 check, initialize visited once for all instances.
11771 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11773 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11775 (vect_model_promotion_demotion_cost): Likewise.
11776 (vectorizable_bswap): Guard cost modeling with !slp_node
11777 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11779 (vectorizable_call): Likewise.
11780 (vectorizable_conversion): Likewise.
11781 (vectorizable_assignment): Likewise.
11782 (vectorizable_shift): Likewise.
11783 (vectorizable_operation): Likewise.
11784 (vectorizable_store): Likewise.
11785 (vectorizable_load): Likewise.
11786 (vectorizable_condition): Likewise.
11787 (vectorizable_comparison): Likewise.
11789 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11792 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11793 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11794 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11796 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11798 PR middle-end/83665
11799 * params.def (inline-min-speedup): Increase from 8 to 15.
11800 (max-inline-insns-auto): Decrease from 40 to 30.
11801 * ipa-split.c (consider_split): Add some buffer for function to
11802 be considered inlining candidate.
11803 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11806 2018-02-12 Richard Biener <rguenther@suse.de>
11808 PR tree-optimization/84037
11809 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11810 matched stmts if we cannot swap the non-matched ones.
11812 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11814 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11815 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11816 _mm_maskz_scalef_round_ss): New intrinsics.
11817 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11818 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11819 __builtin_ia32_scalefss_round): Remove.
11820 (__builtin_ia32_scalefsd_mask_round,
11821 __builtin_ia32_scalefss_mask_round): New intrinsics.
11822 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11823 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11824 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11825 "<round_constraint>")): Changed to ...
11826 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11827 "<round_scalar_constraint>")): ... this.
11828 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11829 %0, %1, %2<round_op3>}"): Changed to ...
11830 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11831 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11832 %2<round_scalar_mask_op3>}"): ... this.
11833 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11835 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11837 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11838 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11839 (_mm_maskz_sqrt_round_ss): New intrinsics.
11840 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11841 (__builtin_ia32_sqrtsd_mask_round)
11842 (__builtin_ia32_sqrtss_mask_round): New builtins.
11843 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11844 (__builtin_ia32_sqrtss_round): Remove.
11845 (__builtin_ia32_sqrtsd_mask_round)
11846 (__builtin_ia32_sqrtss_mask_round): New builtins.
11847 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11848 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11849 ((match_operand:VF_128 1 "vector_operand"
11850 "xBm,<round_constraint>")): Changed to ...
11851 ((match_operand:VF_128 1 "vector_operand"
11852 "xBm,<round_scalar_constraint>")): ... this.
11853 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11854 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11855 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11856 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11857 %<iptr>1<round_scalar_mask_op3>}): ... this.
11858 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11859 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11861 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11864 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11865 Cast vec_cmpeq result to correct type.
11866 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11867 Cast vec_cmpgt result to correct type.
11869 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11871 * final.c (final_scan_insn_1): Renamed from...
11872 (final_scan_insn): ... this. New wrapper, to recover
11873 seen from the outermost call in recursive ones.
11874 * config/sparc/sparc.c (output_return): Drop seen from call.
11875 (output_sibcall): Likewise.
11876 * config/visium/visium.c (output_branch): Likewise.
11878 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11880 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11883 2018-02-10 Alan Modra <amodra@gmail.com>
11886 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11887 Specify LR as an input.
11889 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11892 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11893 remove_member_access_dummy_vars): New functions.
11894 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11895 lower_omp_1, execute_lower_omp): Use them.
11897 PR rtl-optimization/84308
11898 * shrink-wrap.c (spread_components): Release todo vector.
11900 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11902 PR rtl-optimization/57193
11903 * ira-color.c (struct allocno_color_data): Add member
11904 conflict_allocno_hard_prefs.
11905 (update_conflict_allocno_hard_prefs): New.
11906 (bucket_allocno_compare_func): Add a preference based on
11907 conflict_allocno_hard_prefs.
11908 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11909 (color_allocnos): Remove a dead code. Initiate
11910 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11912 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11915 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11916 constraint from =wa to wa. Avoid a subreg on the output operand,
11917 instead use a pseudo and subreg it in a move.
11918 (p9_xxbrd_<mode>): Changed to ...
11919 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11920 (p9_xxbrd_v2df): New expander.
11921 (p9_xxbrw_<mode>): Changed to ...
11922 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11923 (p9_xxbrw_v4sf): New expander.
11925 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11927 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11929 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11932 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11933 multiply in 32-bit mode.
11934 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11935 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11938 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11940 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11941 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11942 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11943 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11945 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11948 * dwarf2out.c (is_trivial_indirect_ref): New function.
11949 (dwarf2out_late_global_decl): Do not generate a location
11950 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11951 and that are not defined in the current unit.
11953 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11955 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11956 instead of a libcall for UNORDERED.
11958 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11961 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11962 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11964 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11967 * config/s390/s390.c (s390_set_current_function): Invoke
11968 s390_indirect_branch_settings also if fndecl didn't change.
11970 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11972 * config/rs6000/rs6000.md (blockage): Set length to zero.
11974 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11976 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11978 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11981 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11982 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11986 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11987 PARALLEL incoming that failed vt_get_decl_and_offset check.
11989 PR middle-end/84237
11990 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11991 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11993 (get_variable_section): For decls in named .bss* sections pass true as
11994 second argument to bss_initializer_p.
11996 2018-02-09 Marek Polacek <polacek@redhat.com>
11997 Jakub Jelinek <jakub@redhat.com>
12000 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12001 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12002 Sync some changes from cxx_fold_indirect_ref.
12004 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12006 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12008 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12009 (BLOCK_INLINE_ENTRY_LABEL): New.
12010 (dwarf2out_var_location): Disregard inline entry markers.
12011 (inline_entry_data): New struct.
12012 (inline_entry_data_hasher): New hashtable type.
12013 (inline_entry_data_hasher::hash): New.
12014 (inline_entry_data_hasher::equal): New.
12015 (inline_entry_data_table): New variable.
12016 (add_high_low_attributes): Add DW_AT_entry_pc and
12017 DW_AT_GNU_entry_view attributes if a pending entry is found
12018 in inline_entry_data_table. Add old entry_pc attribute only
12019 if debug nonbinding markers are disabled.
12020 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12021 markers are enabled.
12022 (block_within_block_p, dwarf2out_inline_entry): New.
12023 (dwarf2out_finish): Check that no entries remained in
12024 inline_entry_data_table.
12025 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12026 (final_scan_insn, notice_source_line): Likewise.
12027 (rest_of_clean_state): Skip inline entry markers.
12028 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12030 * gimple.c (gimple_build_debug_inline_entry): New.
12031 * gimple.h (enum gimple_debug_subcode): Add
12032 GIMPLE_DEBUG_INLINE_ENTRY.
12033 (gimple_build_debug_inline_entry): Declare.
12034 (gimple_debug_inline_entry_p): New.
12035 (gimple_debug_nonbind_marker_p): Adjust.
12036 * insn-notes.def (INLINE_ENTRY): New.
12037 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12038 inline entry marker notes.
12039 (print_insn): Likewise.
12040 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12041 (INSN_DEBUG_MARKER_KIND): Likewise.
12042 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12043 * tree-inline.c (expand_call_inline): Build and insert
12044 debug_inline_entry stmt.
12045 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12046 inline entry blocks early, if nonbind markers are enabled.
12047 (dump_scope_block): Dump fragment info.
12048 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12049 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12050 (gimple_build_debug_inline_entry): New.
12051 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12052 Enable/disable inline entry points too.
12053 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12054 (DEBUG_INSN): Describe inline entry markers.
12056 * common.opt (gvariable-location-views): New.
12057 (gvariable-location-views=incompat5): New.
12058 * config.in: Rebuilt.
12059 * configure: Rebuilt.
12060 * configure.ac: Test assembler for view support.
12061 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12062 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12063 * dwarf2out.c (var_loc_view): New typedef.
12064 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12065 (dwarf2out_locviews_in_attribute): New.
12066 (dwarf2out_locviews_in_loclist): New.
12067 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12068 (enum dw_line_info_opcode): Add LI_adv_address.
12069 (struct dw_line_info_table): Add view.
12070 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12071 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12072 (zero_view_p): New variable.
12073 (ZERO_VIEW_P): New macro.
12074 (output_asm_line_debug_info): New.
12075 (struct var_loc_node): Add view.
12076 (add_AT_view_list, AT_loc_list): New.
12077 (add_var_loc_to_decl): Add view param. Test it against last.
12078 (new_loc_list): Add view params. Record them.
12079 (AT_loc_list_ptr): Handle loc and view lists.
12080 (view_list_to_loc_list_val_node): New.
12081 (print_dw_val): Handle dw_val_class_view_list.
12082 (size_of_die): Likewise.
12083 (value_format): Likewise.
12084 (loc_list_has_views): New.
12085 (gen_llsym): Set vl_symbol too.
12086 (maybe_gen_llsym, skip_loc_list_entry): New.
12087 (dwarf2out_maybe_output_loclist_view_pair): New.
12088 (output_loc_list): Output view list or entries too.
12089 (output_view_list_offset): New.
12090 (output_die): Handle dw_val_class_view_list.
12091 (output_dwarf_version): New.
12092 (output_compilation_unit_header): Use it.
12093 (output_skeleton_debug_sections): Likewise.
12094 (output_rnglists, output_line_info): Likewise.
12095 (output_pubnames, output_aranges): Update version comments.
12096 (output_one_line_info_table): Output view numbers in asm comments.
12097 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12098 Call maybe_gen_llsym.
12099 (loc_list_from_tree_1): Adjust.
12100 (add_AT_location_description): Create view list attribute if
12101 needed, check it's absent otherwise.
12102 (convert_cfa_to_fb_loc_list): Adjust.
12103 (maybe_emit_file): Call output_asm_line_debug_info for test.
12104 (dwarf2out_var_location): Reset views as needed. Precompute
12105 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12106 attribute. Set view.
12107 (new_line_info_table): Reset next view.
12108 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12109 (dwarf2out_source_line): Likewise. Output view resets and labels to
12110 the assembler, or select appropriate line info opcodes.
12111 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12112 (optimize_string_length): Catch it. Adjust.
12113 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12114 dw_val_class_view_list, and remove it if no longer needed.
12115 (hash_loc_list): Hash view numbers.
12116 (loc_list_hasher::equal): Compare them.
12117 (optimize_location_lists): Check whether a view list symbol is
12118 needed, and whether the locview attribute is present, and
12119 whether they match. Remove the locview attribute if no longer
12121 (index_location_lists): Call skip_loc_list_entry for test.
12122 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12123 Use output_dwarf_version.
12124 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12125 (struct dw_val_node): Add val_view_list.
12126 * final.c (SEEN_NEXT_VIEW): New.
12127 (set_next_view_needed): New.
12128 (clear_next_view_needed): New.
12129 (maybe_output_next_view): New.
12130 (final_start_function): Rename to...
12131 (final_start_function_1): ... this. Take pointer to FIRST,
12132 add SEEN parameter. Emit param bindings in the initial view.
12133 (final_start_function): Reintroduce SEEN-less interface.
12134 (final): Rename to...
12135 (final_1): ... this. Take SEEN parameter. Output final pending
12136 next view at the end.
12137 (final): Reintroduce seen-less interface.
12138 (final_scan_insn): Output pending next view before switching
12139 sections or ending a block. Mark the next view as needed when
12140 outputting variable locations. Notify debug backend of section
12141 changes, and of location view changes.
12142 (rest_of_handle_final): Adjust.
12143 * toplev.c (process_options): Autodetect value for debug variable
12144 location views option. Warn on incompat5 without -gdwarf-5.
12145 * doc/invoke.texi (gvariable-location-views): New.
12146 (gvariable-location-views=incompat5): New.
12147 (gno-variable-location-views): New.
12149 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12151 PR tree-optimization/84136
12152 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12153 that the result of find_edge is non-NULL.
12155 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12158 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12159 storing integer register in SImode. Fix cost of 256 and 512
12160 byte aligned SSE register store.
12162 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12164 * config/i386/i386.c (ix86_multiplication_cost): Fix
12165 multiplication cost for TARGET_AVX512DQ.
12167 2018-02-08 Marek Polacek <polacek@redhat.com>
12169 PR tree-optimization/84238
12170 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12173 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12175 PR tree-optimization/84265
12176 * tree-vect-stmts.c (vectorizable_store): Don't treat
12177 VMAT_CONTIGUOUS accesses as grouped.
12178 (vectorizable_load): Likewise.
12180 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12182 PR tree-optimization/81635
12183 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12184 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12185 (test_round_for_mask): New functions.
12186 (wide_int_cc_tests): Call test_round_for_mask.
12187 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12188 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12189 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12190 range returned by get_range_info.
12192 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12195 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12196 * symtab.c: Include builtins.h
12197 (symtab_node::output_to_lto_symbol_table_p): Move here
12198 from lto-streamer-out.c:output_symbol_p.
12199 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12200 (output_symbol_p): Move all logic to symtab.c
12201 (produce_symtab): Update.
12203 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12205 * config/s390/s390-opts.h (enum indirect_branch): Define.
12206 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12207 (s390_indirect_branch_via_thunk)
12208 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12209 (enum s390_indirect_branch_type): Define.
12210 * config/s390/s390.c (struct s390_frame_layout, struct
12211 machine_function): Remove.
12212 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12213 (indirect_branch_table_label_no, indirect_branch_table_name):
12215 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12216 (enum s390_indirect_branch_option): Define.
12217 (s390_return_addr_from_memory): New function.
12218 (s390_handle_string_attribute): New function.
12219 (s390_attribute_table): Add new attribute handler.
12220 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12221 (s390_indirect_branch_via_thunk): New function.
12222 (s390_indirect_branch_via_inline_thunk): New function.
12223 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12224 sibling call optimization for non z10 compiles.
12225 (s390_emit_call): Force indirect branch target to be a single
12226 register. Add r1 clobber for non-z10 compiles.
12227 (s390_emit_epilogue): Emit return jump via return_use expander.
12228 (s390_reorg): Handle JUMP_INSNs as execute targets.
12229 (s390_option_override_internal): Perform validity checks for the
12230 new command line options.
12231 (s390_indirect_branch_attrvalue): New function.
12232 (s390_indirect_branch_settings): New function.
12233 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12234 (s390_output_indirect_thunk_function): New function.
12235 (s390_code_end): Implement target hook.
12236 (s390_case_values_threshold): Implement target hook.
12237 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12239 * config/s390/s390.h (struct s390_frame_layout)
12240 (struct machine_function): Move here from s390.c.
12241 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12242 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12243 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12244 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12245 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12246 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12247 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12248 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12249 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12250 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12251 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12252 (mnemonic attribute): Add values which aren't recognized
12254 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12255 pattern for branch conversion. Fix mnemonic attribute.
12256 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12257 indirect branch via thunk if requested.
12258 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12259 ("*indirect_jump"): Disable for branch conversion using out of
12261 ("indirect_jump_via_thunk<mode>_z10")
12262 ("indirect_jump_via_thunk<mode>")
12263 ("indirect_jump_via_inlinethunk<mode>_z10")
12264 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12265 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12266 ("casesi_jump_via_inlinethunk<mode>_z10")
12267 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12268 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12269 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12270 ("*indirect2_jump"): Disable for branch conversion.
12271 ("casesi_jump"): Turn into expander and expand patterns for branch
12273 ("return_use"): New expander.
12274 ("*return"): Emit return via thunk and rename it to ...
12275 ("*return<mode>"): ... this one.
12276 * config/s390/s390.opt: Add new options and and enum for the
12279 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12281 * lra-constraints.c (match_reload): Unconditionally use
12282 gen_lowpart_SUBREG, rather than selecting between that
12283 and equivalent gen_rtx_SUBREG code.
12285 2018-02-08 Richard Biener <rguenther@suse.de>
12287 PR tree-optimization/84233
12288 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12289 changed flag instead of boguously re-using phi_inserted.
12291 2018-02-08 Martin Jambor <mjambor@suse.cz>
12293 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12294 static local variables.
12296 2018-02-08 Richard Biener <rguenther@suse.de>
12298 PR tree-optimization/84278
12299 * tree-vect-stmts.c (vectorizable_store): When looking for
12300 smaller vector types to perform grouped strided loads/stores
12301 make sure the mode is supported by the target.
12302 (vectorizable_load): Likewise.
12304 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12306 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12307 Increase LDP/STP opportunities by adding adjacent callee-saves.
12309 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12311 PR rtl-optimization/84068
12312 PR rtl-optimization/83459
12313 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12315 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12317 PR tree-optimization/84224
12318 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12319 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12320 non-zero arguments.
12322 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12325 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12326 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12327 count, remove one USE.
12329 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12331 * doc/install.texi (Configuration): Document the
12332 --with-long-double-format={ibm,ieee} PowerPC configuration
12336 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12337 Convert from define_expand to be define_insn_and_split. Rework
12338 float/double/_Float128 conversions to QI/HI/SImode to work with
12339 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12340 conversions to QI/HImode types did a store and then a load to
12341 truncate the value. For conversions to VSX registers, don't split
12342 the insn, instead emit the code directly. Use the code iterator
12343 any_fix to combine signed and unsigned conversions.
12344 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12345 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12346 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12347 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12348 (fix_<mode>di2_hw): Likewise.
12349 (fixuns_<mode>di2_hw): Likewise.
12350 (fix_<mode>si2_hw): Likewise.
12351 (fixuns_<mode>si2_hw): Likewise.
12352 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12353 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12354 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12355 fix<uns>_trunc<SFDF:mode>si2_p8.
12356 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12357 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12358 (fix<uns>_<mode>_mem): Likewise.
12359 (fctiw<u>z_<mode>_mem): Likewise.
12360 (fix<uns>_<mode>_mem): Likewise.
12361 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12362 the register allocator from doing a direct move to the GPRs to do
12363 a store, and instead use the ISA 3.0 store byte/half-word from
12364 vector register instruction. For IEEE 128-bit floating point,
12365 also optimize stores of 32-bit ints.
12366 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12368 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12370 * genextract.c (push_pathstr_operand): New function to support
12372 (walk_rtx): Call push_pathstr_operand.
12373 (print_path): Support [a-zA-Z].
12375 2018-02-07 Richard Biener <rguenther@suse.de>
12377 PR tree-optimization/84037
12378 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12379 (cse_and_gimplify_to_preheader): Declare.
12380 (vect_get_place_in_interleaving_chain): Likewise.
12381 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12383 (_loop_vec_info::~_loop_vec_info): Delete it.
12384 (cse_and_gimplify_to_preheader): New function.
12385 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12386 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12387 (vectorizable_load): Likewise. For grouped stores always base
12388 the IV on the first element.
12389 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12390 condition before gimplifying.
12392 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12394 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12395 *DIV_EXPR and *MOD_EXPR.
12397 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12400 * config/i386/i386.c (ix86_option_override_internal): Mask out
12401 the CF_SET bit when checking -fcf-protection.
12403 2018-02-07 Tom de Vries <tom@codesourcery.com>
12406 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12409 2018-02-07 Richard Biener <rguenther@suse.de>
12411 PR tree-optimization/84204
12412 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12415 PR tree-optimization/84205
12416 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12417 special-case isl_ast_op_zdiv_r.
12419 PR tree-optimization/84223
12420 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12421 Only add conditions from within the region.
12422 (gather_bbs::after_dom_children): Adjust.
12424 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12427 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12428 * config/avr/avr.md: Only post-reload split REG-REG moves if
12429 either register is GENERAL_REG_P.
12431 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12433 PR tree-optimization/84235
12434 * tree-ssa-scopedtables.c
12435 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12436 if the subtraction is performed in floating point type where NaNs are
12437 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12438 build 1. Formatting fix.
12440 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12443 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12444 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12445 and skip it regardless of bb boundaries. Use CALL_P macro,
12446 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12447 unnecessarily, formatting fix.
12449 2018-02-06 Michael Collison <michael.collison@arm.com>
12451 * config/arm/thumb2.md:
12452 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12453 (*thumb_mov_notscc): Ditto.
12455 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12458 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12459 unsigned_fix, not "s".
12461 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12463 * configure.ac (gcc_fn_eh_frame_ro): New function.
12464 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12465 correct .eh_frame permissions.
12466 * configure: Regenerate.
12468 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12470 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12471 irrelevant options.
12473 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12475 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12476 Display warning message for -mno-speculate-indirect-jumps.
12478 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12480 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12482 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12484 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12486 PR tree-optimization/84225
12487 * tree-eh.c (find_trapping_overflow): Only call
12488 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12490 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12493 * config/i386/i386.c: Reimplement the check of possible options
12494 -mibt/-mshstk conbination. Change error messages.
12495 * doc/invoke.texi: Fix a typo: remove extra '='.
12497 2018-02-06 Marek Polacek <polacek@redhat.com>
12499 PR tree-optimization/84228
12500 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12502 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12505 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12506 emitted arch directives.
12507 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12508 __ARM_FEATURE_COPROC before changing architectures.
12510 2018-02-06 Richard Biener <rguenther@suse.de>
12512 * config/i386/i386.c (print_reg): Fix typo.
12513 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12515 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12517 * configure: Regenerate.
12519 2018-02-05 Martin Sebor <msebor@redhat.com>
12521 PR tree-optimization/83369
12522 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12525 2018-02-05 Martin Liska <mliska@suse.cz>
12527 * doc/invoke.texi: Cherry-pick upstream r323995.
12529 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12531 * ira.c (ira_init_register_move_cost): Adjust comment.
12533 2018-02-05 Martin Liska <mliska@suse.cz>
12535 PR gcov-profile/84137
12536 * doc/gcov.texi: Fix typo in documentation.
12538 2018-02-05 Martin Liska <mliska@suse.cz>
12540 PR gcov-profile/83879
12541 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12542 using dlopen functionality.
12544 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12546 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12547 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12548 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12549 _mm_maskz_range_round_ss): New intrinsics.
12550 (__builtin_ia32_rangesd128_round)
12551 (__builtin_ia32_rangess128_round): Remove.
12552 (__builtin_ia32_rangesd128_mask_round,
12553 __builtin_ia32_rangess128_mask_round): New builtins.
12554 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12555 __builtin_ia32_rangess128_round): Remove.
12556 (__builtin_ia32_rangesd128_mask_round,
12557 __builtin_ia32_rangess128_mask_round): New builtins.
12558 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12559 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12560 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12561 "<round_saeonly_constraint>")): Changed to ...
12562 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12563 "<round_saeonly_scalar_constraint>")): ... this.
12564 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12565 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12566 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12567 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12568 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12570 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12572 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12574 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12575 Remove all values except native, 8540 and 8548.
12577 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12579 * config/i386/i386.c (ix86_output_function_return): Pass
12580 INVALID_REGNUM, instead of -1, as invalid register number to
12581 indirect_thunk_name and output_indirect_thunk.
12583 2018-02-02 Julia Koval <julia.koval@intel.com>
12585 * config.gcc: Add -march=icelake.
12586 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12587 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12588 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12589 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12590 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12591 (processor_target_table): Add icelake.
12592 (ix86_option_override_internal): Handle new PTAs.
12593 (get_builtin_code_for_version): Handle icelake.
12594 (M_INTEL_COREI7_ICELAKE): New.
12595 (fold_builtin_cpu): Handle icelake.
12596 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12597 * doc/invoke.texi: Add -march=icelake.
12599 2018-02-02 Julia Koval <julia.koval@intel.com>
12601 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12602 to wide_int_bitmask.
12603 * wide-int-bitmask.h: New.
12605 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12608 * config/i386/i386.md: Replace Pmode with word_mode in
12609 builtin_setjmp_setup and builtin_longjmp to support x32.
12611 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12615 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12617 (rs6000_supported_cpu_names): New static variable.
12618 (linux_cpu_translation_table): Likewise.
12619 (elf_platform) <cpu>: Define new static variable and use it.
12620 Translate kernel AT_PLATFORM name to canonical name if needed.
12621 Error if platform name is unknown.
12623 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12626 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12628 2018-02-01 Jeff Law <law@redhat.com>
12631 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12632 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12633 the scratch if RELEASE_VIA_POP is false.
12634 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12635 If we have to save a temporary register, decrement SIZE appropriately.
12636 Pass new arguments to release_scratch_register_on_entry.
12637 (ix86_adjust_stack_and_probe): Likewise.
12638 (ix86_emit_probe_stack_range): Pass new arguments to
12639 release_scratch_register_on_entry.
12641 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12643 PR rtl-optimization/84157
12644 * combine.c (change_zero_ext): Use REG_P predicate in
12645 front of HARD_REGISTER_P predicate.
12647 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12649 * config/avr/avr.c (avr_option_override): Move disabling of
12650 -fdelete-null-pointer-checks to...
12651 * common/config/avr/avr-common.c (avr_option_optimization_table):
12654 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12656 PR tree-optimization/81635
12657 * tree-data-ref.c (split_constant_offset_1): For types that
12658 wrap on overflow, try to use range info to prove that wrapping
12661 2018-02-01 Renlin Li <renlin.li@arm.com>
12664 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12665 TAILCALL_ADDR_REGS.
12666 (aarch64_register_move_cost): Likewise.
12667 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12668 TAILCALL_ADDR_REGS.
12669 (REG_CLASS_NAMES): Likewise.
12670 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12671 TAILCALL_ADDR_REGS. Remove IP registers.
12672 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12674 2018-02-01 Richard Biener <rguenther@suse.de>
12676 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12677 for specifying RPO order and allow NULL for that.
12678 * domwalk.c (dom_walker::dom_walker): Likewise.
12679 (dom_walker::walk): Handle NULL RPO order.
12680 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12682 (rewrite_update_dom_walker): Likewise.
12683 (mark_def_dom_walker): Likewise.
12685 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12687 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12688 (aarch64_maybe_expand_sve_subreg_move): Declare.
12689 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12690 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12692 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12693 that are semantically a reverse operation.
12694 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12695 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12696 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12698 (aarch64_can_change_mode_class): For big-endian, forbid changes
12699 between two SVE modes if they have different element sizes.
12701 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12703 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12704 the TImode handling for big-endian targets.
12706 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12708 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12709 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12711 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12712 Remove BSWAP handing for big-endian targets and use the form of
12713 LD1RQ appropariate for the mode.
12715 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12717 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12718 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12719 duplicated element.
12721 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12724 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12725 check for operands that need to go through aarch64_sve_reload_be.
12727 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12729 PR tree-optimization/81661
12730 PR tree-optimization/84117
12731 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12732 * tree-eh.c: Include gimplify.h.
12733 (find_trapping_overflow, replace_trapping_overflow,
12734 rewrite_to_non_trapping_overflow): New functions.
12735 * tree-vect-loop.c: Include tree-eh.h.
12736 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12737 * tree-data-ref.c: Include tree-eh.h.
12738 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12740 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12742 PR rtl-optimization/84123
12743 * combine.c (change_zero_ext): Check if hard register satisfies
12744 can_change_dest_mode before calling gen_lowpart_SUBREG.
12746 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12749 * ira.c (ira_init_register_move_cost): Remove assert.
12751 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12753 PR rtl-optimization/84071
12754 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12755 * doc/tm.texi: Regenerate.
12757 2018-01-31 Richard Biener <rguenther@suse.de>
12759 PR tree-optimization/84132
12760 * tree-data-ref.c (analyze_miv_subscript): Properly
12761 check whether evolution_function_is_affine_multivariate_p
12762 before calling gcd_of_steps_may_divide_p.
12764 2018-01-31 Julia Koval <julia.koval@intel.com>
12767 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12768 * config/i386/i386.md (rdpid_rex64) New.
12769 (rdpid): Make 32bit only.
12771 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12774 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12775 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12777 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12780 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12782 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12784 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12786 PR rtl-optimization/84071
12787 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12788 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12790 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12792 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12793 (arc_attribute_table): Add 'aux' attribute.
12794 (arc_in_small_data_p): Consider aux like variables.
12795 (arc_is_aux_reg_p): New function.
12796 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12797 (arc_get_aux_arg): New function.
12798 (prepare_move_operands): Handle aux-register access.
12799 (arc_handle_aux_attribute): New function.
12800 * doc/extend.texi (ARC Variable attributes): Add subsection.
12802 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12804 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12805 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12806 (arc_attribute_table): Add 'uncached' attribute.
12807 (arc_print_operand): Print '.di' flag for uncached memory
12809 (arc_in_small_data_p): Do not consider for small data the uncached
12811 (arc_is_uncached_mem_p): New function.
12812 * config/arc/predicates.md (compact_store_memory_operand): Check
12813 for uncached memory accesses.
12814 (nonvol_nonimm_operand): Likewise.
12815 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12817 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12820 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12821 falign-loops=): Add Optimization flag.
12823 2018-01-30 Jeff Law <law@redhat.com>
12826 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12827 INT_REGISTERS_SAVED. Check it prior to calling
12828 get_scratch_register_on_entry.
12829 (ix86_adjust_stack_and_probe): Similarly.
12830 (ix86_emit_probe_stack_range): Similarly.
12831 (ix86_expand_prologue): Corresponding changes.
12833 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12836 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12837 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12839 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12842 * lra-constraints.c (curr_insn_transform): Process AND in the
12845 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12847 PR rtl-optimization/83986
12848 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12849 dependence against last_pending_memory_flush in addition to
12850 pending_jump_insns.
12852 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12854 PR tree-optimization/81611
12855 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12858 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12861 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12864 2018-01-30 Richard Biener <rguenther@suse.de>
12865 Jakub Jelinek <jakub@redhat.com>
12867 PR tree-optimization/84111
12868 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12869 inner loops added during recursion, as they don't have up-to-date
12872 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12875 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12876 (can_inline_edge_by_limits_p): ... here.
12877 (can_early_inline_edge_p, check_callers,
12878 update_caller_keys, update_callee_keys, recursive_inlining,
12879 add_new_edges_to_heap, speculation_useful_p,
12880 inline_small_functions,
12881 inline_small_functions, flatten_function,
12882 inline_to_all_callers_1): Update.
12884 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12886 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12889 2018-01-30 Richard Biener <rguenther@suse.de>
12891 PR tree-optimization/83008
12892 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12893 invariant and constant vector uses in stmts when they need
12894 more than one stmt.
12896 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12899 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12900 * configure: Regenerate.
12902 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12904 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12906 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12907 Use gen_rtx_REG rather than gen_lowpart.
12909 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12911 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12912 rather than 0 when creating partial subregs.
12914 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12916 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12919 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12922 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12923 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12924 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12925 flags. This restores the settings used before the 2017-07-24.
12926 Turning off pre increment/decrement/modify allows IVOPTS to
12927 optimize DF/SF loops where the index is an int.
12929 2018-01-29 Richard Biener <rguenther@suse.de>
12930 Kelvin Nilsen <kelvin@gcc.gnu.org>
12933 * tree-vect-stmts.c (vectorizable_call): Don't call
12934 targetm.vectorize_builtin_md_vectorized_function if callee is
12937 2018-01-22 Carl Love <cel@us.ibm.com>
12939 * doc/extend.tex: Fix typo in second arg in
12940 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12942 2018-01-29 Richard Biener <rguenther@suse.de>
12944 PR tree-optimization/84086
12945 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12946 (flush_ssaname_freelist): When SSA names were released reset
12947 the SCEV hash table.
12949 2018-01-29 Richard Biener <rguenther@suse.de>
12951 PR tree-optimization/84057
12952 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12953 removed paths when removing edges.
12955 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12957 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12958 -mfunction-return=@var{choice}.
12960 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12962 PR diagnostic/84034
12963 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12964 Handle CR like TAB.
12965 (layout::print_source_line): Likewise.
12966 (test_get_line_width_without_trailing_whitespace): Add test cases.
12968 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12970 PR middle-end/84040
12971 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12974 2018-01-26 Jim Wilson <jimw@sifive.com>
12976 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12978 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12981 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12983 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12984 and CMP + SUB-immediate -> SUBS.
12986 2018-01-26 Martin Sebor <msebor@redhat.com>
12988 PR tree-optimization/83896
12989 * tree-ssa-strlen.c (get_string_len): Rename...
12990 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12991 Avoid assuming length is constant.
12992 (handle_char_store): Use HOST_WIDE_INT for string length.
12994 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12997 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12998 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13000 2018-01-26 Richard Biener <rguenther@suse.de>
13002 PR rtl-optimization/84003
13003 * dse.c (record_store): Only record redundant stores when
13004 the earlier store aliases at least all accesses the later one does.
13006 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13008 PR rtl-optimization/83985
13009 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13010 REG_CFA_RESTORE insns.
13011 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13012 REG_CFA_RESTORE insns here.
13015 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13016 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13018 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13020 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13021 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13022 (arc_init): Likewise.
13023 (arc_override_options): Likewise.
13024 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13026 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13028 * config/arc/arc.h (TARGET_DBNZ): Define.
13029 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13030 properly set the tune attribute.
13031 (dbnz): Use TARGET_DBNZ guard.
13032 * config/arc/arc.opt (mtune): Add core3 option.
13034 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13036 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13037 recognize new pic like addresses.
13038 (arc_delegitimize_address): Clean up.
13040 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13042 * config/arc/arc-arches.def: Option mrf16 valid for all
13044 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13045 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13046 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13047 * config/arc/arc-tables.opt: Regenerate.
13048 * config/arc/arc.c (arc_conditional_register_usage): Handle
13049 reduced register file case.
13050 (arc_file_start): Set must have build attributes.
13051 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13052 mrf16 option value.
13053 * config/arc/arc.opt (mrf16): Add new option.
13054 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13055 * config/arc/genmultilib.awk: Handle new mrf16 option.
13056 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13057 * config/arc/t-multilib: Regenerate.
13058 * doc/invoke.texi (ARC Options): Document mrf16 option.
13060 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13062 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13063 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13064 (arc_attribute_table): Add 'secure_call' attribute.
13065 (arc_print_operand): Print secure call operand.
13066 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13068 (arc_is_secure_call_p): New function. * config/arc/arc.md
13069 (call_i): Add support for sjli instruction.
13070 (call_value_i): Likewise.
13071 * config/arc/constraints.md (Csc): New constraint.
13073 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13074 John Eric Martin <John.Martin@emmicro-us.com>
13076 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13077 * config/arc/arc.c (_arc_jli_section): New struct.
13078 (arc_jli_section): New type.
13079 (rc_jli_sections): New static variable.
13080 (arc_handle_jli_attribute): New function.
13081 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13082 (arc_file_end): New function.
13083 (TARGET_ASM_FILE_END): Define.
13084 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13085 (arc_add_jli_section): New function.
13086 (jli_call_scan): Likewise.
13087 (arc_reorg): Call jli_call_scan.
13088 (arc_output_addsi): Remove 'S' from printing asm operand.
13089 (arc_is_jli_call_p): New function.
13090 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13092 (movhi_insn): Likewise.
13093 (movsi_insn): Likewise.
13094 (movsi_set_cc_insn): Likewise.
13095 (loadqi_update): Likewise.
13096 (load_zeroextendqisi_update): Likewise.
13097 (load_signextendqisi_update): Likewise.
13098 (loadhi_update): Likewise.
13099 (load_zeroextendhisi_update): Likewise.
13100 (load_signextendhisi_update): Likewise.
13101 (loadsi_update): Likewise.
13102 (loadsf_update): Likewise.
13103 (movsicc_insn): Likewise.
13104 (bset_insn): Likewise.
13105 (bxor_insn): Likewise.
13106 (bclr_insn): Likewise.
13107 (bmsk_insn): Likewise.
13108 (bicsi3_insn): Likewise.
13109 (cmpsi_cc_c_insn): Likewise.
13110 (movsi_ne): Likewise.
13111 (movsi_cond_exec): Likewise.
13112 (clrsbsi2): Likewise.
13113 (norm_f): Likewise.
13123 (call_i): Remove 'S' asm letter, add jli instruction.
13124 (call_value_i): Likewise.
13125 * config/arc/arc.op (mjli-always): New option.
13126 * config/arc/constraints.md (Cji): New constraint.
13127 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13129 (subsf3_fpx): Likewise.
13130 (mulsf3_fpx): Likewise.
13131 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13133 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13134 function attrbutes.
13135 * doc/invoke.texi (ARC): Document mjli-always option.
13137 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13139 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13140 avoid addition with 0 and use incw and decw where possible.
13142 2018-01-26 Richard Biener <rguenther@suse.de>
13144 PR tree-optimization/81082
13145 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13146 association if it requires casting to unsigned.
13147 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13148 from fold_plusminus_mult_expr to catch important cases late when
13149 range info is available.
13151 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13153 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13154 * configure.ac (hidden_linkonce): New test.
13155 * configure: Regenerate.
13156 * config.in: Regenerate.
13158 2018-01-26 Julia Koval <julia.koval@intel.com>
13160 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13161 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13162 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13163 _mm_mask_bitshuffle_epi64_mask): Fix type.
13164 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13165 USI_FTYPE_V4DI_V4DI_USI): Remove.
13166 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13167 __builtin_ia32_vpshufbitqmb256_mask,
13168 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13169 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13170 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13172 2018-01-26 Alan Modra <amodra@gmail.com>
13175 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13176 UNSPEC_VBPERMQ. Sort other unspecs.
13178 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13180 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13182 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13184 PR middle-end/83055
13185 * predict.c (drop_profile): Do not push/pop cfun; update also
13187 (handle_missing_profiles): Fix logic looking for zero profiles.
13189 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13191 PR middle-end/83977
13192 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13193 on functions with #pragma omp declare simd or functions with simd
13195 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13196 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13197 Remove trailing \n from warning_at calls.
13199 2018-01-25 Tom de Vries <tom@codesourcery.com>
13202 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13203 for neutered workers.
13205 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13208 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13209 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13211 2018-01-24 Jeff Law <law@redhat.com>
13214 * i386.c (get_probe_interval): Move to earlier point.
13215 (ix86_compute_frame_layout): If -fstack-clash-protection and
13216 the frame is larger than the probe interval, then use pushes
13217 to save registers rather than reg->mem moves.
13218 (ix86_expand_prologue): Remove conditional for int_registers_saved
13221 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13224 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13225 min/max for never referenced object.
13227 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13229 PR middle-end/83977
13230 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13232 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13233 attributes from DECL_ATTRIBUTES (decl) without affecting
13234 DECL_ATTRIBUTES (current_function_decl).
13235 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13236 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13238 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13240 PR tree-optimization/83979
13241 * fold-const.c (fold_comparison): Use constant_boolean_node
13242 instead of boolean_{true,false}_node.
13244 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13246 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13249 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13251 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13252 Simplify the clause that sets the length attribute.
13253 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13254 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13255 clause that sets the length attribute.
13256 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13258 2018-01-24 Tom de Vries <tom@codesourcery.com>
13261 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13262 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13263 Add strict parameter.
13264 (prevent_branch_around_nothing): Insert dummy insn between branch to
13265 label and label with no ptx insn inbetween.
13266 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13268 2018-01-24 Tom de Vries <tom@codesourcery.com>
13271 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13272 for neutered threads in warp.
13273 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13275 2018-01-24 Richard Biener <rguenther@suse.de>
13277 PR tree-optimization/83176
13278 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13281 2018-01-24 Richard Biener <rguenther@suse.de>
13283 PR tree-optimization/82819
13284 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13285 code generating pluses that are no-ops in the target precision.
13287 2018-01-24 Richard Biener <rguenther@suse.de>
13289 PR middle-end/84000
13290 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13292 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13294 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13295 to merge probabilities.
13296 * predict.c (probably_never_executed): Also mark as cold functions
13297 with global 0 profile and guessed local profile.
13298 * profile-count.c (profile_probability::combine_with_count): New
13300 * profile-count.h (profile_probability::operator*,
13301 profile_probability::operator*=, profile_probability::operator/,
13302 profile_probability::operator/=): Reduce precision to adjusted
13303 and set value to guessed on contradictory divisions.
13304 (profile_probability::combine_with_freq): Remove.
13305 (profile_probability::combine_wiht_count): Declare.
13306 (profile_count::force_nonzero):: Set to adjusted.
13307 (profile_count::probability_in):: Set quality to adjusted.
13308 * tree-ssa-tail-merge.c (replace_block_by): Use
13309 combine_with_count.
13311 2018-01-23 Andrew Waterman <andrew@sifive.com>
13312 Jim Wilson <jimw@sifive.com>
13314 * config/riscv/riscv.c (riscv_stack_boundary): New.
13315 (riscv_option_override): Set riscv_stack_boundary. Handle
13316 riscv_preferred_stack_boundary_arg.
13317 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13318 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13319 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13320 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13321 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13322 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13324 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13327 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13328 of struct ix86_frame.
13329 (ix86_expand_epilogue): Likewise. Add a local variable for
13330 the reg_save_offset field in struct ix86_frame.
13332 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13334 PR tree-optimization/82604
13335 * tree-loop-distribution.c (enum partition_kind): New enum item
13336 PKIND_PARTIAL_MEMSET.
13337 (partition_builtin_p): Support above new enum item.
13338 (generate_code_for_partition): Ditto.
13339 (compute_access_range): Differentiate cases that equality can be
13340 proven at all loops, the innermost loops or no loops.
13341 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13342 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13343 (finalize_partitions, distribute_loop): Don't fuse partition of
13344 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13345 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13346 parloop is enabled.
13348 2018-01-23 Martin Liska <mliska@suse.cz>
13350 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13351 order to ignore the predictor.
13352 (PRED_POLYMORPHIC_CALL): Likewise.
13353 (PRED_RECURSIVE_CALL): Likewise.
13355 2018-01-23 Martin Liska <mliska@suse.cz>
13357 * tree-profile.c (tree_profiling): Print function header to
13358 aware reader which function we are working on.
13359 * value-prof.c (gimple_find_values_to_profile): Do not print
13360 not interesting value histograms.
13362 2018-01-23 Martin Liska <mliska@suse.cz>
13364 * profile-count.h (enum profile_quality): Add
13365 profile_uninitialized as the first value. Do not number values
13366 as they are zero based.
13367 (profile_count::verify): Update sanity check.
13368 (profile_probability::verify): Likewise.
13370 2018-01-23 Nathan Sidwell <nathan@acm.org>
13372 * doc/invoke.texi (ffor-scope): Deprecate.
13374 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13376 PR tree-optimization/83510
13377 * domwalk.c (set_all_edges_as_executable): New function.
13378 (dom_walker::dom_walker): Convert bool param
13379 "skip_unreachable_blocks" to enum reachability. Move setup of
13380 edge flags to set_all_edges_as_executable and only do it when
13381 reachability is REACHABLE_BLOCKS.
13382 * domwalk.h (enum dom_walker::reachability): New enum.
13383 (dom_walker::dom_walker): Convert bool param
13384 "skip_unreachable_blocks" to enum reachability.
13385 (set_all_edges_as_executable): New decl.
13386 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13387 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13389 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13390 but converting true to REACHABLE_BLOCKS.
13391 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13393 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13394 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13395 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13397 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13398 if check_all_array_refs will be called.
13400 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13402 * tree.c (selftest::test_location_wrappers): Add more test
13405 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13407 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13408 (selftest::test_bit_in_range): Likewise.
13410 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13413 * doc/sourcebuild.texi (vect_float): Say that the selector
13414 only describes the situation when -funsafe-math-optimizations is on.
13415 (vect_float_strict): Document.
13417 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13419 PR tree-optimization/83965
13420 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13421 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13422 instead of checking only for a reduction.
13423 (vect_recog_widen_sum_pattern): Likewise.
13425 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13427 * predict.c (probably_never_executed): Only use precise profile info.
13428 (compute_function_frequency): Skip after inlining hack since we now
13429 have quality checking.
13431 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13433 * profile-count.h (profile_probability::very_unlikely,
13434 profile_probability::unlikely, profile_probability::even): Set
13435 precision to guessed.
13437 2018-01-23 Richard Biener <rguenther@suse.de>
13439 PR tree-optimization/83963
13440 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13441 Properly terminate dominator walk when crossing the exit edge not
13442 when visiting its source block.
13444 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13447 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13448 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13450 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13452 PR tree-optimization/83957
13453 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13454 semicolon after for body surrounded by braces.
13456 PR tree-optimization/83081
13457 * profile-count.h (profile_probability::split): New method.
13458 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13459 Use profile_probability::split.
13460 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13461 when splitting a single conditional jump into 2.
13463 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13465 PR tree-optimization/69452
13466 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13469 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13471 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13472 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13473 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13475 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13477 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13479 * config/rl78/rl78.md (movdi): New define_expand.
13480 * config/rl78/rl78.c (rl78_split_movdi): New function.
13482 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13485 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13487 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13488 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13489 128-bit to produce an UNSPEC move to get the double word with the
13490 signbit and then a shift directly to do signbit.
13491 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13492 implementation with a new version that just does either a direct
13493 move or a regular move. Move memory interface to separate insns.
13494 Move insns so they are next to the expander.
13495 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13496 with signbit move. Split big and little endian case.
13497 (signbit<mode>2_dm_mem_le): Likewise.
13498 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13499 (signbit<mode>2_dm2): Likewise.
13501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13503 * config/rl78/rl78.md (anddi3): New define_expand.
13505 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13507 * config/rl78/rl78.md (umindi3): New define_expand.
13509 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13511 * config/rl78/rl78.md (smindi3): New define_expand.
13513 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13515 * config/rl78/rl78.md (smaxdi3): New define_expand.
13517 2018-01-22 Carl Love <cel@us.ibm.com>
13519 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13520 LVX_V1TI): Add macro expansion.
13521 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13522 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13523 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13524 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13525 Change check to determine if the instruction is a byte reversing
13526 entry. Fix typo in comment.
13527 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13528 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13529 Add def_builtin calls for new builtins.
13530 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13531 Add define_insn expansion.
13533 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13535 * config/rl78/rl78.md (umaxdi3): New define_expand.
13537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13539 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13540 for non-QImode registers.
13542 2018-01-22 Richard Biener <rguenther@suse.de>
13544 PR tree-optimization/83963
13545 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13546 including the loop exit block.
13547 (scop_detection::merge_sese): Likewise.
13548 (scop_detection::add_scop): Do it here instead.
13550 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13552 * doc/sourcebuild.texi (arm_softfloat): Document.
13554 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13557 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13558 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13559 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13561 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13562 David Edelsohn <dje.gcc@gmail.com>
13565 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13566 Change "crset eq" to "crset 2".
13567 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13568 (*call_indirect_aix<mode>_nospec): Likewise.
13569 (*call_value_indirect_aix<mode>_nospec): Likewise.
13570 (*call_indirect_elfv2<mode>_nospec): Likewise.
13571 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13572 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13573 change assembly output from . to $.
13574 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13575 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13576 (*tablejump<mode>_internal1_nospec): Likewise.
13578 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13581 * config/sh/sh_optimize_sett_clrt.cc:
13582 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13584 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13586 PR tree-optimization/83940
13587 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13588 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13589 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13590 use it to pass back the definition type.
13591 (vect_check_store_rhs): Likewise rhs_dt_out.
13592 (vect_build_gather_load_calls): Add a mask_dt argument and use
13593 it instead of a call to vect_is_simple_use.
13594 (vectorizable_store): Update calls to vect_check_load_store_mask
13595 and vect_check_store_rhs. Use the dt returned by the latter instead
13596 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13597 instead of calls to vect_is_simple_use. Pass the scalar rather
13598 than the vector operand to vect_is_simple_use when handling
13599 second and subsequent copies of an rhs value.
13600 (vectorizable_load): Update calls to vect_check_load_store_mask
13601 and vect_build_gather_load_calls. Use the cached mask_dt and
13602 gs_info.offset_dt instead of calls to vect_is_simple_use.
13604 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13606 PR middle-end/83945
13607 * tree-emutls.c: Include gimplify.h.
13608 (lower_emutls_2): New function.
13609 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13610 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13611 it before further processing.
13614 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13615 UINTVAL (trueop1) instead of INTVAL (op1).
13617 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13621 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13622 INCOMING_FRAME_SP_OFFSET if not defined.
13623 (scan_trace): Add ENTRY argument. If true and
13624 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13625 emit a note to adjust the CFA offset.
13626 (create_cfi_notes): Adjust scan_trace callers.
13627 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13628 INCOMING_FRAME_SP_OFFSET in the CIE.
13629 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13630 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13632 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13633 * doc/tm.texi: Regenerated.
13635 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13637 PR rtl-optimization/83147
13638 * lra-constraints.c (remove_inheritance_pseudos): Use
13639 lra_substitute_pseudo_within_insn.
13641 2018-01-19 Tom de Vries <tom@codesourcery.com>
13642 Cesar Philippidis <cesar@codesourcery.com>
13645 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13647 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13650 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13651 spaces for function labels.
13653 2018-01-19 Martin Liska <mliska@suse.cz>
13655 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13656 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13657 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13658 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13659 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13660 (PRED_CONST_RETURN): Change from 69 to 65.
13661 (PRED_NULL_RETURN): Change from 91 to 71.
13662 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13663 (PRED_LOOP_GUARD): Change from 66 to 73.
13665 2018-01-19 Martin Liska <mliska@suse.cz>
13667 * predict.c (predict_insn_def): Add new assert.
13668 (struct branch_predictor): Change type to signed integer.
13669 (test_prediction_value_range): Amend test to cover
13670 PROB_UNINITIALIZED.
13671 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13672 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13673 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13674 (PRED_LOOP_IV_COMPARE): Likewise.
13675 * predict.h (PROB_UNINITIALIZED): Define new constant.
13677 2018-01-19 Martin Liska <mliska@suse.cz>
13679 * predict.c (dump_prediction): Add new format for
13680 analyze_brprob.py script which is enabled with -details
13682 * profile-count.h (precise_p): New function.
13684 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13686 PR tree-optimization/83922
13687 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13688 there are no statements that need masking.
13689 (vect_active_double_reduction_p): New function.
13690 (vect_analyze_loop_operations): Use it when handling phis that
13691 are not in the loop header.
13693 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13695 PR tree-optimization/83914
13696 * tree-vect-loop.c (vectorizable_induction): Don't convert
13697 init_expr or apply the peeling adjustment for inductions
13698 that are nested within the vectorized loop.
13700 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13702 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13705 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13709 * function.h (gimplify_parameters): Add gimple_seq * argument.
13710 * function.c: Include gimple.h and options.h.
13711 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13712 for the added local temporaries if needed.
13713 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13714 if there are any parameter cleanups, wrap whole body into a
13715 try/finally with the cleanups.
13717 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13720 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13721 Use GET_MODE_CLASS for scalar floating point.
13723 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13727 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13728 Fix call of call_cgraph_insertion_hooks.
13730 2018-01-18 Martin Sebor <msebor@redhat.com>
13732 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13734 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13737 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13740 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13743 * common.opt: (-ffile-prefix-map): New option.
13744 * opts.c (common_handle_option): Defer it.
13745 * opts-global.c (handle_common_deferred_options): Handle it.
13746 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13747 * file-prefix-map.h: New file.
13748 (remap_debug_filename, add_debug_prefix_map): ...here.
13749 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13750 * final.c (debug_prefix_map, add_debug_prefix_map
13751 remap_debug_filename): Move to...
13752 * file-prefix-map.c: New file.
13753 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13754 generalize, get rid of alloca(), use strrchr() instead of strchr().
13755 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13756 Implement in terms of add_prefix_map().
13757 (remap_macro_filename, remap_debug_filename): Implement in term of
13759 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13760 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13761 * dbxout.c: Include file-prefix-map.h.
13762 * varasm.c: Likewise.
13763 * vmsdbgout.c: Likewise.
13764 * xcoffout.c: Likewise.
13765 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13766 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13767 * doc/invoke.texi (-ffile-prefix-map): Document.
13768 (-fdebug-prefix-map): Update description.
13770 2018-01-18 Martin Liska <mliska@suse.cz>
13772 * config/i386/i386.c (indirect_thunk_name): Document that also
13774 (output_indirect_thunk): Document why both instructions
13775 (pause and lfence) are generated.
13777 2018-01-18 Richard Biener <rguenther@suse.de>
13779 PR tree-optimization/83887
13780 * graphite-scop-detection.c
13781 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13782 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13783 (scop_detection::merge_sese): Re-implement with a flood-fill
13784 algorithm that properly finds a SESE region if it exists.
13786 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13789 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13790 pointer_diff optimizations use view_convert instead of convert.
13792 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13794 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13795 Generate different code for -mno-speculate-indirect-jumps.
13796 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13797 (*call_indirect_aix<mode>): Disable for
13798 -mno-speculate-indirect-jumps.
13799 (*call_indirect_aix<mode>_nospec): New define_insn.
13800 (*call_value_indirect_aix<mode>): Disable for
13801 -mno-speculate-indirect-jumps.
13802 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13803 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13804 -mno-speculate-indirect-jumps.
13805 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13807 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13809 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13810 long double type, set the flags for noting the default long double
13811 type, even if we don't pass or return a long double type.
13813 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13816 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13819 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13821 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13822 support for merge[hl].
13823 (fold_mergehl_helper): New helper function.
13824 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13825 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13826 (altivec_vmrglw_direct): Add xxmrglw insn.
13828 2018-01-17 Andrew Waterman <andrew@sifive.com>
13830 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13831 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13833 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13836 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13837 call the lto_location_cache before reading the
13838 DECL_SOURCE_LOCATION of the types.
13840 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13841 Richard Sandiford <richard.sandiford@linaro.org>
13843 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13844 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13845 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13846 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13847 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13849 * config/aarch64/constraints.md (aarch64_movti_operand):
13851 * config/aarch64/predicates.md (Uti): Add new constraint.
13853 2018-01-17 Carl Love <cel@us.ibm.com>
13855 * config/rs6000/vsx.md (define_expand xl_len_r,
13856 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13857 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13859 (define_expand, define_insn): Move the shift left from the
13860 define_insn to the define_expand for lxvl and stxvl instructions.
13861 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13862 and XL_LEN_R definitions to PURE.
13864 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13866 * config/i386/i386.c (indirect_thunk_name): Declare regno
13867 as unsigned int. Compare regno with INVALID_REGNUM.
13868 (output_indirect_thunk): Ditto.
13869 (output_indirect_thunk_function): Ditto.
13870 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13871 in the call to output_indirect_thunk_function.
13873 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13875 PR middle-end/83884
13876 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13877 rather than the size of inner_type to determine the stack slot size
13878 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13880 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13883 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13886 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13888 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13889 endian Linux systems to optionally enable multilibs for selecting
13890 the long double type if the user configured an explicit type.
13891 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13892 have no long double multilibs if not defined.
13893 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13894 warn if the user used -mabi={ieee,ibm}longdouble and we built
13895 multilibs for long double.
13896 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13897 appropriate multilib option.
13898 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13900 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13901 for building long double multilibs.
13902 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13904 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13906 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13909 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13911 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13914 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13917 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13920 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13922 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13923 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13925 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13927 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13928 different rtl trees depending on TARGET_64BIT.
13929 (rs6000_gen_lvx): Likewise.
13931 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13933 * config/visium/visium.md (nop): Tweak comment.
13934 (hazard_nop): Likewise.
13936 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13938 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13939 -mspeculate-indirect-jumps.
13940 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13941 for -mno-speculate-indirect-jumps.
13942 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13943 (*call_value_indirect_elfv2<mode>): Disable for
13944 -mno-speculate-indirect-jumps.
13945 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13946 (indirect_jump): Emit different RTL for
13947 -mno-speculate-indirect-jumps.
13948 (*indirect_jump<mode>): Disable for
13949 -mno-speculate-indirect-jumps.
13950 (*indirect_jump<mode>_nospec): New define_insn.
13951 (tablejump): Emit different RTL for
13952 -mno-speculate-indirect-jumps.
13953 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13954 (tablejumpsi_nospec): New define_expand.
13955 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13956 (tablejumpdi_nospec): New define_expand.
13957 (*tablejump<mode>_internal1): Disable for
13958 -mno-speculate-indirect-jumps.
13959 (*tablejump<mode>_internal1_nospec): New define_insn.
13960 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13963 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13965 * caller-save.c (insert_save): Drop unnecessary parameter. All
13968 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13969 Richard Biener <rguenth@suse.de>
13972 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13973 return early, inline manually is_gimple_sizepos. Make sure if we
13974 call gimplify_expr we don't end up with a gimple constant.
13975 * tree.c (variably_modified_type_p): Don't return true for
13976 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13977 * gimplify.h (is_gimple_sizepos): Remove.
13979 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13981 PR tree-optimization/83857
13982 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13983 vectorizable_live_operation for pure SLP statements.
13984 (vectorizable_live_operation): Handle PHIs.
13986 2018-01-16 Richard Biener <rguenther@suse.de>
13988 PR tree-optimization/83867
13989 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13990 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13992 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13995 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13996 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13997 If off is not INTEGER_CST, issue a may not be aligned warning
13998 rather than isn't aligned. Use isn%'t rather than isn't.
13999 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14001 <case MULT_EXPR>: Improve the case when bottom and one of the
14002 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14003 operand, in that case check if the other operand is multiple of
14004 bottom divided by the INTEGER_CST operand.
14006 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14009 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14010 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14011 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14012 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14013 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14014 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14015 (pa_function_arg_size): New function.
14017 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14019 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14020 in a separate statement.
14022 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14024 PR tree-optimization/83847
14025 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14026 group gathers and scatters.
14028 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14030 PR rtl-optimization/86620
14031 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14033 PR rtl-optimization/83213
14034 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14035 to last if both are JUMP_INSNs.
14037 PR tree-optimization/83843
14038 * gimple-ssa-store-merging.c
14039 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14040 store_immediate_info for bswap/nop orig_stores.
14042 2018-01-15 Andrew Waterman <andrew@sifive.com>
14044 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14046 <UDIV>: Increase cost if !TARGET_DIV.
14048 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14050 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14051 (define_attr "cr_logical_3op"): New.
14052 (cceq_ior_compare): Adjust.
14053 (cceq_ior_compare_complement): Adjust.
14054 (*cceq_rev_compare): Adjust.
14055 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14056 (is_cracked_insn): Adjust.
14057 (insn_must_be_first_in_group): Adjust.
14058 * config/rs6000/40x.md: Adjust.
14059 * config/rs6000/440.md: Adjust.
14060 * config/rs6000/476.md: Adjust.
14061 * config/rs6000/601.md: Adjust.
14062 * config/rs6000/603.md: Adjust.
14063 * config/rs6000/6xx.md: Adjust.
14064 * config/rs6000/7450.md: Adjust.
14065 * config/rs6000/7xx.md: Adjust.
14066 * config/rs6000/8540.md: Adjust.
14067 * config/rs6000/cell.md: Adjust.
14068 * config/rs6000/e300c2c3.md: Adjust.
14069 * config/rs6000/e500mc.md: Adjust.
14070 * config/rs6000/e500mc64.md: Adjust.
14071 * config/rs6000/e5500.md: Adjust.
14072 * config/rs6000/e6500.md: Adjust.
14073 * config/rs6000/mpc.md: Adjust.
14074 * config/rs6000/power4.md: Adjust.
14075 * config/rs6000/power5.md: Adjust.
14076 * config/rs6000/power6.md: Adjust.
14077 * config/rs6000/power7.md: Adjust.
14078 * config/rs6000/power8.md: Adjust.
14079 * config/rs6000/power9.md: Adjust.
14080 * config/rs6000/rs64.md: Adjust.
14081 * config/rs6000/titan.md: Adjust.
14083 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14085 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14086 ix86_indirect_branch_register logic.
14088 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14090 * config/i386/constraints.md (Bs): Update
14091 ix86_indirect_branch_register check. Don't check
14092 ix86_indirect_branch_register with GOT_memory_operand.
14094 * config/i386/predicates.md (GOT_memory_operand): Don't check
14095 ix86_indirect_branch_register here.
14096 (GOT32_symbol_operand): Likewise.
14098 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14100 * config/i386/predicates.md (constant_call_address_operand):
14101 Rewrite ix86_indirect_branch_register logic.
14102 (sibcall_insn_operand): Likewise.
14104 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14106 * config/i386/constraints.md (Bs): Replace
14107 ix86_indirect_branch_thunk_register with
14108 ix86_indirect_branch_register.
14110 * config/i386/i386.md (indirect_jump): Likewise.
14111 (tablejump): Likewise.
14112 (*sibcall_memory): Likewise.
14113 (*sibcall_value_memory): Likewise.
14114 Peepholes of indirect call and jump via memory: Likewise.
14115 * config/i386/i386.opt: Likewise.
14116 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14117 (GOT_memory_operand): Likewise.
14118 (call_insn_operand): Likewise.
14119 (sibcall_insn_operand): Likewise.
14120 (GOT32_symbol_operand): Likewise.
14122 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14124 PR middle-end/83837
14125 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14126 type rather than type addr's type points to.
14127 (expand_omp_atomic_mutex): Likewise.
14128 (expand_omp_atomic): Likewise.
14130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14133 * config/i386/i386.c (output_indirect_thunk_function): Use
14134 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14135 for __x86_return_thunk.
14137 2018-01-15 Richard Biener <rguenther@suse.de>
14139 PR middle-end/83850
14140 * expmed.c (extract_bit_field_1): Fix typo.
14142 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14145 * config/arm/iterators.md (VF): New mode iterator.
14146 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14147 Remove integer-related logic from pattern.
14148 (neon_vabd<mode>_3): Likewise.
14150 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14152 PR middle-end/82694
14153 * common.opt (fstrict-overflow): No longer an alias.
14154 (fwrapv-pointer): New option.
14155 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14156 also for pointer types based on flag_wrapv_pointer.
14157 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14158 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14159 opts->x_flag_wrapv got set.
14160 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14161 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14162 POINTER_TYPE_OVERFLOW_UNDEFINED.
14163 * match.pd: Likewise in address comparison pattern.
14164 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14166 2018-01-15 Richard Biener <rguenther@suse.de>
14169 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14170 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14171 Reset type names to their identifier if their TYPE_DECL doesn't
14172 have linkage (and thus is used for ODR and devirt).
14173 (save_debug_info_for_decl): Remove.
14174 (save_debug_info_for_type): Likewise.
14175 (add_tree_to_fld_list): Adjust.
14176 * tree-pretty-print.c (dump_generic_node): Make dumping of
14177 type names more robust.
14179 2018-01-15 Richard Biener <rguenther@suse.de>
14181 * BASE-VER: Bump to 8.0.1.
14183 2018-01-14 Martin Sebor <msebor@redhat.com>
14186 * builtins.c (check_access): Avoid warning when the no-warning bit
14189 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14191 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14192 * ira-color (allocno_hard_regs_compare): Likewise.
14194 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14197 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14198 Use .pushsection/.popsection.
14200 2018-01-14 Martin Sebor <msebor@redhat.com>
14203 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14205 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14207 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14208 entry from extra_headers.
14209 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14210 extra_headers, make the list bitwise identical to the i?86-*-* one.
14212 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14214 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14215 -mcmodel=large with -mindirect-branch=thunk,
14216 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14217 -mfunction-return=thunk-extern.
14218 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14219 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14220 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14222 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14224 * config/i386/i386.c (print_reg): Print the name of the full
14225 integer register without '%'.
14226 (ix86_print_operand): Handle 'V'.
14227 * doc/extend.texi: Document 'V' modifier.
14229 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14231 * config/i386/constraints.md (Bs): Disallow memory operand for
14232 -mindirect-branch-register.
14234 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14235 (GOT_memory_operand): Likewise.
14236 (call_insn_operand): Likewise.
14237 (sibcall_insn_operand): Likewise.
14238 (GOT32_symbol_operand): Likewise.
14239 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14240 for -mindirect-branch-register.
14241 (tablejump): Likewise.
14242 (*sibcall_memory): Likewise.
14243 (*sibcall_value_memory): Likewise.
14244 Disallow peepholes of indirect call and jump via memory for
14245 -mindirect-branch-register.
14246 (*call_pop): Replace m with Bw.
14247 (*call_value_pop): Likewise.
14248 (*sibcall_pop_memory): Replace m with Bs.
14249 * config/i386/i386.opt (mindirect-branch-register): New option.
14250 * doc/invoke.texi: Document -mindirect-branch-register option.
14252 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14254 * config/i386/i386-protos.h (ix86_output_function_return): New.
14255 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14256 set function_return_type.
14257 (indirect_thunk_name): Add ret_p to indicate thunk for function
14259 (output_indirect_thunk_function): Pass false to
14260 indirect_thunk_name.
14261 (ix86_output_indirect_branch_via_reg): Likewise.
14262 (ix86_output_indirect_branch_via_push): Likewise.
14263 (output_indirect_thunk_function): Create alias for function
14264 return thunk if regno < 0.
14265 (ix86_output_function_return): New function.
14266 (ix86_handle_fndecl_attribute): Handle function_return.
14267 (ix86_attribute_table): Add function_return.
14268 * config/i386/i386.h (machine_function): Add
14269 function_return_type.
14270 * config/i386/i386.md (simple_return_internal): Use
14271 ix86_output_function_return.
14272 (simple_return_internal_long): Likewise.
14273 * config/i386/i386.opt (mfunction-return=): New option.
14274 (indirect_branch): Mention -mfunction-return=.
14275 * doc/extend.texi: Document function_return function attribute.
14276 * doc/invoke.texi: Document -mfunction-return= option.
14278 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14280 * config/i386/i386-opts.h (indirect_branch): New.
14281 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14282 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14283 with local indirect jump when converting indirect call and jump.
14284 (ix86_set_indirect_branch_type): New.
14285 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14286 (indirectlabelno): New.
14287 (indirect_thunk_needed): Likewise.
14288 (indirect_thunk_bnd_needed): Likewise.
14289 (indirect_thunks_used): Likewise.
14290 (indirect_thunks_bnd_used): Likewise.
14291 (INDIRECT_LABEL): Likewise.
14292 (indirect_thunk_name): Likewise.
14293 (output_indirect_thunk): Likewise.
14294 (output_indirect_thunk_function): Likewise.
14295 (ix86_output_indirect_branch_via_reg): Likewise.
14296 (ix86_output_indirect_branch_via_push): Likewise.
14297 (ix86_output_indirect_branch): Likewise.
14298 (ix86_output_indirect_jmp): Likewise.
14299 (ix86_code_end): Call output_indirect_thunk_function if needed.
14300 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14302 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14303 (ix86_attribute_table): Add indirect_branch.
14304 * config/i386/i386.h (machine_function): Add indirect_branch_type
14305 and has_local_indirect_jump.
14306 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14308 (tablejump): Likewise.
14309 (*indirect_jump): Use ix86_output_indirect_jmp.
14310 (*tablejump_1): Likewise.
14311 (simple_return_indirect_internal): Likewise.
14312 * config/i386/i386.opt (mindirect-branch=): New option.
14313 (indirect_branch): New.
14316 (thunk-inline): Likewise.
14317 (thunk-extern): Likewise.
14318 * doc/extend.texi: Document indirect_branch function attribute.
14319 * doc/invoke.texi: Document -mindirect-branch= option.
14321 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14324 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14326 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14328 * ipa-inline.c (want_inline_small_function_p): Return false if
14329 inlining has already failed with CIF_FINAL_ERROR.
14330 (update_caller_keys): Call want_inline_small_function_p before
14332 (update_callee_keys): Likewise.
14334 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14336 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14338 (rs6000_quadword_masked_address_p): Likewise.
14339 (quad_aligned_load_p): Likewise.
14340 (quad_aligned_store_p): Likewise.
14341 (const_load_sequence_p): Add comment to describe the outer-most loop.
14342 (mimic_memory_attributes_and_flags): New function.
14343 (rs6000_gen_stvx): Likewise.
14344 (replace_swapped_aligned_store): Likewise.
14345 (rs6000_gen_lvx): Likewise.
14346 (replace_swapped_aligned_load): Likewise.
14347 (replace_swapped_load_constant): Capitalize argument name in
14348 comment describing this function.
14349 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14350 and stores that access quad-word aligned addresses and replace
14351 with stvx or lvx instructions when appropriate.
14352 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14353 New function prototype.
14354 (rs6000_quadword_masked_address_p): Likewise.
14355 (rs6000_gen_lvx): Likewise.
14356 (rs6000_gen_stvx): Likewise.
14357 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14358 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14359 when memory address is aligned.
14360 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14361 this split to select lvx instruction when memory address is aligned.
14362 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14363 instruction when memory address is aligned.
14364 (*vsx_le_perm_load_v16qi): Likewise.
14365 (four unnamed splitters): Modify to select the stvx instruction
14366 when memory is aligned.
14368 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14370 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14371 which appears in the queue multiple times.
14373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * tree-vectorizer.h (vec_lower_bound): New structure.
14378 (_loop_vec_info): Add check_nonzero and lower_bounds.
14379 (LOOP_VINFO_CHECK_NONZERO): New macro.
14380 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14381 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14382 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14383 fields. Make seg_len the distance travelled, not including the
14385 (dr_direction_indicator): Declare.
14386 (dr_zero_step_indicator): Likewise.
14387 (dr_known_forward_stride_p): Likewise.
14388 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14390 (runtime_alias_check_p): Allow runtime alias checks with
14392 (operator ==): Compare access_size and align.
14393 (prune_runtime_alias_test_list): Rework for new distinction between
14394 the access_size and seg_len.
14395 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14397 (get_segment_min_max): New function.
14398 (create_intersect_range_checks): Use it.
14399 (dr_step_indicator): New function.
14400 (dr_direction_indicator): Likewise.
14401 (dr_zero_step_indicator): Likewise.
14402 (dr_known_forward_stride_p): Likewise.
14403 * tree-loop-distribution.c (data_ref_segment_size): Return
14404 DR_STEP * (niters - 1).
14405 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14407 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14408 (vect_preserves_scalar_order_p): New function, split out from...
14409 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14410 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14411 (vect_vfa_access_size): New function.
14412 (vect_vfa_align): Likewise.
14413 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14414 (dump_lower_bound): New function.
14415 (vect_check_lower_bound): Likewise.
14416 (vect_small_gap_p): Likewise.
14417 (vectorizable_with_step_bound_p): Likewise.
14418 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14419 depencies if the vectorization factor is 1. Convert the checks
14420 for nonzero steps into checks on the bounds of DR_STEP. Try using
14421 a bunds check for variable steps if the minimum required step is
14422 relatively small. Update calls to the dr_with_seg_len
14423 constructor and to vect_compile_time_alias.
14424 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14426 (vect_loop_versioning): Call it.
14427 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14429 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14432 Alan Hayward <alan.hayward@arm.com>
14433 David Sherwood <david.sherwood@arm.com>
14435 * doc/sourcebuild.texi (vect_scatter_store): Document.
14436 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14438 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14440 * genopinit.c (main): Add supports_vec_scatter_store and
14441 supports_vec_scatter_store_cached to target_optabs.
14442 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14443 IFN_MASK_SCATTER_STORE.
14444 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14446 * internal-fn.h (internal_store_fn_p): Declare.
14447 (internal_fn_stored_value_index): Likewise.
14448 * internal-fn.c (scatter_store_direct): New macro.
14449 (expand_scatter_store_optab_fn): New function.
14450 (direct_scatter_store_optab_supported_p): New macro.
14451 (internal_store_fn_p): New function.
14452 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14453 IFN_MASK_SCATTER_STORE.
14454 (internal_fn_mask_index): Likewise.
14455 (internal_fn_stored_value_index): New function.
14456 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14457 for scatter stores.
14458 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14459 * optabs-query.c (supports_vec_scatter_store_p): New function.
14460 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14461 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14462 true for scatter stores.
14463 (vect_gather_scatter_fn_p): Handle scatter stores too.
14464 (vect_check_gather_scatter): Consider using scatter stores if
14465 supports_vec_scatter_store_p.
14466 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14467 scatter stores too.
14468 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14469 internal_fn_stored_value_index.
14470 (check_load_store_masking): Handle scatter stores too.
14471 (vect_get_store_rhs): Make public.
14472 (vectorizable_call): Use internal_store_fn_p.
14473 (vectorizable_store): Handle scatter store internal functions.
14474 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14475 when deciding whether the end of the group has been reached.
14476 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14477 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14478 (mask_scatter_store<mode>): New insns.
14480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14481 Alan Hayward <alan.hayward@arm.com>
14482 David Sherwood <david.sherwood@arm.com>
14484 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14485 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14486 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14488 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14489 Use vect_truncate_gather_scatter_offset if we can't treat the
14490 operation as a normal gather load or scatter store.
14491 (get_group_load_store_type): Take the gather_scatter_info
14492 as argument. Try using a gather load or scatter store for
14493 single-element groups.
14494 (get_load_store_type): Update calls to get_group_load_store_type
14495 and vect_use_strided_gather_scatters_p.
14497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14498 Alan Hayward <alan.hayward@arm.com>
14499 David Sherwood <david.sherwood@arm.com>
14501 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14502 optional tree argument.
14503 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14505 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14506 but continue to use the current value as a fallback.
14507 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14508 to compare the updates.
14509 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14510 (get_load_store_type): Use it when handling a strided access.
14511 (vect_get_strided_load_store_ops): New function.
14512 (vect_get_data_ptr_increment): Likewise.
14513 (vectorizable_load): Handle strided gather loads. Always pass
14514 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14517 Alan Hayward <alan.hayward@arm.com>
14518 David Sherwood <david.sherwood@arm.com>
14520 * doc/md.texi (gather_load@var{m}): Document.
14521 (mask_gather_load@var{m}): Likewise.
14522 * genopinit.c (main): Add supports_vec_gather_load and
14523 supports_vec_gather_load_cached to target_optabs.
14524 * optabs-tree.c (init_tree_optimization_optabs): Use
14525 ggc_cleared_alloc to allocate target_optabs.
14526 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14527 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14529 * internal-fn.h (internal_load_fn_p): Declare.
14530 (internal_gather_scatter_fn_p): Likewise.
14531 (internal_fn_mask_index): Likewise.
14532 (internal_gather_scatter_fn_supported_p): Likewise.
14533 * internal-fn.c (gather_load_direct): New macro.
14534 (expand_gather_load_optab_fn): New function.
14535 (direct_gather_load_optab_supported_p): New macro.
14536 (direct_internal_fn_optab): New function.
14537 (internal_load_fn_p): Likewise.
14538 (internal_gather_scatter_fn_p): Likewise.
14539 (internal_fn_mask_index): Likewise.
14540 (internal_gather_scatter_fn_supported_p): Likewise.
14541 * optabs-query.c (supports_at_least_one_mode_p): New function.
14542 (supports_vec_gather_load_p): Likewise.
14543 * optabs-query.h (supports_vec_gather_load_p): Declare.
14544 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14545 and memory_type field.
14546 (NUM_PATTERNS): Bump to 15.
14547 * tree-vect-data-refs.c: Include internal-fn.h.
14548 (vect_gather_scatter_fn_p): New function.
14549 (vect_describe_gather_scatter_call): Likewise.
14550 (vect_check_gather_scatter): Try using internal functions for
14551 gather loads. Recognize existing calls to a gather load function.
14552 (vect_analyze_data_refs): Consider using gather loads if
14553 supports_vec_gather_load_p.
14554 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14555 (vect_get_gather_scatter_offset_type): Likewise.
14556 (vect_convert_mask_for_vectype): Likewise.
14557 (vect_add_conversion_to_patterm): Likewise.
14558 (vect_try_gather_scatter_pattern): Likewise.
14559 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14560 (vect_vect_recog_func_ptrs): Add it.
14561 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14562 internal_fn_mask_index and internal_gather_scatter_fn_p.
14563 (check_load_store_masking): Take the gather_scatter_info as an
14564 argument and handle gather loads.
14565 (vect_get_gather_scatter_ops): New function.
14566 (vectorizable_call): Check internal_load_fn_p.
14567 (vectorizable_load): Likewise. Handle gather load internal
14569 (vectorizable_store): Update call to check_load_store_masking.
14570 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14571 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14572 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14573 (aarch64_gather_scale_operand_d): New predicates.
14574 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14575 (mask_gather_load<mode>): New insns.
14577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14578 Alan Hayward <alan.hayward@arm.com>
14579 David Sherwood <david.sherwood@arm.com>
14581 * optabs.def (fold_left_plus_optab): New optab.
14582 * doc/md.texi (fold_left_plus_@var{m}): Document.
14583 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14584 * internal-fn.c (fold_left_direct): Define.
14585 (expand_fold_left_optab_fn): Likewise.
14586 (direct_fold_left_optab_supported_p): Likewise.
14587 * fold-const-call.c (fold_const_fold_left): New function.
14588 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14589 * tree-parloops.c (valid_reduction_p): New function.
14590 (gather_scalar_reductions): Use it.
14591 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14592 (vect_finish_replace_stmt): Declare.
14593 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14594 (needs_fold_left_reduction_p): New function, split out from...
14595 (vect_is_simple_reduction): ...here. Accept reductions that
14596 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14597 (vect_force_simple_reduction): Also store the reduction type in
14598 the assignment's STMT_VINFO_REDUC_TYPE.
14599 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14600 (merge_with_identity): New function.
14601 (vect_expand_fold_left): Likewise.
14602 (vectorize_fold_left_reduction): Likewise.
14603 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14604 scalar phi in place for it. Check for target support and reject
14605 cases that would reassociate the operation. Defer the transform
14606 phase to vectorize_fold_left_reduction.
14607 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14608 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14609 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14613 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14614 call to ifc_temp_var.
14616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14617 Alan Hayward <alan.hayward@arm.com>
14618 David Sherwood <david.sherwood@arm.com>
14620 * target.def (legitimize_address_displacement): Take the original
14621 offset as a poly_int.
14622 * targhooks.h (default_legitimize_address_displacement): Update
14624 * targhooks.c (default_legitimize_address_displacement): Likewise.
14625 * doc/tm.texi: Regenerate.
14626 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14627 as an argument, moving assert of ad->disp == ad->disp_term to...
14628 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14629 Try calling targetm.legitimize_address_displacement before expanding
14630 the address rather than afterwards, and adjust for the new interface.
14631 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14632 Match the new hook interface. Handle SVE addresses.
14633 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14634 new hook interface.
14636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14638 * Makefile.in (OBJS): Add early-remat.o.
14639 * target.def (select_early_remat_modes): New hook.
14640 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14641 * doc/tm.texi: Regenerate.
14642 * targhooks.h (default_select_early_remat_modes): Declare.
14643 * targhooks.c (default_select_early_remat_modes): New function.
14644 * timevar.def (TV_EARLY_REMAT): New timevar.
14645 * passes.def (pass_early_remat): New pass.
14646 * tree-pass.h (make_pass_early_remat): Declare.
14647 * early-remat.c: New file.
14648 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14650 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14653 Alan Hayward <alan.hayward@arm.com>
14654 David Sherwood <david.sherwood@arm.com>
14656 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14657 vfm1 with a bound_epilog parameter.
14658 (vect_do_peeling): Update calls accordingly, and move the prologue
14659 call earlier in the function. Treat the base bound_epilog as 0 for
14660 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14661 this base when peeling for gaps.
14662 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14663 with fully-masked loops.
14664 (vect_estimate_min_profitable_iters): Handle the single peeled
14665 iteration in that case.
14667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14668 Alan Hayward <alan.hayward@arm.com>
14669 David Sherwood <david.sherwood@arm.com>
14671 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14672 single-element interleaving even if the size is not a power of 2.
14673 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14674 accesses for single-element interleaving if the group size is
14677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14678 Alan Hayward <alan.hayward@arm.com>
14679 David Sherwood <david.sherwood@arm.com>
14681 * doc/md.texi (fold_extract_last_@var{m}): Document.
14682 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14683 * optabs.def (fold_extract_last_optab): New optab.
14684 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14685 * internal-fn.c (fold_extract_direct): New macro.
14686 (expand_fold_extract_optab_fn): Likewise.
14687 (direct_fold_extract_optab_supported_p): Likewise.
14688 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14689 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14690 EXTRACT_LAST_REDUCTION.
14691 (get_initial_def_for_reduction): Do not create an initial vector
14692 for EXTRACT_LAST_REDUCTION reductions.
14693 (vectorizable_reduction): Leave the scalar phi in place for
14694 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14695 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14696 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14697 transform phase to vectorizable_condition.
14698 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14700 (vect_finish_stmt_generation): ...here.
14701 (vect_finish_replace_stmt): New function.
14702 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14703 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14705 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14708 Alan Hayward <alan.hayward@arm.com>
14709 David Sherwood <david.sherwood@arm.com>
14711 * doc/md.texi (extract_last_@var{m}): Document.
14712 * optabs.def (extract_last_optab): New optab.
14713 * internal-fn.def (EXTRACT_LAST): New internal function.
14714 * internal-fn.c (cond_unary_direct): New macro.
14715 (expand_cond_unary_optab_fn): Likewise.
14716 (direct_cond_unary_optab_supported_p): Likewise.
14717 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14718 loops using EXTRACT_LAST.
14719 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14720 (extract_last_<mode>): ...this optab.
14721 (vec_extract<mode><Vel>): Update accordingly.
14723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14724 Alan Hayward <alan.hayward@arm.com>
14725 David Sherwood <david.sherwood@arm.com>
14727 * target.def (empty_mask_is_expensive): New hook.
14728 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14729 * doc/tm.texi: Regenerate.
14730 * targhooks.h (default_empty_mask_is_expensive): Declare.
14731 * targhooks.c (default_empty_mask_is_expensive): New function.
14732 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14733 if the target says that empty masks are expensive.
14734 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14736 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14739 Alan Hayward <alan.hayward@arm.com>
14740 David Sherwood <david.sherwood@arm.com>
14742 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14743 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14744 (vect_use_loop_mask_for_alignment_p): New function.
14745 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14746 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14747 niters_skip argument. Make sure that the first niters_skip elements
14748 of the first iteration are inactive.
14749 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14750 Update call to vect_set_loop_masks_directly.
14751 (get_misalign_in_elems): New function, split out from...
14752 (vect_gen_prolog_loop_niters): ...here.
14753 (vect_update_init_of_dr): Take a code argument that specifies whether
14754 the adjustment should be added or subtracted.
14755 (vect_update_init_of_drs): Likewise.
14756 (vect_prepare_for_masked_peels): New function.
14757 (vect_do_peeling): Skip prologue peeling if we're using a mask
14758 instead. Update call to vect_update_inits_of_drs.
14759 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14761 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14762 alignment. Do not include the number of peeled iterations in
14763 the minimum threshold in that case.
14764 (vectorizable_induction): Adjust the start value down by
14765 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14766 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14767 Take the number of skipped iterations into account when calculating
14769 * tree-vect-stmts.c (vect_gen_while_not): New function.
14771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14772 Alan Hayward <alan.hayward@arm.com>
14773 David Sherwood <david.sherwood@arm.com>
14775 * doc/sourcebuild.texi (vect_fully_masked): Document.
14776 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14777 default value to 0.
14778 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14780 (vect_analyze_loop_2): ...here. Don't check the vectorization
14781 factor against the number of loop iterations if the loop is
14784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14785 Alan Hayward <alan.hayward@arm.com>
14786 David Sherwood <david.sherwood@arm.com>
14788 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14789 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14790 (dump_groups): Update accordingly.
14791 (iv_use::mem_type): New member variable.
14792 (address_p): New function.
14793 (record_use): Add a mem_type argument and initialize the new
14795 (record_group_use): Add a mem_type argument. Use address_p.
14796 Remove obsolete null checks of base_object. Update call to record_use.
14797 (find_interesting_uses_op): Update call to record_group_use.
14798 (find_interesting_uses_cond): Likewise.
14799 (find_interesting_uses_address): Likewise.
14800 (get_mem_type_for_internal_fn): New function.
14801 (find_address_like_use): Likewise.
14802 (find_interesting_uses_stmt): Try find_address_like_use before
14803 calling find_interesting_uses_op.
14804 (addr_offset_valid_p): Use the iv mem_type field as the type
14805 of the addressed memory.
14806 (add_autoinc_candidates): Likewise.
14807 (get_address_cost): Likewise.
14808 (split_small_address_groups_p): Use address_p.
14809 (split_address_groups): Likewise.
14810 (add_iv_candidate_for_use): Likewise.
14811 (autoinc_possible_for_pair): Likewise.
14812 (rewrite_groups): Likewise.
14813 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14814 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14815 (get_alias_ptr_type_for_ptr_address): New function.
14816 (rewrite_use_address): Rewrite address uses in calls that were
14817 identified by find_address_like_use.
14819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14820 Alan Hayward <alan.hayward@arm.com>
14821 David Sherwood <david.sherwood@arm.com>
14823 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14825 * gimple-expr.h (is_gimple_addressable: Likewise.
14826 * gimple-expr.c (is_gimple_address): Likewise.
14827 * internal-fn.c (expand_call_mem_ref): New function.
14828 (expand_mask_load_optab_fn): Use it.
14829 (expand_mask_store_optab_fn): Likewise.
14831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14832 Alan Hayward <alan.hayward@arm.com>
14833 David Sherwood <david.sherwood@arm.com>
14835 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14836 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14837 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14838 (cond_umax@var{mode}): Document.
14839 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14840 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14841 (cond_umin_optab, cond_umax_optab): New optabs.
14842 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14843 (COND_IOR, COND_XOR): New internal functions.
14844 * internal-fn.h (get_conditional_internal_fn): Declare.
14845 * internal-fn.c (cond_binary_direct): New macro.
14846 (expand_cond_binary_optab_fn): Likewise.
14847 (direct_cond_binary_optab_supported_p): Likewise.
14848 (get_conditional_internal_fn): New function.
14849 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14850 Cope with reduction statements that are vectorized as calls rather
14852 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14853 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14854 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14855 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14856 (UNSPEC_COND_EOR): New unspecs.
14857 (optab): Add mappings for them.
14858 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14859 (sve_int_op, sve_fp_op): New int attributes.
14861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14862 Alan Hayward <alan.hayward@arm.com>
14863 David Sherwood <david.sherwood@arm.com>
14865 * optabs.def (while_ult_optab): New optab.
14866 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14867 * internal-fn.def (WHILE_ULT): New internal function.
14868 * internal-fn.h (direct_internal_fn_supported_p): New override
14869 that takes two types as argument.
14870 * internal-fn.c (while_direct): New macro.
14871 (expand_while_optab_fn): New function.
14872 (convert_optab_supported_p): Likewise.
14873 (direct_while_optab_supported_p): New macro.
14874 * wide-int.h (wi::udiv_ceil): New function.
14875 * tree-vectorizer.h (rgroup_masks): New structure.
14876 (vec_loop_masks): New typedef.
14877 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14878 and fully_masked_p.
14879 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14880 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14881 (vect_max_vf): New function.
14882 (slpeel_make_loop_iterate_ntimes): Delete.
14883 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14884 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14885 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14886 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14887 internal-fn.h, stor-layout.h and optabs-query.h.
14888 (vect_set_loop_mask): New function.
14889 (add_preheader_seq): Likewise.
14890 (add_header_seq): Likewise.
14891 (interleave_supported_p): Likewise.
14892 (vect_maybe_permute_loop_masks): Likewise.
14893 (vect_set_loop_masks_directly): Likewise.
14894 (vect_set_loop_condition_masked): Likewise.
14895 (vect_set_loop_condition_unmasked): New function, split out from
14896 slpeel_make_loop_iterate_ntimes.
14897 (slpeel_make_loop_iterate_ntimes): Rename to..
14898 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14899 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14900 (vect_do_peeling): Update call accordingly.
14901 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14903 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14904 mask_compare_type, can_fully_mask_p and fully_masked_p.
14905 (release_vec_loop_masks): New function.
14906 (_loop_vec_info): Use it to free the loop masks.
14907 (can_produce_all_loop_masks_p): New function.
14908 (vect_get_max_nscalars_per_iter): Likewise.
14909 (vect_verify_full_masking): Likewise.
14910 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14911 retries, and free the mask rgroups before retrying. Check loop-wide
14912 reasons for disallowing fully-masked loops. Make the final decision
14913 about whether use a fully-masked loop or not.
14914 (vect_estimate_min_profitable_iters): Do not assume that peeling
14915 for the number of iterations will be needed for fully-masked loops.
14916 (vectorizable_reduction): Disable fully-masked loops.
14917 (vectorizable_live_operation): Likewise.
14918 (vect_halve_mask_nunits): New function.
14919 (vect_double_mask_nunits): Likewise.
14920 (vect_record_loop_mask): Likewise.
14921 (vect_get_loop_mask): Likewise.
14922 (vect_transform_loop): Handle the case in which the final loop
14923 iteration might handle a partial vector. Call vect_set_loop_condition
14924 instead of slpeel_make_loop_iterate_ntimes.
14925 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14926 (check_load_store_masking): New function.
14927 (prepare_load_store_mask): Likewise.
14928 (vectorizable_store): Handle fully-masked loops.
14929 (vectorizable_load): Likewise.
14930 (supportable_widening_operation): Use vect_halve_mask_nunits for
14932 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14933 (vect_gen_while): New function.
14934 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14935 (aarch64_uqdec<mode>): New insn.
14937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14938 Alan Hayward <alan.hayward@arm.com>
14939 David Sherwood <david.sherwood@arm.com>
14941 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14942 (reduc_xor_scal_optab): New optabs.
14943 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14944 (reduc_xor_scal_@var{m}): Document.
14945 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14946 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14947 internal functions.
14948 * fold-const-call.c (fold_const_call): Handle them.
14949 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14950 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14951 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14952 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14953 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14954 (UNSPEC_XORV): New unspecs.
14955 (optab): Add entries for them.
14956 (BITWISEV): New int iterator.
14957 (bit_reduc_op): New int attributes.
14959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14960 Alan Hayward <alan.hayward@arm.com>
14961 David Sherwood <david.sherwood@arm.com>
14963 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14964 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14965 * optabs.def (vec_shl_insert_optab): New optab.
14966 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14967 (duplicate_and_interleave): Likewise.
14968 * tree-vect-loop.c: Include internal-fn.h.
14969 (neutral_op_for_slp_reduction): New function, split out from
14970 get_initial_defs_for_reduction.
14971 (get_initial_def_for_reduction): Handle option 2 for variable-length
14972 vectors by loading the neutral value into a vector and then shifting
14973 the initial value into element 0.
14974 (get_initial_defs_for_reduction): Replace the code argument with
14975 the neutral value calculated by neutral_op_for_slp_reduction.
14976 Use gimple_build_vector for constant-length vectors.
14977 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14978 but the first group_size elements have a neutral value.
14979 Use duplicate_and_interleave otherwise.
14980 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14981 Update call to get_initial_defs_for_reduction. Handle SLP
14982 reductions for variable-length vectors by creating one vector
14983 result for each scalar result, with the elements associated
14984 with other scalar results stubbed out with the neutral value.
14985 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14986 Require IFN_VEC_SHL_INSERT for double reductions on
14987 variable-length vectors, or SLP reductions that have
14988 a neutral value. Require can_duplicate_and_interleave_p
14989 support for variable-length unchained SLP reductions if there
14990 is no neutral value, such as for MIN/MAX reductions. Also require
14991 the number of vector elements to be a multiple of the number of
14992 SLP statements when doing variable-length unchained SLP reductions.
14993 Update call to vect_create_epilog_for_reduction.
14994 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14995 and remove initial values.
14996 (duplicate_and_interleave): Make public.
14997 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14998 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15001 Alan Hayward <alan.hayward@arm.com>
15002 David Sherwood <david.sherwood@arm.com>
15004 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15005 (can_duplicate_and_interleave_p): New function.
15006 (vect_get_and_check_slp_defs): Take the vector of statements
15007 rather than just the current one. Remove excess parentheses.
15008 Restriction rejectinon of vect_constant_def and vect_external_def
15009 for variable-length vectors to boolean types, or types for which
15010 can_duplicate_and_interleave_p is false.
15011 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15012 (duplicate_and_interleave): New function.
15013 (vect_get_constant_vectors): Use gimple_build_vector for
15014 constant-length vectors and suitable variable-length constant
15015 vectors. Use duplicate_and_interleave for other variable-length
15016 vectors. Don't defer the update when inserting new statements.
15018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15019 Alan Hayward <alan.hayward@arm.com>
15020 David Sherwood <david.sherwood@arm.com>
15022 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15023 min_profitable_iters doesn't go negative.
15025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15026 Alan Hayward <alan.hayward@arm.com>
15027 David Sherwood <david.sherwood@arm.com>
15029 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15030 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15031 * optabs.def (vec_mask_load_lanes_optab): New optab.
15032 (vec_mask_store_lanes_optab): Likewise.
15033 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15034 (MASK_STORE_LANES): Likewise.
15035 * internal-fn.c (mask_load_lanes_direct): New macro.
15036 (mask_store_lanes_direct): Likewise.
15037 (expand_mask_load_optab_fn): Handle masked operations.
15038 (expand_mask_load_lanes_optab_fn): New macro.
15039 (expand_mask_store_optab_fn): Handle masked operations.
15040 (expand_mask_store_lanes_optab_fn): New macro.
15041 (direct_mask_load_lanes_optab_supported_p): Likewise.
15042 (direct_mask_store_lanes_optab_supported_p): Likewise.
15043 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15045 (vect_load_lanes_supported): Likewise.
15046 * tree-vect-data-refs.c (strip_conversion): New function.
15047 (can_group_stmts_p): Likewise.
15048 (vect_analyze_data_ref_accesses): Use it instead of checking
15049 for a pair of assignments.
15050 (vect_store_lanes_supported): Take a masked_p parameter.
15051 (vect_load_lanes_supported): Likewise.
15052 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15053 vect_store_lanes_supported and vect_load_lanes_supported.
15054 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15055 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15056 parameter. Don't allow gaps for masked accesses.
15057 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15058 and vect_load_lanes_supported.
15059 (get_load_store_type): Take a masked_p parameter and update
15060 call to get_group_load_store_type.
15061 (vectorizable_store): Update call to get_load_store_type.
15062 Handle IFN_MASK_STORE_LANES.
15063 (vectorizable_load): Update call to get_load_store_type.
15064 Handle IFN_MASK_LOAD_LANES.
15066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15067 Alan Hayward <alan.hayward@arm.com>
15068 David Sherwood <david.sherwood@arm.com>
15070 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15072 * config/aarch64/aarch64-protos.h
15073 (aarch64_sve_struct_memory_operand_p): Declare.
15074 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15075 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15076 (VPRED, vpred): Handle SVE structure modes.
15077 * config/aarch64/constraints.md (Utx): New constraint.
15078 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15079 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15080 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15081 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15082 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15083 structure modes. Split into pieces after RA.
15084 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15085 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15087 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15088 SVE structure modes.
15089 (aarch64_classify_address): Likewise.
15090 (sizetochar): Move earlier in file.
15091 (aarch64_print_operand): Handle SVE register lists.
15092 (aarch64_array_mode): New function.
15093 (aarch64_sve_struct_memory_operand_p): Likewise.
15094 (TARGET_ARRAY_MODE): Redefine.
15096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15097 Alan Hayward <alan.hayward@arm.com>
15098 David Sherwood <david.sherwood@arm.com>
15100 * target.def (array_mode): New target hook.
15101 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15102 * doc/tm.texi: Regenerate.
15103 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15104 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15105 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15106 targetm.array_mode.
15107 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15111 Alan Hayward <alan.hayward@arm.com>
15112 David Sherwood <david.sherwood@arm.com>
15114 * fold-const.c (fold_binary_loc): Check the argument types
15115 rather than the result type when testing for a vector operation.
15117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15119 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15120 * doc/tm.texi: Regenerate.
15122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15123 Alan Hayward <alan.hayward@arm.com>
15124 David Sherwood <david.sherwood@arm.com>
15126 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15127 (sve): Document new AArch64 extension.
15128 * doc/md.texi (w): Extend the description of the AArch64
15129 constraint to include SVE vectors.
15130 (Upl, Upa): Document new AArch64 predicate constraints.
15131 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15133 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15134 (msve-vector-bits=): New option.
15135 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15136 SVE when these are disabled.
15137 (sve): New extension.
15138 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15139 modes. Adjust their number of units based on aarch64_sve_vg.
15140 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15141 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15142 aarch64_addr_query_type.
15143 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15144 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15145 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15146 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15147 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15148 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15149 (aarch64_simd_imm_zero_p): Delete.
15150 (aarch64_check_zero_based_sve_index_immediate): Declare.
15151 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15152 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15153 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15154 (aarch64_sve_float_mul_immediate_p): Likewise.
15155 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15156 rather than an rtx.
15157 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15158 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15159 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15160 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15161 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15162 (aarch64_regmode_natural_size): Likewise.
15163 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15164 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15166 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15167 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15168 for VG and the SVE predicate registers.
15169 (V_ALIASES): Add a "z"-prefixed alias.
15170 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15171 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15172 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15173 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15174 (REG_CLASS_NAMES): Add entries for them.
15175 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15176 and the predicate registers.
15177 (aarch64_sve_vg): Declare.
15178 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15179 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15180 (REGMODE_NATURAL_SIZE): Define.
15181 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15183 * config/aarch64/aarch64.c: Include cfgrtl.h.
15184 (simd_immediate_info): Add a constructor for series vectors,
15185 and an associated step field.
15186 (aarch64_sve_vg): New variable.
15187 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15188 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15189 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15190 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15191 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15192 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15193 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15194 (aarch64_get_mask_mode): New functions.
15195 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15196 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15197 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15198 predicate modes and predicate registers. Explicitly restrict
15199 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15200 to store a vector mode if it is recognized by
15201 aarch64_classify_vector_mode.
15202 (aarch64_regmode_natural_size): New function.
15203 (aarch64_hard_regno_caller_save_mode): Return the original mode
15205 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15206 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15207 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15208 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15210 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15211 does not overlap dest if the function is frame-related. Handle
15213 (aarch64_split_add_offset): New function.
15214 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15215 them aarch64_add_offset.
15216 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15217 and update call to aarch64_sub_sp.
15218 (aarch64_add_cfa_expression): New function.
15219 (aarch64_expand_prologue): Pass extra temporary registers to the
15220 functions above. Handle the case in which we need to emit new
15221 DW_CFA_expressions for registers that were originally saved
15222 relative to the stack pointer, but now have to be expressed
15223 relative to the frame pointer.
15224 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15226 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15227 IP0 and IP1 values for SVE frames.
15228 (aarch64_expand_vec_series): New function.
15229 (aarch64_expand_sve_widened_duplicate): Likewise.
15230 (aarch64_expand_sve_const_vector): Likewise.
15231 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15232 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15233 into the register, rather than emitting a SET directly.
15234 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15235 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15236 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15237 (offset_9bit_signed_scaled_p): New functions.
15238 (aarch64_replicate_bitmask_imm): New function.
15239 (aarch64_bitmask_imm): Use it.
15240 (aarch64_cannot_force_const_mem): Reject expressions involving
15241 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15242 (aarch64_classify_index): Handle SVE indices, by requiring
15243 a plain register index with a scale that matches the element size.
15244 (aarch64_classify_address): Handle SVE addresses. Assert that
15245 the mode of the address is VOIDmode or an integer mode.
15246 Update call to aarch64_classify_symbol.
15247 (aarch64_classify_symbolic_expression): Update call to
15248 aarch64_classify_symbol.
15249 (aarch64_const_vec_all_in_range_p): New function.
15250 (aarch64_print_vector_float_operand): Likewise.
15251 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15252 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15253 and the FP immediates 1.0 and 0.5.
15254 (aarch64_print_address_internal): Handle SVE addresses.
15255 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15256 (aarch64_regno_regclass): Handle predicate registers.
15257 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15259 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15260 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15261 (aarch64_convert_sve_vector_bits): New function.
15262 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15263 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15264 rather than an rtx.
15265 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15266 Handle SVE vector and predicate modes. Accept VL-based constants
15267 that need only one temporary register, and VL offsets that require
15268 no temporary registers.
15269 (aarch64_conditional_register_usage): Mark the predicate registers
15270 as fixed if SVE isn't available.
15271 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15272 Return true for SVE vector and predicate modes.
15273 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15274 rather than an unsigned int. Handle SVE modes.
15275 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15277 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15279 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15280 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15281 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15282 (aarch64_sve_float_mul_immediate_p): New functions.
15283 (aarch64_sve_valid_immediate): New function.
15284 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15285 Explicitly reject structure modes. Check for INDEX constants.
15286 Handle PTRUE and PFALSE constants.
15287 (aarch64_check_zero_based_sve_index_immediate): New function.
15288 (aarch64_simd_imm_zero_p): Delete.
15289 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15290 vector modes. Accept constants in the range of CNT[BHWD].
15291 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15292 ask for an Advanced SIMD mode.
15293 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15294 (aarch64_simd_vector_alignment): Handle SVE predicates.
15295 (aarch64_vectorize_preferred_vector_alignment): New function.
15296 (aarch64_simd_vector_alignment_reachable): Use it instead of
15298 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15299 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15301 (MAX_VECT_LEN): Delete.
15302 (expand_vec_perm_d): Add a vec_flags field.
15303 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15304 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15305 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15307 (aarch64_evpc_rev): Rename to...
15308 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15309 (aarch64_evpc_rev_global): New function.
15310 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15311 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15313 (aarch64_evpc_sve_tbl): New function.
15314 (aarch64_expand_vec_perm_const_1): Update after rename of
15315 aarch64_evpc_rev. Handle SVE permutes too, trying
15316 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15317 than aarch64_evpc_tbl.
15318 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15319 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15320 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15321 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15322 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15323 (aarch64_expand_sve_vcond): New functions.
15324 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15325 of aarch64_vector_mode_p.
15326 (aarch64_dwarf_poly_indeterminate_value): New function.
15327 (aarch64_compute_pressure_classes): Likewise.
15328 (aarch64_can_change_mode_class): Likewise.
15329 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15330 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15331 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15332 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15333 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15334 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15335 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15336 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15338 (Dn, Dl, Dr): Accept const as well as const_vector.
15339 (Dz): Likewise. Compare against CONST0_RTX.
15340 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15341 of "vector" where appropriate.
15342 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15343 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15344 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15345 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15346 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15347 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15348 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15349 (v_int_equiv): Extend to SVE modes.
15350 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15352 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15353 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15354 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15355 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15356 (SVE_COND_FP_CMP): New int iterators.
15357 (perm_hilo): Handle the new unpack unspecs.
15358 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15360 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15361 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15362 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15363 (aarch64_equality_operator, aarch64_constant_vector_operand)
15364 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15365 (aarch64_sve_nonimmediate_operand): Likewise.
15366 (aarch64_sve_general_operand): Likewise.
15367 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15368 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15369 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15370 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15371 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15372 (aarch64_sve_float_arith_immediate): Likewise.
15373 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15374 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15375 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15376 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15377 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15378 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15379 (aarch64_sve_float_arith_operand): Likewise.
15380 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15381 (aarch64_sve_float_mul_operand): Likewise.
15382 (aarch64_sve_vec_perm_operand): Likewise.
15383 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15384 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15385 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15386 as well as const_vector.
15387 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15388 in file. Use CONST0_RTX and CONSTM1_RTX.
15389 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15390 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15391 Use aarch64_simd_imm_zero.
15392 * config/aarch64/aarch64-sve.md: New file.
15393 * config/aarch64/aarch64.md: Include it.
15394 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15395 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15396 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15397 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15398 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15399 (sve): New attribute.
15400 (enabled): Disable instructions with the sve attribute unless
15402 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15403 aarch64_expand_mov_immediate.
15404 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15405 CNT[BHSD] immediates.
15406 (movti): Split CONST_POLY_INT moves into two halves.
15407 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15408 Split additions that need a temporary here if the destination
15409 is the stack pointer.
15410 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15411 (*add<mode>3_poly_1): New instruction.
15412 (set_clobber_cc): New expander.
15414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15416 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15417 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15418 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15419 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15420 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15421 Change innermode from fixed_mode_size to machine_mode.
15422 (simplify_subreg): Update call accordingly. Handle a constant-sized
15423 subreg of a variable-length CONST_VECTOR.
15425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15426 Alan Hayward <alan.hayward@arm.com>
15427 David Sherwood <david.sherwood@arm.com>
15429 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15430 (add_offset_to_base): New function, split out from...
15431 (create_mem_ref): ...here. When handling a scale other than 1,
15432 check first whether the address is valid without the offset.
15433 Add it into the base if so, leaving the index and scale as-is.
15435 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15438 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15439 fold_for_warn before checking if arg2 is INTEGER_CST.
15441 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15443 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15444 (store_multiple_operation): Delete.
15445 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15446 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15447 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15448 guarded by TARGET_STRING.
15449 (rs6000_output_load_multiple): Delete.
15450 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15451 OPTION_MASK_STRING / TARGET_STRING handling.
15452 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15453 (const rs6000_opt_masks) <"string">: Change mask to 0.
15454 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15455 (MASK_STRING): Delete.
15456 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15458 (load_multiple): Delete.
15465 (store_multiple): Delete.
15472 (movmemsi_8reg): Delete.
15473 (corresponding unnamed define_insn): Delete.
15474 (movmemsi_6reg): Delete.
15475 (corresponding unnamed define_insn): Delete.
15476 (movmemsi_4reg): Delete.
15477 (corresponding unnamed define_insn): Delete.
15478 (movmemsi_2reg): Delete.
15479 (corresponding unnamed define_insn): Delete.
15480 (movmemsi_1reg): Delete.
15481 (corresponding unnamed define_insn): Delete.
15482 * config/rs6000/rs6000.opt (mno-string): New.
15483 (mstring): Replace by deprecation warning stub.
15484 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15486 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15488 * regrename.c (regrename_do_replace): If replacing the same
15489 reg multiple times, try to reuse last created gen_raw_REG.
15492 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15493 main to workaround a bug in GDB.
15495 2018-01-12 Tom de Vries <tom@codesourcery.com>
15498 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15500 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15502 PR rtl-optimization/80481
15503 * ira-color.c (get_cap_member): New function.
15504 (allocnos_conflict_by_live_ranges_p): Use it.
15505 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15506 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15508 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15511 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15512 (*saddl_se_1): Ditto.
15513 (*ssubsi_1): Ditto.
15514 (*ssubl_se_1): Ditto.
15516 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15518 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15519 rather than wi::to_widest for DR_INITs.
15520 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15521 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15522 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15524 (vect_analyze_group_access_1): Note that here.
15526 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15528 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15529 polynomial type sizes.
15531 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15533 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15534 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15535 (gimple_add_tmp_var): Likewise.
15537 2018-01-12 Martin Liska <mliska@suse.cz>
15539 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15540 (gimple_alloc_sizes): Likewise.
15541 (dump_gimple_statistics): Use PRIu64 in printf format.
15542 * gimple.h: Change uint64_t to int.
15544 2018-01-12 Martin Liska <mliska@suse.cz>
15546 * tree-core.h: Use uint64_t instead of int.
15547 * tree.c (tree_node_counts): Likewise.
15548 (tree_node_sizes): Likewise.
15549 (dump_tree_statistics): Use PRIu64 in printf format.
15551 2018-01-12 Martin Liska <mliska@suse.cz>
15553 * Makefile.in: As qsort_chk is implemented in vec.c, add
15554 vec.o to linkage of gencfn-macros.
15555 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15556 passing the info to record_node_allocation_statistics.
15557 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15559 * ggc-common.c (struct ggc_usage): Add operator== and use
15560 it in operator< and compare function.
15561 * mem-stats.h (struct mem_usage): Likewise.
15562 * vec.c (struct vec_usage): Remove operator< and compare
15563 function. Can be simply inherited.
15565 2018-01-12 Martin Jambor <mjambor@suse.cz>
15568 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15569 * tree-ssa-math-opts.c: Include domwalk.h.
15570 (convert_mult_to_fma_1): New function.
15571 (fma_transformation_info): New type.
15572 (fma_deferring_state): Likewise.
15573 (cancel_fma_deferring): New function.
15574 (result_of_phi): Likewise.
15575 (last_fma_candidate_feeds_initial_phi): Likewise.
15576 (convert_mult_to_fma): Added deferring logic, split actual
15577 transformation to convert_mult_to_fma_1.
15578 (math_opts_dom_walker): New type.
15579 (math_opts_dom_walker::after_dom_children): New method, body moved
15580 here from pass_optimize_widening_mul::execute, added deferring logic
15582 (pass_optimize_widening_mul::execute): Moved most of code to
15583 math_opts_dom_walker::after_dom_children.
15584 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15585 * config/i386/i386.c (ix86_option_override_internal): Added
15586 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15588 2018-01-12 Richard Biener <rguenther@suse.de>
15591 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15592 inline instance vars.
15594 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15597 * config/rx/rx.c (rx_is_restricted_memory_address):
15598 Handle SUBREG case.
15600 2018-01-12 Richard Biener <rguenther@suse.de>
15602 PR tree-optimization/80846
15603 * target.def (split_reduction): New target hook.
15604 * targhooks.c (default_split_reduction): New function.
15605 * targhooks.h (default_split_reduction): Declare.
15606 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15607 target requests first reduce vectors by combining low and high
15609 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15610 (get_vectype_for_scalar_type_and_size): Export.
15611 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15612 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15613 * doc/tm.texi: Regenerate.
15614 * config/i386/i386.c (ix86_split_reduction): Implement
15615 TARGET_VECTORIZE_SPLIT_REDUCTION.
15617 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15620 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15621 in PIC mode except for TARGET_VXWORKS_RTP.
15622 * config/sparc/sparc.c: Include cfgrtl.h.
15623 (TARGET_INIT_PIC_REG): Define.
15624 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15625 (sparc_pic_register_p): New predicate.
15626 (sparc_legitimate_address_p): Use it.
15627 (sparc_legitimize_pic_address): Likewise.
15628 (sparc_delegitimize_address): Likewise.
15629 (sparc_mode_dependent_address_p): Likewise.
15630 (gen_load_pcrel_sym): Remove 4th parameter.
15631 (load_got_register): Adjust call to above. Remove obsolete stuff.
15632 (sparc_expand_prologue): Do not call load_got_register here.
15633 (sparc_flat_expand_prologue): Likewise.
15634 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15635 (sparc_use_pseudo_pic_reg): New function.
15636 (sparc_init_pic_reg): Likewise.
15637 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15638 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15640 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15642 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15643 Add item for branch_cost.
15645 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15647 PR rtl-optimization/83565
15648 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15649 not extend the result to a larger mode for rotate operations.
15650 (num_sign_bit_copies1): Likewise.
15652 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15655 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15657 Use values-Xc.o for -pedantic.
15658 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15660 2018-01-12 Martin Liska <mliska@suse.cz>
15663 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15665 (possible_polymorphic_call_targets): Use it.
15666 (ipa_devirt): Likewise.
15668 2018-01-12 Martin Liska <mliska@suse.cz>
15670 * profile-count.h (enum profile_quality): Use 0 as invalid
15671 enum value of profile_quality.
15673 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15675 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15676 -mext-string options.
15678 2018-01-12 Richard Biener <rguenther@suse.de>
15680 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15681 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15682 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15684 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15686 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15688 * configure.ac (--with-long-double-format): Add support for the
15689 configuration option to change the default long double format on
15691 * config.gcc (powerpc*-linux*-*): Likewise.
15692 * configure: Regenerate.
15693 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15694 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15695 used without modification.
15697 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15699 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15700 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15701 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15702 MISC_BUILTIN_SPEC_BARRIER.
15703 (rs6000_init_builtins): Likewise.
15704 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15706 (speculation_barrier): New define_insn.
15707 * doc/extend.texi: Document __builtin_speculation_barrier.
15709 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15712 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15713 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15714 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15716 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15717 integral modes instead of "ss" and "sd".
15718 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15719 vectors with 32-bit and 64-bit elements.
15720 (vecdupssescalarmodesuffix): New mode attribute.
15721 (vec_dup<mode>): Use it.
15723 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15726 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15727 frame if argument is passed on stack.
15729 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15732 * ree.c (combine_reaching_defs): Optimize also
15733 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15734 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15736 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15738 PR middle-end/83189
15739 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15741 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15743 PR middle-end/83718
15744 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15745 after they are computed.
15747 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15749 PR tree-optimization/83695
15750 * gimple-loop-linterchange.cc
15751 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15752 reset cached scev information after interchange.
15753 (pass_linterchange::execute): Remove call to scev_reset_htab.
15755 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15757 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15758 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15759 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15760 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15761 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15762 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15763 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15764 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15765 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15766 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15767 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15768 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15769 (V_lane_reg): Likewise.
15770 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15772 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15773 (vfmal_lane_low<mode>_intrinsic,
15774 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15775 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15776 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15777 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15778 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15779 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15781 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15783 * config/arm/arm-cpus.in (fp16fml): New feature.
15784 (ALL_SIMD): Add fp16fml.
15785 (armv8.2-a): Add fp16fml as an option.
15786 (armv8.3-a): Likewise.
15787 (armv8.4-a): Add fp16fml as part of fp16.
15788 * config/arm/arm.h (TARGET_FP16FML): Define.
15789 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15791 * config/arm/arm-modes.def (V2HF): Define.
15792 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15793 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15794 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15795 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15796 vfmsl_low, vfmsl_high): New set of builtins.
15797 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15798 (vfml_op): New code attribute.
15799 (VFMLHALVES): New int iterator.
15800 (VFML, VFMLSEL): New mode attributes.
15801 (V_reg): Define mapping for V2HF.
15802 (V_hi, V_lo): New mode attributes.
15803 (VF_constraint): Likewise.
15804 (vfml_half, vfml_half_selector): New int attributes.
15805 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15807 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15808 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15810 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15811 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15812 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15813 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15815 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15816 Document new effective target and option set.
15818 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15820 * config/arm/arm-cpus.in (armv8_4): New feature.
15821 (ARMv8_4a): New fgroup.
15822 (armv8.4-a): New arch.
15823 * config/arm/arm-tables.opt: Regenerate.
15824 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15825 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15826 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15827 Add matching rules for -march=armv8.4-a and extensions.
15828 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15830 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15833 * config/rx/rx.md (BW): New mode attribute.
15834 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15836 2018-01-11 Richard Biener <rguenther@suse.de>
15838 PR tree-optimization/83435
15839 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15840 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15841 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15844 Alan Hayward <alan.hayward@arm.com>
15845 David Sherwood <david.sherwood@arm.com>
15847 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15849 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15850 (aarch64_print_address_internal): Use it to check for a zero offset.
15852 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15853 Alan Hayward <alan.hayward@arm.com>
15854 David Sherwood <david.sherwood@arm.com>
15856 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15857 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15858 Return a poly_int64 rather than a HOST_WIDE_INT.
15859 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15860 rather than a HOST_WIDE_INT.
15861 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15862 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15863 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15864 final_offset from HOST_WIDE_INT to poly_int64.
15865 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15866 to_constant when getting the number of units in an Advanced SIMD
15868 (aarch64_builtin_vectorized_function): Check for a constant number
15870 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15872 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15873 attribute instead of GET_MODE_NUNITS.
15874 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15875 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15876 GET_MODE_SIZE for fixed-size registers.
15877 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15878 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15879 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15880 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15881 (aarch64_print_operand, aarch64_print_address_internal)
15882 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15883 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15884 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15885 Handle polynomial GET_MODE_SIZE.
15886 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15887 wider than SImode without modification.
15888 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15889 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15890 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15891 passing and returning SVE modes.
15892 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15893 rather than GEN_INT.
15894 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15895 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15896 (aarch64_allocate_and_probe_stack_space): Likewise.
15897 (aarch64_layout_frame): Cope with polynomial offsets.
15898 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15899 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15900 polynomial offsets.
15901 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15902 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15903 poly_int64 rather than a HOST_WIDE_INT.
15904 (aarch64_get_separate_components, aarch64_process_components)
15905 (aarch64_expand_prologue, aarch64_expand_epilogue)
15906 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15907 (aarch64_anchor_offset): New function, split out from...
15908 (aarch64_legitimize_address): ...here.
15909 (aarch64_builtin_vectorization_cost): Handle polynomial
15910 TYPE_VECTOR_SUBPARTS.
15911 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15913 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15914 number of elements from the PARALLEL rather than the mode.
15915 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15916 rather than GET_MODE_BITSIZE.
15917 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15918 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15919 (aarch64_expand_vec_perm_const_1): Handle polynomial
15920 d->perm.length () and d->perm elements.
15921 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15922 Apply to_constant to d->perm elements.
15923 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15924 polynomial CONST_VECTOR_NUNITS.
15925 (aarch64_move_pointer): Take amount as a poly_int64 rather
15927 (aarch64_progress_pointer): Avoid temporary variable.
15928 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15929 the mode attribute instead of GET_MODE.
15931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15932 Alan Hayward <alan.hayward@arm.com>
15933 David Sherwood <david.sherwood@arm.com>
15935 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15936 x exists before using it.
15937 (aarch64_add_constant_internal): Rename to...
15938 (aarch64_add_offset_1): ...this. Replace regnum with separate
15939 src and dest rtxes. Handle the case in which they're different,
15940 including when the offset is zero. Replace scratchreg with an rtx.
15941 Use 2 additions if there is no spare register into which we can
15942 move a 16-bit constant.
15943 (aarch64_add_constant): Delete.
15944 (aarch64_add_offset): Replace reg with separate src and dest
15945 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15946 Use aarch64_add_offset_1.
15947 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15948 an rtx rather than an int. Take the delta as a poly_int64
15949 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15950 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15951 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15952 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15953 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15954 and aarch64_add_sp.
15955 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15956 aarch64_add_constant.
15958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15960 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15961 Use scalar_float_mode.
15963 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15965 * config/aarch64/aarch64-simd.md
15966 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15967 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15968 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15969 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15970 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15971 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15972 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15973 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15974 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15975 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15977 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15980 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15981 targ_options->x_arm_arch_string is non NULL.
15983 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15985 * config/aarch64/aarch64.h
15986 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15988 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15991 * expmed.c (emit_store_flag_force): Swap if const op0
15992 and change VOIDmode to mode of op0.
15994 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15996 PR rtl-optimization/83761
15997 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15998 than bytes to mode_for_size.
16000 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16002 PR middle-end/83189
16003 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16004 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16007 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16009 PR middle-end/83575
16010 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16011 when in layout mode.
16012 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16013 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16016 2018-01-10 Michael Collison <michael.collison@arm.com>
16018 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16019 * config/aarch64/aarch64-option-extension.def: Add
16020 AARCH64_OPT_EXTENSION of 'fp16fml'.
16021 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16022 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16023 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16024 * config/aarch64/constraints.md (Ui7): New constraint.
16025 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16026 (VFMLA_SEL_W): Ditto.
16029 (VFMLA16_LOW): New int iterator.
16030 (VFMLA16_HIGH): Ditto.
16031 (UNSPEC_FMLAL): New unspec.
16032 (UNSPEC_FMLSL): Ditto.
16033 (UNSPEC_FMLAL2): Ditto.
16034 (UNSPEC_FMLSL2): Ditto.
16035 (f16mac): New code attribute.
16036 * config/aarch64/aarch64-simd-builtins.def
16037 (aarch64_fmlal_lowv2sf): Ditto.
16038 (aarch64_fmlsl_lowv2sf): Ditto.
16039 (aarch64_fmlalq_lowv4sf): Ditto.
16040 (aarch64_fmlslq_lowv4sf): Ditto.
16041 (aarch64_fmlal_highv2sf): Ditto.
16042 (aarch64_fmlsl_highv2sf): Ditto.
16043 (aarch64_fmlalq_highv4sf): Ditto.
16044 (aarch64_fmlslq_highv4sf): Ditto.
16045 (aarch64_fmlal_lane_lowv2sf): Ditto.
16046 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16047 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16048 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16049 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16050 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16051 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16052 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16053 (aarch64_fmlal_lane_highv2sf): Ditto.
16054 (aarch64_fmlsl_lane_highv2sf): Ditto.
16055 (aarch64_fmlal_laneq_highv2sf): Ditto.
16056 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16057 (aarch64_fmlalq_lane_highv4sf): Ditto.
16058 (aarch64_fmlsl_lane_highv4sf): Ditto.
16059 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16060 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16061 * config/aarch64/aarch64-simd.md:
16062 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16063 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16064 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16065 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16066 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16067 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16068 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16069 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16070 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16071 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16072 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16073 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16074 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16075 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16076 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16077 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16078 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16079 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16080 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16081 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16082 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16083 (vfmlsl_low_u32): Ditto.
16084 (vfmlalq_low_u32): Ditto.
16085 (vfmlslq_low_u32): Ditto.
16086 (vfmlal_high_u32): Ditto.
16087 (vfmlsl_high_u32): Ditto.
16088 (vfmlalq_high_u32): Ditto.
16089 (vfmlslq_high_u32): Ditto.
16090 (vfmlal_lane_low_u32): Ditto.
16091 (vfmlsl_lane_low_u32): Ditto.
16092 (vfmlal_laneq_low_u32): Ditto.
16093 (vfmlsl_laneq_low_u32): Ditto.
16094 (vfmlalq_lane_low_u32): Ditto.
16095 (vfmlslq_lane_low_u32): Ditto.
16096 (vfmlalq_laneq_low_u32): Ditto.
16097 (vfmlslq_laneq_low_u32): Ditto.
16098 (vfmlal_lane_high_u32): Ditto.
16099 (vfmlsl_lane_high_u32): Ditto.
16100 (vfmlal_laneq_high_u32): Ditto.
16101 (vfmlsl_laneq_high_u32): Ditto.
16102 (vfmlalq_lane_high_u32): Ditto.
16103 (vfmlslq_lane_high_u32): Ditto.
16104 (vfmlalq_laneq_high_u32): Ditto.
16105 (vfmlslq_laneq_high_u32): Ditto.
16106 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16107 (AARCH64_FL_FOR_ARCH8_4): New.
16108 (AARCH64_ISA_F16FML): New ISA flag.
16109 (TARGET_F16FML): New feature flag for fp16fml.
16110 (doc/invoke.texi): Document new fp16fml option.
16112 2018-01-10 Michael Collison <michael.collison@arm.com>
16114 * config/aarch64/aarch64-builtins.c:
16115 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16116 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16117 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16118 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16119 (AARCH64_ISA_SHA3): New ISA flag.
16120 (TARGET_SHA3): New feature flag for sha3.
16121 * config/aarch64/iterators.md (sha512_op): New int attribute.
16122 (CRYPTO_SHA512): New int iterator.
16123 (UNSPEC_SHA512H): New unspec.
16124 (UNSPEC_SHA512H2): Ditto.
16125 (UNSPEC_SHA512SU0): Ditto.
16126 (UNSPEC_SHA512SU1): Ditto.
16127 * config/aarch64/aarch64-simd-builtins.def
16128 (aarch64_crypto_sha512hqv2di): New builtin.
16129 (aarch64_crypto_sha512h2qv2di): Ditto.
16130 (aarch64_crypto_sha512su0qv2di): Ditto.
16131 (aarch64_crypto_sha512su1qv2di): Ditto.
16132 (aarch64_eor3qv8hi): Ditto.
16133 (aarch64_rax1qv2di): Ditto.
16134 (aarch64_xarqv2di): Ditto.
16135 (aarch64_bcaxqv8hi): Ditto.
16136 * config/aarch64/aarch64-simd.md:
16137 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16138 (aarch64_crypto_sha512su0qv2di): Ditto.
16139 (aarch64_crypto_sha512su1qv2di): Ditto.
16140 (aarch64_eor3qv8hi): Ditto.
16141 (aarch64_rax1qv2di): Ditto.
16142 (aarch64_xarqv2di): Ditto.
16143 (aarch64_bcaxqv8hi): Ditto.
16144 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16145 (vsha512h2q_u64): Ditto.
16146 (vsha512su0q_u64): Ditto.
16147 (vsha512su1q_u64): Ditto.
16148 (veor3q_u16): Ditto.
16149 (vrax1q_u64): Ditto.
16150 (vxarq_u64): Ditto.
16151 (vbcaxq_u16): Ditto.
16152 * config/arm/types.md (crypto_sha512): New type attribute.
16153 (crypto_sha3): Ditto.
16154 (doc/invoke.texi): Document new sha3 option.
16156 2018-01-10 Michael Collison <michael.collison@arm.com>
16158 * config/aarch64/aarch64-builtins.c:
16159 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16160 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16161 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16162 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16163 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16164 (AARCH64_ISA_SM4): New ISA flag.
16165 (TARGET_SM4): New feature flag for sm4.
16166 * config/aarch64/aarch64-simd-builtins.def
16167 (aarch64_sm3ss1qv4si): Ditto.
16168 (aarch64_sm3tt1aq4si): Ditto.
16169 (aarch64_sm3tt1bq4si): Ditto.
16170 (aarch64_sm3tt2aq4si): Ditto.
16171 (aarch64_sm3tt2bq4si): Ditto.
16172 (aarch64_sm3partw1qv4si): Ditto.
16173 (aarch64_sm3partw2qv4si): Ditto.
16174 (aarch64_sm4eqv4si): Ditto.
16175 (aarch64_sm4ekeyqv4si): Ditto.
16176 * config/aarch64/aarch64-simd.md:
16177 (aarch64_sm3ss1qv4si): Ditto.
16178 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16179 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16180 (aarch64_sm4eqv4si): Ditto.
16181 (aarch64_sm4ekeyqv4si): Ditto.
16182 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16183 (sm3part_op): Ditto.
16184 (CRYPTO_SM3TT): Ditto.
16185 (CRYPTO_SM3PART): Ditto.
16186 (UNSPEC_SM3SS1): New unspec.
16187 (UNSPEC_SM3TT1A): Ditto.
16188 (UNSPEC_SM3TT1B): Ditto.
16189 (UNSPEC_SM3TT2A): Ditto.
16190 (UNSPEC_SM3TT2B): Ditto.
16191 (UNSPEC_SM3PARTW1): Ditto.
16192 (UNSPEC_SM3PARTW2): Ditto.
16193 (UNSPEC_SM4E): Ditto.
16194 (UNSPEC_SM4EKEY): Ditto.
16195 * config/aarch64/constraints.md (Ui2): New constraint.
16196 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16197 * config/arm/types.md (crypto_sm3): New type attribute.
16198 (crypto_sm4): Ditto.
16199 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16200 (vsm3tt1aq_u32): Ditto.
16201 (vsm3tt1bq_u32): Ditto.
16202 (vsm3tt2aq_u32): Ditto.
16203 (vsm3tt2bq_u32): Ditto.
16204 (vsm3partw1q_u32): Ditto.
16205 (vsm3partw2q_u32): Ditto.
16206 (vsm4eq_u32): Ditto.
16207 (vsm4ekeyq_u32): Ditto.
16208 (doc/invoke.texi): Document new sm4 option.
16210 2018-01-10 Michael Collison <michael.collison@arm.com>
16212 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16213 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16214 (AARCH64_FL_FOR_ARCH8_4): New.
16215 (AARCH64_FL_V8_4): New flag.
16216 (doc/invoke.texi): Document new armv8.4-a option.
16218 2018-01-10 Michael Collison <michael.collison@arm.com>
16220 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16221 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16222 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16223 * config/aarch64/aarch64-option-extension.def: Add
16224 AARCH64_OPT_EXTENSION of 'sha2'.
16225 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16226 (crypto): Disable sha2 and aes if crypto disabled.
16227 (crypto): Enable aes and sha2 if enabled.
16228 (simd): Disable sha2 and aes if simd disabled.
16229 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16231 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16232 (TARGET_SHA2): New feature flag for sha2.
16233 (TARGET_AES): New feature flag for aes.
16234 * config/aarch64/aarch64-simd.md:
16235 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16236 conditional on TARGET_AES.
16237 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16238 (aarch64_crypto_sha1hsi): Make pattern conditional
16240 (aarch64_crypto_sha1hv4si): Ditto.
16241 (aarch64_be_crypto_sha1hv4si): Ditto.
16242 (aarch64_crypto_sha1su1v4si): Ditto.
16243 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16244 (aarch64_crypto_sha1su0v4si): Ditto.
16245 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16246 (aarch64_crypto_sha256su0v4si): Ditto.
16247 (aarch64_crypto_sha256su1v4si): Ditto.
16248 (doc/invoke.texi): Document new aes and sha2 options.
16250 2018-01-10 Martin Sebor <msebor@redhat.com>
16252 PR tree-optimization/83781
16253 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16256 2018-01-11 Martin Sebor <msebor@gmail.com>
16257 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16259 PR tree-optimization/83501
16260 PR tree-optimization/81703
16262 * tree-ssa-strlen.c (get_string_cst): Rename...
16263 (get_string_len): ...to this. Handle global constants.
16264 (handle_char_store): Adjust.
16266 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16267 Jim Wilson <jimw@sifive.com>
16269 * config/riscv/riscv-protos.h (riscv_output_return): New.
16270 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16271 (riscv_attribute_table, riscv_output_return),
16272 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16273 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16274 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16276 (riscv_expand_prologue): Add early return for naked function.
16277 (riscv_expand_epilogue): Likewise.
16278 (riscv_function_ok_for_sibcall): Return false for naked function.
16279 (riscv_set_current_function): New.
16280 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16281 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16282 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16283 * doc/extend.texi (RISC-V Function Attributes): New.
16285 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16287 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16288 check for 128-bit long double before checking TCmode.
16289 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16290 128-bit long doubles before checking TFmode or TCmode.
16291 (FLOAT128_IBM_P): Likewise.
16293 2018-01-10 Martin Sebor <msebor@redhat.com>
16295 PR tree-optimization/83671
16296 * builtins.c (c_strlen): Unconditionally return zero for the empty
16298 Use -Warray-bounds for warnings.
16299 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16300 for non-constant array indices with COMPONENT_REF, arrays of
16301 arrays, and pointers to arrays.
16302 (gimple_fold_builtin_strlen): Determine and set length range for
16303 non-constant character arrays.
16305 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16307 PR middle-end/81897
16308 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16311 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16313 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16315 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16318 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16319 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16320 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16321 indexed_or_indirect_operand predicate.
16322 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16323 (*vsx_le_perm_load_v8hi): Likewise.
16324 (*vsx_le_perm_load_v16qi): Likewise.
16325 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16326 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16327 (*vsx_le_perm_store_v8hi): Likewise.
16328 (*vsx_le_perm_store_v16qi): Likewise.
16329 (eight unnamed splitters): Likewise.
16331 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16333 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16334 * config/rs6000/emmintrin.h: Likewise.
16335 * config/rs6000/mmintrin.h: Likewise.
16336 * config/rs6000/xmmintrin.h: Likewise.
16338 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16341 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16343 * tree.c (tree_nop_conversion): Return true for location wrapper
16345 (maybe_wrap_with_location): New function.
16346 (selftest::check_strip_nops): New function.
16347 (selftest::test_location_wrappers): New function.
16348 (selftest::tree_c_tests): Call it.
16349 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16350 (maybe_wrap_with_location): New decl.
16351 (EXPR_LOCATION_WRAPPER_P): New macro.
16352 (location_wrapper_p): New inline function.
16353 (tree_strip_any_location_wrapper): New inline function.
16355 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16358 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16359 stack_realign_offset for the largest alignment of stack slot
16361 (ix86_find_max_used_stack_alignment): New function.
16362 (ix86_finalize_stack_frame_flags): Use it. Set
16363 max_used_stack_alignment if we don't realign stack.
16364 * config/i386/i386.h (machine_function): Add
16365 max_used_stack_alignment.
16367 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16369 * config/arm/arm.opt (-mbranch-cost): New option.
16370 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16373 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16376 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16377 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16379 2018-01-10 Richard Biener <rguenther@suse.de>
16382 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16383 early out so it also covers the case where we have a non-NULL
16386 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16388 PR tree-optimization/83753
16389 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16390 for non-strided grouped accesses if the number of elements is 1.
16392 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16395 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16396 * i386.h (TARGET_USE_GATHER): Define.
16397 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16399 2018-01-10 Martin Liska <mliska@suse.cz>
16402 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16403 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16405 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16406 CLEANUP_NO_PARTITIONING is not set.
16408 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16410 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16411 for vectors, as a partial revert of r254296.
16412 * rtl.h (const_vec_p): Delete.
16413 (const_vec_duplicate_p): Don't test for vector CONSTs.
16414 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16415 * expmed.c (make_tree): Likewise.
16418 * common.md (E, F): Use CONSTANT_P instead of checking for
16420 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16421 checking for CONST_VECTOR.
16423 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16425 PR middle-end/83575
16426 * predict.c (force_edge_cold): Handle in more sane way edges
16427 with no prediction.
16429 2018-01-09 Carl Love <cel@us.ibm.com>
16431 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16433 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16434 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16435 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16436 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16437 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16438 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16439 * config/rs6000/rs6000-protos.h: Add extern defition for
16440 rs6000_generate_float2_double_code.
16441 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16443 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16444 (float2_v2df): Add define_expand.
16446 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16449 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16450 op_mode in the force_to_mode call.
16452 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16454 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16455 instead of checking each element individually.
16456 (aarch64_evpc_uzp): Likewise.
16457 (aarch64_evpc_zip): Likewise.
16458 (aarch64_evpc_ext): Likewise.
16459 (aarch64_evpc_rev): Likewise.
16460 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16461 instead of checking each element individually. Return true without
16463 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16464 whether all selected elements come from the same input, instead of
16465 checking each element individually. Remove calls to gen_rtx_REG,
16466 start_sequence and end_sequence and instead assert that no rtl is
16469 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16471 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16472 order of HIGH and CONST checks.
16474 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16476 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16477 if the destination isn't an SSA_NAME.
16479 2018-01-09 Richard Biener <rguenther@suse.de>
16481 PR tree-optimization/83668
16482 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16484 (canonicalize_loop_form): ... here, renamed from ...
16485 (canonicalize_loop_closed_ssa_form): ... this and amended to
16486 swap successor edges for loop exit blocks to make us use
16487 the RPO order we need for initial schedule generation.
16489 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16491 PR tree-optimization/64811
16492 * match.pd: When optimizing comparisons with Inf, avoid
16493 introducing or losing exceptions from comparisons with NaN.
16495 2018-01-09 Martin Liska <mliska@suse.cz>
16498 * asan.c (shadow_mem_size): Add gcc_assert.
16500 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16502 Don't save registers in main().
16505 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16506 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16507 * config/avr/avr.c (avr_set_current_function): Don't error if
16508 naked, OS_task or OS_main are specified at the same time.
16509 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16511 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16513 * common/config/avr/avr-common.c (avr_option_optimization_table):
16514 Switch on -mmain-is-OS_task for optimizing compilations.
16516 2018-01-09 Richard Biener <rguenther@suse.de>
16518 PR tree-optimization/83572
16519 * graphite.c: Include cfganal.h.
16520 (graphite_transform_loops): Connect infinite loops to exit
16521 and remove fake edges at the end.
16523 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16525 * ipa-inline.c (edge_badness): Revert accidental checkin.
16527 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16530 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16531 symbols; not inline clones.
16533 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16536 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16537 hard registers. Formatting fixes.
16539 PR preprocessor/83722
16540 * gcc.c (try_generate_repro): Pass
16541 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16542 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16545 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16546 Kito Cheng <kito.cheng@gmail.com>
16548 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16549 (riscv_leaf_function_p): Delete.
16550 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16552 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16554 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16556 (do_ifelse): New function.
16557 (do_isel): New function.
16558 (do_sub3): New function.
16559 (do_add3): New function.
16560 (do_load_mask_compare): New function.
16561 (do_overlap_load_compare): New function.
16562 (expand_compare_loop): New function.
16563 (expand_block_compare): Call expand_compare_loop() when appropriate.
16564 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16565 option description.
16566 (-mblock-compare-inline-loop-limit): New option.
16568 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16571 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16572 Reverse order of second and third operands in first alternative.
16573 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16574 of first and second elements in UNSPEC_VPERMR vector.
16575 (altivec_expand_vec_perm_le): Likewise.
16577 2018-01-08 Jeff Law <law@redhat.com>
16579 PR rtl-optimizatin/81308
16580 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16581 (process_switch): If group_case_labels makes a change, then set
16583 (pass_convert_switch::execute): If a switch is converted, then
16584 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16586 PR rtl-optimization/81308
16587 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16590 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16592 PR target/83663 - Revert r255946
16593 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16594 generation for cases where splatting a value is not useful.
16595 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16596 across a vec_duplicate and a paradoxical subreg forming a vector
16597 mode to a vec_concat.
16599 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16601 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16602 -march=armv8.3-a variants.
16603 * config/arm/t-multilib: Likewise.
16604 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16606 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16608 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16610 (cceq_ior_compare_complement): Give it a name so I can use it, and
16611 change boolean_or_operator predicate to boolean_operator so it can
16612 be used to generate a crand.
16613 (eqne): New code iterator.
16614 (bd/bd_neg): New code_attrs.
16615 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16616 a single define_insn.
16617 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16618 decrement (bdnzt/bdnzf/bdzt/bdzf).
16619 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16620 with the new names of the branch decrement patterns, and added the
16621 names of the branch decrement conditional patterns.
16623 2018-01-08 Richard Biener <rguenther@suse.de>
16625 PR tree-optimization/83563
16626 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16629 2018-01-08 Richard Biener <rguenther@suse.de>
16631 PR middle-end/83713
16632 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16634 2018-01-08 Richard Biener <rguenther@suse.de>
16636 PR tree-optimization/83685
16637 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16638 references to abnormals.
16640 2018-01-08 Richard Biener <rguenther@suse.de>
16643 * dwarf2out.c (output_indirect_strings): Handle empty
16644 skeleton_debug_str_hash.
16645 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16647 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16649 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16650 (emit_store_direct): Likewise.
16651 (arc_trampoline_adjust_address): Likewise.
16652 (arc_asm_trampoline_template): New function.
16653 (arc_initialize_trampoline): Use asm_trampoline_template.
16654 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16655 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16656 * config/arc/arc.md (flush_icache): Delete pattern.
16658 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16660 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16661 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16664 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16667 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16668 by not USED_FOR_TARGET.
16669 (make_pass_resolve_sw_modes): Likewise.
16671 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16673 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16676 2018-01-08 Richard Biener <rguenther@suse.de>
16678 PR middle-end/83580
16679 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16681 2018-01-08 Richard Biener <rguenther@suse.de>
16683 PR middle-end/83517
16684 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16686 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16688 PR middle-end/81897
16689 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16690 basic blocks with a small number of successors.
16691 (convert_control_dep_chain_into_preds): Improve handling of
16693 (dump_predicates): Split apart into...
16694 (dump_pred_chain): ...here...
16695 (dump_pred_info): ...and here.
16696 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16697 (can_chain_union_be_invalidated_p): Improve check for invalidation
16699 (uninit_uses_cannot_happen): Avoid unnecessary if
16700 convert_control_dep_chain_into_preds yielded nothing.
16702 2018-01-06 Martin Sebor <msebor@redhat.com>
16704 PR tree-optimization/83640
16705 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16706 subtracting negative offset from size.
16707 (builtin_access::overlap): Adjust offset bounds of the access to fall
16708 within the size of the object if possible.
16710 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16712 PR rtl-optimization/83699
16713 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16714 extract_bit_field_as_subreg to cases in which the extracted
16715 value is also a vector.
16717 * lra-constraints.c (process_alt_operands): Test for the equivalence
16718 substitutions when detecting a possible reload cycle.
16720 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16723 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16724 by default if flag_selective_schedling{,2}. Formatting fixes.
16726 PR rtl-optimization/83682
16727 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16728 if it has non-VECTOR_MODE element mode.
16729 (vec_duplicate_p): Likewise.
16731 PR middle-end/83694
16732 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16733 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16735 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16738 * config/i386/i386-builtin.def
16739 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16740 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16741 Require also OPTION_MASK_ISA_AVX512F in addition to
16742 OPTION_MASK_ISA_GFNI.
16743 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16744 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16745 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16746 to OPTION_MASK_ISA_GFNI.
16747 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16748 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16749 OPTION_MASK_ISA_AVX512BW.
16750 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16751 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16752 addition to OPTION_MASK_ISA_GFNI.
16753 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16754 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16755 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16756 to OPTION_MASK_ISA_GFNI.
16757 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16758 a requirement for all ISAs rather than any of them with a few
16760 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16762 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16763 bitmasks to be enabled with 3 exceptions, instead of requiring any
16764 enabled ISA with lots of exceptions.
16765 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16766 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16767 Change avx512bw in isa attribute to avx512f.
16768 * config/i386/sgxintrin.h: Add license boilerplate.
16769 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16770 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16771 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16772 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16774 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16775 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16776 temporarily sse2 rather than sse if not enabled already.
16779 * config/i386/sse.md (VI248_VLBW): Rename to ...
16780 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16781 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16782 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16783 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16784 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16785 mode iterator instead of VI248_VLBW.
16787 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16789 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16790 (record_modified): Skip clobbers; add debug output.
16791 (param_change_prob): Use sreal frequencies.
16793 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16795 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16796 punt for user-aligned variables.
16798 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16800 * tree-chrec.c (chrec_contains_symbols): Return true for
16803 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16806 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16807 of (x|y) == x for BICS pattern.
16809 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16811 PR tree-optimization/83605
16812 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16813 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16816 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16818 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16819 * config/epiphany/rtems.h: New file.
16821 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16822 Uros Bizjak <ubizjak@gmail.com>
16825 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16826 QIreg_operand instead of register_operand predicate.
16827 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16828 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16829 comments instead of -fmitigate[-_]rop.
16831 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16834 * cgraphunit.c (symbol_table::compile): Switch to text_section
16835 before calling assembly_start debug hook.
16836 * run-rtl-passes.c (run_rtl_passes): Likewise.
16839 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16841 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16842 range_int_cst_p rather than !symbolic_range_p before calling
16843 extract_range_from_multiplicative_op_1.
16845 2018-01-04 Jeff Law <law@redhat.com>
16847 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16848 redundant test in assertion.
16850 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16852 * doc/rtl.texi: Document machine_mode wrapper classes.
16854 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16856 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16857 using tree_to_uhwi.
16859 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16861 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16862 the VEC_PERM_EXPR fold to fail.
16864 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16867 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16868 to switched_sections.
16870 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16873 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16874 test for d.testing.
16876 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16879 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16880 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16882 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16885 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16886 is BLKmode and bitpos not zero or mode change is needed.
16888 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16891 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16894 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16897 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16898 instead of MULT rtx. Update all corresponding splitters.
16899 (*saddl_se): Ditto.
16900 (*ssub<modesuffix>): Ditto.
16901 (*ssubl_se): Ditto.
16902 (*cmp_sadd_di): Update split patterns.
16903 (*cmp_sadd_si): Ditto.
16904 (*cmp_sadd_sidi): Ditto.
16905 (*cmp_ssub_di): Ditto.
16906 (*cmp_ssub_si): Ditto.
16907 (*cmp_ssub_sidi): Ditto.
16908 * config/alpha/predicates.md (const23_operand): New predicate.
16909 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16910 Look for ASHIFT, not MULT inner operand.
16911 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16913 2018-01-04 Martin Liska <mliska@suse.cz>
16915 PR gcov-profile/83669
16916 * gcov.c (output_intermediate_file): Add version to intermediate
16918 * doc/gcov.texi: Document new field 'version' in intermediate
16919 file format. Fix location of '-k' option of gcov command.
16921 2018-01-04 Martin Liska <mliska@suse.cz>
16924 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16926 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16928 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16930 2018-01-03 Martin Sebor <msebor@redhat.com>
16932 PR tree-optimization/83655
16933 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16934 checking calls with invalid arguments.
16936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16938 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16939 (vectorizable_mask_load_store): Delete.
16940 (vectorizable_call): Return false for masked loads and stores.
16941 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16942 instead of gimple_assign_rhs1.
16943 (vectorizable_load): Handle IFN_MASK_LOAD.
16944 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16948 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16950 (vectorizable_mask_load_store): ...here.
16951 (vectorizable_load): ...and here.
16953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16955 * tree-vect-stmts.c (vect_build_all_ones_mask)
16956 (vect_build_zero_merge_argument): New functions, split out from...
16957 (vectorizable_load): ...here.
16959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16961 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16963 (vectorizable_mask_load_store): ...here.
16964 (vectorizable_store): ...and here.
16966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16968 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16970 (vectorizable_mask_load_store): ...here.
16972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16974 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16975 (vect_model_store_cost): Take a vec_load_store_type instead of a
16977 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16978 (vect_model_store_cost): Take a vec_load_store_type instead of a
16980 (vectorizable_mask_load_store): Update accordingly.
16981 (vectorizable_store): Likewise.
16982 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16986 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16987 IFN_MASK_LOAD calls here rather than...
16988 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16991 Alan Hayward <alan.hayward@arm.com>
16992 David Sherwood <david.sherwood@arm.com>
16994 * expmed.c (extract_bit_field_1): For vector extracts,
16995 fall back to extract_bit_field_as_subreg if vec_extract
16998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16999 Alan Hayward <alan.hayward@arm.com>
17000 David Sherwood <david.sherwood@arm.com>
17002 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17003 they are variable or constant sized.
17004 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17005 slots for constant-sized data.
17007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17008 Alan Hayward <alan.hayward@arm.com>
17009 David Sherwood <david.sherwood@arm.com>
17011 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17012 handling COND_EXPRs with boolean comparisons, try to find a better
17013 basis for the mask type than the boolean itself.
17015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17017 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17018 is calculated and how it can be overridden.
17019 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17020 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17022 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17026 Alan Hayward <alan.hayward@arm.com>
17027 David Sherwood <david.sherwood@arm.com>
17029 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17030 Remove the mode argument.
17031 (aarch64_simd_valid_immediate): Remove the mode and inverse
17033 * config/aarch64/iterators.md (bitsize): New iterator.
17034 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17035 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17036 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17037 aarch64_simd_valid_immediate.
17038 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17039 (aarch64_reg_or_bic_imm): Likewise.
17040 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17041 with an insn_type enum and msl with a modifier_type enum.
17042 Replace element_width with a scalar_mode. Change the shift
17043 to unsigned int. Add constructors for scalar_float_mode and
17044 scalar_int_mode elements.
17045 (aarch64_vect_float_const_representable_p): Delete.
17046 (aarch64_can_const_movi_rtx_p)
17047 (aarch64_simd_scalar_immediate_valid_for_move)
17048 (aarch64_simd_make_constant): Update call to
17049 aarch64_simd_valid_immediate.
17050 (aarch64_advsimd_valid_immediate_hs): New function.
17051 (aarch64_advsimd_valid_immediate): Likewise.
17052 (aarch64_simd_valid_immediate): Remove mode and inverse
17053 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17054 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17055 and aarch64_float_const_representable_p on the result.
17056 (aarch64_output_simd_mov_immediate): Remove mode argument.
17057 Update call to aarch64_simd_valid_immediate and use of
17058 simd_immediate_info.
17059 (aarch64_output_scalar_simd_mov_immediate): Update call
17062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17063 Alan Hayward <alan.hayward@arm.com>
17064 David Sherwood <david.sherwood@arm.com>
17066 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17067 (mode_nunits): Likewise CONST_MODE_NUNITS.
17068 * machmode.def (ADJUST_NUNITS): Document.
17069 * genmodes.c (mode_data::need_nunits_adj): New field.
17070 (blank_mode): Update accordingly.
17071 (adj_nunits): New variable.
17072 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17074 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17075 listed in adj_nunits.
17076 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17077 listed in adj_nunits. Don't emit case statements for such modes.
17078 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17079 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17080 nothing if adj_nunits is nonnull.
17081 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17082 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17083 (emit_mode_fbit): Update use of print_maybe_const_decl.
17084 (emit_move_size): Likewise. Treat the array as non-const
17086 (emit_mode_adjustments): Handle adj_nunits.
17088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17090 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17091 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17092 (VECTOR_MODES): Use it.
17093 (make_vector_modes): Take the prefix as an argument.
17095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17096 Alan Hayward <alan.hayward@arm.com>
17097 David Sherwood <david.sherwood@arm.com>
17099 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17100 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17101 for MODE_VECTOR_BOOL.
17102 * machmode.def (VECTOR_BOOL_MODE): Document.
17103 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17104 (make_vector_bool_mode): New function.
17105 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17107 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17108 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17110 * stor-layout.c (int_mode_for_mode): Likewise.
17111 * tree.c (build_vector_type_for_mode): Likewise.
17112 * varasm.c (output_constant_pool_2): Likewise.
17113 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17114 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17115 for MODE_VECTOR_BOOL.
17116 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17117 of mode class checks.
17118 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17119 instead of a list of mode class checks.
17120 (expand_vector_scalar_condition): Likewise.
17121 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17124 Alan Hayward <alan.hayward@arm.com>
17125 David Sherwood <david.sherwood@arm.com>
17127 * machmode.h (mode_size): Change from unsigned short to
17129 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17130 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17131 or if measurement_type is not polynomial.
17132 (fixed_size_mode::includes_p): Check for constant-sized modes.
17133 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17134 return a poly_uint16 rather than an unsigned short.
17135 (emit_mode_size): Change the type of mode_size from unsigned short
17136 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17137 (emit_mode_adjustments): Cope with polynomial vector sizes.
17138 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17140 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17142 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17143 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17144 * caller-save.c (setup_save_areas): Likewise.
17145 (replace_reg_with_saved_mem): Likewise.
17146 * calls.c (emit_library_call_value_1): Likewise.
17147 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17148 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17149 (gen_lowpart_for_combine): Likewise.
17150 * convert.c (convert_to_integer_1): Likewise.
17151 * cse.c (equiv_constant, cse_insn): Likewise.
17152 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17153 (cselib_subst_to_values): Likewise.
17154 * dce.c (word_dce_process_block): Likewise.
17155 * df-problems.c (df_word_lr_mark_ref): Likewise.
17156 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17157 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17158 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17159 (rtl_for_decl_location): Likewise.
17160 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17161 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17162 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17163 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17164 (expand_expr_real_1): Likewise.
17165 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17166 (pad_below): Likewise.
17167 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17168 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17169 * ira.c (get_subreg_tracking_sizes): Likewise.
17170 * ira-build.c (ira_create_allocno_objects): Likewise.
17171 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17172 (ira_sort_regnos_for_alter_reg): Likewise.
17173 * ira-costs.c (record_operand_costs): Likewise.
17174 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17175 (resolve_simple_move): Likewise.
17176 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17177 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17178 (lra_constraints): Likewise.
17179 (CONST_POOL_OK_P): Reject variable-sized modes.
17180 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17181 (add_pseudo_to_slot, lra_spill): Likewise.
17182 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17183 * optabs-query.c (get_best_extraction_insn): Likewise.
17184 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17185 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17186 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17187 * recog.c (offsettable_address_addr_space_p): Likewise.
17188 * regcprop.c (maybe_mode_change): Likewise.
17189 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17190 * regrename.c (build_def_use): Likewise.
17191 * regstat.c (dump_reg_info): Likewise.
17192 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17193 (find_reloads, find_reloads_subreg_address): Likewise.
17194 * reload1.c (eliminate_regs_1): Likewise.
17195 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17196 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17197 (simplify_binary_operation_1, simplify_subreg): Likewise.
17198 * targhooks.c (default_function_arg_padding): Likewise.
17199 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17200 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17201 (verify_gimple_assign_ternary): Likewise.
17202 * tree-inline.c (estimate_move_cost): Likewise.
17203 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17204 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17205 (get_address_cost_ainc): Likewise.
17206 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17207 (vect_supportable_dr_alignment): Likewise.
17208 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17209 (vectorizable_reduction): Likewise.
17210 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17211 (vectorizable_operation, vectorizable_load): Likewise.
17212 * tree.c (build_same_sized_truth_vector_type): Likewise.
17213 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17214 * var-tracking.c (emit_note_insn_var_location): Likewise.
17215 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17216 (ADDR_VEC_ALIGN): Likewise.
17218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17219 Alan Hayward <alan.hayward@arm.com>
17220 David Sherwood <david.sherwood@arm.com>
17222 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17224 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17225 or if measurement_type is polynomial.
17226 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17227 * combine.c (make_extraction): Likewise.
17228 * dse.c (find_shift_sequence): Likewise.
17229 * dwarf2out.c (mem_loc_descriptor): Likewise.
17230 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17231 (extract_bit_field, extract_low_bits): Likewise.
17232 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17233 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17234 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17235 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17236 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17237 * reload.c (find_reloads): Likewise.
17238 * reload1.c (alter_reg): Likewise.
17239 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17240 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17241 * tree-if-conv.c (predicate_mem_writes): Likewise.
17242 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17243 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17244 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17245 * valtrack.c (dead_debug_insert_temp): Likewise.
17246 * varasm.c (mergeable_constant_section): Likewise.
17247 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17250 Alan Hayward <alan.hayward@arm.com>
17251 David Sherwood <david.sherwood@arm.com>
17253 * expr.c (expand_assignment): Cope with polynomial mode sizes
17254 when assigning to a CONCAT.
17256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17257 Alan Hayward <alan.hayward@arm.com>
17258 David Sherwood <david.sherwood@arm.com>
17260 * machmode.h (mode_precision): Change from unsigned short to
17262 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17264 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17265 or if measurement_type is not polynomial.
17266 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17267 in which the mode is already known to be a scalar_int_mode.
17268 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17269 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17271 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17272 for GET_MODE_PRECISION.
17273 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17274 for GET_MODE_PRECISION.
17275 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17277 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17278 (expand_field_assignment, make_extraction): Likewise.
17279 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17280 (get_last_value): Likewise.
17281 * convert.c (convert_to_integer_1): Likewise.
17282 * cse.c (cse_insn): Likewise.
17283 * expr.c (expand_expr_real_1): Likewise.
17284 * lra-constraints.c (simplify_operand_subreg): Likewise.
17285 * optabs-query.c (can_atomic_load_p): Likewise.
17286 * optabs.c (expand_atomic_load): Likewise.
17287 (expand_atomic_store): Likewise.
17288 * ree.c (combine_reaching_defs): Likewise.
17289 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17290 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17291 * tree.h (type_has_mode_precision_p): Likewise.
17292 * ubsan.c (instrument_si_overflow): Likewise.
17294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17295 Alan Hayward <alan.hayward@arm.com>
17296 David Sherwood <david.sherwood@arm.com>
17298 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17299 polynomial numbers of units.
17300 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17301 (valid_vector_subparts_p): New function.
17302 (build_vector_type): Remove temporary shim and take the number
17303 of units as a poly_uint64 rather than an int.
17304 (build_opaque_vector_type): Take the number of units as a
17305 poly_uint64 rather than an int.
17306 * tree.c (build_vector_from_ctor): Handle polynomial
17307 TYPE_VECTOR_SUBPARTS.
17308 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17309 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17310 (build_vector_from_val): If the number of units is variable,
17311 use build_vec_duplicate_cst for constant operands and
17312 VEC_DUPLICATE_EXPR otherwise.
17313 (make_vector_type): Remove temporary is_constant ().
17314 (build_vector_type, build_opaque_vector_type): Take the number of
17315 units as a poly_uint64 rather than an int.
17316 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17318 * cfgexpand.c (expand_debug_expr): Likewise.
17319 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17320 (store_constructor, expand_expr_real_1): Likewise.
17321 (const_scalar_mask_from_tree): Likewise.
17322 * fold-const-call.c (fold_const_reduction): Likewise.
17323 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17324 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17325 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17326 (fold_relational_const): Likewise.
17327 (native_interpret_vector): Likewise. Change the size from an
17328 int to an unsigned int.
17329 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17330 TYPE_VECTOR_SUBPARTS.
17331 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17332 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17333 duplicating a non-constant operand into a variable-length vector.
17334 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17335 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17336 * ipa-icf.c (sem_variable::equals): Likewise.
17337 * match.pd: Likewise.
17338 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17339 * print-tree.c (print_node): Likewise.
17340 * stor-layout.c (layout_type): Likewise.
17341 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17342 * tree-cfg.c (verify_gimple_comparison): Likewise.
17343 (verify_gimple_assign_binary): Likewise.
17344 (verify_gimple_assign_ternary): Likewise.
17345 (verify_gimple_assign_single): Likewise.
17346 * tree-pretty-print.c (dump_generic_node): Likewise.
17347 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17348 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17349 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17350 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17351 (vect_shift_permute_load_chain): Likewise.
17352 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17353 (expand_vector_condition, optimize_vector_constructor): Likewise.
17354 (lower_vec_perm, get_compute_type): Likewise.
17355 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17356 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17357 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17358 (vect_recog_mask_conversion_pattern): Likewise.
17359 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17360 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17361 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17362 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17363 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17364 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17365 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17366 (supportable_widening_operation): Likewise.
17367 (supportable_narrowing_operation): Likewise.
17368 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17370 * varasm.c (output_constant): Likewise.
17372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17373 Alan Hayward <alan.hayward@arm.com>
17374 David Sherwood <david.sherwood@arm.com>
17376 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17377 so that both the length == 3 and length != 3 cases set up their
17378 own permute vectors. Add comments explaining why we know the
17379 number of elements is constant.
17380 (vect_permute_load_chain): Likewise.
17382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17383 Alan Hayward <alan.hayward@arm.com>
17384 David Sherwood <david.sherwood@arm.com>
17386 * machmode.h (mode_nunits): Change from unsigned char to
17388 (ONLY_FIXED_SIZE_MODES): New macro.
17389 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17390 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17391 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17393 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17394 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17395 or if measurement_type is not polynomial.
17396 * genmodes.c (ZERO_COEFFS): New macro.
17397 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17399 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17400 Use ZERO_COEFFS when emitting initializers.
17401 * data-streamer.h (bp_pack_poly_value): New function.
17402 (bp_unpack_poly_value): Likewise.
17403 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17404 for GET_MODE_NUNITS.
17405 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17406 for GET_MODE_NUNITS.
17407 * tree.c (make_vector_type): Remove temporary shim and make
17408 the real function take the number of units as a poly_uint64
17409 rather than an int.
17410 (build_vector_type_for_mode): Handle polynomial nunits.
17411 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17412 * emit-rtl.c (const_vec_series_p_1): Likewise.
17413 (gen_rtx_CONST_VECTOR): Likewise.
17414 * fold-const.c (test_vec_duplicate_folding): Likewise.
17415 * genrecog.c (validate_pattern): Likewise.
17416 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17417 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17418 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17419 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17420 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17421 * rtlanal.c (subreg_get_info): Likewise.
17422 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17423 (vect_grouped_load_supported): Likewise.
17424 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17425 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17426 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17427 (simplify_const_unary_operation, simplify_binary_operation_1)
17428 (simplify_const_binary_operation, simplify_ternary_operation)
17429 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17430 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17431 instead of CONST_VECTOR_NUNITS.
17432 * varasm.c (output_constant_pool_2): Likewise.
17433 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17434 explicit-encoded elements in the XVEC for variable-length vectors.
17436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17438 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17441 Alan Hayward <alan.hayward@arm.com>
17442 David Sherwood <david.sherwood@arm.com>
17444 * coretypes.h (fixed_size_mode): Declare.
17445 (fixed_size_mode_pod): New typedef.
17446 * builtins.h (target_builtins::x_apply_args_mode)
17447 (target_builtins::x_apply_result_mode): Change type to
17448 fixed_size_mode_pod.
17449 * builtins.c (apply_args_size, apply_result_size, result_vector)
17450 (expand_builtin_apply_args_1, expand_builtin_apply)
17451 (expand_builtin_return): Update accordingly.
17453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17455 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17456 * cselib.c (cselib_hash_rtx): Likewise.
17457 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17458 CONST_VECTOR encoding.
17460 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17461 Jeff Law <law@redhat.com>
17464 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17465 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17466 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17467 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17470 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17471 explicitly probe *sp in a noreturn function if there were any callee
17472 register saves or frame pointer is needed.
17474 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17477 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17478 BLKmode for ternary, binary or unary expressions.
17481 * var-tracking.c (delete_vta_debug_insn): New inline function.
17482 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17483 insns from get_insns () to NULL instead of each bb separately.
17484 Use delete_vta_debug_insn. No longer static.
17485 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17486 delete_vta_debug_insns callers.
17487 * rtl.h (delete_vta_debug_insns): Declare.
17488 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17489 instead of variable_tracking_main.
17491 2018-01-03 Martin Sebor <msebor@redhat.com>
17493 PR tree-optimization/83603
17494 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17495 arguments past the endof the argument list in functions declared
17496 without a prototype.
17497 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17498 Avoid checking when arguments are null.
17500 2018-01-03 Martin Sebor <msebor@redhat.com>
17503 * doc/extend.texi (attribute const): Fix a typo.
17504 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17505 issuing -Wsuggest-attribute for void functions.
17507 2018-01-03 Martin Sebor <msebor@redhat.com>
17509 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17510 offset_int::from instead of wide_int::to_shwi.
17511 (maybe_diag_overlap): Remove assertion.
17512 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17513 * gimple-ssa-sprintf.c (format_directive): Same.
17514 (parse_directive): Same.
17515 (sprintf_dom_walker::compute_format_length): Same.
17516 (try_substitute_return_value): Same.
17518 2018-01-03 Jeff Law <law@redhat.com>
17520 PR middle-end/83654
17521 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17522 non-constant residual for zero at runtime and avoid probing in
17523 that case. Reorganize code for trailing problem to mirror handling
17526 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17528 PR tree-optimization/83501
17529 * tree-ssa-strlen.c (get_string_cst): New.
17530 (handle_char_store): Call get_string_cst.
17532 2018-01-03 Martin Liska <mliska@suse.cz>
17534 PR tree-optimization/83593
17535 * tree-ssa-strlen.c: Include tree-cfg.h.
17536 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17537 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17538 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17540 (strlen_dom_walker::before_dom_children): Call
17541 gimple_purge_dead_eh_edges. Dump tranformation with details
17543 (strlen_dom_walker::before_dom_children): Update call by adding
17544 new argument cleanup_eh.
17545 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17547 2018-01-03 Martin Liska <mliska@suse.cz>
17550 * cif-code.def (VARIADIC_THUNK): New enum value.
17551 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17554 2018-01-03 Jan Beulich <jbeulich@suse.com>
17556 * sse.md (mov<mode>_internal): Tighten condition for when to use
17557 vmovdqu<ssescalarsize> for TI and OI modes.
17559 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17561 Update copyright years.
17563 2018-01-03 Martin Liska <mliska@suse.cz>
17566 * ipa-visibility.c (function_and_variable_visibility): Skip
17567 functions with noipa attribure.
17569 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17571 * gcc.c (process_command): Update copyright notice dates.
17572 * gcov-dump.c (print_version): Ditto.
17573 * gcov.c (print_version): Ditto.
17574 * gcov-tool.c (print_version): Ditto.
17575 * gengtype.c (create_file): Ditto.
17576 * doc/cpp.texi: Bump @copying's copyright year.
17577 * doc/cppinternals.texi: Ditto.
17578 * doc/gcc.texi: Ditto.
17579 * doc/gccint.texi: Ditto.
17580 * doc/gcov.texi: Ditto.
17581 * doc/install.texi: Ditto.
17582 * doc/invoke.texi: Ditto.
17584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17586 * vector-builder.h (vector_builder::m_full_nelts): Change from
17587 unsigned int to poly_uint64.
17588 (vector_builder::full_nelts): Update prototype accordingly.
17589 (vector_builder::new_vector): Likewise.
17590 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17591 (vector_builder::operator ==): Likewise.
17592 (vector_builder::finalize): Likewise.
17593 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17594 Take the number of elements as a poly_uint64 rather than an
17596 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17597 from unsigned int to poly_uint64.
17598 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17599 (vec_perm_indices::new_vector): Likewise.
17600 (vec_perm_indices::length): Likewise.
17601 (vec_perm_indices::nelts_per_input): Likewise.
17602 (vec_perm_indices::input_nelts): Likewise.
17603 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17604 number of elements per input as a poly_uint64 rather than an
17605 unsigned int. Use the original encoding for variable-length
17606 vectors, rather than clamping each individual element.
17607 For the second and subsequent elements in each pattern,
17608 clamp the step and base before clamping their sum.
17609 (vec_perm_indices::series_p): Handle polynomial element counts.
17610 (vec_perm_indices::all_in_range_p): Likewise.
17611 (vec_perm_indices_to_tree): Likewise.
17612 (vec_perm_indices_to_rtx): Likewise.
17613 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17614 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17615 (tree_vector_builder::new_binary_operation): Handle polynomial
17616 element counts. Return false if we need to know the number
17617 of elements at compile time.
17618 * fold-const.c (fold_vec_perm): Punt if the number of elements
17619 isn't known at compile time.
17621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17623 * vec-perm-indices.h (vec_perm_builder): Change element type
17624 from HOST_WIDE_INT to poly_int64.
17625 (vec_perm_indices::element_type): Update accordingly.
17626 (vec_perm_indices::clamp): Handle polynomial element_types.
17627 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17628 (vec_perm_indices::all_in_range_p): Likewise.
17629 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17631 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17632 polynomial vec_perm_indices element types.
17633 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17634 * fold-const.c (fold_vec_perm): Likewise.
17635 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17636 * tree-vect-generic.c (lower_vec_perm): Likewise.
17637 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17638 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17639 element type to HOST_WIDE_INT.
17641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17642 Alan Hayward <alan.hayward@arm.com>
17643 David Sherwood <david.sherwood@arm.com>
17645 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17646 rather than an int. Use plus_constant.
17647 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17648 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17651 Alan Hayward <alan.hayward@arm.com>
17652 David Sherwood <david.sherwood@arm.com>
17654 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17655 a HOST_WIDE_INT to a poly_int64.
17657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17658 Alan Hayward <alan.hayward@arm.com>
17659 David Sherwood <david.sherwood@arm.com>
17661 * calls.c (load_register_parameters): Cope with polynomial
17662 mode sizes. Require a constant size for BLKmode parameters
17663 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17664 forces a parameter to be padded at the lsb end in order to
17665 fill a complete number of words, require the parameter size
17666 to be ordered wrt UNITS_PER_WORD.
17668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17669 Alan Hayward <alan.hayward@arm.com>
17670 David Sherwood <david.sherwood@arm.com>
17672 * reload1.c (spill_stack_slot_width): Change element type
17673 from unsigned int to poly_uint64_pod.
17674 (alter_reg): Treat mode sizes as polynomial.
17676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17677 Alan Hayward <alan.hayward@arm.com>
17678 David Sherwood <david.sherwood@arm.com>
17680 * reload.c (complex_word_subreg_p): New function.
17681 (reload_inner_reg_of_subreg, push_reload): Use it.
17683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17684 Alan Hayward <alan.hayward@arm.com>
17685 David Sherwood <david.sherwood@arm.com>
17687 * lra-constraints.c (process_alt_operands): Reject matched
17688 operands whose sizes aren't ordered.
17689 (match_reload): Refer to this check here.
17691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17692 Alan Hayward <alan.hayward@arm.com>
17693 David Sherwood <david.sherwood@arm.com>
17695 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17696 that the mode size is in the set {1, 2, 4, 8, 16}.
17698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17699 Alan Hayward <alan.hayward@arm.com>
17700 David Sherwood <david.sherwood@arm.com>
17702 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17703 Use plus_constant instead of gen_rtx_PLUS.
17705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17706 Alan Hayward <alan.hayward@arm.com>
17707 David Sherwood <david.sherwood@arm.com>
17709 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17710 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17711 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17712 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17713 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17714 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17715 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17716 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17717 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17718 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17720 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17721 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17722 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17723 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17724 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17725 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17726 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17727 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17728 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17729 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17731 * expr.c (emit_move_resolve_push): Treat the input and result
17732 of PUSH_ROUNDING as a poly_int64.
17733 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17734 (emit_push_insn): Likewise.
17735 * lra-eliminations.c (mark_not_eliminable): Likewise.
17736 * recog.c (push_operand): Likewise.
17737 * reload1.c (elimination_effects): Likewise.
17738 * rtlanal.c (nonzero_bits1): Likewise.
17739 * calls.c (store_one_arg): Likewise. Require the padding to be
17740 known at compile time.
17742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17743 Alan Hayward <alan.hayward@arm.com>
17744 David Sherwood <david.sherwood@arm.com>
17746 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17747 Use plus_constant instead of gen_rtx_PLUS.
17749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17750 Alan Hayward <alan.hayward@arm.com>
17751 David Sherwood <david.sherwood@arm.com>
17753 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17754 rather than an int.
17756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17757 Alan Hayward <alan.hayward@arm.com>
17758 David Sherwood <david.sherwood@arm.com>
17760 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17761 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17762 via stack temporaries. Treat the mode size as polynomial too.
17764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17765 Alan Hayward <alan.hayward@arm.com>
17766 David Sherwood <david.sherwood@arm.com>
17768 * expr.c (expand_expr_real_2): When handling conversions involving
17769 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17770 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17771 as a poly_uint64 too.
17773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17774 Alan Hayward <alan.hayward@arm.com>
17775 David Sherwood <david.sherwood@arm.com>
17777 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17780 Alan Hayward <alan.hayward@arm.com>
17781 David Sherwood <david.sherwood@arm.com>
17783 * combine.c (can_change_dest_mode): Handle polynomial
17784 REGMODE_NATURAL_SIZE.
17785 * expmed.c (store_bit_field_1): Likewise.
17786 * expr.c (store_constructor): Likewise.
17787 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17788 and polynomial REGMODE_NATURAL_SIZE.
17789 (gen_lowpart_common): Likewise.
17790 * reginfo.c (record_subregs_of_mode): Likewise.
17791 * rtlanal.c (read_modify_subreg_p): Likewise.
17793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17794 Alan Hayward <alan.hayward@arm.com>
17795 David Sherwood <david.sherwood@arm.com>
17797 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17798 numbers of elements.
17800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17801 Alan Hayward <alan.hayward@arm.com>
17802 David Sherwood <david.sherwood@arm.com>
17804 * match.pd: Cope with polynomial numbers of vector elements.
17806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17807 Alan Hayward <alan.hayward@arm.com>
17808 David Sherwood <david.sherwood@arm.com>
17810 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17811 in a POINTER_PLUS_EXPR.
17813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17814 Alan Hayward <alan.hayward@arm.com>
17815 David Sherwood <david.sherwood@arm.com>
17817 * omp-simd-clone.c (simd_clone_subparts): New function.
17818 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17819 (ipa_simd_modify_function_body): Likewise.
17821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17822 Alan Hayward <alan.hayward@arm.com>
17823 David Sherwood <david.sherwood@arm.com>
17825 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17826 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17827 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17828 (expand_vector_condition, vector_element): Likewise.
17829 (subparts_gt): New function.
17830 (get_compute_type): Use subparts_gt.
17831 (count_type_subparts): Delete.
17832 (expand_vector_operations_1): Use subparts_gt instead of
17833 count_type_subparts.
17835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17836 Alan Hayward <alan.hayward@arm.com>
17837 David Sherwood <david.sherwood@arm.com>
17839 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17840 (vect_compile_time_alias): ...this new function. Do the calculation
17841 on poly_ints rather than trees.
17842 (vect_prune_runtime_alias_test_list): Update call accordingly.
17844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17845 Alan Hayward <alan.hayward@arm.com>
17846 David Sherwood <david.sherwood@arm.com>
17848 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17850 (vect_schedule_slp_instance): Likewise.
17852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17853 Alan Hayward <alan.hayward@arm.com>
17854 David Sherwood <david.sherwood@arm.com>
17856 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17857 constant and extern definitions for variable-length vectors.
17858 (vect_get_constant_vectors): Note that the number of units
17859 is known to be constant.
17861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17862 Alan Hayward <alan.hayward@arm.com>
17863 David Sherwood <david.sherwood@arm.com>
17865 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17866 of units as polynomial. Choose between WIDE and NARROW based
17869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17870 Alan Hayward <alan.hayward@arm.com>
17871 David Sherwood <david.sherwood@arm.com>
17873 * tree-vect-stmts.c (simd_clone_subparts): New function.
17874 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17877 Alan Hayward <alan.hayward@arm.com>
17878 David Sherwood <david.sherwood@arm.com>
17880 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17881 vectors as polynomial. Use build_index_vector for
17882 IFN_GOMP_SIMD_LANE.
17884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17885 Alan Hayward <alan.hayward@arm.com>
17886 David Sherwood <david.sherwood@arm.com>
17888 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17889 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17890 for variable-length vectors.
17891 (vectorizable_mask_load_store): Treat the number of units as
17892 polynomial, asserting that it is constant if the condition has
17893 already been enforced.
17894 (vectorizable_store, vectorizable_load): Likewise.
17896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17897 Alan Hayward <alan.hayward@arm.com>
17898 David Sherwood <david.sherwood@arm.com>
17900 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17901 of units as polynomial. Punt if we can't tell at compile time
17902 which vector contains the final result.
17904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17905 Alan Hayward <alan.hayward@arm.com>
17906 David Sherwood <david.sherwood@arm.com>
17908 * tree-vect-loop.c (vectorizable_induction): Treat the number
17909 of units as polynomial. Punt on SLP inductions. Use an integer
17910 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17911 cast of such a series for variable-length floating-point
17914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17915 Alan Hayward <alan.hayward@arm.com>
17916 David Sherwood <david.sherwood@arm.com>
17918 * tree.h (build_index_vector): Declare.
17919 * tree.c (build_index_vector): New function.
17920 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17921 of units as polynomial, forcibly converting it to a constant if
17922 vectorizable_reduction has already enforced the condition.
17923 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17924 to create a {1,2,3,...} vector.
17925 (vectorizable_reduction): Treat the number of units as polynomial.
17926 Choose vectype_in based on the largest scalar element size rather
17927 than the smallest number of units. Enforce the restrictions
17930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17931 Alan Hayward <alan.hayward@arm.com>
17932 David Sherwood <david.sherwood@arm.com>
17934 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17935 number of units as polynomial.
17937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17938 Alan Hayward <alan.hayward@arm.com>
17939 David Sherwood <david.sherwood@arm.com>
17941 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17942 * target.def (autovectorize_vector_sizes): Return the vector sizes
17943 by pointer, using vector_sizes rather than a bitmask.
17944 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17945 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17946 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17948 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17949 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17950 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17951 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17952 * omp-general.c (omp_max_vf): Likewise.
17953 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17954 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17955 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17956 * tree-vect-slp.c (vect_slp_bb): Likewise.
17957 * doc/tm.texi: Regenerate.
17958 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17960 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17961 the vector size as a poly_uint64 rather than an unsigned int.
17962 (current_vector_size): Change from an unsigned int to a poly_uint64.
17963 (get_vectype_for_scalar_type): Update accordingly.
17964 * tree.h (build_truth_vector_type): Take the size and number of
17965 units as a poly_uint64 rather than an unsigned int.
17966 (build_vector_type): Add a temporary overload that takes
17967 the number of units as a poly_uint64 rather than an unsigned int.
17968 * tree.c (make_vector_type): Likewise.
17969 (build_truth_vector_type): Take the number of units as a poly_uint64
17970 rather than an unsigned int.
17972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17973 Alan Hayward <alan.hayward@arm.com>
17974 David Sherwood <david.sherwood@arm.com>
17976 * target.def (get_mask_mode): Take the number of units and length
17977 as poly_uint64s rather than unsigned ints.
17978 * targhooks.h (default_get_mask_mode): Update accordingly.
17979 * targhooks.c (default_get_mask_mode): Likewise.
17980 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17981 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17988 * omp-general.c (omp_max_vf): Likewise.
17989 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17990 (expand_omp_simd): Handle polynomial safelen.
17991 * omp-low.c (omplow_simd_context): Add a default constructor.
17992 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17993 (lower_rec_simd_input_clauses): Update accordingly.
17994 (lower_rec_input_clauses): Likewise.
17996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17997 Alan Hayward <alan.hayward@arm.com>
17998 David Sherwood <david.sherwood@arm.com>
18000 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18001 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18002 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18003 (vect_analyze_slp_cost): Likewise.
18004 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18005 (vect_model_load_cost): Likewise.
18007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18008 Alan Hayward <alan.hayward@arm.com>
18009 David Sherwood <david.sherwood@arm.com>
18011 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18012 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18013 from an unsigned int * to a poly_uint64_pod *.
18014 (calculate_unrolling_factor): New function.
18015 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
18022 from an unsigned int to a poly_uint64.
18023 (_loop_vec_info::slp_unrolling_factor): Likewise.
18024 (_loop_vec_info::vectorization_factor): Change from an int
18026 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18027 (vect_get_num_vectors): New function.
18028 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18029 (vect_get_num_copies): Use vect_get_num_vectors.
18030 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18031 to an unsigned int *.
18032 (vect_analyze_data_refs): Change min_vf from an int * to a
18034 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18035 than an unsigned HOST_WIDE_INT.
18036 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18037 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18038 to an unsigned int *.
18039 (vect_analyze_data_ref_dependences): Likewise.
18040 (vect_compute_data_ref_alignment): Handle polynomial vf.
18041 (vect_enhance_data_refs_alignment): Likewise.
18042 (vect_prune_runtime_alias_test_list): Likewise.
18043 (vect_shift_permute_load_chain): Likewise.
18044 (vect_supportable_dr_alignment): Likewise.
18045 (dependence_distance_ge_vf): Take the vectorization factor as a
18046 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18047 (vect_analyze_data_refs): Change min_vf from an int * to a
18049 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18050 vfm1 as a poly_uint64 rather than an int. Make the same change
18051 for the returned bound_scalar.
18052 (vect_gen_vector_loop_niters): Handle polynomial vf.
18053 (vect_do_peeling): Likewise. Update call to
18054 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18055 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18057 * tree-vect-loop.c (vect_determine_vectorization_factor)
18058 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18059 (vect_get_known_peeling_cost): Likewise.
18060 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18061 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18062 (vect_transform_loop): Likewise. Use the lowest possible VF when
18063 updating the upper bounds of the loop.
18064 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18065 rather than an int.
18066 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18067 polynomial unroll factors.
18068 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18069 (vect_make_slp_decision): Likewise.
18070 (vect_supported_load_permutation_p): Likewise, and polynomial
18072 (vect_analyze_slp_cost): Handle polynomial vf.
18073 (vect_slp_analyze_node_operations): Likewise.
18074 (vect_slp_analyze_bb_1): Likewise.
18075 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18076 than an unsigned HOST_WIDE_INT.
18077 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18078 (vectorizable_load): Handle polynomial vf.
18079 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18081 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
18088 and try to fold one pair.
18090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18092 * tree-vect-loop-manip.c: Include gimple-fold.h.
18093 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18094 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18095 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18096 Add a path that uses a step of VF instead of 1, but disable it
18098 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18099 and niters_no_overflow parameters. Update calls to
18100 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18101 Create a new SSA name if the latter choses to use a ste other
18102 than zero, and return it via niters_vector_mult_vf_var.
18103 * tree-vect-loop.c (vect_transform_loop): Update calls to
18104 vect_do_peeling, vect_gen_vector_loop_niters and
18105 slpeel_make_loop_iterate_ntimes.
18106 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18107 (vect_gen_vector_loop_niters): Update declarations after above changes.
18109 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18111 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18112 128-bit round to integer instructions.
18113 (ceil<mode>2): Likewise.
18114 (btrunc<mode>2): Likewise.
18115 (round<mode>2): Likewise.
18117 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18119 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18120 unaligned VSX load/store on P8/P9.
18121 (expand_block_clear): Allow the use of unaligned VSX
18122 load/store on P8/P9.
18124 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18126 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18128 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18129 swap associated with both a load and a store.
18131 2018-01-02 Andrew Waterman <andrew@sifive.com>
18133 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18134 * config/riscv/riscv.md (clear_cache): Use it.
18136 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18138 * web.c: Remove out-of-date comment.
18140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18142 * expr.c (fixup_args_size_notes): Check that any existing
18143 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18144 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18145 (emit_single_push_insn): ...here.
18147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18149 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18150 (const_vector_encoded_nelts): New function.
18151 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18152 (const_vector_int_elt, const_vector_elt): Declare.
18153 * emit-rtl.c (const_vector_int_elt_1): New function.
18154 (const_vector_elt): Likewise.
18155 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18156 of CONST_VECTOR_ELT.
18158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18160 * expr.c: Include rtx-vector-builder.h.
18161 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18162 directly on the tree encoding.
18163 (const_vector_from_tree): Likewise.
18164 * optabs.c: Include rtx-vector-builder.h.
18165 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18166 sequence of "u" values.
18167 * vec-perm-indices.c: Include rtx-vector-builder.h.
18168 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18169 directly on the vec_perm_indices encoding.
18171 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18173 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18174 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18175 * rtx-vector-builder.h: New file.
18176 * rtx-vector-builder.c: Likewise.
18177 * rtl.h (rtx_def::u2): Add a const_vector field.
18178 (CONST_VECTOR_NPATTERNS): New macro.
18179 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18180 (CONST_VECTOR_DUPLICATE_P): Likewise.
18181 (CONST_VECTOR_STEPPED_P): Likewise.
18182 (CONST_VECTOR_ENCODED_ELT): Likewise.
18183 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18184 (unwrap_const_vec_duplicate): Likewise.
18185 (const_vec_series_p): Check for a non-duplicated vector encoding.
18186 Say that the function only returns true for integer vectors.
18187 * emit-rtl.c: Include rtx-vector-builder.h.
18188 (gen_const_vec_duplicate_1): Delete.
18189 (gen_const_vector): Call gen_const_vec_duplicate instead of
18190 gen_const_vec_duplicate_1.
18191 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18192 (gen_const_vec_duplicate): Use rtx_vector_builder.
18193 (gen_const_vec_series): Likewise.
18194 (gen_rtx_CONST_VECTOR): Likewise.
18195 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18196 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18197 Build a new vector rather than modifying a CONST_VECTOR in-place.
18198 (handle_special_swappables): Update call accordingly.
18199 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18200 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18201 Build a new vector rather than modifying a CONST_VECTOR in-place.
18202 (handle_special_swappables): Update call accordingly.
18204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18206 * simplify-rtx.c (simplify_const_binary_operation): Use
18207 CONST_VECTOR_ELT instead of XVECEXP.
18209 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18211 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18212 the selector elements to be different from the data elements
18213 if the selector is a VECTOR_CST.
18214 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18215 ssizetype for the selector.
18217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18219 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18220 before testing each element individually.
18221 * tree-vect-generic.c (lower_vec_perm): Likewise.
18223 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18225 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18226 * selftest-run-tests.c (selftest::run_tests): Call it.
18227 * vector-builder.h (vector_builder::operator ==): New function.
18228 (vector_builder::operator !=): Likewise.
18229 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18230 (vec_perm_indices::all_from_input_p): New function.
18231 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18232 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18233 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18234 instead of reading the VECTOR_CST directly. Detect whether both
18235 vector inputs are the same before constructing the vec_perm_indices,
18236 and update the number of inputs argument accordingly. Use the
18237 utility functions added above. Only construct sel2 if we need to.
18239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18241 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18242 the broadcast of the low byte.
18243 (expand_mult_highpart): Use an explicit encoding for the permutes.
18244 * optabs-query.c (can_mult_highpart_p): Likewise.
18245 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18246 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18247 (vectorizable_bswap): Likewise.
18248 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18249 explicit encoding for the power-of-2 permutes.
18250 (vect_permute_store_chain): Likewise.
18251 (vect_grouped_load_supported): Likewise.
18252 (vect_permute_load_chain): Likewise.
18254 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18256 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18257 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18258 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18259 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18260 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18261 (vect_gen_perm_mask_any): Likewise.
18263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18265 * int-vector-builder.h: New file.
18266 * vec-perm-indices.h: Include int-vector-builder.h.
18267 (vec_perm_indices): Redefine as an int_vector_builder.
18268 (auto_vec_perm_indices): Delete.
18269 (vec_perm_builder): Redefine as a stand-alone class.
18270 (vec_perm_indices::vec_perm_indices): New function.
18271 (vec_perm_indices::clamp): Likewise.
18272 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18273 (vec_perm_indices::new_vector): New function.
18274 (vec_perm_indices::new_expanded_vector): Update for new
18275 vec_perm_indices class.
18276 (vec_perm_indices::rotate_inputs): New function.
18277 (vec_perm_indices::all_in_range_p): Operate directly on the
18278 encoded form, without computing elided elements.
18279 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18280 encoding. Update for new vec_perm_indices class.
18281 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18282 the given vec_perm_builder.
18283 (expand_vec_perm_var): Update vec_perm_builder constructor.
18284 (expand_mult_highpart): Use vec_perm_builder instead of
18285 auto_vec_perm_indices.
18286 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18287 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18288 or double series encoding as appropriate.
18289 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18290 vec_perm_indices instead of auto_vec_perm_indices.
18291 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18292 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18293 (vect_permute_store_chain): Likewise.
18294 (vect_grouped_load_supported): Likewise.
18295 (vect_permute_load_chain): Likewise.
18296 (vect_shift_permute_load_chain): Likewise.
18297 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18298 (vect_transform_slp_perm_load): Likewise.
18299 (vect_schedule_slp_instance): Likewise.
18300 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18301 (vectorizable_mask_load_store): Likewise.
18302 (vectorizable_bswap): Likewise.
18303 (vectorizable_store): Likewise.
18304 (vectorizable_load): Likewise.
18305 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18306 vec_perm_indices instead of auto_vec_perm_indices. Use
18307 tree_to_vec_perm_builder to read the vector from a tree.
18308 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18309 vec_perm_builder instead of a vec_perm_indices.
18310 (have_whole_vector_shift): Use vec_perm_builder and
18311 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18312 truncation to calc_vec_perm_mask_for_shift.
18313 (vect_create_epilog_for_reduction): Likewise.
18314 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18315 from auto_vec_perm_indices to vec_perm_indices.
18316 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18317 instead of changing individual elements.
18318 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18319 the vector in d.perm.
18320 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18321 from auto_vec_perm_indices to vec_perm_indices.
18322 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18323 instead of changing individual elements.
18324 (arm_vectorize_vec_perm_const): Use new_vector to install
18325 the vector in d.perm.
18326 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18327 Update vec_perm_builder constructor.
18328 (rs6000_expand_interleave): Likewise.
18329 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18330 (rs6000_expand_interleave): Likewise.
18332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18334 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18335 to qimode could truncate the indices.
18336 * optabs.c (expand_vec_perm_var): Likewise.
18338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18340 * Makefile.in (OBJS): Add vec-perm-indices.o.
18341 * vec-perm-indices.h: New file.
18342 * vec-perm-indices.c: Likewise.
18343 * target.h (vec_perm_indices): Replace with a forward class
18345 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18346 * optabs.h: Include vec-perm-indices.h.
18347 (expand_vec_perm): Delete.
18348 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18349 (expand_vec_perm_const): Declare.
18350 * target.def (vec_perm_const_ok): Replace with...
18351 (vec_perm_const): ...this new hook.
18352 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18353 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18354 * doc/tm.texi: Regenerate.
18355 * optabs.def (vec_perm_const): Delete.
18356 * doc/md.texi (vec_perm_const): Likewise.
18357 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18358 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18359 expand_vec_perm for constant permutation vectors. Assert that
18360 the mode of variable permutation vectors is the integer equivalent
18361 of the mode that is being permuted.
18362 * optabs-query.h (selector_fits_mode_p): Declare.
18363 * optabs-query.c: Include vec-perm-indices.h.
18364 (selector_fits_mode_p): New function.
18365 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18366 is defined, instead of checking whether the vec_perm_const_optab
18367 exists. Use targetm.vectorize.vec_perm_const instead of
18368 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18369 fit in the vector mode before using a variable permute.
18370 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18371 vec_perm_indices instead of an rtx.
18372 (expand_vec_perm): Replace with...
18373 (expand_vec_perm_const): ...this new function. Take the selector
18374 as a vec_perm_indices rather than an rtx. Also take the mode of
18375 the selector. Update call to shift_amt_for_vec_perm_mask.
18376 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18377 Use vec_perm_indices::new_expanded_vector to expand the original
18378 selector into bytes. Check whether the indices fit in the vector
18379 mode before using a variable permute.
18380 (expand_vec_perm_var): Make global.
18381 (expand_mult_highpart): Use expand_vec_perm_const.
18382 * fold-const.c: Includes vec-perm-indices.h.
18383 * tree-ssa-forwprop.c: Likewise.
18384 * tree-vect-data-refs.c: Likewise.
18385 * tree-vect-generic.c: Likewise.
18386 * tree-vect-loop.c: Likewise.
18387 * tree-vect-slp.c: Likewise.
18388 * tree-vect-stmts.c: Likewise.
18389 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18391 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18392 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18393 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18394 (aarch64_vectorize_vec_perm_const): ...this new function.
18395 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18396 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18397 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18398 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18399 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18400 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18401 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18403 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18404 check for NEON modes.
18405 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18406 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18407 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18408 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18410 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18411 the old VEC_PERM_CONST conditions.
18412 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18413 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18414 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18415 (ia64_vectorize_vec_perm_const_ok): Merge into...
18416 (ia64_vectorize_vec_perm_const): ...this new function.
18417 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18418 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18419 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18420 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18421 * config/mips/mips.c (mips_expand_vec_perm_const)
18422 (mips_vectorize_vec_perm_const_ok): Merge into...
18423 (mips_vectorize_vec_perm_const): ...this new function.
18424 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18425 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18426 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18427 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18428 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18429 (rs6000_expand_vec_perm_const): Delete.
18430 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18433 (altivec_expand_vec_perm_const_le): Take each operand individually.
18434 Operate on constant selectors rather than rtxes.
18435 (altivec_expand_vec_perm_const): Likewise. Update call to
18436 altivec_expand_vec_perm_const_le.
18437 (rs6000_expand_vec_perm_const): Delete.
18438 (rs6000_vectorize_vec_perm_const_ok): Delete.
18439 (rs6000_vectorize_vec_perm_const): New function.
18440 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18441 an element count and rtx array.
18442 (rs6000_expand_extract_even): Update call accordingly.
18443 (rs6000_expand_interleave): Likewise.
18444 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18445 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18446 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18447 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18448 (rs6000_expand_vec_perm_const): Delete.
18449 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18450 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18451 (altivec_expand_vec_perm_const_le): Take each operand individually.
18452 Operate on constant selectors rather than rtxes.
18453 (altivec_expand_vec_perm_const): Likewise. Update call to
18454 altivec_expand_vec_perm_const_le.
18455 (rs6000_expand_vec_perm_const): Delete.
18456 (rs6000_vectorize_vec_perm_const_ok): Delete.
18457 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18458 reference to the SPE evmerge intructions.
18459 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18460 an element count and rtx array.
18461 (rs6000_expand_extract_even): Update call accordingly.
18462 (rs6000_expand_interleave): Likewise.
18463 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18464 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18466 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18470 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18471 vector mode and that that mode matches the mode of the data
18473 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18474 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18475 directly using expand_vec_perm_1 when forcing selectors into
18477 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18479 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18481 * optabs-query.h (can_vec_perm_p): Delete.
18482 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18483 * optabs-query.c (can_vec_perm_p): Split into...
18484 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18485 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18486 particular selector is valid.
18487 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18488 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18489 (vect_grouped_load_supported): Likewise.
18490 (vect_shift_permute_load_chain): Likewise.
18491 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18492 (vect_transform_slp_perm_load): Likewise.
18493 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18494 (vectorizable_bswap): Likewise.
18495 (vect_gen_perm_mask_checked): Likewise.
18496 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18497 implementations of variable permutation vectors into account
18498 when deciding which selector to use.
18499 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18500 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18501 with a false third argument.
18502 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18503 to test whether the constant selector is valid and can_vec_perm_var_p
18504 to test whether a variable selector is valid.
18506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18508 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18509 * optabs-query.c (can_vec_perm_p): Likewise.
18510 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18511 instead of vec_perm_indices.
18512 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18513 (vect_gen_perm_mask_checked): Likewise,
18514 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18515 (vect_gen_perm_mask_checked): Likewise,
18517 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18519 * optabs-query.h (qimode_for_vec_perm): Declare.
18520 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18521 (qimode_for_vec_perm): ...this new function.
18522 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18524 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18526 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18527 does not have a conditional at the top.
18529 2018-01-02 Richard Biener <rguenther@suse.de>
18531 * ipa-inline.c (big_speedup_p): Fix expression.
18533 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18536 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18539 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18543 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18544 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18545 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18546 cond_taken_branch_cost 3->4.
18548 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18550 PR tree-optimization/83581
18551 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18552 TODO_cleanup_cfg if any changes have been made.
18554 PR middle-end/83608
18555 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18556 convert_modes if target mode has the right side, but different mode
18559 PR middle-end/83609
18560 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18561 last argument when extracting from CONCAT. If either from_real or
18562 from_imag is NULL, use expansion through memory. If result is not
18563 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18564 the parts directly to inner mode, if even that fails, use expansion
18567 PR middle-end/83623
18568 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18569 check for bswap in mode rather than HImode and use that in expand_unop
18572 Copyright (C) 2018 Free Software Foundation, Inc.
18574 Copying and distribution of this file, with or without modification,
18575 are permitted in any medium without royalty provided the copyright
18576 notice and this notice are preserved.