1 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
5 (vect_recog_rotate_pattern): ...here.
6 (vect_convert_input): Try to insert casts of invariants in the
8 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14 vector type. If given, install it in the new statement's
16 (vect_element_precision): New function.
17 (vect_unpromoted_value): New struct.
18 (vect_unpromoted_value::vect_unpromoted_value): New function.
19 (vect_unpromoted_value::set_op): Likewise.
20 (vect_look_through_possible_promotion): Likewise.
21 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
22 (vect_widened_op_tree, vect_convert_input): Likewise.
23 (vect_convert_inputs, vect_convert_output): Likewise.
24 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
25 to handle the optional cast of the multiplication result and
26 vect_widened_op_tree to detect the widened multiplication itself.
27 Do not require the input and output of promotion casts to have
28 the same sign, but base the signedness of the operation on the
29 input rather than the result. If the pattern includes two
30 promotions, check that those promotions have the same sign.
31 Do not restrict the MULT_EXPR handling to a double-width result;
32 handle quadruple-width results and wider. Use vect_convert_inputs
33 to convert the inputs to the common type.
34 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
35 to handle the optional cast of the ABS result. Also allow a sign
36 change or a sign extension between the ABS and MINUS.
37 Use vect_widened_op_tree to detect the widened subtraction and use
38 vect_convert_inputs to convert the inputs to the common type.
39 (vect_handle_widen_op_by_const): Delete.
40 (vect_recog_widen_op_pattern): New function.
41 (vect_recog_widen_mult_pattern): Use it.
42 (vect_recog_widen_shift_pattern): Likewise.
43 (vect_recog_widen_sum_pattern): Use
44 vect_look_through_possible_promotion to handle the promoted
47 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
49 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
50 the containing gimple_seq *.
51 * gimple-iterator.h (gsi_for_stmt): Declare it.
52 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
53 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
54 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
55 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
56 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
58 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
60 (vect_mark_pattern_stmts): ...here. Handle cases in which the
61 statement being replaced is part of an existing pattern
62 definition sequence, inserting the new pattern statements before
64 (vect_pattern_recog_1): Don't return a bool. If the statement
65 is already part of a pattern, instead apply pattern matching
66 to the pattern definition statements. Don't clear the
67 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
68 (vect_pattern_recog): Don't break after the first match;
69 continue processing the pattern definition statements instead.
70 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
72 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
74 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
75 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
76 (vect_recog_widen_sum_pattern): Use it.
78 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
80 * tree-vect-loop.c (vectorizable_reduction): Assert that the
81 phi is not a pattern statement and has not been replaced by
83 * tree-vect-patterns.c (type_conversion_p): Don't check
84 STMT_VINFO_IN_PATTERN_P.
85 (vect_recog_vector_vector_shift_pattern): Likewise.
86 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
87 the pattern statement rather than the original statement; check
88 directly for a WIDEN_MULT_EXPR here.
89 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
90 vect_is_simple_use to return the pattern statement rather
91 than the original statement; use is_pattern_stmt_p to check
92 for such a pattern statement.
93 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
94 to return the pattern statement rather than the original statement;
95 don't do the same transformation here.
96 (vect_is_simple_use): If the defining statement has been replaced
97 by a pattern statement, return the pattern statement instead.
98 Remove the corresponding (local) transformation from the vectype
101 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
103 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
104 end and default to null.
105 * tree-vect-loop.c (vect_create_epilog_for_reduction)
106 (vectorizable_reduction): Update calls accordingly, dropping the
107 gimple ** argument if the passed-back statement isn't needed.
108 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
109 (vect_recog_rotate_pattern): Likewise.
110 (vect_recog_mask_conversion_pattern): Likewise.
111 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
112 (vect_mask_constant_operand_p): Likewise.
113 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
114 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
115 (get_group_load_store_type, get_load_store_type): Likewise.
116 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
117 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
118 (vectorizable_conversion, vectorizable_assignment): Likewise.
119 (vectorizable_shift, vectorizable_operation): Likewise.
120 (vectorizable_store, vect_is_simple_cond): Likewise.
121 (vectorizable_condition, vectorizable_comparison): Likewise.
122 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
123 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
124 and move it to the end. Cope with null def_stmt_outs.
126 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
128 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
130 2018-06-29 Jeff Law <law@redhat.com>
132 * config/v850/v850.c (v850_legitimate_address_p): Handle large
133 displacements for TARGET_V850E2V3 and newer.
134 (TARGET_LRA_P): Remove. Defaults to LRA now.
135 * config/v850/v850.md (sign23byte_load): Remove.
136 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
137 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
139 2018-06-29 Martin Liska <mliska@suse.cz>
142 * coverage.c (coverage_init): Mangle full path name.
143 * doc/invoke.texi: Document the change.
144 * gcov-io.c (mangle_path): New.
145 * gcov-io.h (mangle_path): Likewise.
146 * gcov.c (mangle_name): Use mangle_path for path mangling.
148 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
150 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
151 if starting source register is not even.
153 2018-06-29 Martin Liska <mliska@suse.cz>
155 PR tree-optimization/86263
156 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
157 Make edge redirection.
159 2018-06-29 David Malcolm <dmalcolm@redhat.com>
161 * dumpfile.c (dump_loc): Add indentation based on scope depth.
162 (dump_scope_depth): New variable.
163 (get_dump_scope_depth): New function.
164 (dump_begin_scope): New function.
165 (dump_end_scope): New function.
166 * dumpfile.h (get_dump_scope_depth): New declaration.
167 (dump_begin_scope): New declaration.
168 (dump_end_scope): New declaration.
169 (class auto_dump_scope): New class.
170 (AUTO_DUMP_SCOPE): New macro.
171 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
174 2018-06-29 Richard Biener <rguenther@suse.de>
176 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
177 compute_all_dependences succeeds.
178 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
179 exceed --param loop-max-datarefs-for-datadeps.
181 2018-06-29 Jakub Jelinek <jakub@redhat.com>
183 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
185 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
188 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
189 alternative 0 in preferred_for_speed attribute.
191 2018-06-28 Paul Koning <ni1d@arrl.net>
193 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
194 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
195 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
196 * config/pdp11/pdp11.md: Correct "length" attribute calculation
197 for shift insn patterns.
199 2018-06-28 David Malcolm <dmalcolm@redhat.com>
201 * cgraph.c (cgraph_node::get_body): Replace assignments to
202 "dump_file" with calls to set_dump_file.
203 * dumpfile.c (alt_dump_file): Make static, and group with...
204 (alt_flags): ...this definition.
205 (dumps_are_enabled): New variable.
206 (refresh_dumps_are_enabled): New function.
207 (set_dump_file): New function.
208 (set_alt_dump_file): New function.
209 (gcc::dump_manager::dump_start): Replace assignments to
210 "dump_file" and "alt_dump_file" with calls to set_dump_file and
212 (gcc::dump_manager::dump_finish): Likewise.
213 * dumpfile.h (alt_dump_file): Delete decl.
214 (dumps_are_enabled): New variable decl.
215 (set_dump_file): New function decl.
216 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
218 * tree-nested.c (lower_nested_functions): Replace assignments to
219 "dump_file" with calls to set_dump_file.
221 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
223 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
224 goto_locus of each outgoing edge of each basic block.
226 2018-06-28 Richard Biener <rguenther@suse.de>
228 * dwarf2out.c (decl_scope_table): Remove.
229 (push_decl_scope): Likewise.
230 (pop_decl_scope): Likewise.
231 (gen_type_die_for_member): Do not call push/pop_decl_scope.
232 (gen_struct_or_union_type_die): Likewise.
233 (gen_tagged_type_die): Likewise.
234 (dwarf2out_init): Do not initialize decl_scope_table.
235 (dwarf2out_c_finalize): Do not free it.
237 2018-06-28 Richard Biener <rguenther@suse.de>
239 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
240 deciding whether to not re-use a DIE.
242 2018-06-28 Richard Biener <rguenther@suse.de>
244 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
245 DW_AT_abstract_origin attribute.
247 2018-06-28 Martin Liska <mliska@suse.cz>
249 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
250 Use newly introduced constants.
251 * tree-switch-conversion.h (struct jump_table_cluster):
252 Define max_ratio_for_size and max_ratio_for_speed.
254 2018-06-28 Martin Liska <mliska@suse.cz>
256 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
257 Add new checking assert to catch invalid state.
258 (jump_table_cluster::can_be_handled): Handle single case
260 (jump_table_cluster::is_beneficial): Bail out for such case.
261 (bit_test_cluster::find_bit_tests):
262 Add new checking assert to catch invalid state.
263 (bit_test_cluster::can_be_handled): Handle single case
265 (bit_test_cluster::is_beneficial): Bail out for such case.
266 (switch_decision_tree::analyze_switch_statement):
269 2018-06-28 Martin Liska <mliska@suse.cz>
271 * common.opt: Introduce -completion option.
272 * gcc.c (driver_handle_option): Handle it.
273 (driver::main): Print completions if completion
275 * opt-suggestions.c (option_proposer::get_completions):
277 (option_proposer::suggest_completion): Likewise.
278 (option_proposer::find_param_completions): Likewise.
279 (verify_autocompletions): Likewise.
280 (test_completion_valid_options): Likewise.
281 (test_completion_valid_params): Likewise.
282 (in_completion_p): Likewise.
283 (empty_completion_p): Likewise.
284 (test_completion_partial_match): Likewise.
285 (test_completion_garbage): Likewise.
286 (opt_proposer_c_tests): Likewise.
287 * opt-suggestions.h: Declare new functions.
288 * opts.c (common_handle_option): Handle OPT__completion_.
289 * selftest-run-tests.c (selftest::run_tests): Add
290 opt_proposer_c_tests.
291 * selftest.c (assert_str_startswith): New.
292 * selftest.h (assert_str_startswith): Likewise.
293 (opt_proposer_c_tests): New.
294 (ASSERT_STR_STARTSWITH): Likewise.
296 2018-06-28 Martin Liska <mliska@suse.cz>
298 * Makefile.in: Add opt-suggestions.o.
299 * gcc-main.c: Include opt-suggestions.h.
300 * gcc.c (driver::driver): Likewise.
301 (driver::~driver): Remove m_option_suggestions.
302 (driver::build_option_suggestions): Moved to option_proposer.
303 (driver::suggest_option): Likewise.
304 (driver::handle_unrecognized_options): Use option_proposer.
305 * gcc.h (class driver): Add new memver m_option_proposer.
306 * opt-suggestions.c: New file.
307 * opt-suggestions.h: New file.
309 2018-06-28 Martin Liska <mliska@suse.cz>
311 * vec.h (class auto_string_vec): New (moved from auto_argvec).
312 (auto_string_vec::~auto_string_vec): Likewise.
314 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
316 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
317 statement if id->reset_location is true.
318 (copy_edges_for_bb): Do not set goto_locus on the new edges if
319 id->reset_location is true.
320 (copy_phis_for_bb): Force input_location on the arguments if
321 id->reset_location is true.
322 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
323 is set on the function to be inlined.
324 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
325 prevent_decl_creation_for_types fields up and add reset_location field.
327 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
329 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
331 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
333 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
335 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
336 hard registers for the clobbered pseudo.
338 2018-06-27 Paul Koning <ni1d@arrl.net>
340 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
341 mutually exclusive options.
342 * config/pdp11/constraints.md (h): New constraint.
343 (O): Update definition to match shift code generation.
345 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
347 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
349 (output_jump): Change arguments.
350 (pdp11_fixed_cc_regs): New function.
351 (pdp11_cc_mode): Ditto.
352 (pdp11_expand_shift): Ditto.
353 (pdp11_assemble_shift): Ditto.
354 (pdp11_small_shift): Ditto.
355 (pdp11_branch_cost): Remove.
356 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
358 (pdp11_register_move_cost): Update for CC registers.
359 (pdp11_rtx_costs): Add case for LSHIFTRT.
360 (pdp11_output_jump): Add CCNZ mode conditional branches.
361 (notice_update_cc_on_set): Remove.
362 (pdp11_cc_mode): New function.
363 (simple_memory_operand): Correct pre/post decrement case.
364 (no_side_effect_operand): New function.
365 (pdp11_regno_reg_class): Add CC_REGS class.
366 (pdp11_fixed_cc_regs): New function.
367 (pdp11_small_shift): New function.
368 (pdp11_expand_shift): New function to expand shift insns.
369 (pdp11_assemble_shift): New function to output shifts.
370 (pdp11_branch_cost): Remove.
371 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
372 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
374 (PTRDIFF_TYPE): Ditto.
375 (ADJUST_INSN_LENGTH): New macro.
376 (FIXED_REGISTERS): Add CC registers.
377 (CALL_USED_REGISTERS): Ditto.
379 (REG_CLASS_NAMES): Ditto.
380 (REG_CLASS_CONTENTS): Ditto.
381 (SELECT_CC_MODE): Use new function.
382 (TARGET_FLAGS_REGNUM): New macro.
383 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
384 (cc0_reg_rtx): Remove.
385 (CC_STATUS_MDEP): Remove.
386 (CC_STATUS_MDEFP_INIT): Remove.
388 (NOTICE_UPDATE_CC): Remove.
389 (REGISTER_NAMES): Add CC registers.
390 (BRANCH_COST): Change to constant 1.
391 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
393 * config/pdp11/pdp11.opt (mbcopy): Remove.
394 (mbcopy-builtin): Remove.
395 (mbranch-cheap): Remove.
396 (mbranch-expensive): Remove.
397 * config/pdp11/predicates.md (expand_shift_operand): Update to
398 match shift code generation.
399 (ccnz_operator): New predicate.
400 * doc/invoke.texi (PDP-11 Options): Remove deleted options
401 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
402 Remove non-existent option -mabshi, -mno-abshi. Document mutually
404 * doc/md.texi (PDP-11): Document new D and h constraints. Update
405 description of O constraint.
407 2018-06-27 Jeff Law <law@redhat.com>
408 Austin Law <austinklaw@gmail.com>
410 * config/v850/v850.md (addsi3_set_flags): New pattern.
411 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
412 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
413 (zero_extendhisi2_v850_set_flags): Likewise.
414 (zero_extendqisi2_v850_set_flags): Likewise.
415 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
416 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
417 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
419 * config/v850/v850-protos.h (notice_update_cc): Remove.
420 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
421 (v850_print_operand): Handle 'D' and "d".
422 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
423 Add handling of arithmetic/logical operations compared against zero.
424 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
425 Do not look at v850_compare_op, instead get mode from last argument.
426 (v850_gen_compare): Remove
427 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
428 after reload for prologue insns.
429 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
431 (construct_save_jarl): Likewise.
432 (TARGET_FLAGS_REGNUM): Define.
433 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
434 (NOTICE_UPDATE_CC): Remove.
435 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
436 than cc0. Conditionalize on reload_completed.
437 (cmpsi_insn, setfcc_insn): Likewise.
438 (tst1 splitter): Turn into define_and_split which sets the flags
440 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
441 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
442 (cstoresf4, cstoredf4): Clobber the flags.
443 (cmpsi, cmpsf, cmpdf): Remove expanders.
444 (setf_insn): Remove pattern.
445 (addsi3): Turn into define_and_split which clobbers the flags after
446 reload and a suitable pattern (addsi3_clobber_flags) for use after
448 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
449 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
450 (ashrsi3, ashrsi3_v850e2): Likewise.
451 (bins): Clobber the flags.
452 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
453 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
454 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
455 (call_value_internal_short, call_value_internal_long): Likewise.
456 (callt_save_interrupt, callt_return_interrupt): Likewise.
457 (save_interrupt, return_interrupt): Likewise.
458 (callt_save_all_interrupt, save_all_interrupt): Likewise.
459 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
460 (restore_all_interrupt, _restore_all_interrupt): Likewise.
461 (All FP comparisons): Only allow after reload has completed.
463 (divh, divhu): Tweak output template.
464 (branch_z_normal, branch_z_invert): Remove
465 (branch_nz_normal, branch_nz_invert): Likewise.
466 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
468 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
469 * config/v850/v850.c (notice_update_cc): Remove.
470 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
471 (CC_NO_CARRY): Likewise.
472 (NOTICE_UPDATE_CC): Define to nothing.
473 * config/v850/v850.md: Remove block comment on cc0 handling
474 Remove "cc" attribute from all patterns. Remove cc_status handling
475 from all patterns. Minor formatting fixes.
477 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
479 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
480 (cortex-a76.cortex-a55): Likewise.
481 * config/aarch64/aarch64-tune.md: Regenerate.
482 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
483 cortex-a76.cortex-a55.
485 2018-06-27 Jeff Law <law@redhat.com>
487 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
488 (MULTILIB_DIRNAMES): Similarly.
490 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
492 * gimple.h (gimple_return_retbnd): Delete.
493 (gimple_return_set_retbnd): Likewise.
494 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
495 gimple_return_set_retbnd.
496 * gimple-pretty-print.c (dump_gimple_return): Remove call to
497 gimple_return_retbnd and adjust.
498 * tree-inline.h (struct copy_body_data): Remove retbnd field.
499 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
500 Explicitly return NULL in a couple more cases. Move assertion
501 on debug statements and remove unreachable code.
502 (reset_debug_binding): Do not test id->retbnd.
503 (expand_call_inline): Do not set it.
505 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
507 * configure.ac: Add --disable-gcov option.
508 * configure: Regenerate.
509 * Makefile.in: Honour @enable_gcov@.
510 * doc/install.texi: Document --disable-gcov.
512 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
514 * config/arm/arm-cpus.in (cortex-a76): New entry.
515 (cortex-a76.cortex-a55): Likewise.
516 * config/arm/arm-tables.opt: Regenerate.
517 * config/arm/arm-tune.md: Likewise.
518 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
519 * doc/invoke.texi (ARM Options): Document cortex-a76 and
520 cortex-a76.cortex-a55.
522 2018-06-27 Tamar Christina <tamar.christina@arm.com>
525 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
527 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
529 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
531 (EPILOGUE_USES): Likewise.
533 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
535 * tree-inline.c (remap_location): New function extracted from...
536 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
537 (copy_phis_for_bb): ...here. Call remap_location.
538 (copy_cfg_body): Adjust call to copy_edges_for_bb.
540 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
542 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
543 unaligned vsx for 16B memset.
545 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
548 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
549 ieee128_float_type_node to long_double_type_node unless
550 TARGET_LONG_DOUBLE_128 is set.
552 2018-06-26 David Malcolm <dmalcolm@redhat.com>
554 * cfgloop.c (get_loop_location): Convert return type from
555 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
556 by implicit construction from rtx_insn *, and using
557 dump_user_location_t::from_function_decl for the fallback case.
558 * cfgloop.h (get_loop_location): Convert return type from
559 location_t to dump_user_location_t.
560 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
561 dump_printf_loc to pass in a dump_location_t rather than a
562 location_t, via the gimple stmt.
563 * coverage.c (get_coverage_counts): Update calls to
564 dump_printf_loc to pass in dump_location_t rather than a
566 * doc/optinfo.texi (Dump types): Convert example of
567 dump_printf_loc from taking "locus" to taking "insn". Update
568 description of the "_loc" calls to cover dump_location_t.
569 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
571 (dump_user_location_t::dump_user_location_t): New constructors,
572 from gimple *stmt and rtx_insn *.
573 (dump_user_location_t::from_function_decl): New function.
574 (dump_loc): Make static.
575 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
576 const dump_location_t &.
577 (dump_generic_expr_loc): Delete.
578 (dump_printf_loc): Convert param "loc" from location_t to
579 const dump_location_t &.
580 (selftest::test_impl_location): New function.
581 (selftest::dumpfile_c_tests): New function.
582 * dumpfile.h: Include "profile-count.h".
583 (class dump_user_location_t): New class.
584 (struct dump_impl_location_t): New struct.
585 (class dump_location_t): New class.
586 (dump_printf_loc): Convert 2nd param from source_location to
587 const dump_location_t &.
588 (dump_generic_expr_loc): Delete.
589 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
590 const dump_location_t &.
591 * gimple-fold.c (fold_gimple_assign): Update call to
592 dump_printf_loc to pass in a dump_location_t rather than a
593 location_t, via the gimple stmt.
594 (gimple_fold_call): Likewise.
595 * gimple-loop-interchange.cc
596 (loop_cand::analyze_iloop_reduction_var): Update for change to
597 check_reduction_path.
598 (tree_loop_interchange::interchange): Update for change to
600 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
601 change in return-type of find_loop_location.
602 (graphite_regenerate_ast_isl): Likewise.
603 * graphite-optimize-isl.c (optimize_isl): Likewise.
604 * graphite.c (graphite_transform_loops): Likewise.
605 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
606 pass in a dump_location_t rather than a location_t, via the
608 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
609 * ipa.c (walk_polymorphic_call_targets): Likewise.
610 * loop-unroll.c (report_unroll): Convert "locus" param from
611 location_t to dump_location_t.
612 (decide_unrolling): Update for change to get_loop_location's
614 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
615 location_t to dump_user_location_t.
616 (grid_find_single_omp_among_assignments_1): Updates calls to
617 dump_printf_loc to pass in a dump_location_t rather than a
618 location_t, via the gimple stmt.
619 (grid_parallel_clauses_gridifiable): Convert "tloc" from
620 location_t to dump_location_t. Updates calls to dump_printf_loc
621 to pass in a dump_location_t rather than a location_t, via the
623 (grid_inner_loop_gridifiable_p): Likewise.
624 (grid_dist_follows_simple_pattern): Likewise.
625 (grid_gfor_follows_tiling_pattern): Likewise.
626 (grid_target_follows_gridifiable_pattern): Likewise.
627 (grid_attempt_target_gridification): Convert initialization
628 of local "grid" from memset to zero-initialization; FIXME: does
629 this require C++11? Update call to dump_printf_loc to pass in a
630 optinfo_location rather than a location_t, via the gimple stmt.
631 * profile.c (read_profile_edge_counts): Updates call to
632 dump_printf_loc to pass in a dump_location_t rather than a
634 (compute_branch_probabilities): Likewise.
635 * selftest-run-tests.c (selftest::run_tests): Call
637 * selftest.h (dumpfile_c_tests): New decl.
638 * tree-loop-distribution.c (pass_loop_distribution::execute):
639 Update for change in return type of find_loop_location.
640 * tree-parloops.c (parallelize_loops): Likewise.
641 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
642 "locus" from location_t to dump_user_location_t.
643 (canonicalize_loop_induction_variables): Likewise.
644 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
645 for change in return type of find_loop_location.
646 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
647 to dump_printf_loc to pass in a dump_location_t rather than a
648 location_t, via the stmt.
649 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
651 * tree-vect-loop-manip.c (find_loop_location): Convert return
652 type from source_location to dump_user_location_t.
653 (vect_do_peeling): Update for above change.
654 (vect_loop_versioning): Update for change in type of
656 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
657 from location_t to dump_user_location_t.
658 (vect_estimate_min_profitable_iters): Update for change in type
660 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
661 location_t to dump_location_t.
662 (vect_slp_bb): Update for change in type of vect_location.
663 * tree-vectorizer.c (vect_location): Convert from source_location
664 to dump_user_location_t.
665 (try_vectorize_loop_1): Update for change in vect_location's type.
666 (vectorize_loops): Likewise.
667 (increase_alignment): Likewise.
668 * tree-vectorizer.h (vect_location): Convert from source_location
669 to dump_user_location_t.
670 (find_loop_location): Convert return type from source_location to
671 dump_user_location_t.
672 (check_reduction_path): Convert 1st param from location_t to
673 dump_user_location_t.
674 * value-prof.c (check_counter): Update call to dump_printf_loc to
675 pass in a dump_user_location_t rather than a location_t; update
676 call to error_at for change in type of "locus".
677 (check_ic_target): Update call to dump_printf_loc to
678 pass in a dump_user_location_t rather than a location_t, via the
681 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
683 * config/s390/s390.h (enum processor_flags): Do not use
684 default tune parameter when -march was specified.
686 2018-06-26 Jakub Jelinek <jakub@redhat.com>
689 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
690 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
693 2018-06-26 Richard Biener <rguenther@suse.de>
695 PR tree-optimization/86287
697 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
698 (vect_analyze_loop): Initialize n_stmts.
700 2018-06-26 Richard Biener <rguenther@suse.de>
703 * fold-const.c (fold_convertible_p): Pointer extension
706 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
709 * dwarf2out.c (loc_list_has_views): Adjust comments.
710 (dw_loc_list): Split single cross-partition range with
713 2018-06-25 Jeff Law <law@redhat.com>
715 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
716 on -mbig-switch by default.
718 * config/v850/predicates.md (const_float_1_operand): Fix match_code
720 (const_float_0_operand): Remove unused predicate.
721 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
722 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
723 (recipsf2): New expander. Original pattern now called
725 (recipdf2, recipdf2_insn): Similarly.
726 (rsqrtsf2, rsqrtsf2_insn): Similarly
727 (rsqrtdf2, rsqrtdf2_insn): Similarly
729 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
731 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
732 Simplify logic for FreeBSD (twice).
734 2018-06-25 Martin Sebor <msebor@redhat.com>
736 PR tree-optimization/86204
737 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
738 a strnlen result if it's less than the length of the string.
740 2018-06-25 Martin Sebor <msebor@redhat.com>
742 PR tree-optimization/85700
743 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
744 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
745 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
747 2018-06-25 Martin Sebor <msebor@redhat.com>
749 * doc/extend.texi (Zero-length arrays): Update and clarify.
751 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
753 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
754 added IEEE/IBM long double multilib support on PowerPC little
755 endian Linux systems.
756 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
757 (MULTILIB_DEFAULTS): Likewise.
758 * config/rs6000/rs6000.c (rs6000_option_override_internal):
760 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
761 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
762 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
764 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
767 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
768 (REORDER_45): Likewise.
770 2018-06-25 Jeff Law <law@redhat.com>
772 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
773 dividend to 32 bits. Adjust length.
774 (udivmodhi4): Cleanup output template. Fix length.
776 2018-06-25 Carl Love <cel@us.ibm.com>
778 * config/rs6000/vsx.md: Change word selector to prefered location.
780 2018-06-25 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/86304
783 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
784 epilogue-if-converted loops as well.
786 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
788 * lto-section-out.c (lto_begin_section): Do not print section
789 name for noaddr and unnumbered dumps.
791 2018-06-25 Richard Biener <rguenther@suse.de>
793 * tree-vectorizer.h (struct vec_info_shared): New structure
794 with parts split out from struct vec_info and loop_nest from
795 struct _loop_vec_info.
796 (struct vec_info): Adjust accordingly.
797 (struct _loop_vec_info): Likewise.
798 (LOOP_VINFO_LOOP_NEST): Adjust.
799 (LOOP_VINFO_DATAREFS): Likewise.
800 (LOOP_VINFO_DDRS): Likewise.
801 (struct _bb_vec_info): Likewise.
802 (BB_VINFO_DATAREFS): Likewise.
803 (BB_VINFO_DDRS): Likewise.
804 (struct _stmt_vec_info): Add dr_aux member.
805 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
806 (DR_MISALIGNMENT_UNINITIALIZED): New.
807 (set_dr_misalignment): Adjust.
808 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
809 (vect_analyze_loop): Adjust prototype.
810 (vect_analyze_loop_form): Likewise.
811 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
812 Compute dependences lazily.
813 (vect_record_base_alignments): Use shared datarefs/ddrs.
814 (vect_verify_datarefs_alignment): Likewise.
815 (vect_analyze_data_refs_alignment): Likewise.
816 (vect_analyze_data_ref_accesses): Likewise.
817 (vect_analyze_data_refs): Likewise.
818 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
819 constructor parameter for shared part.
820 (vect_analyze_loop_form): Pass in shared part and adjust.
821 (vect_analyze_loop_2): Pass in storage for the number of
822 stmts. Move loop nest finding to the caller. Compute
824 (vect_analyze_loop): Pass in shared part.
825 (vect_transform_loop): Verify shared datarefs are unchanged.
826 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
827 constructor parameter for shared part.
828 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
829 (vect_slp_bb): Verify shared datarefs are unchanged before
831 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
833 (new_stmt_vec_info): Initialize DR_AUX misalignment to
834 DR_MISALIGNMENT_UNINITIALIZED.
835 * tree-vectorizer.c (vec_info::vec_info): Add constructor
836 parameter for shared part.
837 (vec_info::~vec_info): Adjust.
838 (vec_info_shared::vec_info_shared): New.
839 (vec_info_shared::~vec_info_shared): Likewise.
840 (vec_info_shared::save_datarefs): Likewise.
841 (vec_info_shared::check_datarefs): Likewise.
842 (try_vectorize_loop_1): Construct shared part live for analyses
843 of a single loop for multiple vector sizes.
844 * tree-parloops.c (gather_scalar_reductions): Adjust.
846 2018-06-25 Richard Biener <rguenther@suse.de>
848 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
849 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
850 (vect_analyze_data_refs): Remove similar code from here and
851 simplify accordingly.
853 2018-06-25 Richard Biener <rguenther@suse.de>
855 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
856 for reverse storage order accesses rather than asserting
857 they cannot happen here.
859 2018-06-25 Tom de Vries <tdevries@suse.de>
862 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
863 Use data16 instead of .byte for insn prefix.
865 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
868 * parser.c (make_char_string_pack): Pass this literal chars
869 through cpp_interpret_string.
870 (cp_parser_userdef_numeric_literal): Check the result of
871 make_char_string_pack.
873 2018-06-24 Maya Rashish <coypu@sdf.org>
875 * ginclude/stddef.h: Simplify conditions around avoiding
876 re-definition of __size_t.
878 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
880 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
881 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
883 2018-06-22 Maya Rashish <coypu@sdf.org>
885 * doc/invoke.texi (mno-fancy-math-387): Update for changes
886 made to OpenBSD and NetBSD through the years.
888 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
890 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
891 behavior of vec_pack (vector double, vector double) to match
892 behavior of vec_float2 (vector double, vector double).
894 2018-06-22 Olivier Hainque <hainque@adacore.com>
896 * gimplify.c (gimplify_function_tree): Prevent creation
897 of a trampoline for the address of the current function
898 passed to entry/exit instrumentation hooks.
900 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
903 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
906 2018-06-22 Martin Liska <mliska@suse.cz>
908 PR tree-optimization/86263
909 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
910 Bail out if is_enabled is false.
911 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
913 (jump_table_cluster::is_enabled): New function.
915 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
917 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
918 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
919 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
920 (lto_input_ts_binfo_tree_pointers): Likewise.
921 * tree-streamer-out.c (streamer_write_tree_bitfields,
922 write_ts_binfo_tree_pointers): Likewise.
923 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
925 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
927 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
929 2018-06-22 Martin Liska <mliska@suse.cz>
931 * symbol-summary.h (get): Make it pure and inline move
932 functionality from ::get function.
933 (get): Remove and inline into ::get and ::get_create.
934 (get_create): Move code from ::get function.
936 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
939 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
940 -x assembler-with-cpp.
942 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
944 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
945 _FILE_OFFSET_BITS=64 for C++.
947 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
949 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
950 conversion insn that shows up when pr85657-3.c is compiled using
951 IEEE 128-bit long double.
952 (neg<mode>2_internal): Use the correct mode to check whether the
953 mode is IBM extended.
954 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
955 multiply and divide external functions from being created more
958 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
960 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
962 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
963 the edge can be forwarded.
964 (cfg_layout_merge_blocks): Likewise.
966 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
968 * except.c (finish_eh_generation): Commit edge insertions only after
969 the EH edges have been redirected from post-landing to landing pads.
971 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
973 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
974 create_tmp_var_for to create the FRAME decl.
975 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
977 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
979 * tree-inline.c (copy_edges_for_bb): Minor tweak.
980 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
981 debug statement when resetting its value.
982 (expand_call_inline): Copy the locus of the call onto the assignment
983 of the return value, if any. Use local variable in more cases.
985 2018-06-21 Martin Liska <mliska@suse.cz>
987 * ipa-pure-const.c (propagate_nothrow): Use
988 funct_state_summaries->get.
989 (dump_malloc_lattice): Likewise.
990 (propagate_malloc): Likewise.
992 2018-06-21 Richard Biener <rguenther@suse.de>
994 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
995 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
996 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
998 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
999 BLOCK_ABSTRACT_ORIGIN unconditionally.
1001 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1003 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1005 * ipa-reference.c (ipa_reference_c_finalize): Delete
1006 ipa_ref_opt_sum_summaries and set it to NULL.
1008 2018-06-21 Tom de Vries <tdevries@suse.de>
1010 PR tree-optimization/85859
1011 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1012 test with comment from bb_no_side_effects_p.
1014 2018-06-21 Richard Biener <rguenther@suse.de>
1016 PR tree-optimization/86232
1017 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1018 max for constant niter.
1020 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1022 * config/aarch64/aarch64-simd.md
1023 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1025 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1027 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1028 Make opernads of the unspec commutative.
1030 2018-06-21 Richard Biener <rguenther@suse.de>
1032 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1033 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1034 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1035 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1037 (vect_record_base_alignments): Do not record base alignment
1038 for gathers or scatters.
1039 (vect_compute_data_ref_alignment): Drop return value that is always
1040 true. Bail out early for gathers or scatters.
1041 (vect_enhance_data_refs_alignment): Bail out early for gathers
1043 (vect_find_same_alignment_drs): Likewise.
1044 (vect_analyze_data_refs_alignment): Remove dead code.
1045 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1046 (vect_analyze_data_refs): For possible gathers or scatters do
1047 not create an alternate DR, just check their possible validity
1048 and mark them. Adjust DECL_NONALIASED handling to not rely
1050 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1051 update inits of gathers or scatters.
1052 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1053 Also copy gather/scatter flag to pattern vinfo.
1055 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1057 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1058 behavior of vec_packsu (vector unsigned long long, vector unsigned
1059 long long) to match behavior of vec_packs with same signature.
1061 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1062 Thomas Schwinge <thomas@codesourcery.com>
1063 Cesar Philippidis <cesar@codesourcery.com>
1065 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1066 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1067 (gimplify_adjust_omp_clauses): Likewise.
1068 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1069 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1070 (gimplify_omp_target_update): Update handling of acc update and
1072 * omp-low.c (install_var_field): Remove unused parameter
1073 base_pointers_restrict.
1074 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1075 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1077 (omp_target_base_pointers_restrict_p): Delete.
1078 (scan_omp_target): Update call to scan_sharing_clauses.
1079 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1081 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1082 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1083 (convert_local_omp_clauses): Likewise.
1084 * tree-pretty-print.c (dump_omp_clause): Likewise.
1085 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1087 (omp_clause_code_name): Likewise.
1089 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1092 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1095 PR tree-optimization/86231
1096 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1097 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1099 2018-06-20 Tom de Vries <tdevries@suse.de>
1101 PR tree-optimization/86097
1102 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1103 iv type if signedness of iv type is not the same as that of *nit.
1105 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1107 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1108 EDGE_EH edges, verify they are all EDGE_EH.
1110 2018-06-20 Maya Rashish <coypu@sdf.org>
1112 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1114 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1116 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1117 * config/aarch64/aarch64.c (xgene1_tunings): Add
1118 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1119 (aarch64_mode_valid_for_sched_fusion_p):
1120 Allow 16-byte modes.
1121 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1122 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1124 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1126 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1127 * config/aarch64/iterators.md (VQ2): New mode iterator.
1129 2018-06-20 Martin Liska <mliska@suse.cz>
1131 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1132 Change default ratio from 10 to 8.
1134 2018-06-20 Martin Liska <mliska@suse.cz>
1136 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1138 (bit_test_cluster::find_bit_tests): Likewise.
1139 (switch_decision_tree::analyze_switch_statement): Find clusters.
1140 * tree-switch-conversion.h (struct jump_table_cluster): Document
1143 2018-06-20 Martin Liska <mliska@suse.cz>
1145 * tree-switch-conversion.c (switch_conversion::collect):
1146 Record m_uniq property.
1147 (switch_conversion::expand): Bail out for special conditions.
1148 (group_cluster::~group_cluster): New.
1149 (group_cluster::group_cluster): Likewise.
1150 (group_cluster::dump): Likewise.
1151 (jump_table_cluster::emit): New.
1152 (switch_decision_tree::fix_phi_operands_for_edges): New.
1153 (struct case_node): Remove struct.
1154 (jump_table_cluster::can_be_handled): New.
1155 (case_values_threshold): Moved to header.
1156 (reset_out_edges_aux): Likewise.
1157 (jump_table_cluster::is_beneficial): New.
1158 (bit_test_cluster::can_be_handled): Likewise.
1159 (add_case_node): Remove.
1160 (bit_test_cluster::is_beneficial): New.
1161 (case_bit_test::cmp): New.
1162 (bit_test_cluster::emit): New.
1163 (expand_switch_as_decision_tree_p): Remove.
1164 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1165 (fix_phi_operands_for_edge): Likewise.
1166 (switch_decision_tree::analyze_switch_statement): New.
1167 (compute_cases_per_edge): Move ...
1168 (switch_decision_tree::compute_cases_per_edge): ... here.
1169 (try_switch_expansion): Likewise.
1170 (switch_decision_tree::try_switch_expansion): Likewise.
1171 (record_phi_operand_mapping): Likewise.
1172 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1173 (emit_case_decision_tree): Likewise.
1174 (switch_decision_tree::emit): Likewise.
1175 (balance_case_nodes): Likewise.
1176 (switch_decision_tree::balance_case_nodes): Likewise.
1177 (dump_case_nodes): Likewise.
1178 (switch_decision_tree::dump_case_nodes): Likewise.
1179 (emit_jump): Likewise.
1180 (switch_decision_tree::emit_jump): Likewise.
1181 (emit_cmp_and_jump_insns): Likewise.
1182 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1183 (emit_case_nodes): Likewise.
1184 (switch_decision_tree::emit_case_nodes): Likewise.
1185 (conditional_probability): Remove.
1186 * tree-switch-conversion.h (enum cluster_type): New.
1188 (struct cluster): Likewise.
1189 (cluster::cluster): Likewise.
1190 (struct simple_cluster): Likewise.
1191 (simple_cluster::simple_cluster): Likewise.
1192 (struct group_cluster): Likewise.
1193 (struct jump_table_cluster): Likewise.
1194 (struct bit_test_cluster): Likewise.
1195 (struct min_cluster_item): Likewise.
1196 (struct case_tree_node): Likewise.
1197 (case_tree_node::case_tree_node): Likewise.
1198 (jump_table_cluster::case_values_threshold): Likewise.
1199 (struct case_bit_test): Likewise.
1200 (struct switch_decision_tree): Likewise.
1201 (struct switch_conversion): Likewise.
1202 (switch_decision_tree::reset_out_edges_aux): Likewise.
1204 2018-06-20 Martin Liska <mliska@suse.cz>
1206 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1207 (hoist_edge_and_branch_if_true): Likewise.
1208 (expand_switch_using_bit_tests_p): Likewise.
1209 (struct case_bit_test): Likewise.
1210 (case_bit_test_cmp): Likewise.
1211 (emit_case_bit_tests): Likewise.
1212 (switch_conversion::switch_conversion): New class.
1213 (struct switch_conv_info): Remove old struct.
1214 (collect_switch_conv_info): More to ...
1215 (switch_conversion::collect): ... this.
1216 (check_range): Likewise.
1217 (switch_conversion::check_range): Likewise.
1218 (check_all_empty_except_final): Likewise.
1219 (switch_conversion::check_all_empty_except_final): Likewise.
1220 (check_final_bb): Likewise.
1221 (switch_conversion::check_final_bb): Likewise.
1222 (create_temp_arrays): Likewise.
1223 (switch_conversion::create_temp_arrays): Likewise.
1224 (free_temp_arrays): Likewise.
1225 (gather_default_values): Likewise.
1226 (switch_conversion::gather_default_values): Likewise.
1227 (build_constructors): Likewise.
1228 (switch_conversion::build_constructors): Likewise.
1229 (constructor_contains_same_values_p): Likewise.
1230 (switch_conversion::contains_same_values_p): Likewise.
1231 (array_value_type): Likewise.
1232 (switch_conversion::array_value_type): Likewise.
1233 (build_one_array): Likewise.
1234 (switch_conversion::build_one_array): Likewise.
1235 (build_arrays): Likewise.
1236 (switch_conversion::build_arrays): Likewise.
1237 (gen_def_assigns): Likewise.
1238 (switch_conversion::gen_def_assigns): Likewise.
1239 (prune_bbs): Likewise.
1240 (switch_conversion::prune_bbs): Likewise.
1241 (fix_phi_nodes): Likewise.
1242 (switch_conversion::fix_phi_nodes): Likewise.
1243 (gen_inbound_check): Likewise.
1244 (switch_conversion::gen_inbound_check): Likewise.
1245 (process_switch): Use the newly created class.
1246 (switch_conversion::expand): New.
1247 (switch_conversion::~switch_conversion): New.
1248 * tree-switch-conversion.h: New file.
1250 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1252 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1253 tree-vect-patterns.c.
1254 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1255 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1256 (vect_recog_sad_pattern): Likewise.
1257 (vect_recog_widen_sum_pattern): Likewise.
1258 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1259 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1260 (vect_recog_rotate_pattern): Likewise.
1261 (vect_recog_mult_pattern): Likewise.
1262 (vect_recog_vector_vector_shift_pattern): Likewise.
1263 (vect_recog_divmod_pattern): Likewise.
1264 (vect_recog_mixed_size_cond_pattern): Likewise.
1265 (vect_recog_bool_pattern): Likewise.
1266 (vect_recog_mask_conversion_pattern): Likewise.
1267 (vect_try_gather_scatter_pattern): Likewise.
1268 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1269 (vect_recog_over_widening_pattern): Likewise.
1270 (vect_recog_gather_scatter_pattern): Likewise.
1271 (vect_recog_func_ptr): Move from tree-vectorizer.h
1272 (vect_vect_recog_func_ptrs): Move further down the file.
1273 (vect_recog_func): Likewise. Remove the third argument.
1274 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1275 (vect_pattern_recog_1): Expect the pattern function to do any
1276 necessary target tests. Also expect it to provide a vector type.
1277 Remove the type_in handling.
1279 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1281 * tree-vect-patterns.c (vect_pattern_detected): New function.
1282 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1283 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1284 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1285 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1286 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1287 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1288 (vect_recog_mask_conversion_pattern)
1289 (vect_try_gather_scatter_pattern): Likewise.
1291 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1293 * tree-vect-patterns.c (vect_get_internal_def): New function.
1294 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1295 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1296 (search_type_for_mask_1): Use it.
1298 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1300 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1301 redundant WIDEN_SUM_EXPR handling.
1302 (vect_recog_sad_pattern): Likewise.
1304 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1306 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1307 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1308 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1309 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1310 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1312 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1314 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1315 use the stmt_vec_info of the original bb statement for the
1316 new zero assignment, even if the call is part of a pattern.
1318 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1320 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1321 that the sequence is attached to the original statement rather
1322 than the pattern statement.
1323 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1324 PATTERN_DEF_SEQ from the original statement rather than
1325 the main pattern statement.
1326 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1327 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1328 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1330 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1332 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1333 definition statements before the early exit for statements that aren't
1335 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1337 (vect_transform_loop): ...here. Process pattern definition
1338 statements without first checking whether the main pattern
1339 statement is live or relevant.
1341 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1343 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1344 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1346 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1348 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1349 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1350 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1351 (expand_block_compare): Change select_block_compare_mode call.
1352 (expand_strncmp_align_check): Use new functions, fix comment.
1353 (emit_final_str_compare_gpr): New function.
1354 (expand_strn_compare): Refactor and clean up code.
1355 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1357 2018-06-19 Tony Reix <tony.reix@atos.com>
1358 Damien Bergamini <damien.bergamini@atos.com>
1359 David Edelsohn <dje.gcc@gmail.com>
1361 * collect2.c (static_obj): New variable.
1362 (static_libs): New variable.
1363 (is_in_list): Uncomment declaration.
1364 (main): Track AIX libraries linked statically.
1365 (is_in_list): Uncomment definition.
1366 (scan_prog_file): Don't add AIX shared libraries initializer
1367 to constructor list if linking statically.
1369 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1371 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1373 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1375 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1377 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1380 2018-06-19 Martin Liska <mliska@suse.cz>
1382 * config/i386/i386.c (ix86_can_inline_p): Do not use
1383 ipa_fn_summaries::get_create.
1384 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1386 (devirtualization_time_bonus): Likewise.
1387 (ipcp_propagate_stage): Likewise.
1388 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1389 (edge_set_predicate): Likewise.
1390 (evaluate_conditions_for_known_args): Likewise.
1391 (evaluate_properties_for_edge): Likewise.
1392 (ipa_call_summary::reset): Tranform to ...
1393 (ipa_call_summary::~ipa_call_summary): ... this.
1394 (ipa_fn_summary::reset): Transform to ...
1395 (ipa_fn_summary::~ipa_fn_summary): ... this.
1396 (ipa_fn_summary_t::remove): Rename to ...
1397 (ipa_fn_summary_t::remove_callees): ... this.
1398 (ipa_fn_summary_t::duplicate): Use placement new
1399 instead of memory copy.
1400 (ipa_call_summary_t::duplicate): Likewise.
1401 (ipa_call_summary_t::remove): Remove.
1402 (dump_ipa_call_summary): Change get_create to get.
1403 (ipa_dump_fn_summary): Dump only when summary exists.
1404 (analyze_function_body): Use symbol_summary::get instead
1406 (compute_fn_summary): Likewise.
1407 (estimate_edge_devirt_benefit): Likewise.
1408 (estimate_edge_size_and_time): Likewise.
1409 (inline_update_callee_summaries): Likewise.
1410 (remap_edge_change_prob): Likewise.
1411 (remap_edge_summaries): Likewise.
1412 (ipa_merge_fn_summary_after_inlining): Likewise.
1413 (write_ipa_call_summary): Likewise.
1414 (ipa_fn_summary_write): Likewise.
1415 (ipa_free_fn_summary): Likewise.
1416 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1417 (struct ipa_call_summary): Likewise.
1418 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1420 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1421 (estimate_size_after_inlining): Likewise.
1422 (estimate_growth): Likewise.
1423 (growth_likely_positive): Likewise.
1424 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1425 (inline_call): Likewise.
1426 * ipa-inline.c (caller_growth_limits): Likewise.
1427 (can_inline_edge_p): Likewise.
1428 (can_inline_edge_by_limits_p): Likewise.
1429 (compute_uninlined_call_time): Likewise.
1430 (compute_inlined_call_time): Likewise.
1431 (want_inline_small_function_p): Likewise.
1432 (edge_badness): Likewise.
1433 (update_caller_keys): Likewise.
1434 (update_callee_keys): Likewise.
1435 (inline_small_functions): Likewise.
1436 (inline_to_all_callers_1): Likewise.
1437 (dump_overall_stats): Likewise.
1438 (early_inline_small_functions): Likewise.
1439 (early_inliner): Likewise.
1440 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1441 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1442 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1443 * ipa-split.c (execute_split_functions): Likewise.
1444 * symbol-summary.h: Likewise.
1445 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1447 2018-06-19 Richard Biener <rguenther@suse.de>
1449 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1450 (vectorize_loops): ... here. Fix dbgcnt handling.
1451 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1453 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1456 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1457 ieee128 argument takes up only one (vector) register, not two (floating
1460 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1462 * gimplify.c (gimplify_init_constructor): Really never clear for an
1463 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1465 2018-06-19 Richard Biener <rguenther@suse.de>
1467 PR tree-optimization/86179
1468 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1469 after failed recognition.
1471 2018-06-18 Martin Sebor <msebor@redhat.com>
1474 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1475 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1476 Handle integer subtraction.
1477 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1478 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1480 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1482 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1483 param from rtx to rtx_insn *.
1484 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1486 (frv_ifcvt_modify_insn): Likwise.
1487 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1488 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1489 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1490 as_a <rtx_insn *> cast to local "unprotected_region" once
1491 it's been established that it's not NULL or pc_rtx.
1492 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1493 param "sethi" from rtx to rtx_insn *.
1494 (nds32_group_float_insns): Likewise for param "insn".
1495 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1497 (vax_output_int_subtract): Likewise.
1498 * config/vax/vax.c (vax_output_int_add): Likewise for param
1500 (vax_output_int_subtract): Likewise.
1501 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1502 (emit_pattern_after): Likewise for param "after".
1503 (emit_insn_after): Likewise.
1504 (emit_jump_insn_after): Likewise.
1505 (emit_call_insn_after): Likewise.
1506 (emit_debug_insn_after): Likewise.
1507 (emit_pattern_before): Likewise for param "before".
1508 (emit_insn_before): Likewise.
1509 (emit_jump_insn_before): Likewise.
1510 * final.c (get_insn_template): Likewise for param "insn", removing
1512 * output.h (get_insn_template): Likewise for 2nd param.
1513 * rtl.h (emit_insn_before): Likewise.
1514 (emit_jump_insn_before): Likewise.
1515 (emit_debug_insn_before_noloc): Likewise.
1516 (emit_insn_after): Likewise.
1517 (emit_jump_insn_after): Likewise.
1518 (emit_call_insn_after): Likewise.
1519 (emit_debug_insn_after): Likewise.
1520 (set_insn_deleted): Likewise for param.
1522 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1525 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1526 floating point modes, so that IFmode is numerically greater than
1527 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1528 to declare the ordering. This prevents IFmode from being
1529 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1530 machine. Include rs6000-modes.h to share the fractional values
1531 between genmodes* and the rest of the compiler.
1535 * config/rs6000/rs6000-modes.h: New file.
1536 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1537 meaning of rs6000_long_double_size so that 126..128 selects an
1538 appropriate 128-bit floating point type.
1539 (rs6000_option_override_internal): Likewise.
1540 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1541 (TARGET_LONG_DOUBLE_128): Change the meaning of
1542 rs6000_long_double_size so that 126..128 selects an appropriate
1543 128-bit floating point type.
1544 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1545 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1546 source and destination to match the standard usage.
1547 (truncifkf2): Likewise.
1548 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1549 ISA 2.07 to use an explicit clobber, instead of passing in a
1551 (copysign<mode>3_soft): Likewise.
1553 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1555 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1556 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1557 (vect_slp_analyze_instance_dependence): Likewise.
1558 (vect_enhance_data_refs_alignment): Likewise.
1559 (vect_analyze_data_refs_alignment): Likewise.
1560 (vect_slp_analyze_and_verify_instance_alignment
1561 (vect_analyze_data_ref_accesses): Likewise.
1562 (vect_prune_runtime_alias_test_list): Likewise.
1563 (vect_analyze_data_refs): Likewise.
1564 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1565 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1566 (vect_analyze_scalar_cycles_1): Likewise.
1567 (vect_get_loop_niters): Likewise.
1568 (vect_analyze_loop_form_1): Likewise.
1569 (vect_update_vf_for_slp): Likewise.
1570 (vect_analyze_loop_operations): Likewise.
1571 (vect_analyze_loop): Likewise.
1572 (vectorizable_induction): Likewise.
1573 (vect_transform_loop): Likewise.
1574 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1575 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1576 (vect_make_slp_decision): Likewise.
1577 (vect_detect_hybrid_slp): Likewise.
1578 (vect_slp_analyze_operations): Likewise.
1579 (vect_slp_bb): Likewise.
1580 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1581 (vectorizable_bswap): Likewise.
1582 (vectorizable_call): Likewise.
1583 (vectorizable_simd_clone_call): Likewise.
1584 (vectorizable_conversion): Likewise.
1585 (vectorizable_assignment): Likewise.
1586 (vectorizable_shift): Likewise.
1587 (vectorizable_operation): Likewise.
1588 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1590 2018-06-18 Martin Sebor <msebor@redhat.com>
1592 PR tree-optimization/81384
1593 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1594 * builtins.c (expand_builtin_strnlen): New function.
1595 (expand_builtin): Call it.
1596 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1597 * builtins.def (BUILT_IN_STRNLEN): New.
1598 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1599 Warn for bounds in excess of maximum object size.
1600 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1601 single-value ranges. Handle strnlen.
1602 (handle_builtin_strlen): Handle strnlen.
1603 (strlen_check_and_optimize_stmt): Same.
1604 * doc/extend.texi (Other Builtins): Document strnlen.
1606 2018-06-18 Maya Rashish <coypu@sdf.org>
1608 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1609 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1610 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1612 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1614 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1616 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1618 * tree.c (escaped_string::escape): Replace cast to char * by
1619 const_cast<char *> (unescaped).
1621 2018-06-18 Nick Clifton <nickc@redhat.com>
1624 * tree.c (escaped_string): New class. Converts an unescaped
1625 string into its escaped equivalent.
1626 (warn_deprecated_use): Use the new class to convert the
1627 deprecation message, if present.
1628 (test_escaped_strings): New self test.
1629 (test_c_tests): Add test_escaped_strings.
1630 * doc/extend.texi (deprecated): Add a note that the
1631 deprecation message is affected by the -fmessage-length
1632 option, and that control characters will be escaped.
1633 (#pragma GCC error): Document this pragma.
1634 (#pragma GCC warning): Likewise.
1635 * doc/invoke.texi (-fmessage-length): Document this option's
1636 effect on the #warning and #error preprocessor directives and
1637 the deprecated attribute.
1639 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1641 * tree.c (decl_value_expr_lookup): Revert latest change.
1642 (decl_value_expr_insert): Likewise.
1644 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1646 * gimplify.c (nonlocal_vlas): Delete.
1647 (nonlocal_vla_vars): Likewise.
1648 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1650 (gimplify_body): Do not create and destroy nonlocal_vlas.
1651 * tree-nested.c: Include diagnostic.h.
1652 (use_pointer_in_frame): Tweak.
1653 (lookup_field_for_decl): Add assertion and declare the transformation.
1654 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1655 internal error when the reference is in a wrong context. Do not
1656 create a debug decl by default.
1657 (note_nonlocal_block_vlas): Delete.
1658 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1659 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1660 create a debug decl by default.
1661 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1662 call is in a wrong context.
1663 (fixup_vla_decls): New function.
1664 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1665 debug variables were created.
1666 * tree.c (decl_value_expr_lookup): Add checking assertion.
1667 (decl_value_expr_insert): Likewise.
1669 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1672 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1673 * tree-scalar-evolution.c (interpret_expr): Likewise.
1674 (expression_expensive_p): Likewise.
1675 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1676 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1677 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1678 (ssa_defined_by_minus_one_stmt_p): New.
1680 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1683 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1684 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1685 * dojump.c (do_jump): Likewise.
1686 * expr.c (expand_expr_real_2): Check operand type's sign.
1687 * fold-const.c (const_unop): Handle ABSU_EXPR.
1688 (fold_abs_const): Likewise.
1689 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1690 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1691 (strip_sign_op_1): Likesise.
1692 * match.pd: Add new pattern to generate ABSU_EXPR.
1693 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1694 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1695 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1696 * tree-inline.c (estimate_operator_cost): Likewise.
1697 * tree-pretty-print.c (dump_generic_node): Likewise.
1698 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1699 * tree.def (ABSU_EXPR): New.
1701 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1704 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1705 documented as preserved for backward compatibility only.
1706 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1708 PR rtl-optimization/86108
1709 * bb-reorder.c (create_forwarder_block): Renamed to ...
1710 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1711 jump from new landing pad to the second part.
1712 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1715 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1718 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1719 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1720 Only call store_expr for halves if the mode is the same.
1723 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1724 Fix up comment formatting.
1726 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1728 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1729 (typed_splay_tree::closure,
1730 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1731 (typed_splay_tree::typed_splay_tree,
1732 typed_splay_tree::operator =): Declared private.
1733 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1734 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1735 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1736 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1737 typed_splay_tree::splay_tree_splay,
1738 typed_splay_tree::splay_tree_foreach_helper,
1739 typed_splay_tree::splay_tree_insert,
1740 typed_splay_tree::splay_tree_remove,
1741 typed_splay_tree::splay_tree_lookup,
1742 typed_splay_tree::splay_tree_predecessor,
1743 typed_splay_tree::splay_tree_successor,
1744 typed_splay_tree::splay_tree_min,
1745 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1746 (typed_splay_tree::root, typed_splay_tree::comp,
1747 typed_splay_tree::delete_key,
1748 typed_splay_tree::delete_value): New data members.
1749 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1750 typed_splay_tree::remove.
1752 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1754 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1755 -mginv and -mno-ginv to the assembler.
1756 * config/mips/mips.opt (-mcrc): New option.
1758 * doc/invoke.text (-mcrc): Document.
1761 2018-06-15 Nick Clifton <nickc@redhat.com>
1764 * tree.c (escaped_string): New class. Converts an unescaped
1765 string into its escaped equivalent.
1766 (warn_deprecated_use): Use the new class to convert the
1767 deprecation message, if present.
1768 (test_escaped_strings): New self test.
1769 (test_c_tests): Add test_escaped_strings.
1770 * doc/extend.texi (deprecated): Add a note that the
1771 deprecation message is affected by the -fmessage-length
1772 option, and that control characters will be escaped.
1773 (#pragma GCC error): Document this pragma.
1774 (#pragma GCC warning): Likewise.
1775 * doc/invoke.texi (-fmessage-length): Document this option's
1776 effect on the #warning and #error preprocessor directives and
1777 the deprecated attribute.
1779 2018-06-15 Richard Biener <rguenther@suse.de>
1781 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1782 here, also noting vector size used.
1783 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1784 size used in MSG_OPTIMIZED_LOCATIONS dump.
1785 (pass_slp_vectorize::execute): Adjust.
1787 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1790 * config/arc/arc.c (arc_return_address_register): Fix
1793 2018-06-15 Richard Biener <rguenther@suse.de>
1796 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1797 leave useless conversion stripping to force_gimple_operand_gsi.
1798 (gimplify_build2): Likewise.
1799 (gimplify_build1): Likewise.
1801 2018-06-15 Richard Biener <rguenther@suse.de>
1804 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1805 before adjusting their block.
1807 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1809 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1810 multilibs for *-*-rtems*.
1811 * config/riscv/t-rtems: New file.
1813 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1816 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1817 unsigned_type_for returns NULL.
1820 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1821 subregs of multi-word pseudos unless the float mode has word size.
1823 2018-06-14 Richard Biener <rguenther@suse.de>
1826 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1827 duplicate and harmful type_hash_canon.
1828 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1830 2018-06-14 Richard Biener <rguenther@suse.de>
1833 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1836 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1838 * config/rtems.h (STDINT_LONG32): Define.
1840 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1841 Prachi Godbole <prachi.godbole@imgtec.com>
1843 * config/mips/mips-cpus.def: Define P6600.
1844 * config/mips/mips-tables.opt: Regenerate.
1845 * config/mips/mips.c (mips_ucbranch_type): New enum.
1846 (mips_rtx_cost_data): Add support for P6600.
1847 (mips_issue_rate): Likewise.
1848 (mips_multipass_dfa_lookahead): Likewise.
1849 (mips_avoid_hazard): Likewise.
1850 (mips_reorg_process_insns): Likewise.
1851 (mips_classify_branch_p6600): New function.
1852 * config/mips/mips.h (TUNE_P6600): New define.
1853 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1854 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1855 * config/mips/mips.md: Include p6600.md.
1856 (processor): Add p6600.
1857 * config/mips/p6600.md: New file.
1858 * doc/invoke.texi: Add p6600 to supported architectures.
1860 2018-06-13 Martin Sebor <msebor@redhat.com>
1862 PR tree-optimization/86114
1863 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1865 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1867 2018-06-13 Richard Biener <rguenther@suse.de>
1869 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1870 Properly set vector type of the intermediate stmt.
1871 * tree-vect-stmts.c (vectorizable_operation): The destination
1872 var always has vectype_out type.
1874 2018-06-13 Jeff Law <law@redhat.com>
1876 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1877 integer 0 for argument to print_rtl_with_bb.
1878 (rl78_reorg): Likewise.
1880 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1882 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1883 from rtx to rtx_insn *.
1884 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1886 (add_sched_insns_for_speculation): Likewise for local "target",
1887 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1888 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1889 from rtx_insn ** to rtx_code_label **.
1890 (reorg_emit_nops): Likewise.
1891 (c6x_reorg): Likewise for local "call_labels".
1892 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1894 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1895 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1896 the loops over LABEL_REFS.
1897 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1899 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1900 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1901 (split_branches): Strengthen local "olabel" from rtx to
1902 rtx_insn *, adding a safe_as_a cast.
1903 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1905 (add_insn_after): Likewise for first two params.
1906 (add_insn_before): Likewise.
1907 (remove_insn): Likewise for param.
1908 (emit_pattern_before_noloc): Likewise for second and third params.
1909 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1910 (emit_call_insn_before_noloc): Likewise.
1911 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1913 (emit_barrier_before): Likewise.
1914 (emit_label_before): Strengthen "label" param from "rtx" to
1915 "rtx_code_label *". Strengthen "before" param from "rtx" to
1917 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1919 (emit_pattern_after_noloc): Likewise.
1920 (emit_insn_after_noloc): Likewise.
1921 (emit_jump_insn_after_noloc): Likewise.
1922 (emit_call_insn_after_noloc): Likewise.
1923 (emit_debug_insn_after_noloc): Likewise.
1924 (emit_barrier_after): Likewise.
1925 (emit_label_after): Likewise for both params.
1926 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1927 "loc" param from "int" to "location_t".
1928 (emit_insn_after_setloc): Likewise.
1929 (emit_jump_insn_after_setloc): Likewise.
1930 (emit_call_insn_after_setloc): Likewise.
1931 (emit_debug_insn_after_setloc): Likewise.
1932 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1933 "loc" param from "int" to "location_t".
1934 (emit_pattern_before): Convert NULL_RTX to NULL.
1935 (emit_insn_before_setloc): Convert "loc" param from "int" to
1937 (emit_jump_insn_before_setloc): Likewise.
1938 (emit_call_insn_before_setloc): Likewise.
1939 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1940 rtx_insn *. Convert "loc" param from "int" to "location_t".
1941 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1942 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1943 Convert 3rd param from "int" to "location_t".
1944 (emit_barrier_before, emit_barrier_after, next_real_insn):
1945 Strengthen param from rtx to rtx_insn *.
1946 (emit_label_before): Strengthen 1st param from "rtx" to
1947 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1949 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1950 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1951 Strengthen 2nd param from "rtx" to "rtx_insn *".
1952 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1953 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1954 Likewise. Convert 3rd param from "int" to "location_t".
1955 (emit_label_after): Strengthen 1st param from "rtx" to
1957 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1959 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1960 from "rtx" to "rtx_insn *".
1962 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1964 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1965 bodies streamed in with -Q.
1966 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1967 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1968 * gimple-streamer-out.c: Include gimple-pretty-print.h
1969 (output_bb): Dump stmts streamed.
1970 * lto-section-out.c: Include print-tree.h
1971 (lto_begin_section): Dump sections created.
1972 (lto_output_decl_index): Dump decl encoded.
1973 * lto-streamer-out.c: Include print-tree.h
1974 (create_output_block): Dump output block created.
1975 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1976 (output_function): Dump function output.
1977 (output_constructor): Dump constructor streamed.
1978 (write_global_stream): Output indexes encoded.
1979 (produce_asm_for_decls): Dump streams encoded.
1980 * lto-streamer.c (streamer_dump_file): New global var.
1981 * lto-streamer.h (streamer_dump_file): Declare.
1982 * passes.c (ipa_write_summaries): Initialize streamer dump.
1983 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1986 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1989 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1990 offsets for register save directives. Emit a second batch of save
1991 directives, if need be, when the function accesses prior frames.
1993 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1995 * config/arc/fpu.md (fmasf4): Force operand to register.
1996 (fnmasf4): Likewise.
1998 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2000 * config/arc/arc-protos.h (arc_pad_return): Remove.
2001 * config/arc/arc.c (machine_function): Remove force_short_suffix
2003 (arc_print_operand): Adjust printing of '&'.
2004 (arc_verify_short): Remove conditional printing of short suffix.
2005 (arc_final_prescan_insn): Remove reference to size_reason.
2006 (pad_return): New function.
2007 (arc_reorg): Call pad_return.
2008 (arc_pad_return): Remove.
2009 (arc_init_machine_status): Remove reference to force_short_suffix.
2010 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2011 (attr length): When attribute iscompact is true force to 2
2012 regardless; in the case of maybe check if we want to force the
2013 instruction to have 4 bytes length.
2014 (nopv): Change it to generate 4 byte long nop as well.
2015 (blockage): New pattern.
2016 (simple_return): Remove call to arc_pad_return.
2017 (p_return_i): Likewise.
2019 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2021 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2023 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2025 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2028 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2030 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2031 for ARC700 and ARCv2.
2033 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2036 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2037 operands[2] instead of operands[1].
2040 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2042 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2043 case, check whether the outer register overlaps an unallocatable
2044 register, not just whether it fits the required class.
2046 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2048 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2050 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2051 (memrefs_conflict_p): Likewise.
2052 (init_alias_analysis): Likewise.
2053 * cfgexpand.c (expand_debug_expr): Likewise.
2054 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2055 * cse.c (fold_rtx): Likewise.
2056 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2057 * expr.c (emit_block_move_hints): Likewise.
2058 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2059 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2060 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2061 (emit_group_store): Likewise.
2062 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2063 to read the PRE/POST_MODIFY increment.
2064 * calls.c (store_one_arg): Use strip_offset.
2065 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2067 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2069 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2070 (simplify_binary_operation_1): Extend CONST_INT handling to
2072 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2073 than a HOST_WIDE_INT.
2074 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2076 (adjust_mems, add_stores): Update accodingly.
2077 (vt_canonicalize_addr): Track polynomial offsets.
2078 (emit_note_insn_var_location): Likewise.
2079 (vt_add_function_parameter): Likewise.
2080 (vt_initialize): Likewise.
2082 2018-06-12 Jeff Law <law@redhat.com>
2084 * config.gcc (alpha*-*-freebsd*): Remove.
2085 * config/alpha/freebsd.h: Remove.
2087 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2090 * spellcheck-tree.c (levenshtein_distance): Rename to...
2091 (get_edit_distance): ...this, and update for underlying renaming.
2092 * spellcheck-tree.h (levenshtein_distance): Rename to...
2093 (get_edit_distance): ...this.
2094 * spellcheck.c (levenshtein_distance): Rename to...
2095 (get_edit_distance): ...this. Convert from Levenshtein distance
2096 to Damerau-Levenshtein distance by supporting transpositions of
2097 adjacent characters. Rename "v1" to "v_next" and "v0" to
2099 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2100 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2101 update for underlying renaming.
2102 (selftest::levenshtein_distance_unit_test): Rename to...
2103 (selftest::test_get_edit_distance_unit): ...this, and update for
2104 underlying renaming.
2105 (selftest::test_find_closest_string): Add example from PR 69968
2106 where transposition helps
2107 (selftest::test_metric_conditions): Update for renaming.
2108 (selftest::test_metric_conditions): Likewise.
2109 (selftest::spellcheck_c_tests): Likewise.
2110 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2111 (get_edit_distance): ...this.
2112 (best_match::consider): Update for renaming.
2114 2018-06-12 Martin Sebor <msebor@redhat.com>
2116 PR tree-optimization/85259
2117 * builtins.c (compute_objsize): Handle constant offsets.
2118 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2119 true iff a warning has been issued.
2120 * gimple.h (gimple_nonartificial_location): New function.
2121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2122 gimple_nonartificial_location and handle -Wno-system-headers.
2123 (handle_builtin_stxncpy): Same.
2125 2018-06-12 Martin Sebor <msebor@redhat.com>
2128 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2130 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2132 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2133 BUILTIN_VEC_XST entries for pointer to double and long long.
2135 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2138 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2140 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2143 2018-06-12 Martin Liska <mliska@suse.cz>
2145 * doc/options.texi: Document IntegerRange.
2147 2018-06-12 Martin Liska <mliska@suse.cz>
2149 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2150 * opt-functions.awk: Handle Deprecated flag.
2151 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2153 (read_cmdline_option): Report warning for a deprecated option.
2154 * opts.h (struct cl_option): Add new field cl_deprecated.
2155 (CL_ERR_DEPRECATED): New.
2157 2018-06-12 Martin Liska <mliska@suse.cz>
2159 * doc/options.texi: Document Deprecated option flag.
2161 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2163 * config/arc/arc-arch.h (arc_extras): New enum.
2164 (arc_cpu_t):Add field extra.
2165 (arc_cpu_types): Consider the extras.
2166 * config/arc/arc-cpus.def: Add extras info.
2167 * config/arc/arc-opts.h (processor_type): Consider extra field.
2168 * config/arc/arc.c (arc_override_options): Handle extra field.
2170 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2172 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2173 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2175 * config/arc/arc.c (arc_init): Update pic warning.
2176 (irq_range): Update irq range parsing warnings.
2177 (arc_override_options): Update various warning messages.
2178 (arc_handle_aux_attribute): Likewise.
2180 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2182 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2184 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2186 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2187 compiler messages directives.
2189 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2191 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2192 * config/mips/mips-tables.opt: Regenerate.
2193 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2195 * doc/invoke.texi: Document -march=i6500.
2197 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2199 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2200 (i6400_gpmul): Add cpu_unit.
2201 (i6400_gpdiv): Likewise.
2202 (i6400_msa_add_d): Update reservations.
2203 (i6400_msa_int_add) Likewise.
2204 (i6400_msa_short_logic3) Likewise.
2205 (i6400_msa_short_logic2) Likewise.
2206 (i6400_msa_short_logic) Likewise.
2207 (i6400_msa_move) Likewise.
2208 (i6400_msa_cmp) Likewise.
2209 (i6400_msa_short_float2) Likewise.
2210 (i6400_msa_div_d) Likewise.
2211 (i6400_msa_long_logic1) Likewise.
2212 (i6400_msa_long_logic2) Likewise.
2213 (i6400_msa_mult) Likewise.
2214 (i6400_msa_long_float2) Likewise.
2215 (i6400_msa_long_float4) Likewise.
2216 (i6400_msa_long_float5) Likewise.
2217 (i6400_msa_long_float8) Likewise.
2218 (i6400_fpu_fadd): Include frint type.
2219 (i6400_fpu_store): New define_insn_reservation.
2220 (i6400_fpu_load): Likewise.
2221 (i6400_fpu_move): Likewise.
2222 (i6400_fpu_fcmp): Likewise.
2223 (i6400_fpu_fmadd): Likewise.
2224 (i6400_int_mult): Include imul3nc type and update reservation.
2225 (i6400_int_div): Include idiv3 type and update reservation.
2226 (i6400_int_load): Update to check type not move_type.
2227 (i6400_int_store): Likewise.
2228 (i6400_int_prefetch): Set zero latency.
2230 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2232 * gcc.c: Document new %@{...} sequence.
2233 (LINK_COMMAND_SPEC): Use it for the -L switches.
2234 (cpp_unique_options): Use it for the -I switches.
2235 (at_file_argbuf): New global variable.
2236 (in_at_file): Likewise.
2237 (alloc_args): Create at_file_argbuf.
2238 (clear_args): Truncate at_file_argbuf.
2239 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2240 (open_at_file): New function.
2241 (close_at_file): Likewise.
2242 (create_at_file): Delete.
2243 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2246 (validate_switches_from_spec): Deal with %@{...} sequence.
2247 (validate_switches): Likewise.
2248 (driver::finalize): Call clear_args.
2250 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2252 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2254 2018-06-11 Martin Sebor <msebor@redhat.com>
2256 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2257 (Wno-class-memaccess): Add @opindex.
2258 (Wno-templates, Wno-multiple-inheritance): Same.
2259 (Wno-virtual-inheritance, Wno-namespaces): Same.
2260 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2261 (Wno-format-overflow, Wno-format-truncation): Same.
2262 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2263 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2264 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2265 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2266 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2267 (Wno-misspelled-isr): Same.
2269 2018-06-11 Martin Sebor <msebor@redhat.com>
2271 * PR tree-optimization/86083
2272 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2274 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2276 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2278 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2281 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2282 on the correct operand.
2283 (*movdi_internal64): Ditto.
2285 2018-06-11 Martin Liska <mliska@suse.cz>
2287 PR tree-optimization/86089
2288 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2290 2018-06-11 Julia Koval <julia.koval@intel.com>
2292 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2293 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2294 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2296 2018-06-11 Olivier Hainque <hainque@adacore.com>
2298 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2299 for Ada with strict dwarf2.
2301 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2304 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2307 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2309 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2311 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2313 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2316 2018-06-08 Martin Liska <mliska@suse.cz>
2318 * tree-cfg.h (debug_function): Fix argument type to match
2321 2018-06-08 Martin Liska <mliska@suse.cz>
2323 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2324 Remove usage of MPX-related (and removed) fields.
2325 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2327 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2329 * cfg.c (debug): Use TDF_NONE rather than 0.
2330 * cfghooks.c (debug): Likewise.
2331 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2332 (struct dump_option_value_info): Convert to...
2333 (struct kv_pair): ...this template type.
2334 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2336 (optinfo_verbosity_options): Likewise.
2337 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2339 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2340 than int for "optgroup_flags" param.
2341 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2343 (dump_dec): Likewise.
2344 (dump_finish): Use TDF_NONE rather than 0.
2345 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2346 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2347 than 0. Update for change to option_ptr.
2348 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2349 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2350 0. Update for changes to optinfo_verbosity_options and
2352 (opt_info_switch_p): Convert optgroup_flags from int to
2354 (dump_basic_block): Use dump_flags_t rather than int
2355 for "dump_kind" param.
2356 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2357 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2358 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2359 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2360 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2361 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2362 TDF_NONE): Convert from macros to...
2363 (enum dump_flag): ...this new enum.
2364 (dump_flags_t): Update to use enum.
2365 (operator|, operator&, operator~, operator|=, operator&=):
2366 Implement for dump_flags_t.
2367 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2368 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2369 Convert from macros to...
2370 (enum optgroup_flag): ...this new enum.
2371 (optgroup_flags_t): New typedef.
2372 (operator|, operator|=): Implement for optgroup_flags_t.
2373 (struct dump_file_info): Convert field "alt_flags" to
2374 dump_flags_t. Convert field "optgroup_flags" to
2376 (dump_basic_block): Use dump_flags_t rather than int for param.
2377 (dump_generic_expr_loc): Likewise.
2378 (dump_dec): Likewise.
2379 (dump_register): Convert param "optgroup_flags" to
2381 (opt_info_enable_passes): Likewise.
2382 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2384 * gimple-pretty-print.c (debug): Likewise.
2385 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2386 (merged_store_group::apply_stores): Likewise.
2387 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2388 * gimple.c (verify_gimple_pp): Likewise.
2389 * graphite-poly.c (print_pbb_body): Likewise.
2390 * passes.c (pass_manager::register_one_dump_file): Convert
2391 local "optgroup_flags" to optgroup_flags_t.
2392 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2394 (debug_body): Likewise.
2395 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2396 to optgroup_flags_t.
2397 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2399 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2400 (convert_mult_to_fma): Likewise.
2401 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2402 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2403 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2404 "dump_kind" to dump_flags_t.
2406 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2408 * config/rs6000/rs6000.c (min, max): Delete.
2410 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2412 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2413 -mabi=spe and -mabi=no-spe.
2415 2018-06-08 Martin Liska <mliska@suse.cz>
2417 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2418 where we expect an existing summary.
2420 2018-06-08 Martin Liska <mliska@suse.cz>
2422 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2423 * ipa-inline.h (estimate_edge_growth): Likewise.
2425 2018-06-08 Martin Liska <mliska@suse.cz>
2427 * cgraph.c (function_version_hasher::hash): Use
2428 cgraph_node::get_uid ().
2429 (function_version_hasher::equal):
2430 * cgraph.h (cgraph_node::get_uid): New method.
2431 * ipa-inline.c (update_caller_keys): Use
2432 cgraph_node::get_uid ().
2433 (update_callee_keys): Likewise.
2434 * ipa-utils.c (searchc): Likewise.
2435 (ipa_reduced_postorder): Likewise.
2436 * lto-cgraph.c (input_node): Likewise.
2437 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2438 * symbol-summary.h (symtab_insertion): Likewise.
2439 (symtab_removal): Likewise.
2440 (symtab_duplication): Likewise.
2441 * tree-pretty-print.c (dump_function_header): Likewise.
2442 * tree-sra.c (convert_callers_for_node): Likewise.
2444 2018-06-08 Martin Liska <mliska@suse.cz>
2446 * cgraph.c (symbol_table::create_edge): Always assign a new
2448 (symbol_table::free_edge): Do not recycle numbers.
2449 * cgraph.h (cgraph_edge::get): New method.
2450 * symbol-summary.h (symtab_removal): Use it.
2451 (symtab_duplication): Likewise.
2452 (call_summary::hashable_uid): Remove.
2454 2018-06-08 Martin Liska <mliska@suse.cz>
2456 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2457 (initialize_growth_caches): Remove.
2458 (free_growth_caches): Likewise.
2459 (do_estimate_edge_time): Use edge_growth_cache.
2460 (do_estimate_edge_size): Likewise.
2461 (do_estimate_edge_hints): Likewise.
2462 * ipa-inline.c (reset_edge_caches): Likewise.
2463 (recursive_inlining): Likewise.
2464 (inline_small_functions): Likewise.
2465 * ipa-inline.h (initialize_growth_caches): Remove.
2466 (estimate_edge_size): Likewise.
2467 (estimate_edge_time): Likewise.
2468 (estimate_edge_hints): Likewise.
2469 (reset_edge_growth_cache): Likewise.
2470 * symbol-summary.h (call_summary::remove): New method.
2472 2018-06-08 Martin Liska <mliska@suse.cz>
2474 * ipa-cp.c (class edge_clone_summary): New summary.
2475 (grow_edge_clone_vectors): Remove.
2476 (ipcp_edge_duplication_hook): Remove.
2477 (class edge_clone_summary_t): New call_summary class.
2478 (ipcp_edge_removal_hook): Remove.
2479 (edge_clone_summary_t::duplicate): New function.
2480 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2481 (create_specialized_node): Likewise.
2482 (ipcp_driver): Initialize edge_clone_summaries and do not
2485 2018-06-08 Martin Liska <mliska@suse.cz>
2487 * symbol-summary.h (get): New function.
2488 (call_summary::m_initialize_when_cloning): New class member.
2490 2018-06-08 Martin Liska <mliska@suse.cz>
2492 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2493 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2494 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2495 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2496 (remove_cgraph_node_from_order): Use the removed_nodes set.
2497 (do_per_function_toporder): Likwise.
2498 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2499 instead of summary_uid.
2500 (symtab_removal): Likewise.
2501 (symtab_duplication): Likewise.
2503 2018-06-08 Martin Liska <mliska@suse.cz>
2505 * ipa-cp.c (ipcp_store_bits_results): Use
2506 ipcp_transformation_sum.
2507 (ipcp_store_vr_results): Likewise.
2508 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2510 (ipcp_transformation_initialize): ... this.
2511 (ipa_set_node_agg_value_chain):
2512 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2513 (write_ipcp_transformation_info): Likewise.
2514 (read_ipcp_transformation_info): Likewise.
2515 (ipcp_update_bits): Likewise.
2516 (ipcp_update_vr): Likewise.
2517 (ipcp_transform_function): Likewise.
2518 * ipa-prop.h: Rename ipcp_transformation_summary to
2519 ipcp_transformation.
2520 (class ipcp_transformation_t): New function summary.
2521 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2522 (ipa_get_agg_replacements_for_node): Likewise.
2524 2018-06-08 Martin Liska <mliska@suse.cz>
2526 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2528 (class funct_state_summary_t): New function_summary class.
2529 (has_function_state): Remove.
2530 (get_function_state): Likewise.
2531 (set_function_state): Likewise.
2532 (add_new_function): Likewise.
2533 (funct_state_summary_t::insert): New function.
2534 (duplicate_node_data): Remove.
2535 (remove_node_data): Remove.
2536 (funct_state_summary_t::duplicate): New function.
2537 (register_hooks): Create new funct_state_summaries.
2538 (pure_const_generate_summary): Use it.
2539 (pure_const_write_summary): Likewise.
2540 (pure_const_read_summary): Likewise.
2541 (propagate_pure_const): Likewise.
2542 (propagate_nothrow): Likewise.
2543 (dump_malloc_lattice): Likewise.
2544 (propagate_malloc): Likewise.
2545 (execute): Do not register hooks, just remove summary
2547 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2550 2018-06-08 Martin Liska <mliska@suse.cz>
2552 * ipa-reference.c (remove_node_data): Remove.
2553 (duplicate_node_data): Likewise.
2554 (class ipa_ref_var_info_summary_t): New class.
2555 (class ipa_ref_opt_summary_t): Likewise.
2556 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2557 (get_reference_optimization_summary): Use
2558 ipa_ref_opt_sum_summaries.
2559 (set_reference_vars_info): Remove.
2560 (set_reference_optimization_summary): Likewise.
2561 (ipa_init): Create summaries.
2562 (init_function_info): Use function summary.
2563 (ipa_ref_opt_summary_t::duplicate): New function.
2564 (ipa_ref_opt_summary_t::remove): New function.
2565 (get_read_write_all_from_node): Fix GNU coding style.
2566 (propagate): Use function summary.
2567 (write_node_summary_p): Fix GNU coding style.
2568 (stream_out_bitmap): Likewise.
2569 (ipa_reference_read_optimization_summary): Use function summary.
2570 (ipa_reference_c_finalize): Do not release hooks.
2572 2018-06-08 Martin Liska <mliska@suse.cz>
2574 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2575 (analyze_function_body): Extract multiple calls of get_create.
2576 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2577 * ipa-inline.c (recursive_inlining): Use ::get method.
2578 * ipa-inline.h (estimate_edge_growth): Likewise.
2580 2018-06-08 Martin Liska <mliska@suse.cz>
2582 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2584 (hsa_function_summary::hsa_function_summary): Use the new enum
2586 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2587 * hsa-gen.c (hsa_get_host_function): Likewise.
2588 (get_brig_function_name): Likewise.
2589 * ipa-hsa.c (process_hsa_functions): Likewise.
2590 (ipa_hsa_write_summary): Likewise.
2591 * symbol-summary.h (symtab_duplication): Use ::get function/
2592 (get): New function.
2594 2018-06-08 Martin Liska <mliska@suse.cz>
2596 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2598 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2599 (hsa_register_kernel): Likewise.
2600 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2601 * hsa-gen.c (hsa_get_host_function): Likewise.
2602 (get_brig_function_name): Likewise.
2603 (generate_hsa): Likewise.
2604 (pass_gen_hsail::execute): Likewise.
2605 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2606 (devirtualization_time_bonus): Likewise.
2607 (ipcp_propagate_stage): Likewise.
2608 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2609 (edge_set_predicate): Likewise.
2610 (evaluate_conditions_for_known_args): Likewise.
2611 (evaluate_properties_for_edge): Likewise.
2612 (ipa_fn_summary::reset): Likewise.
2613 (ipa_fn_summary_t::duplicate): Likewise.
2614 (dump_ipa_call_summary): Likewise.
2615 (ipa_dump_fn_summary): Likewise.
2616 (analyze_function_body): Likewise.
2617 (compute_fn_summary): Likewise.
2618 (estimate_edge_devirt_benefit): Likewise.
2619 (estimate_edge_size_and_time): Likewise.
2620 (estimate_calls_size_and_time): Likewise.
2621 (estimate_node_size_and_time): Likewise.
2622 (inline_update_callee_summaries): Likewise.
2623 (remap_edge_change_prob): Likewise.
2624 (remap_edge_summaries): Likewise.
2625 (ipa_merge_fn_summary_after_inlining): Likewise.
2626 (ipa_update_overall_fn_summary): Likewise.
2627 (read_ipa_call_summary): Likewise.
2628 (inline_read_section): Likewise.
2629 (write_ipa_call_summary): Likewise.
2630 (ipa_fn_summary_write): Likewise.
2631 (ipa_free_fn_summary): Likewise.
2632 * ipa-hsa.c (process_hsa_functions): Likewise.
2633 (ipa_hsa_write_summary): Likewise.
2634 (ipa_hsa_read_section): Likewise.
2635 * ipa-icf.c (sem_function::merge): Likewise.
2636 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2637 (do_estimate_edge_time): Likewise.
2638 (estimate_size_after_inlining): Likewise.
2639 (estimate_growth): Likewise.
2640 (growth_likely_positive): Likewise.
2641 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2642 (inline_call): Likewise.
2643 * ipa-inline.c (caller_growth_limits): Likewise.
2644 (can_inline_edge_p): Likewise.
2645 (can_inline_edge_by_limits_p): Likewise.
2646 (compute_uninlined_call_time): Likewise.
2647 (compute_inlined_call_time): Likewise.
2648 (want_inline_small_function_p): Likewise.
2649 (edge_badness): Likewise.
2650 (update_caller_keys): Likewise.
2651 (update_callee_keys): Likewise.
2652 (recursive_inlining): Likewise.
2653 (inline_small_functions): Likewise.
2654 (inline_to_all_callers_1): Likewise.
2655 (dump_overall_stats): Likewise.
2656 (early_inline_small_functions): Likewise.
2657 (early_inliner): Likewise.
2658 * ipa-inline.h (estimate_edge_growth): Likewise.
2659 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2660 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2661 * ipa-prop.h (IPA_NODE_REF): Likewise.
2662 (IPA_EDGE_REF): Likewise.
2663 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2664 (propagate_malloc): Likewise.
2665 * ipa-split.c (execute_split_functions): Likewise.
2666 * symbol-summary.h: Rename get to get_create.
2668 (get_create): Likewise.
2669 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2671 2018-06-08 Martin Liska <mliska@suse.cz>
2673 * symbol-summary.h (release): Move definition out of class
2675 (symtab_removal): Likewise.
2676 (symtab_duplication): Likewise.
2678 2018-06-08 Martin Liska <mliska@suse.cz>
2680 * symbol-summary.h (function_summary): Move constructor
2681 implementation out of class declaration.
2682 (release): Likewise.
2683 (symtab_insertion): Likewise.
2684 (symtab_removal): Likewise.
2685 (symtab_duplication): Likewise.
2688 2018-06-08 Martin Liska <mliska@suse.cz>
2690 * Makefile.in: Remove support for MPX (macros, related functions,
2691 fields in cgraph_node, ...).
2692 * builtin-types.def (BT_BND): Likewise.
2693 (BT_FN_BND_CONST_PTR): Likewise.
2694 (BT_FN_CONST_PTR_BND): Likewise.
2695 (BT_FN_VOID_PTR_BND): Likewise.
2696 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2697 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2698 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2699 (expand_builtin_mempcpy_with_bounds): Likewise.
2700 (expand_builtin_memset_with_bounds): Likewise.
2701 (expand_builtin_memset_args): Likewise.
2702 (std_expand_builtin_va_start): Likewise.
2703 (expand_builtin): Likewise.
2704 (expand_builtin_with_bounds): Likewise.
2705 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2706 (DEF_LIB_BUILTIN_CHKP): Likewise.
2707 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2708 (DEF_CHKP_BUILTIN): Likewise.
2709 (BUILT_IN_MEMCPY): Likewise.
2710 (BUILT_IN_MEMMOVE): Likewise.
2711 (BUILT_IN_MEMPCPY): Likewise.
2712 (BUILT_IN_MEMSET): Likewise.
2713 (BUILT_IN_STPCPY): Likewise.
2714 (BUILT_IN_STRCAT): Likewise.
2715 (BUILT_IN_STRCHR): Likewise.
2716 (BUILT_IN_STRCPY): Likewise.
2717 (BUILT_IN_STRLEN): Likewise.
2718 (BUILT_IN_MEMCPY_CHK): Likewise.
2719 (BUILT_IN_MEMMOVE_CHK): Likewise.
2720 (BUILT_IN_MEMPCPY_CHK): Likewise.
2721 (BUILT_IN_MEMSET_CHK): Likewise.
2722 (BUILT_IN_STPCPY_CHK): Likewise.
2723 (BUILT_IN_STRCAT_CHK): Likewise.
2724 (BUILT_IN_STRCPY_CHK): Likewise.
2725 * calls.c (store_bounds): Likewise.
2726 (emit_call_1): Likewise.
2727 (special_function_p): Likewise.
2728 (maybe_warn_nonstring_arg): Likewise.
2729 (initialize_argument_information): Likewise.
2730 (finalize_must_preallocate): Likewise.
2731 (compute_argument_addresses): Likewise.
2732 (expand_call): Likewise.
2733 * cfgexpand.c (expand_call_stmt): Likewise.
2734 (expand_return): Likewise.
2735 (expand_gimple_stmt_1): Likewise.
2736 (pass_expand::execute): Likewise.
2737 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2738 (cgraph_node::remove): Likewise.
2739 (cgraph_node::dump): Likewise.
2740 (cgraph_node::verify_node): Likewise.
2741 * cgraph.h (chkp_function_instrumented_p): Likewise.
2742 (symtab_node::get_alias_target): Likewise.
2743 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2744 (cgraph_local_p): Likewise.
2745 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2746 (cgraph_edge::rebuild_references): Likewise.
2747 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2748 (walk_polymorphic_call_targets): Likewise.
2749 (cgraph_node::expand_thunk): Likewise.
2750 (symbol_table::output_weakrefs): Likewise.
2751 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2752 (ix86_handle_option): Likewise.
2753 * config/i386/constraints.md: Likewise.
2754 * config/i386/i386-builtin-types.def (BND): Likewise.
2758 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2759 (BDESC_FIRST): Likewise.
2761 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2762 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2763 * config/i386/i386.c (enum reg_class): Likewise.
2764 (ix86_target_string): Likewise.
2765 (ix86_option_override_internal): Likewise.
2766 (ix86_conditional_register_usage): Likewise.
2767 (ix86_valid_target_attribute_inner_p): Likewise.
2768 (ix86_set_indirect_branch_type): Likewise.
2769 (ix86_set_current_function): Likewise.
2770 (ix86_function_arg_regno_p): Likewise.
2771 (init_cumulative_args): Likewise.
2772 (ix86_function_arg_advance): Likewise.
2773 (ix86_function_arg): Likewise.
2774 (ix86_pass_by_reference): Likewise.
2775 (ix86_function_value_regno_p): Likewise.
2776 (ix86_function_value_1): Likewise.
2777 (ix86_function_value_bounds): Likewise.
2778 (ix86_return_in_memory): Likewise.
2779 (ix86_setup_incoming_vararg_bounds): Likewise.
2780 (ix86_va_start): Likewise.
2781 (indirect_thunk_need_prefix): Likewise.
2782 (print_reg): Likewise.
2783 (ix86_print_operand): Likewise.
2784 (ix86_expand_call): Likewise.
2785 (ix86_output_function_return): Likewise.
2786 (reg_encoded_number): Likewise.
2787 (BDESC_VERIFYS): Likewise.
2788 (ix86_init_mpx_builtins): Likewise.
2789 (ix86_init_builtins): Likewise.
2790 (ix86_emit_cmove): Likewise.
2791 (ix86_emit_move_max): Likewise.
2792 (ix86_expand_builtin): Likewise.
2793 (ix86_builtin_mpx_function): Likewise.
2794 (ix86_get_arg_address_for_bt): Likewise.
2795 (ix86_load_bounds): Likewise.
2796 (ix86_store_bounds): Likewise.
2797 (ix86_load_returned_bounds): Likewise.
2798 (ix86_store_returned_bounds): Likewise.
2799 (ix86_class_likely_spilled_p): Likewise.
2800 (ix86_hard_regno_mode_ok): Likewise.
2801 (x86_order_regs_for_local_alloc): Likewise.
2802 (ix86_mitigate_rop): Likewise.
2803 (ix86_bnd_prefixed_insn_p): Likewise.
2804 (ix86_mpx_bound_mode): Likewise.
2805 (ix86_make_bounds_constant): Likewise.
2806 (ix86_initialize_bounds): Likewise.
2807 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2808 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2809 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2810 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2811 (TARGET_CHKP_BOUND_MODE): Likewise.
2812 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2813 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2814 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2815 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2816 * config/i386/i386.h (TARGET_MPX): Likewise.
2817 (TARGET_MPX_P): Likewise.
2818 (VALID_BND_REG_MODE): Likewise.
2819 (FIRST_BND_REG): Likewise.
2820 (LAST_BND_REG): Likewise.
2821 (enum reg_class): Likewise.
2822 (BND_REG_P): Likewise.
2823 (BND_REGNO_P): Likewise.
2824 (BNDmode): Likewise.
2825 (ADJUST_INSN_LENGTH): Likewise.
2826 * config/i386/i386.md: Likewise.
2827 * config/i386/i386.opt: Likewise.
2828 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2829 (defined): Likewise.
2830 (LINK_MPX): Likewise.
2831 (MPX_SPEC): Likewise.
2832 (LIBMPX_SPEC): Likewise.
2833 (LIBMPXWRAPPERS_SPEC): Likewise.
2834 (CHKP_SPEC): Likewise.
2835 * config/i386/predicates.md: Likewise.
2836 * dbxout.c (dbxout_type): Likewise.
2837 * doc/extend.texi: Likewise.
2838 * doc/invoke.texi: Likewise.
2839 * doc/md.texi: Likewise.
2840 * doc/tm.texi: Likewise.
2841 * doc/tm.texi.in: Likewise.
2842 * dwarf2out.c (is_base_type): Likewise.
2843 (gen_formal_types_die): Likewise.
2844 (gen_subprogram_die): Likewise.
2845 (gen_type_die_with_usage): Likewise.
2846 (gen_decl_die): Likewise.
2847 (dwarf2out_late_global_decl): Likewise.
2848 * expr.c (expand_assignment): Likewise.
2849 (emit_storent_insn): Likewise.
2850 (store_expr_with_bounds): Likewise.
2851 (store_expr): Likewise.
2852 (expand_expr_real_1): Likewise.
2853 * expr.h (store_expr_with_bounds): Likewise.
2854 * function.c (use_register_for_decl): Likewise.
2855 (struct bounds_parm_data): Likewise.
2856 (assign_parms_augmented_arg_list): Likewise.
2857 (assign_parm_find_entry_rtl): Likewise.
2858 (assign_parm_is_stack_parm): Likewise.
2859 (assign_parm_load_bounds): Likewise.
2860 (assign_bounds): Likewise.
2861 (assign_parms): Likewise.
2862 (expand_function_start): Likewise.
2863 * gcc.c (CHKP_SPEC): Likewise.
2864 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2865 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2866 (wrestrict_dom_walker::check_call): Likewise.
2867 * gimple.c (gimple_build_call_from_tree): Likewise.
2868 * gimple.h (enum gf_mask): Likewise.
2869 (gimple_call_with_bounds_p): Likewise.
2870 (gimple_call_set_with_bounds): Likewise.
2871 * gimplify.c (gimplify_init_constructor): Likewise.
2872 * ipa-cp.c (initialize_node_lattices): Likewise.
2873 (propagate_constants_across_call): Likewise.
2874 (find_more_scalar_values_for_callers_subset): Likewise.
2875 * ipa-hsa.c (process_hsa_functions): Likewise.
2876 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2877 * ipa-icf.c (sem_function::merge): Likewise.
2878 * ipa-inline.c (early_inliner): Likewise.
2879 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2880 (warn_function_cold): Likewise.
2881 (propagate_pure_const): Likewise.
2882 * ipa-ref.h (enum GTY): Likewise.
2883 * ipa-split.c (find_retbnd): Likewise.
2884 (consider_split): Likewise.
2885 (split_function): Likewise.
2886 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2887 * ipa.c (walk_polymorphic_call_targets): Likewise.
2888 (symbol_table::remove_unreachable_nodes): Likewise.
2889 (process_references): Likewise.
2890 (cgraph_build_static_cdtor_1): Likewise.
2891 * lto-cgraph.c (lto_output_node): Likewise.
2892 (output_refs): Likewise.
2893 (compute_ltrans_boundary): Likewise.
2894 (input_overwrite_node): Likewise.
2895 (input_node): Likewise.
2896 (input_cgraph_1): Likewise.
2897 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2898 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2899 (class pass_chkp_instrumentation_passes): Likewise.
2900 (make_pass_chkp_instrumentation_passes): Likewise.
2901 * passes.def: Likewise.
2902 * rtl.h (struct GTY): Likewise.
2903 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2904 * stor-layout.c (layout_type): Likewise.
2905 * symtab.c: Likewise.
2906 * target.def: Likewise.
2907 * targhooks.c (default_chkp_bound_type): Likewise.
2908 (default_chkp_bound_mode): Likewise.
2909 (default_builtin_chkp_function): Likewise.
2910 (default_chkp_function_value_bounds): Likewise.
2911 (default_chkp_make_bounds_constant): Likewise.
2912 (default_chkp_initialize_bounds): Likewise.
2913 * targhooks.h (default_chkp_bound_type): Likewise.
2914 (default_chkp_bound_mode): Likewise.
2915 (default_builtin_chkp_function): Likewise.
2916 (default_chkp_function_value_bounds): Likewise.
2917 (default_chkp_make_bounds_constant): Likewise.
2918 (default_chkp_initialize_bounds): Likewise.
2919 * toplev.c (compile_file): Likewise.
2920 (process_options): Likewise.
2921 * tree-core.h (DEF_BUILTIN): Likewise.
2922 (DEF_BUILTIN_CHKP): Likewise.
2923 * tree-inline.c (declare_return_variable): Likewise.
2924 (remap_gimple_stmt): Likewise.
2925 (copy_bb): Likewise.
2926 (initialize_inlined_parameters): Likewise.
2927 (expand_call_inline): Likewise.
2928 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2929 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2930 (make_pass_ipa_chkp_produce_thunks): Likewise.
2931 (make_pass_chkp): Likewise.
2932 (make_pass_chkp_opt): Likewise.
2933 (make_pass_chkp_instrumentation_passes): Likewise.
2934 * tree-pretty-print.c (dump_generic_node): Likewise.
2935 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2936 * tree-ssa-dce.c (propagate_necessity): Likewise.
2937 (eliminate_unnecessary_stmts): Likewise.
2938 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2939 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2940 * tree-ssa-sccvn.h: Likewise.
2941 * tree-ssa-strlen.c (get_string_length): Likewise.
2942 (valid_builtin_call): Likewise.
2943 (adjust_last_stmt): Likewise.
2944 (handle_builtin_strchr): Likewise.
2945 (handle_builtin_strcpy): Likewise.
2946 (handle_builtin_stxncpy): Likewise.
2947 (handle_builtin_memcpy): Likewise.
2948 (handle_builtin_strcat): Likewise.
2949 (strlen_check_and_optimize_stmt): Likewise.
2950 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2951 * tree-streamer-in.c: Likewise.
2952 * tree-streamer.c (record_common_node): Likewise.
2953 * tree.c (tree_code_size): Likewise.
2954 (wide_int_to_tree_1): Likewise.
2955 (type_contains_placeholder_1): Likewise.
2956 (build_common_tree_nodes): Likewise.
2957 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2958 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2959 (POINTER_BOUNDS_P): Likewise.
2960 (BOUNDED_TYPE_P): Likewise.
2961 (BOUNDED_P): Likewise.
2962 (CALL_WITH_BOUNDS_P): Likewise.
2963 (pointer_bounds_type_node): Likewise.
2964 * value-prof.c (gimple_ic): Likewise.
2965 * var-tracking.c (vt_add_function_parameters): Likewise.
2966 * varasm.c (make_decl_rtl): Likewise.
2967 (assemble_start_function): Likewise.
2968 (output_constant): Likewise.
2969 (maybe_assemble_visibility): Likewise.
2970 * varpool.c (ctor_for_folding): Likewise.
2971 * chkp-builtins.def: Remove.
2972 * ipa-chkp.c: Remove.
2973 * ipa-chkp.h: Remove.
2974 * rtl-chkp.c: Remove.
2975 * rtl-chkp.h: Remove.
2976 * tree-chkp-opt.c: Remove.
2977 * tree-chkp.c: Remove.
2978 * tree-chkp.h: Remove.
2980 2018-06-07 Carl Love <cel@us.ibm.com>
2982 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2983 vextract_fp_from_shortl): Add BE support.
2985 2018-06-07 Paul Koning <ni1d@arrl.net>
2987 * compare-elim.c (try_merge_compare): Don't merge compare if
2988 address contains a side effect.
2989 (try_eliminate_compare): Likewise.
2991 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2993 * config.gcc: Support "tremont".
2994 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2995 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2997 * config/i386/i386.c (m_TREMONT): Define.
2998 (processor_target_table): Add "tremont".
2999 (PTA_TREMONT): Define.
3000 (ix86_lea_outperforms): Add TARGET_TREMONT.
3001 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3002 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3003 and M_INTEL_GOLDMONT_PLUS.
3004 (fold_builtin_cpu): Add "tremont".
3005 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3006 (ix86_option_override_internal): Add "tremont".
3007 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3008 (processor_type): Add PROCESSOR_TREMONT.
3009 * config/i386/x86-tune.def: Add m_TREMONT.
3010 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3012 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3014 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3015 symbol defined for msp430i* devices to be lower case.
3017 gcc/testsuite/gcc.target/msp430/
3018 * msp430i-device-symbol.c: New test.
3019 * msp430f-device-symbol.c: New test.
3020 * msp430.h: New test header file.
3022 2018-06-07 Richard Biener <rguenther@suse.de>
3024 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3025 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3026 Properly wrap signed arithmetic if overflow wraps.
3028 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3030 PR tree-optimization/69615
3031 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3032 of a cast from a same precision integral SSA_NAME in a bb dominated
3033 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3034 cast to utype if rhs2 has already a compatible type.
3036 2018-06-07 Richard Biener <rguenther@suse.de>
3038 PR tree-optimization/85935
3039 * graphite-scop-detection.c (find_params_in_bb): Analyze
3040 condition operands with respect to the correct loop. Assert
3041 the analysis doesn't fail.
3043 2018-06-04 Carl Love <cel@us.ibm.com>
3045 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3046 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3047 as it is slightly cheaper.
3048 (first_match_or_eos_index_<mode>):
3049 Calculate index using natural element order.
3050 (first_match_index_<mode>):
3051 Calculate index using natural element order.
3052 (first_match_or_eos_index_<mode>):
3053 Calculate index using natural order.
3054 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3055 for BE and LE modes.
3056 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3057 P9V_BUILTIN_VCLZLSBB_V16QI.
3058 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3061 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3063 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3064 indentation and line wrap for many prototypes. Add missing
3065 @smallexample directives around block of prototypes for vec_xl and
3068 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3070 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3071 track if we pass or return IEEE 128-bit floating point.
3072 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3073 C++ mangling that is compatible with GCC 8.1.
3074 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3075 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3076 floating point types.
3077 (rs6000_function_arg_advance_1): Likewise.
3078 (rs6000_mangle_type): Optionally generate mangled names that match
3079 what GCC 8.1 generated for IEEE 128-bit floating point types.
3080 (rs6000_globalize_decl_name): If we have an external function that
3081 passes or returns IEEE 128-bit types, generate a weak reference
3082 from the mangled name used in GCC 8.1 to the current mangled
3084 (rs6000_init_builtins): Make __ibm128 use the long double type if
3085 long double is IBM extended double. Make __float128 use the long
3086 double type if long double is IEEE 128-bit.
3089 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3090 macro for __ibm128 built-in functions.
3091 (PACK_IF): Add __ibm128 pack/unpack functions.
3092 (UNPACK_IF): Likewise.
3093 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3094 enable long double built-in functions if long double is IEEE
3095 128-bit floating point.
3096 (rs6000_invalid_builtin): Update long double built-in function
3098 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3099 functions, adjust the built-in function to use the long double
3100 built-in function if __ibm128 and long double are the same type.
3101 * doc/extend.texi (PowerPC builtins): Update documention for
3102 __builtin_{,un}pack_longdouble. Add documentation for
3103 __builtin_{,un}pack_ibm128.
3105 2018-06-06 Jim Wilson <jimw@sifive.com>
3107 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3108 (struct machine_function): New field interrupt_mode.
3109 (riscv_handle_type_attribute): New function. Add forward declaration.
3110 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3111 (riscv_expand_epilogue): Check interrupt_mode field.
3112 (riscv_set_current_function): Check interrupt attribute args and
3113 set interrupt_mode field.
3114 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3115 (riscv_sret, riscv_uret): New.
3116 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3117 new arguments to interrupt attribute.
3119 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3122 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3123 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3125 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3127 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3128 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3129 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3130 several redundant entries.
3132 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3134 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3135 type from "rtx" to "rtx_insn *".
3136 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3137 for local "call_insn", removing cast.
3138 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3140 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3142 PR tree-optimization/86066
3143 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3144 for BIT_INSERT_EXPR stores.
3146 2018-06-06 Richard Biener <rguenther@suse.de>
3148 PR tree-optimization/86062
3149 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3150 component refs ontop
3151 of to be offsetted base.
3153 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3155 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3156 to be static and remove check on interrupt attribute name.
3158 gcc/testsuite/gcc.target/msp430/
3159 * function-attributes-4.c: New test.
3160 * static-interrupts.c: New test.
3162 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3164 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3165 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3167 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3170 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3172 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3173 Remove second argument, change how error is called.
3174 (aarch64_layout_arg): Remove second argument from
3175 aarch64_err_no_fpadvsimd call.
3176 (aarch64_init_cumulative_args): Ditto.
3177 (aarch64_gimplify_va_arg_expr): Ditto.
3178 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3180 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3182 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3183 (*simple_return_indirect_internal<mode>): Rename from
3184 simple_return_indirect_internal. Use W mode iterator.
3185 (rstorssp): New expander.
3186 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3187 (clrssbsy): New expander.
3188 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3190 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3192 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3194 (cmse_check_pointed_object): Likewise.
3196 2018-06-05 Martin Liska <mliska@suse.cz>
3198 PR gcov-profile/47618
3199 * doc/invoke.texi: Document how -fprofile-dir format
3202 2018-06-05 Richard Biener <rguenther@suse.de>
3204 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3205 removal pretend DOM info isn't available so we do not update
3206 it and only remove edges, not dominated blocks. Actually free
3207 DOM info in case we removed something. Remove unreachable blocks.
3208 (mfb_keep_latches): Work with either DOM info or marked backedges.
3209 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3210 first. Mark backedges if DOM info isn't available.
3211 (Re-)compute DOM info after cleanup_control_flow_pre.
3213 2018-06-05 Richard Biener <rguenther@suse.de>
3215 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3216 (locus_discrim_hasher::hash): Adjust.
3217 (locus_discrim_hasher::equal): Likewise.
3218 (next_discriminator_for_locus): Work on line directly.
3219 (same_line_p): Pass in expanded locus1 as well.
3220 (assign_discriminators): Avoid redundant location expansions.
3222 2018-06-05 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/86046
3225 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3226 if required after clearing TREE_ADDRESSABLE.
3228 2018-06-05 Richard Biener <rguenther@suse.de>
3230 PR tree-optimization/86047
3231 * tree-ssa-loop.c (for_each_index): Glob handling of all
3232 decls and constants and really handle all of them.
3234 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3237 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3238 qualifier_void_pointer and qualifier_const_void_pointer.
3239 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3240 (arm_init_builtins): Handle the above.
3241 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3242 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3243 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3246 2018-06-05 Martin Liska <mliska@suse.cz>
3248 * auto-profile.c (read_autofdo_file): Do not use
3249 gcov_ctr_summary struct.
3250 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3251 * coverage.c (struct counts_entry): Likewise.
3252 (read_counts_file): Read just single summary entry.
3253 (get_coverage_counts): Use gcov_summary struct.
3254 * coverage.h (get_coverage_counts): Likewise.
3255 * gcov-dump.c (dump_working_sets): Likewise.
3256 (tag_summary): Dump just single summary.
3257 * gcov-io.c (gcov_write_summary): Write just histogram
3259 (gcov_read_summary): Read just single summary.
3260 (compute_working_sets): Use gcov_summary struct.
3261 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3262 of GCOV_COUNTERS_SUMMABLE.
3263 (GCOV_COUNTERS_SUMMABLE): Remove.
3264 (GCOV_FIRST_VALUE_COUNTER): Replace with
3265 GCOV_COUNTER_V_INTERVAL.
3266 (struct gcov_ctr_summary): Remove.
3267 (struct gcov_summary): Directly use fields of former
3269 (compute_working_sets): Use gcov_summary struct.
3270 * gcov.c (read_count_file): Do not use ctrs fields.
3271 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3273 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3275 * profile.c: Likewise.
3276 * profile.h: Likewise.
3278 2018-06-05 Martin Liska <mliska@suse.cz>
3280 PR gcov-profile/84846
3281 * gcov.c (output_lines): Print working directory only
3282 in intermediate format.
3284 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3286 * config/s390/s390-builtin-types.def: Add void function type.
3287 * config/s390/s390-builtins.def: Use the function type for the
3290 2018-06-04 Jim Wilson <jimw@sifive.com>
3292 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3294 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3295 and maybe_eh_return. Change regno to unsigned int. Use new args to
3296 handle EH_RETURN_DATA_REGNO registers properly.
3297 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3298 (riscv_expand_epilogue): Update comment. Change argument name and
3299 type. Update code to use new name and type. Pass new args to
3300 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3302 * config/riscv/riscv.md (NORMAL_RETURN): New.
3303 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3304 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3305 (eh_return): Call gen_eh_return_internal and emit barrier.
3306 (eh_return_internal): Call riscv_expand_epilogue.
3308 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3310 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3311 bit_insertion field and declare can_be_merged_into method.
3312 (merged_store_group::can_be_merged_into): New method.
3313 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3314 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3315 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3317 2018-06-04 Richard Biener <rguenther@suse.de>
3319 PR tree-optimization/85955
3320 * builtins.c (fold_builtin_sincos): Convert pointers to
3321 destination to appropriate type before dereferencing.
3323 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3325 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3327 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3329 * expr.c (expand_expr_real_1): Force the operand into memory if
3330 its TYPE_MODE is BLKmode and if there is no integer mode for
3331 the number of bits being extracted.
3333 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3337 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3338 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3340 2018-06-04 Richard Biener <rguenther@suse.de>
3342 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3343 (cleanup_tree_cfg_noloop): ... single caller. Do
3344 start_recording_case_labels later.
3346 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3348 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3349 to _IMMINTRIN_H_INCLUDED.
3350 * config/i386/pconfigintrin.h: Ditto.
3351 * config/i386/waitpkgintrin.h: Ditto.
3352 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3353 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3354 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3355 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3356 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3357 waitpkgintrin.h and cldemoteintrin.h.
3359 2018-06-04 Richard Biener <rguenther@suse.de>
3361 PR tree-optimization/86038
3362 * tracer.c (find_best_successor): Check probability for
3363 being initialized, bail out if not.
3365 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3368 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3369 of bits to ignore when comparing architectures.
3371 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3373 PR tree-optimization/69615
3374 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3375 maximum or minimum of the type, try to merge it also as if
3376 range1 is + [-, x - 1] or + [x + 1, -].
3379 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3381 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3383 PR tree-optimization/86034
3384 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3385 the unsigned bitfield type in a bit insertion sequence if it does not
3386 have a larger precision than the bitfield size.
3387 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3389 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3391 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3393 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3395 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3396 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3397 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3398 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3400 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3403 Disable -fdelete-null-pointer-checks for ELF toolchain.
3405 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3406 Kito Cheng <kito.cheng@gmail.com>
3408 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3409 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3410 (nds32*-*-*): Add glibc and uclibc conditions.
3411 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3412 (TARGET_EXCEPT_UNWIND_INFO): Define.
3413 * config/nds32/elf.h: New file.
3414 * config/nds32/linux.h: New file.
3415 * config/nds32/nds32-elf.opt: New file.
3416 * config/nds32/nds32-linux.opt: New file.
3417 * config/nds32/nds32-fp-as-gp.c
3418 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3419 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3421 (nds32_asm_file_end): Ditto.
3422 (nds32_print_operand): Ditto.
3423 (nds32_insert_attributes): Ditto.
3424 (nds32_init_libfuncs): New function.
3425 (TARGET_HAVE_TLS): Define.
3426 (TARGET_INIT_LIBFUNCS): Define.
3427 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3429 (TARGET_ELF): Apply different mcmodel setting.
3430 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3431 been migrated into elf.h and linux.h files.
3432 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3433 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3434 (mcmodel): The content has been migrated into nds32-elf.opt and
3435 nds32-linux.opt files.
3436 * config/nds32/t-elf: New file.
3437 * config/nds32/t-linux: New file.
3439 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3440 Shiva Chen <shiva0217@gmail.com>
3442 * config/nds32/constants.md (unspec_volatile_element): Add
3443 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3444 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3446 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3447 (make_pass_nds32_fp_as_gp): Declare.
3448 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3450 (nds32_asm_function_end_prologue): Remove unused asm output.
3451 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3452 (nds32_asm_file_start): Output necessary fp_as_gp information.
3453 (nds32_option_override): Adjust register usage.
3454 (nds32_expand_prologue): Consider fp_as_gp situation.
3455 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3456 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3459 (simple_return): Ditto.
3460 (omit_fp_begin): Output special directive for fp_as_gp.
3461 (omit_fp_end): Output special directive for fp_as_gp.
3462 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3463 mforbid-fp-as-gp): New options.
3465 2018-06-01 Mark Wielaard <mark@klomp.org>
3467 * dwarf2out.c (dwarf2out_finish): Remove generation of
3468 DW_AT_loclists_base.
3470 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3472 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3473 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3474 (struct merged_store_group): Add bit_insertion field.
3475 (dump_char_array): Use standard hexadecimal format.
3476 (merged_store_group::merged_store_group): Set bit_insertion to false.
3477 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3478 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3479 also print the mask in the dump file.
3480 (pass_store_merging::gate): Minor tweak.
3481 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3482 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3483 stores with INTEGER_CST stores.
3484 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3485 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3486 and use it throughout. Generate bit insertion sequences if need be.
3487 (pass_store_merging::process_store): Remove redundant condition.
3488 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3490 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3492 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3493 the 128-bit floating point types. Fix function comment.
3495 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3497 * config/aarch64/aarch64-simd.md
3498 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3500 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3503 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3505 PR tree-optimization/85989
3506 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3508 (backprop::intersect_uses): Check it when deciding whether this
3509 is a backedge reference.
3510 (backprop::process_block): Add each phi to m_visited_phis
3511 after visiting it, then clear it at the end.
3513 2018-06-01 Richard Biener <rguenther@suse.de>
3515 * tree-vectorizer.h (vect_dr_stmt): New function.
3516 (vect_get_load_cost): Adjust.
3517 (vect_get_store_cost): Likewise.
3518 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3519 Use vect_dr_stmt instead of DR_SMTT.
3520 (vect_record_base_alignments): Likewise.
3521 (vect_calculate_target_alignment): Likewise.
3522 (vect_compute_data_ref_alignment): Likewise and make static.
3523 (vect_update_misalignment_for_peel): Likewise.
3524 (vect_verify_datarefs_alignment): Likewise.
3525 (vector_alignment_reachable_p): Likewise.
3526 (vect_get_data_access_cost): Likewise. Pass down
3527 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3528 (vect_get_peeling_costs_all_drs): Likewise.
3529 (vect_peeling_hash_get_lowest_cost): Likewise.
3530 (vect_enhance_data_refs_alignment): Likewise.
3531 (vect_find_same_alignment_drs): Likewise.
3532 (vect_analyze_data_refs_alignment): Likewise.
3533 (vect_analyze_group_access_1): Likewise.
3534 (vect_analyze_group_access): Likewise.
3535 (vect_analyze_data_ref_access): Likewise.
3536 (vect_analyze_data_ref_accesses): Likewise.
3537 (vect_vfa_segment_size): Likewise.
3538 (vect_small_gap_p): Likewise.
3539 (vectorizable_with_step_bound_p): Likewise.
3540 (vect_prune_runtime_alias_test_list): Likewise.
3541 (vect_analyze_data_refs): Likewise.
3542 (vect_supportable_dr_alignment): Likewise.
3543 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3544 (vect_gen_prolog_loop_niters): Likewise.
3545 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3546 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3548 (vect_recog_mask_conversion_pattern): Likewise.
3549 (vect_try_gather_scatter_pattern): Likewise.
3550 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3551 to vect_get_store_cost.
3552 (vect_get_store_cost): Get stmt_info instead of DR.
3553 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3554 (vect_get_load_cost): Get stmt_info instead of DR.
3556 2018-06-01 Richard Biener <rguenther@suse.de>
3559 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3560 vars wrapped in MEM_REFs.
3562 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3564 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3565 Fix subreg tests so that we only return a choice between
3566 GENERAL_REGS and FP_REGS if the original classes included both.
3568 2018-06-01 Richard Biener <rguenther@suse.de>
3571 * tree-ssa-structalias.c (get_function_part_constraint):
3572 Handle NULL fi->decl.
3573 (find_func_aliases_for_call): Properly handle indirect
3574 fi from direct call.
3575 (find_func_clobbers): Likewise.
3576 (ipa_pta_execute): Likewise.
3577 (create_variable_info_for): For functions that are ifunc_resolver
3578 resolve to a varinfo that contains the result of the resolver call.
3579 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3582 2018-05-31 Michael Collison <michael.collison@arm.com>
3584 * config/aarch64/aarch64.md:
3585 (*fix_to_zero_extenddfdi2): New pattern.
3586 * gcc.target/aarch64/fix_extend1.c: New testcase.
3588 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3592 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3593 and BUILT_IN_STRNCMP_EQ.
3594 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3595 BUILT_IN_STRNCMP_EQ.
3596 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3597 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3598 (gimple_fold_builtin): Likewise.
3599 * tree-ssa-strlen.c (compute_string_length): New function.
3600 (determine_min_obsize): New function.
3601 (handle_builtin_string_cmp): New function to handle calls to
3602 string compare functions.
3603 (strlen_optimize_stmt): Add handling to builtin string compare
3605 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3606 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3607 * tree.c (build_common_builtin_nodes): Add new defines of
3608 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3610 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3613 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3614 functions with naked attribute.
3616 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3618 * config/i386/sse.md (avx_vec_concat<mode>):
3619 Substitute concat_tg_mode mode attribute with xtg_mode.
3620 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3621 (concat_tg_mode): Remove mode attribute.
3623 2018-05-31 Martin Sebor <msebor@redhat.com>
3626 * calls.c (alloc_max_size): Correct a logic error/typo.
3627 Treat excessive arguments as infinite. Warn for invalid arguments.
3628 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3630 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3633 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3634 and movx for Haswell.
3636 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3637 Cesar Philippidis <cesar@codesourcery.com>
3640 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3641 when emitting error on private/firstprivate reductions.
3642 * omp-low.c (lower_omp_target): Avoid reference-type processing
3643 on pointers for firstprivate clause.
3645 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3647 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3650 * config/aarch64/aarch64-simd.md
3651 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3652 (aarch64_ld1_x3_<mode>): Likewise
3653 (aarch64_st1x2<VALLDIF:mode>): Likewise
3654 (aarch64_st1_x2_<mode>): Likewise
3655 (aarch64_st1x3<VALLDIF:mode>): Likewise
3656 (aarch64_st1_x3_<mode>): Likewise
3657 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3658 (vld1_s8_x3): Likewise.
3659 (vld1_u16_x3): Likewise.
3660 (vld1_s16_x3): Likewise.
3661 (vld1_u32_x3): Likewise.
3662 (vld1_s32_x3): Likewise.
3663 (vld1_u64_x3): Likewise.
3664 (vld1_s64_x3): Likewise.
3665 (vld1_f16_x3): Likewise.
3666 (vld1_f32_x3): Likewise.
3667 (vld1_f64_x3): Likewise.
3668 (vld1_p8_x3): Likewise.
3669 (vld1_p16_x3): Likewise.
3670 (vld1_p64_x3): Likewise.
3671 (vld1q_u8_x3): Likewise.
3672 (vld1q_s8_x3): Likewise.
3673 (vld1q_u16_x3): Likewise.
3674 (vld1q_s16_x3): Likewise.
3675 (vld1q_u32_x3): Likewise.
3676 (vld1q_s32_x3): Likewise.
3677 (vld1q_u64_x3): Likewise.
3678 (vld1q_s64_x3): Likewise.
3679 (vld1q_f16_x3): Likewise.
3680 (vld1q_f32_x3): Likewise.
3681 (vld1q_f64_x3): Likewise.
3682 (vld1q_p8_x3): Likewise.
3683 (vld1q_p16_x3): Likewise.
3684 (vld1q_p64_x3): Likewise.
3685 (vst1_s64_x2): Likewise.
3686 (vst1_u64_x2): Likewise.
3687 (vst1_f64_x2): Likewise.
3688 (vst1_s8_x2): Likewise.
3689 (vst1_p8_x2): Likewise.
3690 (vst1_s16_x2): Likewise.
3691 (vst1_p16_x2): Likewise.
3692 (vst1_s32_x2): Likewise.
3693 (vst1_u8_x2): Likewise.
3694 (vst1_u16_x2): Likewise.
3695 (vst1_u32_x2): Likewise.
3696 (vst1_f16_x2): Likewise.
3697 (vst1_f32_x2): Likewise.
3698 (vst1_p64_x2): Likewise.
3699 (vst1q_s8_x2): Likewise.
3700 (vst1q_p8_x2): Likewise.
3701 (vst1q_s16_x2): Likewise.
3702 (vst1q_p16_x2): Likewise.
3703 (vst1q_s32_x2): Likewise.
3704 (vst1q_s64_x2): Likewise.
3705 (vst1q_u8_x2): Likewise.
3706 (vst1q_u16_x2): Likewise.
3707 (vst1q_u32_x2): Likewise.
3708 (vst1q_u64_x2): Likewise.
3709 (vst1q_f16_x2): Likewise.
3710 (vst1q_f32_x2): Likewise.
3711 (vst1q_f64_x2): Likewise.
3712 (vst1q_p64_x2): Likewise.
3713 (vst1_s64_x3): Likewise.
3714 (vst1_u64_x3): Likewise.
3715 (vst1_f64_x3): Likewise.
3716 (vst1_s8_x3): Likewise.
3717 (vst1_p8_x3): Likewise.
3718 (vst1_s16_x3): Likewise.
3719 (vst1_p16_x3): Likewise.
3720 (vst1_s32_x3): Likewise.
3721 (vst1_u8_x3): Likewise.
3722 (vst1_u16_x3): Likewise.
3723 (vst1_u32_x3): Likewise.
3724 (vst1_f16_x3): Likewise.
3725 (vst1_f32_x3): Likewise.
3726 (vst1_p64_x3): Likewise.
3727 (vst1q_s8_x3): Likewise.
3728 (vst1q_p8_x3): Likewise.
3729 (vst1q_s16_x3): Likewise.
3730 (vst1q_p16_x3): Likewise.
3731 (vst1q_s32_x3): Likewise.
3732 (vst1q_s64_x3): Likewise.
3733 (vst1q_u8_x3): Likewise.
3734 (vst1q_u16_x3): Likewise.
3735 (vst1q_u32_x3): Likewise.
3736 (vst1q_u64_x3): Likewise.
3737 (vst1q_f16_x3): Likewise.
3738 (vst1q_f32_x3): Likewise.
3739 (vst1q_f64_x3): Likewise.
3740 (vst1q_p64_x3): Likewise.
3742 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3744 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3745 user_label_prefix to name.
3747 * tree-core.h: Update comment about the format of NAME string
3748 passed to handler in attribute_spec.
3750 * config/msp430/msp430.md: Remove erroneous subreg expression from
3751 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3752 zero_extend{q,h}isi2.
3754 2018-05-30 Borislav Petkov <bp@suse.de>
3756 * doc/extend.texi: Document some architecture specific
3757 constraints and sort entries.
3759 2018-05-30 Martin Sebor <msebor@redhat.com>
3762 * builtins.c (expand_builtin_stpcpy_1): New function.
3763 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3764 only if the former succeeds.
3766 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3768 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3771 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3773 * doc/invoke.texi (-flinker-output): Document
3775 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3777 * passes.c (ipa_write_summaries): Only modify statements if body
3779 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3780 incrementally linking.
3781 (ipa_passes): Likewise.
3782 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3783 pass down resolution info.
3784 * common.opt (flag_incremental_link): Update info.
3785 * gcc.c (plugin specs): Turn flinker-output=* to
3786 -plugin-opt=-linker-output-known
3787 * toplev.c (compile_file): Also cut compilation when doing incremental
3789 * flag-types. (enum lto_partition_model): Add
3790 LTO_LINKER_OUTPUT_NOLTOREL.
3791 (invoke.texi): Add -flinker-output docs.
3792 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3793 link same way as WPA; do not stream in dead initializers.
3795 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3796 darf2out_register_external_decl): Support incremental link.
3798 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3800 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3802 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3804 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3805 it down to simple_object_copy_lto_debug_sections.
3806 (run_gcc): Determine incremental LTO link time and configure
3807 lto1 into non-wpa mode, disable renaming of debug sections.
3809 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3811 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3812 descriptions of various incorrectly documented functions.
3814 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3817 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3818 address check not strict.
3820 2018-05-30 Richard Biener <rguenther@suse.de>
3822 PR tree-optimization/85964
3823 * tracer.c (better_p): Drop initialized count check, we only
3824 call the function with initialized counts now.
3825 (find_best_successor): Do find a best edge if one
3826 has uninitialized count.
3827 (find_best_predecessor): Likewise. Do BB frequency check only
3828 if count is initialized.
3830 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3832 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3833 (aarch64_ldrstr_offset_compare): New.
3834 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3835 load/store orderings.
3836 (aarch64_gen_adjusted_ldpstp): Likewise.
3838 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3840 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3841 Check for subset of GENERAL_REGS and FP_REGS.
3842 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3845 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3847 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3848 and wi::to_poly_offset. Add the current offset and then check
3849 whether the sum fits, rather than using an unchecked addition of
3850 a checked term. Check for a shwi rather than a uhwi.
3851 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3852 (store_constructor): Use poly_int_tree_p.
3853 (expand_expr_real_1): Likewise.
3854 * function.c (assign_temp): Likewise.
3855 * fold-const.c (const_binop): Use poly_int_tree_p and
3857 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3859 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3860 to_poly_offset for MEM offsets.
3861 * ipa-icf.c (sem_variable::equals): Likewise.
3862 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3863 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3864 wi::to_poly_offset for BIT_FIELD_REF offsets.
3865 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3867 * var-tracking.c (emit_note_insn_var_location): Use
3868 tree_to_poly_uint64.
3870 2018-05-29 Jim Wilson <jimw@sifive.com>
3872 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3874 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3877 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3878 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3880 (sse4_1_round<mode>2): Use nonimmediate_operand
3881 for operand 1 predicate.
3883 2018-05-29 Martin Sebor <msebor@redhat.com>
3884 Richard Biener <rguenther@suse.de>
3887 * calls.c (get_size_range): Call determine_value_range instead
3888 of get_value_range..
3889 * tree-vrp.h (determine_value_range): Declared new function.
3890 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3892 2018-05-29 Richard Biener <rguenther@suse.de>
3894 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3895 sure to use non-pattern stmts for get_earlier_stmt arguments.
3896 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3897 called on pattern stmts.
3898 (get_later_stmt): Likewise.
3900 2018-05-29 Martin Liska <mliska@suse.cz>
3902 PR gcov-profile/85759
3903 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3906 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3908 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3910 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3914 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3915 VEC_PACK_FLOAT_EXPR): New tree codes.
3916 * tree-pretty-print.c (op_code_prio): Handle
3917 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3918 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3919 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3920 * tree-inline.c (estimate_operator_cost): Likewise.
3921 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3922 * fold-const.c (const_binop): Likewise.
3923 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3924 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3925 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3926 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3927 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3928 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3929 * expr.c (expand_expr_real_2): Likewise.
3930 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3931 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3932 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3934 * optabs.c (expand_widen_pattern_expr): For
3935 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3936 sign from result type rather than operand's type.
3937 (expand_binop_directly): For vec_packu_float_optab and
3938 vec_packs_float_optab allow result type to be different from operand's
3940 * optabs-tree.c (optab_for_tree_code): Handle
3941 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3942 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3943 * tree-vect-generic.c (expand_vector_operations_1): Handle
3944 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3945 VEC_PACK_FLOAT_EXPR.
3946 * tree-vect-stmts.c (supportable_widening_operation): Handle
3948 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3949 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3950 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3951 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3952 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3954 (vec_pack<floatprefix>_float_<mode>): New expander.
3955 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3957 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3958 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3959 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3960 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3961 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3963 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3964 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3965 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3966 VEC_PACK_FLOAT_EXPR): Document.
3968 2018-05-29 Richard Biener <rguenther@suse.de>
3970 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3972 (stmt_vec_info_vec): Make pointer.
3973 (init_stmt_vec_info_vec): Remove.
3974 (free_stmt_vec_info_vec): Likewise.
3975 (set_stmt_vec_info_vec): New function.
3976 (free_stmt_vec_infos): Likewise.
3977 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3978 (set_vinfo_for_stmt): Likewise.
3979 (get_earlier_stmt): Likewise.
3980 (get_later_stmt): Likewise.
3981 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3982 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3983 (vec_info::~vec_info): Free stmt_vec_infos.
3984 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3985 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3986 (pass_slp_vectorize::execute): Likewise.
3987 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3988 (free_stmt_vec_info_vec): Likewise.
3989 (set_stmt_vec_info_vec): New function.
3990 (free_stmt_vec_infos): Likewise.
3991 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3992 the global stmt_vec_info_vec.
3993 * tree-parloops.c (gather_scalar_reductions): Use
3994 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3997 2018-05-29 Richard Biener <rguenther@suse.de>
3999 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4001 2018-05-29 Martin Liska <mliska@suse.cz>
4002 David Malcolm <dmalcolm@redhat.com>
4004 * vec.c (test_reverse): New.
4005 (vec_c_tests): Add new test.
4006 * vec.h (vl_ptr>::reverse): New function.
4008 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4010 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4012 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4015 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4017 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4019 2018-05-28 Richard Biener <rguenther@suse.de>
4021 PR tree-optimization/85933
4022 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4023 look at stmts marked as vectorizable.
4025 2018-05-28 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/85934
4028 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4029 vector boolean check before scalar optimization.
4031 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4033 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4036 2018-05-28 Mark Wielaard <mark@klomp.org>
4038 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4039 if it is an expression containing a minus sign.
4041 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4043 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4045 2018-05-27 Paul Koning <ni1d@arrl.net>
4047 * config/pdp11/pdp11.md (truncsihi2): Remove.
4049 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4050 Chung-Ju Wu <jasonwucj@gmail.com>
4052 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4054 (unaligned_store_dw): Ditto.
4055 * config/nds32/nds32-memory-manipulation.c
4056 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4057 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4058 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4059 (emit_setmem_word_loop): Rename to ...
4060 (emit_setmem_doubleword_loop): ... this.
4061 (nds32_gen_dup_4_byte_to_word_value): New function.
4062 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4063 (nds32_expand_setmem_loop): Refine implementation.
4064 (nds32_expand_setmem_loop_v3m): Ditto.
4065 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4068 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4070 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4072 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4074 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4075 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4076 machine->attr_no_prologue_p.
4077 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4078 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4079 (nds32_expand_epilogue): Consider attr_naked_p.
4080 (nds32_expand_epilogue_v3pop): Likewise.
4081 (nds32_can_use_return_insn): Likewise.
4082 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4083 attr_no_prologue_p fields.
4084 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4086 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4089 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4091 * config/i386/sse.md
4092 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4094 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4096 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4098 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4100 (*<floatsuffix>floatv2div2sf2): Rename to ...
4101 (*float<floatunssuffix>v2div2sf2): ... this.
4102 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4103 (float<floatunssuffix>v2div2sf2_mask): ... this.
4104 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4105 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4106 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4108 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4110 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4112 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4114 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4116 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4118 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4119 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4120 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4121 gen_ufix_truncv8dfv8si2.
4122 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4123 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4124 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4125 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4126 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4127 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4128 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4129 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4131 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4135 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4137 2018-05-25 Jim Wilson <jimw@sifive.com>
4139 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4140 * config/riscv/riscv.c (struct machine_function): Add
4141 interrupt_handler_p and attribute_checked_p fields.
4142 (riscv_attribute_table): Add interrupt.
4143 (riscv_interrupt_type_p): New.
4144 (riscv_save_reg_p): Save extra regs for interrupt handler.
4145 (riscv_use_save_libcall): Return false for interrupt handler.
4146 (riscv_first_stack_step): Add forward declaration.
4147 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4148 for interrupt handler with large frame. Use it for saved reg list.
4149 (riscv_expand_prologue): Move flag_stack_usage_info support to
4150 eliminate duplication.
4151 (riscv_expand_epilogue): Generate mret for interrupt handler.
4152 (riscv_epilogue_uses): New.
4153 (riscv_can_use_return_insn): Return false for interrupt handler.
4154 (riscv_function_ok_for_sibcall): Likewise.
4155 (riscv_set_current_function): Add interrupt handler support.
4156 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4157 * config/riscv/riscv.md (UNSPECV_MRET): New.
4159 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4161 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4163 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4165 PR tree-optimization/85712
4166 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4167 this candidate has already been replaced in-situ by a copy.
4169 2018-05-25 Jason Merrill <jason@redhat.com>
4171 PR c++/80485 - inline function non-zero address.
4172 * symtab.c (nonzero_address): Check DECL_COMDAT.
4174 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4177 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4178 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4179 extension of SImode operation. Use const123_operand predicate.
4180 (*saddsi_1): Remove.
4181 (*saddl_se_1): Ditto.
4183 (*ssubl_se_1): Ditto.
4184 * config/alpha/predicates.md (const123_operand): New predicate.
4185 * config/alpha/constraints.md (P): Use IN_RANGE.
4187 2018-05-25 Richard Biener <rguenther@suse.de>
4189 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4191 (ref_maybe_used_by_stmt_p): Likewise.
4192 (stmt_may_clobber_ref_p): Likewise.
4193 (stmt_may_clobber_ref_p_1): Likewise.
4194 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4196 (ref_maybe_used_by_stmt_p): Likewise.
4197 (stmt_may_clobber_ref_p): Likewise.
4198 (stmt_may_clobber_ref_p_1): Likewise.
4199 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4200 the alias oracle to disambiguate DRs with stmts DR analysis
4202 (vect_analyze_data_refs): Do not give up on not analyzable
4203 DRs for BB vectorization. Remove code truncating the dataref
4206 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4209 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4210 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4211 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4213 2018-05-25 Richard Biener <rguenther@suse.de>
4215 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4216 function, combining stmt data ref gathering and fatal analysis
4218 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4219 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4220 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4221 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4222 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4224 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4226 PR tree-optimization/85720
4227 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4228 SCC if all partitions are builtins.
4229 (version_loop_by_alias_check): New parameter. Generate cancelable
4230 runtime alias check if all partitions are builtins.
4231 (distribute_loop): Update call to above function.
4233 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4235 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4236 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4237 (parm_default_def_partition_arg): Ditto.
4238 (set_parm_default_def_partition): Ditto.
4239 (get_parm_default_def_partitions): Ditto and make it static.
4240 (get_undefined_value_partitions): Ditto and make it static.
4241 (remove_ssa_form): Refactor call to init_var_map here.
4242 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4243 computation for loop region.
4244 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4245 (register_default_def): Delete.
4246 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4247 (parm_default_def_partition_arg): Ditto.
4248 (set_parm_default_def_partition): Ditto.
4249 (get_parm_default_def_partitions): Ditto and make it static.
4250 (get_undefined_value_partitions): Ditto and make it static.
4251 (coalesce_with_default, coalesce_with_default): Update comment.
4252 (create_coalesce_list_for_region): New func factored out from
4253 create_outofssa_var_map.
4254 (populate_coalesce_list_for_outofssa): New func factored out from
4255 create_outofssa_var_map and coalesce_ssa_name.
4256 (create_outofssa_var_map): Delete.
4257 (coalesce_ssa_name): Refactor to support live range computation.
4258 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4259 (get_parm_default_def_partitions): Delete.
4260 (get_undefined_value_partitions): Ditto.
4261 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4262 computation for loop region.
4263 (new_tree_live_info, loe_visit_block): Ditto.
4264 (live_worklist, set_var_live_on_entry): Ditto.
4265 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4266 * tree-ssa-live.h (struct _var_map): New fields.
4267 (init_var_map): Change decl.
4268 (region_contains_p): New.
4270 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4272 * tree-ssa-live.h (live_merge_and_clear): Delete.
4274 2018-05-25 Richard Biener <rguenther@suse.de>
4277 * tree-dump.c (dequeue_and_dump): Remove access to removed
4278 operand 2 of a SWITCH_EXPR.
4280 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4282 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4283 multiplication and division.
4284 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4285 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4286 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4287 (cond_udiv_optab, cond_umod_optab): New optabs.
4288 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4289 (IFN_COND_RDIV): New internal functions.
4290 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4291 TRUNC_MOD_EXPR and RDIV_EXPR.
4292 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4293 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4295 (SVE_INT_BINARY): Include mult.
4296 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4297 (optab, sve_int_op): Handle mult.
4298 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4300 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4301 for SVE_INT_BINARY_SD.
4303 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4305 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4306 (optab, sve_int_op): Handle div and udiv.
4307 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4308 for SVE_INT_BINARY_SD.
4309 (*<optab><mode>3): New insn for the same.
4311 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4313 * tree-vect-patterns.c: Include predict.h.
4314 (vect_recog_divmod_pattern): Restrict check for division support
4315 to when optimizing for size.
4317 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4319 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4320 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4321 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4322 (gimple_match_op::set_op): Likewise.
4323 (gimple_resimplify4): Declare.
4324 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4325 (expr::gen_transform): Likewise.
4326 (decision_tree::gen): Generate a simplification routine for 4 operands.
4327 * gimple-match-head.c (gimple_simplify): Add an overload for
4328 4 operands. In the top-level function, handle up to 4 call
4329 arguments and call gimple_resimplify4.
4330 (gimple_resimplify4): New function.
4331 (build_call_internal): Pass a fourth operand.
4332 (maybe_push_to_seq): Likewise.
4333 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4334 Fold VEC_COND_EXPRs of an operation and a default value into
4335 an IFN_COND_* function if possible.
4336 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4338 (SVE_COND_FP_BINARY): Include them.
4339 (optab, sve_fp_op): Handle them.
4340 (SVE_INT_BINARY_REV): New code iterator.
4341 (SVE_COND_FP_BINARY_REV): New int iterator.
4342 (commutative): New int attribute.
4343 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4345 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4347 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4348 (*cond_<optab><mode>): New patterns for reversed operands.
4350 2018-05-25 Richard Biener <rguenther@suse.de>
4352 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4353 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4354 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4355 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4356 * tree-vect-data-refs.c (everywhere): Adjust users.
4357 * tree-vect-loop.c (everywhere): Likewise.
4358 * tree-vect-slp.c (everywhere): Likewise.
4359 * tree-vect-stmts.c (everywhere): Likewise.
4360 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4362 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4364 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4366 (gcc_cv_as_section_exclude): ... this.
4367 Try Solaris as #exclude syntax.
4368 * configure: Regenerate.
4369 * config.in: Regenerate.
4370 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4372 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4373 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4375 * varasm.c (default_elf_asm_named_section): Don't check if
4376 HAVE_GAS_SECTION_EXCLUDE is defined.
4378 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4380 * doc/md.texi: Update the documentation of the cond_* optabs
4381 to mention the new final operand. Fix GET_MODE_NUNITS call.
4382 Describe the scalar case too.
4383 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4384 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4386 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4387 (get_conditional_internal_fn): Update comment.
4388 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4389 accumulator value as a final argument to conditional functions.
4390 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4391 a define_expand and add an "else" operand. Assert for now that
4392 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4393 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4394 (*cond_<optab><mode>): New patterns.
4395 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4396 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4397 (UNSPEC_COND_EOR): Delete.
4398 (optab): Remove associated mappings.
4399 (SVE_INT_BINARY): New code iterator.
4400 (sve_int_op): Remove int attribute and add "minus" to the code
4402 (SVE_COND_INT_OP): Delete.
4403 (SVE_COND_FP_OP): Rename to...
4404 (SVE_COND_FP_BINARY): ...this.
4406 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4408 * optabs.c (can_reuse_operands_p): New function.
4409 (maybe_legitimize_operands): Try to reuse the results for
4412 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4414 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4415 Add {q} suffix to insn mnemonic.
4417 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4419 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4420 (msp430_warn_func_return): New.
4422 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4424 * fold-const.c (tree_nonzero_bits): New function.
4425 * fold-const.h (tree_nonzero_bits): Likewise.
4426 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4427 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4429 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4433 * varasm.c (assemble_alias): Check ifunc_resolver only on
4436 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4439 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4440 when memory input operand is handled.
4442 2018-05-24 Luis Machado <luis.machado@linaro.org>
4444 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4446 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4448 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4450 * match.pd: Delay FMA folds until after vectorization.
4452 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4455 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4456 address check not strict.
4458 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4460 * gimple-match.h (gimple_match_op): New class.
4461 (mprts_hook): Replace parameters with a gimple_match_op *.
4462 (maybe_build_generic_op): Likewise.
4463 (gimple_simplified_result_is_gimple_val): Replace parameters with
4464 a const gimple_match_op *.
4465 (gimple_simplify): Replace code_helper * and tree * parameters with
4466 a gimple_match_op * parameter.
4467 (gimple_resimplify1): Replace code_helper *, tree and tree *
4468 parameters with a gimple_match_op * parameter.
4469 (gimple_resimplify2): Likewise.
4470 (gimple_resimplify3): Likewise.
4471 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4472 parameters with a gimple_match_op * parameter.
4473 * gimple-match-head.c (gimple_simplify): Change prototypes of
4474 auto-generated functions to take a gimple_match_op * instead of
4475 separate code_helper * and tree * parameters. Make the same
4476 change in the top-level overload and update calls to the
4477 gimple_resimplify routines. Update calls to the auto-generated
4478 functions and to maybe_push_res_to_seq in the publicly-facing
4479 operation-specific gimple_simplify overloads.
4480 (gimple_match_op::MAX_NUM_OPS): Define.
4481 (gimple_resimplify1): Replace rcode and ops with a single res_op
4482 parameter. Update call to gimple_simplify.
4483 (gimple_resimplify2): Likewise.
4484 (gimple_resimplify3): Likewise.
4485 (mprts_hook): Replace parameters with a gimple_match_op *.
4486 (maybe_build_generic_op): Likewise.
4487 (build_call_internal): Replace type, nargs and ops with
4488 a gimple_match_op *.
4489 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4490 with a single gimple_match_op *. Update calls to mprts_hook,
4491 build_call_internal and gimple_simplified_result_is_gimple_val.
4492 Factor out code that is common to the tree_code and combined_fn cases.
4493 * genmatch.c (expr::gen_transform): Replace tem_code and
4494 tem_ops with a gimple_match_op called tem_op. Update calls
4495 to the gimple_resimplify functions and maybe_push_res_to_seq.
4496 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4497 res_ops. Update call to the gimple_resimplify functions.
4498 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4499 (decision_tree::gen): Make the functions take a gimple_match_op *
4500 called res_op instead of separate res_code and res_ops parameters.
4501 Update call accordingly.
4502 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4503 and ops with a single res_op parameter. Update calls to
4504 maybe_build_generic_op and maybe_push_res_to_seq.
4505 (fold_stmt_1): Update calls to gimple_simplify and
4506 replace_stmt_with_simplification.
4507 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4508 and gimple_simplified_result_is_gimple_val.
4509 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4511 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4512 with a gimple_match_op *.
4513 (vn_nary_build_or_lookup): Likewise. Update call to
4514 vn_nary_build_or_lookup_1.
4515 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4516 gimple_match_op *. Update calls to the gimple_resimplify routines
4517 and to gimple_simplified_result_is_gimple_val.
4518 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4519 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4520 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4521 (visit_nary_op): Likewise.
4522 (visit_reference_op_load): Likewise.
4524 2018-05-23 Luis Machado <luis.machado@linaro.org>
4526 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4527 modifier for printing the step amount.
4529 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4532 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4535 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4537 * doc/sourcebuild.texi (Endianness): New subsubsection.
4539 2018-05-23 Luis Machado <luis.machado@linaro.org>
4541 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4542 <prefetch_dynamic_strides>: New const bool field.
4543 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4544 prefetch_dynamic_strides.
4545 (exynosm1_prefetch_tune): Likewise.
4546 (thunderxt88_prefetch_tune): Likewise.
4547 (thunderx_prefetch_tune): Likewise.
4548 (thunderx2t99_prefetch_tune): Likewise.
4549 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4551 (aarch64_override_options_internal): Update to set
4552 PARAM_PREFETCH_DYNAMIC_STRIDES.
4553 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4554 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4555 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4556 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4557 prefetch-dynamic-strides setting.
4559 2018-05-23 Luis Machado <luis.machado@linaro.org>
4561 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4562 <minimum_stride>: New const int field.
4563 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4564 minimum_stride field defaulting to -1.
4565 (exynosm1_prefetch_tune): Likewise.
4566 (thunderxt88_prefetch_tune): Likewise.
4567 (thunderx_prefetch_tune): Likewise.
4568 (thunderx2t99_prefetch_tune): Likewise.
4569 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4570 <default_opt_level>: Set to 3.
4571 (aarch64_override_options_internal): Update to set
4572 PARAM_PREFETCH_MINIMUM_STRIDE.
4573 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4574 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4575 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4576 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4577 stride is constant and is below the minimum stride threshold.
4579 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4581 * config/arm/arm-cpus.in (mode26): Delete.
4582 (armv4): Delete mode26 reference.
4583 * config/arm/arm.c (arm_configure_build_target): Delete use of
4586 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4588 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4590 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4591 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4593 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4594 (floatunsdidf2): Ditto.
4596 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4598 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4599 (fixuns_trunc<mode>si2_avx512f): Ditto.
4600 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4601 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4602 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4604 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4606 PR rtl-optimization/79985
4607 * df-scan.c (df_insn_refs_collect): Remove special case for
4608 global registers and asm statements.
4610 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4612 * extend.texi (Global Register Variables): Rewrite the bullet list.
4613 Note that the register is available for allocation. Note that access
4614 via inline asm must use constraints. Add note about async-signal
4615 handlers. Remove paragraph about automagic register selection.
4617 2018-05-23 Richard Biener <rguenther@suse.de>
4619 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4620 of fixed offset from memset VN.
4622 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4624 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4626 (alloc_cand_and_find_basis): Initialize first_interp field.
4627 (slsr_process_mul): Modify first_interp field.
4628 (slsr_process_add): Likewise.
4629 (slsr_process_cast): Modify first_interp field for each new
4631 (slsr_process_copy): Likewise.
4632 (dump_candidate): Dump first_interp field.
4633 (replace_mult_candidate): Process all interpretations, not just
4635 (replace_rhs_if_not_dup): Likewise.
4636 (replace_one_candidate): Likewise.
4638 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4640 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4642 (aarch64_needs_frame_chain): New function.
4643 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4645 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4648 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4649 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4650 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4651 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4652 as true for strict-align.
4653 (aarch64_can_inline_p): Perform checks even when callee has no
4654 attributes to check for strict alignment.
4655 * doc/extend.texi (AArch64 Function Attributes): Document
4657 * doc/invoke.texi: (AArch64 Options): Likewise.
4659 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4661 PR tree-optimization/85853
4662 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4663 the handling of the root of the node to...
4664 (vect_slp_analyze_node_operations_1): ...this new function,
4665 and run the whole thing with the child nodes' def types
4666 set according to their SLP node's def type.
4668 2018-05-23 Richard Biener <rguenther@suse.de>
4671 * tree-data-ref.c (create_runtime_alias_checks): Defer
4672 and ignore overflow warnings.
4674 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4676 PR tree-optimization/85822
4677 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4680 2018-05-23 Richard Biener <rguenther@suse.de>
4682 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4683 memset constants via native_interpret_expr.
4685 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4688 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4690 (cgraph_node::create_alias): Likewise.
4691 (cgraph_node::get_availability): Check ifunc_resolver instead
4692 of looking up ifunc attribute.
4693 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4694 * varasm.c (do_assemble_alias): Likewise.
4695 (assemble_alias): Likewise.
4696 (default_binds_local_p_3): Likewise.
4697 * cgraph.h (cgraph_node): Add ifunc_resolver.
4698 (cgraph_node::only_called_directly_or_aliased_p): Return false
4700 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4702 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4703 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4704 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4705 instead of looking up ifunc attribute.
4707 2018-05-22 Luis Machado <luis.machado@linaro.org>
4709 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4711 2018-05-22 Martin Sebor <msebor@redhat.com>
4714 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4715 only when expasion succeeds.
4716 (expand_builtin_strcmp): Same.
4717 (expand_builtin_strncmp): Same.
4719 2018-05-22 Martin Sebor <msebor@redhat.com>
4721 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4723 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4724 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4726 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4727 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4729 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4730 Add check for legitimate_address.
4731 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4732 (aarch64_swap_ldrstr_operands): New.
4733 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4736 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4737 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4739 * config/aarch64/aarch64.md: New patterns to generate stp
4741 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4742 single words and double words.
4743 (load_pair_sw, load_pair_dw): Likewise.
4744 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4746 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4748 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4749 for different mode ldpstp and add peephole for merged zero stores.
4751 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4753 (aarch64_gen_store_pair): Rename calls to match new patterns.
4754 (aarch64_gen_load_pair): Rename calls to match new patterns.
4755 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4756 (load_pair<DREG:mode><DREG2:mode>): ... This.
4757 (store_pair<mode>): Rename to...
4758 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4759 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4761 (V_INT_EQUIV): Handle SImode.
4762 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4765 2018-05-22 Martin Sebor <msebor@redhat.com>
4768 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4769 or ajust the presumed bound on an operation to avoid unnecessary
4772 2018-05-22 Martin Sebor <msebor@redhat.com>
4774 PR tree-optimization/85826
4775 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4776 assuming that a DECL necesarily has a constant size.
4778 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4781 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4783 2018-05-22 Richard Biener <rguenther@suse.de>
4785 PR tree-optimization/85834
4786 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4787 non-constant and non-zero memset arguments.
4789 2018-05-22 Martin Liska <mliska@suse.cz>
4792 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4794 2018-05-22 Richard Biener <rguenther@suse.de>
4796 PR tree-optimization/85863
4797 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4798 comparisons when vectype is specified.
4799 (vectorizable_condition): Do not specify vectype for
4800 vect_is_simple_cond when SLP vectorizing.
4802 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4805 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4806 define __ibm128 as long double.
4807 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4808 as a distinct type when IEEE 128-bit support is enabled.
4809 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4810 128-bit types to use the correct functions.
4811 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4812 convert between 128-bit floating point types that have different
4813 modes but the same representation, instead of using gen_lowpart to
4815 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4817 (IFKF_reg): New attributes to give the register constraints for
4819 (extend<mode>tf2_internal): New insns to mark an explicit
4820 conversion between 128-bit floating point types that have a
4821 different mode but share the same representation.
4823 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4825 PR tree-optimization/85814
4826 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4827 a null return from get_strinfo when unsharing the next
4828 strinfo in the chain.
4830 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4833 * varasm.c (weak_finish): Clean up weak_decls.
4835 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4837 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4838 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4839 UNSPEC_UADALP values.
4840 * config/aarch64/iterators.md (ABAL): New int iterator.
4843 (sur): Add mappings for the above.
4844 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4846 (aarch64_<sur>abal<mode>_4): Likewise.
4847 (aarch64_<sur>adalp<mode>_3): Likewise.
4848 (<sur>sadv16qi): New define_expand.
4850 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4852 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4853 (*movdf_internal): Ditto.
4854 (*rcpsf2_sse): Ditto.
4855 (*rsqrtsf2_sse): Ditto.
4856 (*sqrt<mode>2_sse): Ditto.
4858 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4860 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4862 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4863 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4864 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4865 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4867 * config/aarch64/arm_neon.h: Likewise.
4868 * config/aarch64/iterators.md (VQ_I): New.
4870 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4872 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4874 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4877 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4879 * config/nds32/constraints.md (S): New constraint.
4880 * config/nds32/nds32.md (call_internal): Use constraint S.
4881 (call_value_internal): Likewise.
4882 (sibcall_internal): Likewise.
4883 (sibcall_value_internal): Likewise.
4885 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4886 Chung-Ju Wu <jasonwucj@gmail.com>
4888 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4891 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4892 Chung-Ju Wu <jasonwucj@gmail.com>
4894 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4895 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4896 (nds32_rtx_costs_impl): Simplify.
4897 (nds32_address_cost_impl): Simplify.
4898 (nds32_init_rtx_costs): New function.
4899 (nds32_rtx_costs_speed_prefer): Likewise.
4900 (nds32_rtx_costs_size_prefer): Likewise.
4901 (nds32_address_cost_speed_prefer): Likewise.
4902 (nds32_address_cost_speed_fwprop): Likewise.
4903 (nds32_address_cost_size_prefer): Likewise.
4904 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4905 * config/nds32/nds32.c (nds32_option_override): Use
4906 nds32_init_rtx_costs function.
4908 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4910 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4911 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4912 (TARGET_PIPELINE_N8): Likewise.
4913 (TARGET_PIPELINE_N10): Likewise.
4914 (TARGET_PIPELINE_N13): Likewise.
4915 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4917 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4919 * config/nds32/nds32-fpu.md: Update copyright year.
4921 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4923 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4925 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4927 * config/nds32/nds32.c
4928 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4929 * config/nds32/nds32.opt (minline-asm-r15): New option.
4931 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4933 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4935 * config/nds32/nds32.md (abssi2): New pattern.
4937 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4939 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4940 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4941 Merge insn pattern from sse_cvtsi2ss<round_name> and
4942 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4943 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4944 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4945 using SWI48 mode iterator.
4946 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4947 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4948 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4949 pattern from sse_cvttss2si<round_saeonly_name>
4950 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4951 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4952 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4953 using SWI48 mode iterator.
4954 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4955 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4956 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4957 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4958 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4959 using SWI48 mode iterator.
4960 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4961 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4962 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4963 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4964 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4965 SWI48 mode iterator.
4966 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4967 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4968 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4969 pattern from sse_cvttsd2si<round_saeonly_name>
4970 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4972 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4974 * config/nds32/nds32-md-auxiliary.c
4975 (nds32_valid_smw_lwm_base_p): Refine.
4976 (nds32_output_smw_single_word): Refine.
4977 (nds32_output_smw_double_word): New.
4978 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4980 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4982 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4983 (nds32_output_stack_pop): Refine.
4984 (nds32_expand_unaligned_load): Refine.
4985 (nds32_expand_unaligned_store): Refine.
4987 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4988 Chung-Ju Wu <jasonwucj@gmail.com>
4990 * config/nds32/constants.md: Add TP_REGNUM constant.
4991 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4992 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4994 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4995 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4996 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4997 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4998 and PIC code generation.
4999 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5001 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5003 * config/nds32/nds32.md: Support TLS and PIC.
5004 * config/nds32/nds32.c: Support TLS and PIC.
5005 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5006 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5009 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5011 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5012 mode with E_ prefix.
5014 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5015 Chung-Ju Wu <jasonwucj@gmail.com>
5017 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5018 * config/nds32/nds32-md-auxiliary.c
5019 (symbolic_reference_mentioned_p): New.
5020 (nds32_legitimize_ict_address): New.
5021 (nds32_expand_ict_move): New.
5022 (nds32_indirect_call_referenced_p): New.
5023 (nds32_symbol_binds_local_p): Delete.
5024 (nds32_long_call_p): Modify.
5025 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5026 * config/nds32/nds32-protos.h
5027 (symbolic_reference_mentioned_p): Declare.
5028 (nds32_legitimize_ict_address): Declare.
5029 (nds32_expand_ict_move): Declare.
5030 (nds32_indirect_call_referenced_p): Declare.
5031 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5032 (nds32_relax_group): Use nds32_ict_const_p as condition.
5033 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5034 (nds32_asm_file_start): Output ict_model directive in asm code.
5035 (nds32_legitimate_address_p): Consider indirect call.
5036 (nds32_print_operand): Consider indirect call.
5037 (nds32_print_operand_address): Consider indirect call.
5038 (nds32_insert_attributes): Handle "indirect_call" attribute.
5039 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5040 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5041 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5042 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5043 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5044 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5045 (TARGET_ICT_MODEL_SMALL): Define.
5046 (TARGET_ICT_MODEL_LARGE): Define.
5047 * config/nds32/nds32.md (movsi): Consider ict model.
5048 (call, call_value): Consider ict model.
5049 (sibcall, sibcall_value): Consider ict model.
5050 * config/nds32/nds32.opt (mict-model): New option.
5051 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5054 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5055 Monk Chiang <sh.chiang04@gmail.com>
5056 Jim Wilson <jimw@sifive.com>
5058 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5059 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5060 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5061 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5062 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5063 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5064 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5065 compute save_libcall_adjustment properly.
5066 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5067 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5068 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5069 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5070 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5071 (ABI_SPEC): Handle mabi=ilp32e.
5072 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5073 (RVE): Add RVE mask.
5074 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5075 <-march>: Add rv32e as an example.
5077 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5080 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5081 (intra_create_variable_infos): Handle C++ constructors.
5083 2018-05-18 Martin Liska <mliska@suse.cz>
5085 * passes.def: Remove a redundant pass.
5087 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5090 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5092 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5094 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5096 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5098 (armv2, armv2a, armv3, armv3m): Delete architectures.
5099 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5100 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5101 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5103 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5104 (*mulsidi3adddi): Likewise.
5105 (mulsidi3): Likewise.
5106 (*mulsidi3_nov6): Likewise.
5107 (umulsidi3): Likewise.
5108 (umulsidi3_nov6): Likewise.
5109 (umaddsidi4): Likewise.
5110 (*umulsidi3adddi): Likewise.
5111 (smulsi3_highpart): Likewise.
5112 (*smulsi3_highpart_nov6): Likewise.
5113 (umulsi3_highpart): Likewise.
5114 (*umulsi3_highpart_nov6): Likewise.
5115 * config/arm/arm.h (arm_arch3m): Delete.
5116 * config/arm/arm.c (arm_arch3m): Delete.
5117 (arm_option_override_internal): Update armv3-related comment.
5118 (arm_configure_build_target): Delete use of isa_bit_mode32.
5119 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5120 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5121 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5123 (mulusa3): Likewise.
5124 * config/arm/arm-protos.h (arm_arch3m): Delete.
5125 * config/arm/arm-tables.opt: Regenerate.
5126 * config/arm/arm-tune.md: Likewise.
5127 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5128 deleted architectures.
5130 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5132 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5133 (armv5t, armv5te): New features.
5134 (ARMv5, ARMv5e): Delete fgroups.
5135 (ARMv5t, ARMv5te): Adjust for above changes.
5137 (armv5, armv5e): Delete arches.
5138 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5140 (*call_reg_arm): Likewise.
5141 (*call_value_reg_armv5): Likewise.
5142 (*call_value_reg_arm): Likewise.
5143 (*call_symbol): Likewise.
5144 (*call_value_symbol): Likewise.
5145 (*sibcall_insn): Likewise.
5146 (*sibcall_value_insn): Likewise.
5148 (prefetch): Likewise.
5149 (define_split and define_peephole2 dependent on arm_arch5):
5151 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5153 (TARGET_ARM_QBIT): Likewise.
5154 (TARGET_DSP_MULTIPLY): Likewise.
5155 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5156 (arm_arch5, arm_arch5e): Delete.
5157 (arm_arch5t, arm_arch5te): Declare.
5158 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5159 (arm_arch5t): Declare.
5160 (arm_option_reconfigure_globals): Update for the above.
5161 (arm_options_perform_arch_sanity_checks): Update comment, replace
5162 use of arm_arch5 with arm_arch5t.
5163 (use_return_insn): Likewise.
5164 (arm_emit_call_insn): Likewise.
5165 (output_return_instruction): Likewise.
5166 (arm_final_prescan_insn): Likewise.
5167 (arm_coproc_builtin_available): Likewise.
5168 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5169 arm_arch5e with arm_arch5t and arm_arch5te.
5170 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5171 (arm_arch5t, arm_arch5te): Declare.
5172 * config/arm/arm-tables.opt: Regenerate.
5173 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5174 * config/arm/t-multilib: Likewise.
5175 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5176 instead of arm_arch5.
5177 (*call_reg_thumb1): Likewise.
5178 (*call_value_reg_thumb1_v5): Likewise.
5179 (*call_value_reg_thumb1): Likewise.
5180 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5182 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5184 2018-05-18 Martin Liska <mliska@suse.cz>
5186 PR gcov-profile/84846
5187 * doc/gcov.texi: Document -t option of gcov tool.
5189 2018-05-18 Martin Liska <mliska@suse.cz>
5191 PR gcov-profile/84846
5192 * gcov.c (print_usage): Add new -t option.
5193 (process_args): Handle the option.
5194 (generate_results): Use stdout as output when requested by
5197 2018-05-18 Martin Liska <mliska@suse.cz>
5199 PR gcov-profile/84846
5200 * coverage.c (coverage_init): Write PWD to .gcno file.
5201 * doc/gcov.texi: Document how working directory is printed.
5202 * gcov-dump.c (dump_gcov_file): Print PWD.
5203 * gcov.c (output_intermediate_file): Likewise.
5204 (read_graph_file): Read PWD string.
5205 (output_lines): Print PWD.
5207 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5210 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5211 for retval and return false if all args to phi are zero.
5213 2018-05-18 Richard Biener <rguenther@suse.de>
5215 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5217 (evrp_dom_walker::before_dom_children): Call it.
5219 2018-05-18 Richard Biener <rguenther@suse.de>
5221 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5222 results when processing array refs with variable index.
5224 2018-05-18 Toon Moene <toon@moene.org>
5226 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5227 directly after that of -floop-interchange. Indicate that both
5228 options are enabled by default when specifying -O3.
5230 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5232 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5233 iterator. Delete separate integer-mode vec_set<mode> expander.
5234 (aarch64_simd_vec_setv2di): Delete.
5235 (vec_setv2di): Delete.
5236 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5237 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5238 the "w, r" alternative.
5240 2018-05-18 Martin Liska <mliska@suse.cz>
5242 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5243 * tree-pass.h (make_pass_lower_switch_O0): New function.
5244 * tree-switch-conversion.c (node_has_low_bound): Remove.
5245 (node_has_high_bound): Likewise.
5246 (node_is_bounded): Likewise.
5247 (class pass_lower_switch): Make it a template type and create
5249 (pass_lower_switch::execute): Add template argument.
5250 (make_pass_lower_switch): New function.
5251 (make_pass_lower_switch_O0): New function.
5252 (do_jump_if_equal): Remove.
5253 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5254 all the hard work to tree optimization passes.
5256 2018-05-18 Martin Liska <mliska@suse.cz>
5258 * dbgcnt.c (limit_low): Renamed from limit.
5259 (limit_high): New variable.
5260 (dbg_cnt_is_enabled): Check for upper limit.
5261 (dbg_cnt): Adjust dumping.
5262 (dbg_cnt_set_limit_by_index): Add new argument for high
5264 (dbg_cnt_set_limit_by_name): Likewise.
5265 (dbg_cnt_process_single_pair): Parse new format.
5266 (dbg_cnt_process_opt): Use strtok.
5267 (dbg_cnt_list_all_counters): Remove 'value' and add
5269 * doc/invoke.texi: Document changes.
5271 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5273 * doc/sourcebuild.texi (scalar_all_fma): Document.
5274 * tree.def (FMA_EXPR): Delete.
5275 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5276 * internal-fn.c (ternary_direct): New macro.
5277 (expand_ternary_optab_fn): Likewise.
5278 (direct_ternary_optab_supported_p): Likewise.
5279 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5280 * builtins.c (fold_builtin_fma): Delete.
5281 (fold_builtin_3): Don't call it.
5282 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5283 * expr.c (expand_expr_real_2): Likewise.
5284 * fold-const.c (operand_equal_p): Likewise.
5285 (fold_ternary_loc): Likewise.
5286 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5287 * gimple.c (DEFTREECODE): Likewise.
5288 * gimplify.c (gimplify_expr): Likewise.
5289 * optabs-tree.c (optab_for_tree_code): Likewise.
5290 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5291 * tree-eh.c (operation_could_trap_p): Likewise.
5292 (stmt_could_throw_1_p): Likewise.
5293 * tree-inline.c (estimate_operator_cost): Likewise.
5294 * tree-pretty-print.c (dump_generic_node): Likewise.
5295 (op_code_prio): Likewise.
5296 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5297 * tree-ssa-operands.c (get_expr_operands): Likewise.
5298 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5299 * fold-const-call.h (fold_fma): Delete.
5300 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5301 CFN_FNMA and CFN_FNMS.
5303 * genmatch.c (combined_fn): New enum.
5304 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5305 (commutative_op): New function.
5306 (commutate): Use it. Handle more than 2 operands.
5307 (dt_operand::gen_gimple_expr): Use commutative_op.
5308 (parser::parse_expr): Allow :c to be used with non-binary
5309 operators if the commutative operand is known.
5310 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5311 CFN_FMS, CFN_FNMA and CFN_FNMS.
5312 (backprop::process_assign_use): Remove FMA_EXPR handling.
5313 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5314 (gen_hsa_fma): New function.
5315 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5316 IFN_FNMA and IFN_FNMS.
5317 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5318 * gimple-fold.h (follow_all_ssa_edges): Declare.
5319 * gimple-fold.c (follow_all_ssa_edges): New function.
5320 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5321 gimple_build interface and use follow_all_ssa_edges to fold the result.
5322 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5323 instead of checking for optabs directly.
5324 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5325 rather than FMA_EXPRs.
5326 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5327 call to IFN_FMA instead of an FMA_EXPR.
5329 2018-05-17 Jim Wilson <jimw@sifive.com>
5331 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5332 sign extended promoted subreg, and the range does not have the sign bit
5333 set, then do a sign extend.
5335 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5336 test, check for sign extended subreg and/or constant operands, and
5337 do a sign extend in that case.
5339 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5341 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5342 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5344 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5345 Change logics_shift_reg to logics_shift_imm.
5346 (thunderx2t99_fp_loadpair_basic): Delete.
5347 (thunderx2t99_fp_storepair_basic): Delete.
5348 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5349 (thunderx2t99_asimd_polynomial): Delete.
5350 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5351 and neon_fp_mul_d_scalar_q.
5352 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5353 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5354 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5355 (thunderx2t99_asimd_lut): Add missing tbl types.
5356 (thunderx2t99_asimd_ext): Delete.
5357 (thunderx2t99_asimd_load1_1_mult): Delete.
5358 (thunderx2t99_asimd_load1_2_mult): Delete.
5359 (thunderx2t99_asimd_load1_ldp): New.
5360 (thunderx2t99_asimd_load1): New.
5361 (thunderx2t99_asimd_load2): Add missing *load2* types.
5362 (thunderx2t99_asimd_load3): New.
5363 (thunderx2t99_asimd_load4): New.
5364 (thunderx2t99_asimd_store1_1_mult): Delete.
5365 (thunderx2t99_asimd_store1_2_mult): Delete.
5366 (thunderx2t99_asimd_store2_mult): Delete.
5367 (thunderx2t99_asimd_store2_onelane): Delete.
5368 (thunderx2t99_asimd_store_stp): New.
5369 (thunderx2t99_asimd_store1): New.
5370 (thunderx2t99_asimd_store2): New.
5371 (thunderx2t99_asimd_store3): New.
5372 (thunderx2t99_asimd_store4): New.
5374 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5376 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5377 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5379 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5380 Segher Boessenkool <segher@kernel.crashing.org>
5383 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5386 2018-05-17 Richard Biener <rguenther@suse.de>
5388 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5389 for pruning loop and prune defs feeding only already visited PHIs.
5391 2018-05-17 Richard Biener <rguenther@suse.de>
5393 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5395 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5396 Richard Biener <rguenther@suse.de>
5398 PR tree-optimization/85793
5399 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5400 for VMAT_ELEMENTWISE.
5402 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5404 * internal-fn.h (lookup_internal_fn): Declare
5405 * internal-fn.c (lookup_internal_fn): New function.
5406 * gimple.c (gimple_build_call_from_tree): Handle calls to
5408 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5409 internal function names.
5410 * tree-pretty-print.c (dump_generic_node): Likewise.
5411 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5413 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5415 * gimple-fold.h (gimple_build): Make the function forms take
5416 combined_fn rather than built_in_function.
5417 (gimple_simplify): Likewise.
5418 * gimple-match-head.c (gimple_simplify): Likewise.
5419 * gimple-fold.c (gimple_build): Likewise.
5420 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5421 rather than gimple_build_call_internal.
5422 (get_initial_defs_for_reduction): Likewise.
5423 (vect_create_epilog_for_reduction): Likewise.
5424 (vectorizable_live_operation): Likewise.
5426 2018-05-17 Martin Liska <mliska@suse.cz>
5428 * gimple-ssa-sprintf.c (format_directive): Do not use
5429 space in between 'G_' and '('.
5431 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5434 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5435 even if the mask is not all ones.
5438 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5440 (ix86_gimple_fold_builtin): Likewise.
5443 * config/i386/i386.c: Include tree-vector-builder.h.
5444 (ix86_vector_shift_count): New function.
5445 (ix86_fold_builtin): Fold shift builtins by scalar count.
5446 (ix86_gimple_fold_builtin): Likewise.
5448 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5449 _mm512_setzero): New intrinsics.
5451 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5452 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5454 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5455 code generation for cases where splatting a value is not useful.
5456 * simplify-rtx.c (simplify_ternary_operation): Simplify
5457 vec_merge across a vec_duplicate and a paradoxical subreg forming
5458 a vector mode to a vec_concat.
5460 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5462 * config.gcc: Support "goldmont-plus".
5463 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5465 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5466 PROCESSOR_GOLDMONT_PLUS.
5467 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5468 (processor_target_table): Add "goldmont-plus".
5469 (PTA_GOLDMONT_PLUS): Define.
5470 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5471 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5472 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5473 (fold_builtin_cpu): Add "goldmont-plus".
5474 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5475 (ix86_option_override_internal): Add "goldmont-plus".
5476 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5477 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5478 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5479 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5481 2018-05-17 Richard Biener <rguenther@suse.de>
5483 PR tree-optimization/85757
5484 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5485 remove defs that only feed that PHI from further processing.
5487 2018-05-16 Jim Wilson <jimw@sifive.com>
5489 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5491 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5493 2018-05-16 Mark Wielaard <mark@klomp.org>
5495 * dwarf2out.c (count_index_strings): New function.
5496 (output_indirect_strings): Call count_index_strings and generate
5497 header for dwarf_version >= 5.
5499 2018-05-16 Mark Wielaard <mark@klomp.org>
5501 * dwarf2out.c (dwarf_FORM): New function.
5502 (set_indirect_string): Use dwarf_FORM.
5503 (reset_indirect_string): Likewise.
5504 (size_of_die): Likewise.
5505 (value_format): Likewise.
5506 (output_die): Likewise.
5507 (add_skeleton_AT_string): Likewise.
5508 (output_macinfo_op): Likewise.
5509 (index_string): Likewise.
5510 (output_index_string_offset): Likewise.
5511 (output_index_string): Likewise.
5512 (count_index_strings): Likewise.
5514 2018-05-16 Carl Love <cel@us.ibm.com>
5516 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5517 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5519 2018-05-16 Martin Jambor <mjambor@suse.cz>
5521 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5522 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5524 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5526 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5527 (fnma<mode>4): Likewise.
5528 (fms<mode>4): Likewise.
5529 (fnms<mode>4): Likewise.
5530 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5531 (aarch64_fnma<mode>4): Likewise.
5532 (aarch64_fms<mode>4): Likewise.
5533 (aarch64_fnms<mode>4): Likewise.
5534 (aarch64_fnmadd<mode>4): Likewise.
5536 2018-05-16 Jason Merrill <jason@redhat.com>
5538 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5540 2018-05-16 Richard Biener <rguenther@suse.de>
5542 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5543 (dump_stmt_cost): Declare.
5544 (add_stmt_cost): Dump cost we add.
5545 (add_stmt_costs): New function.
5546 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5548 (vect_analyze_stmt): Adjust prototype.
5549 (vectorizable_condition): Likewise.
5550 (vectorizable_live_operation): Likewise.
5551 (vectorizable_reduction): Likewise.
5552 (vectorizable_induction): Likewise.
5553 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5554 cost vector to pass to vectorizable_ and record afterwards.
5555 (vect_model_reduction_cost): Take cost vector argument and adjust.
5556 (vect_model_induction_cost): Likewise.
5557 (vectorizable_reduction): Likewise.
5558 (vectorizable_induction): Likewise.
5559 (vectorizable_live_operation): Likewise.
5560 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5561 SLP_TREE_NUMBER_OF_VEC_STMTS.
5562 (vect_analyze_slp_cost_1): Remove.
5563 (vect_analyze_slp_cost): Likewise.
5564 (vect_slp_analyze_node_operations): Take visited args and
5565 a target cost vector. Avoid processing already visited stmt sets.
5566 (vect_slp_analyze_operations): Use a local cost vector to gather
5567 costs and register those of non-discarded instances.
5568 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5569 (vect_schedule_slp_instance): Remove copying of
5570 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5572 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5573 adding cost. Record cost entry location.
5574 (vect_prologue_cost_for_slp_op): Function to compute cost of
5575 a constant or invariant generated for SLP vect in the prologue,
5576 split out from vect_analyze_slp_cost_1.
5577 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5578 (vect_model_promotion_demotion_cost): Likewise.
5579 (vect_model_store_cost): Likewise, make static.
5580 (vect_model_load_cost): Likewise.
5581 (vectorizable_bswap): Add cost vector arg and adjust.
5582 (vectorizable_call): Likewise.
5583 (vectorizable_simd_clone_call): Likewise.
5584 (vectorizable_conversion): Likewise.
5585 (vectorizable_assignment): Likewise.
5586 (vectorizable_shift): Likewise.
5587 (vectorizable_operation): Likewise.
5588 (vectorizable_store): Likewise.
5589 (vectorizable_load): Likewise.
5590 (vectorizable_condition): Likewise.
5591 (vectorizable_comparison): Likewise.
5592 (can_vectorize_live_stmts): Likewise.
5593 (vect_analyze_stmt): Likewise.
5594 (vect_transform_stmt): Adjust calls to vectorizable_*.
5595 * tree-vectorizer.c: Include gimple-pretty-print.h.
5596 (dump_stmt_cost): New function.
5598 2018-05-16 Richard Biener <rguenther@suse.de>
5600 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5601 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5602 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5603 (check_name): New callback.
5604 (dse_classify_store): Track cycles via a visited bitmap of PHI
5605 defs and simplify handling of in-loop and across loop dead stores
5606 and properly fail for loop-variant refs. Handle byte-tracking with
5607 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5610 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5612 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5613 (vect_get_mask_type_for_stmt): Likewise.
5614 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5616 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5617 to determine the statement's vector type and the vector type that
5618 should be used for calculating nunits. Deal with cases in which
5619 the type has to be deferred.
5620 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5621 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5622 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5623 (vect_determine_vf_for_stmt): New functions, split out from...
5624 (vect_determine_vectorization_factor): ...here.
5625 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5626 (vect_get_mask_type_for_stmt): New functions, split out from
5627 vect_determine_vectorization_factor.
5629 2018-05-16 Richard Biener <rguenther@suse.de>
5631 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5632 verify the [VEC_]COND_EXPR embedded comparison.
5634 2018-05-15 Martin Sebor <msebor@redhat.com>
5636 PR tree-optimization/85753
5637 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5638 RECORD_TYPE in addition to ARRAY_TYPE.
5640 2018-05-15 Martin Sebor <msebor@redhat.com>
5643 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5645 2018-05-15 Richard Biener <rguenther@suse.de>
5647 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5648 add by_clobber_p one. Change algorithm to collect all defs
5649 representing uses we need to walk and try reducing them to
5650 a single one before failing.
5651 (dse_dom_walker::dse_optimize_stmt): Adjust.
5653 2018-05-13 Mark Wielaard <mark@klomp.org>
5655 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5656 (size_of_loc_descr): Likewise.
5657 (output_loc_operands): Likewise.
5658 (output_loc_operands_raw): Likewise.
5659 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5660 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5661 (hash_loc_operands): Likewise.
5662 (compare_loc_operands): Likewise.
5664 2018-05-14 Mark Wielaard <mark@klomp.org>
5666 * dwarf2out.c (count_index_addrs): New function.
5667 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5669 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5671 PR tree-optimization/83648
5672 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5673 return value as malloc candidate.
5675 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5678 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5679 param as true in call to suggest_attribute.
5681 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5683 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5684 -mreadonly-in-sdata.
5686 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5688 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5690 (aarch64_crypto_aesd_fused): Likewise.
5692 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5694 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5695 (movsi_aarch64): Likewise.
5696 (load_pairsi): Likewise.
5697 (load_pairdi): Likewise.
5698 (store_pairsi): Likewise.
5699 (store_pairdi): Likewise.
5700 (load_pairsf): Likewise.
5701 (load_pairdf): Likewise.
5702 (store_pairsf): Likewise.
5703 (store_pairdf): Likewise.
5704 (zero_extend): Likewise.
5705 (trunc): Swap alternatives.
5706 (fcvt_target): Add '?' to prefer w over r.
5708 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5711 * config/i386/i386.md: Disallow non-commutative arithmetics in
5712 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5713 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5714 in the peephole2 before it.
5716 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5718 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5719 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5720 (ix86_handle_option): Handle -mcldemote.
5721 * config.gcc: New header.
5722 * config/i386/cldemoteintrin.h: New file.
5723 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5724 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5726 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5727 OPTION_MASK_ISA_CLDEMOTE.
5728 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5729 (ix86_valid_target_attribute_inner_p): Ditto.
5730 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5731 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5732 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5733 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5734 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5736 * config/i386/i386.opt: Add -mcldemote.
5737 * config/i386/x86intrin.h: New header.
5738 * doc/invoke.texi: Add -mcldemote.
5740 2018-05-14 Richard Biener <rguenther@suse.de>
5742 * doc/match-and-simplify.texi: Adjust :s documentation.
5744 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5746 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5747 intended memcpy size.
5748 (REORDER_45): Likewise.
5750 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5752 * sort.cc: New file.
5753 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5754 * vec.c (qsort_chk): Use gcc_qsort.
5755 * Makefile.in (OBJS-libcommon): Add sort.o.
5756 (build/sort.o): New target. Use it...
5757 (BUILD_RTL): ... here, and...
5758 (build/gencfn-macros): ... here, and...
5759 (build/genmatch): ... here.
5761 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5762 Chung-Ju Wu <jasonwucj@gmail.com>
5764 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5765 * config/nds32/nds32-graywolf.md: New file.
5766 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5767 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5769 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5770 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5771 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5772 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5773 * config/nds32/pipelines.md: Include n15 settings.
5775 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5776 Chung-Ju Wu <jasonwucj@gmail.com>
5778 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5779 * config/nds32/nds32-n13.md: New file.
5780 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5783 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5784 * config/nds32/nds32.md (pipeline_model): Add n13.
5785 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5786 * config/nds32/pipelines.md: Include n13 settings.
5788 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5789 Chung-Ju Wu <jasonwucj@gmail.com>
5791 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5792 * config/nds32/nds32-n10.md: New file.
5793 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5794 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5796 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5797 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5798 * config/nds32/nds32.md (pipeline_model): Add n10.
5799 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5800 * config/nds32/pipelines.md: Include n10 settings.
5802 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5803 Kito Cheng <kito.cheng@gmail.com>
5804 Chung-Ju Wu <jasonwucj@gmail.com>
5806 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5807 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5808 Add enum values for DSP extension instructions.
5809 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5811 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5812 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5814 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5815 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5816 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5817 * config/nds32/nds32-intrinsic.md: Likewise.
5818 * config/nds32/nds32_intrinsic.h: Likewise.
5819 * config/nds32/nds32-md-auxiliary.c: Likewise.
5820 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5821 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5822 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5823 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5824 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5825 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5826 TYPE_DMAC in switch statement.
5827 * config/nds32/nds32.c: New checking and implementation for DSP
5828 extension instructions.
5829 * config/nds32/nds32.h: Likewise.
5830 * config/nds32/nds32.md: Likewise.
5831 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5832 * config/nds32/predicates.md: Implement new predicates for DSP
5835 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5837 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5838 Reformat alternatives and attributes so it is easier to identify
5839 which constraints/attributes go with which instruction.
5840 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5841 (mov<mode>_softfloat32, FMOVE64): Likewise.
5842 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5843 (mov<mode>_softfloat64, FMOVE64): Likewise.
5845 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5847 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5849 (Basic PowerPC Built-in Functions): The new name of the
5850 subsection previously known as "PowerPC Built-in Functions".
5851 (Basic PowerPC Built-in Functions Available on all Configurations):
5853 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5854 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5855 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5856 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5858 2018-05-11 Martin Jambor <mjambor@suse.cz>
5861 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5864 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5867 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5869 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5871 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5872 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5873 (ix86_handle_option): Handle -mwaitpkg.
5874 * config.gcc: New header.
5875 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5876 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5877 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5879 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5880 OPTION_MASK_ISA_WAITPKG.
5881 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5882 (ix86_option_override_internal): Add PTA_WAITPKG.
5883 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5884 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5885 IX86_BUILTIN_TPAUSE.
5886 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5887 __builtin_ia32_umwait and __builtin_ia32_tpause.
5888 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5889 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5890 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5891 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5892 UNSPECV_TPAUSE): New.
5893 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5894 * config/i386/i386.opt: Add -mwaitpkg.
5895 * config/i386/waitpkgintrin.h: New file.
5896 * config/i386/x86intrin.h: New header.
5897 * doc/invoke.texi: Add -mwaitpkg.
5899 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5902 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5904 (cortex-m0): Use armv6s-m isa.
5905 (cortex-m0plus): Likewise.
5906 (cortex-m1): Likewise.
5907 (cortex-m0.small-multiply): Likewise.
5908 (cortex-m0plus.small-multiply): Likewise.
5909 (cortex-m1.small-multiply): Likewise.
5911 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5912 Jakub Jelinek <jakub@redhat.com>
5914 PR tree-optimization/85692
5915 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5916 source permute as well.
5918 2018-05-11 Martin Liska <mliska@suse.cz>
5921 * doc/extend.texi: Document LLVM style format for no_sanitize
5924 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5926 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5927 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5928 (mode_supports_vsx_dform_quad): Likewise.
5929 (mode_supports_vmx_dform): Move these functions to be next to the
5930 other mode_supports functions.
5931 (mode_supports_dq_form): Likewise.
5932 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5933 mode_supports_dq_form.
5934 (reg_offset_addressing_ok_p): Likewise.
5935 (offsettable_ok_by_alignment): Likewise.
5936 (rs6000_legitimate_offset_address_p): Likewise.
5937 (legitimate_lo_sum_address_p): Likewise.
5938 (rs6000_legitimize_address): Likewise.
5939 (rs6000_legitimize_reload_address): Likewise.
5940 (rs6000_secondary_reload_inner): Likewise.
5941 (rs6000_preferred_reload_class): Likewise.
5942 (rs6000_output_move_128bit): Likewise.
5944 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5946 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5947 Generate SImode target register for null target.
5948 <case IX86_BUILTIN_XGETBV>: Ditto.
5949 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5950 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5952 2018-05-10 Carl Love <cel@us.ibm.com>
5954 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5955 dcbtt and dcbtstt if operands[2] is 0.
5957 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5960 * config/i386/sse.md (usadv64qi): New expander.
5962 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5964 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5965 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5966 -maltivec=be support.
5967 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5968 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5969 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5970 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5971 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5972 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5973 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5974 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5975 altivec_vsumsws): Adjust.
5976 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5977 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5978 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5980 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5981 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5982 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5983 (altivec_lve<VI_char>x): Delete expand.
5984 (*altivec_lve<VI_char>x_internal): Rename to...
5985 (altivec_lve<VI_char>x): ... this.
5986 (altivec_lvxl_<mode>): Delete expand.
5987 (*altivec_lvxl_<mode>_internal): Rename to ...
5988 (altivec_lvxl_<mode>): ... this.
5989 (altivec_stvxl_<mode>): Delete expand.
5990 (*altivec_stvxl_<mode>_internal): Rename to ...
5991 (altivec_stvxl_<mode>): ... this.
5992 (altivec_stve<VI_char>x): Delete expand.
5993 (*altivec_stve<VI_char>x_internal): Rename to ...
5994 (altivec_stve<VI_char>x): ... this.
5995 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5996 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5997 reduc_plus_scal_<mode>): Adjust.
5998 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6000 (rs6000_cpu_cpp_builtins): Adjust.
6001 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6002 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6003 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6004 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6005 -maltivec=be support.
6006 (rs6000_split_vec_extract_var): Adjust.
6007 (rs6000_split_v4si_init): Adjust.
6008 (swap_selector_for_mode): Delete.
6009 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6010 altivec_expand_stvex_be): Delete.
6011 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6012 -maltivec=be support.
6013 (rs6000_gimple_fold_builtin): Ditto.
6014 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6016 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6017 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6018 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6019 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6020 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6021 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6022 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6023 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6024 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6025 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6026 anonymous split): Adjust.
6027 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6028 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6030 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6032 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6033 when --with-gxx-include-dir is also specified.
6034 * configure: Regenerate.
6036 2018-05-09 Jim Wilson <jimw@sifive.com>
6039 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6040 * config/riscv/t-withmultilib: New.
6041 * config/riscv/withmultilib.h: New.
6042 * doc/install.texi: Document RISC-V --with-multilib-list support.
6044 2018-05-09 Richard Biener <rguenther@suse.de>
6046 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6048 (vect_bb_vectorization_profitable_p): Adjust. Compute
6049 actual scalar cost using the cost vector and the add_stmt_cost
6052 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6054 PR rtl-optimization/85645
6055 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6056 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6058 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6060 PR rtl-optimization/85645
6061 * shrink-wrap.c (spread_components): Return a boolean saying if
6062 anything was changed.
6063 (try_shrink_wrapping_separate): Iterate spread_components until
6064 nothing changes anymore.
6066 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6068 PR rtl-optimization/85645
6069 * regrename.c (build_def_use): Also kill the chains that include the
6070 destination of a REG_CFA_REGISTER note.
6072 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6074 PR rtl-optimization/85645
6075 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6076 insn that has a REG_CFA_REGISTER note.
6078 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6080 * cfgexpand.c (expand_clobber): New function.
6081 (expand_gimple_stmt_1): Use it.
6082 * tree-vect-stmts.c (vect_clobber_variable): New function,
6084 (vectorizable_simd_clone_call): ...here.
6085 (vectorizable_store): Emit a clobber either side of an
6086 IFN_STORE_LANES sequence.
6087 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6089 2018-05-09 Tom de Vries <tom@codesourcery.com>
6092 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6093 (define_insn "trap_if_false"): Add exit after trap.
6095 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6097 PR rtl-optimization/85638
6098 * bb-reorder.c: Include common/common-target.h.
6099 (create_forwarder_block): New function extracted from...
6100 (fix_up_crossing_landing_pad): ...here. Rename into...
6101 (dw2_fix_up_crossing_landing_pad): ...this.
6102 (sjlj_fix_up_crossing_landing_pad): New function.
6103 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6104 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6105 from both partitions and exit the loop after one iteration.
6107 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6110 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6112 (Basic PowerPC Built-in Functions): The new name of the
6113 subsection previously known as "PowerPC Built-in Functions".
6114 (Basic PowerPC Built-in Functions Available on all Configurations):
6116 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6118 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6120 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6122 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6125 2018-05-08 Jim Wilson <jimw@sifive.com>
6127 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6128 (LD_EMUL_SUFFIX): New.
6129 (LINK_SPEC): Use it.
6131 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6133 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6135 (Basic PowerPC Built-in Functions): The new name of the
6136 subsection previously known as "PowerPC Built-in Functions".
6137 (Basic PowerPC Built-in Functions Available on all Configurations):
6139 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6141 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6143 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6145 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6148 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6151 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6152 after cmpelim optimization.
6154 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6156 * config.gcc: Support "goldmont".
6157 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6158 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6160 * config/i386/i386.c (m_GOLDMONT): Define.
6161 (processor_target_table): Add "goldmont".
6162 (PTA_GOLDMONT): Define.
6163 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6164 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6165 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6166 (fold_builtin_cpu): Add "goldmont".
6167 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6168 (ix86_option_override_internal): Add "goldmont".
6169 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6170 (processor_type): Add PROCESSOR_GOLDMONT.
6171 * config/i386/i386.md: Add CPU "glm".
6172 * config/i386/glm.md: New file.
6173 * config/i386/x86-tune.def: Add m_GOLDMONT.
6174 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6176 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6179 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6181 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6182 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6183 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6186 * config/i386/i386.c (ix86_fold_builtin): Handle
6187 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6190 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6191 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6193 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6196 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6197 (check_arch): Likewise.
6198 (check_fpu): Return the result rather than printing it.
6199 (end arch): Fix operator precedence.
6200 (end cpu): Likewise.
6201 (END): Print the result from check_fpu.
6203 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6204 Alan Hayward <alan.hayward@arm.com>
6205 David Sherwood <david.sherwood@arm.com>
6207 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6208 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6209 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6210 (*fcmuo<mode>_and): New patterns.
6212 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6214 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6215 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6216 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6217 (cmp_op, sve_imm_con): New code attributes.
6218 (SVE_COND_INT_CMP, imm_con): Delete.
6219 (cmp_op): Remove above unspecs from int attribute.
6220 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6222 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6223 comparison-specific unspecs.
6224 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6225 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6226 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6227 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6228 (*vec_fcm<cmp_op><mode>): Rename to...
6229 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6230 (*vec_fcmuo<mode>): Rename to...
6231 (*fcmuo<mode>): ...this and adjust likewise.
6232 (*pred_fcm<cmp_op><mode>): New pattern.
6233 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6234 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6236 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6238 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6239 (aarch64_emit_sve_predicated_cond): New function.
6240 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6241 (aarch64_emit_unspec_cond_or): Replace with...
6242 (aarch64_emit_sve_or_conds): ...this new function. Use
6243 aarch64_emit_sve_ptrue_op for the individual comparisons and
6244 aarch64_emit_binop to OR them together.
6245 (aarch64_emit_inverted_unspec_cond): Replace with...
6246 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6247 aarch64_emit_sve_ptrue_op for the comparison and
6248 aarch64_emit_unop to invert the result.
6249 (aarch64_expand_sve_vec_cmp_float): Update after the above
6250 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6252 2018-05-07 Nathan Sidwell <nathan@acm.org>
6254 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6255 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6256 (Backwards Compatibility): Likewise.
6258 2018-05-07 Luis Machado <luis.machado@linaro.org>
6262 2018-05-07 Luis Machado <luis.machado@linaro.org>
6264 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6265 <prefetch_dynamic_strides>: New const bool field.
6266 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6267 prefetch_dynamic_strides.
6268 (exynosm1_prefetch_tune): Likewise.
6269 (thunderxt88_prefetch_tune): Likewise.
6270 (thunderx_prefetch_tune): Likewise.
6271 (thunderx2t99_prefetch_tune): Likewise.
6272 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6274 (aarch64_override_options_internal): Update to set
6275 PARAM_PREFETCH_DYNAMIC_STRIDES.
6276 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6277 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6278 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6279 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6280 prefetch-dynamic-strides setting.
6282 2018-05-07 Luis Machado <luis.machado@linaro.org>
6284 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6285 <minimum_stride>: New const int field.
6286 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6287 minimum_stride field.
6288 (exynosm1_prefetch_tune): Likewise.
6289 (thunderxt88_prefetch_tune): Likewise.
6290 (thunderx_prefetch_tune): Likewise.
6291 (thunderx2t99_prefetch_tune): Likewise.
6292 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6293 (aarch64_override_options_internal): Update to set
6294 PARAM_PREFETCH_MINIMUM_STRIDE.
6295 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6296 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6297 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6298 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6299 stride is constant and is below the minimum stride threshold.
6301 2018-05-07 Luis Machado <luis.machado@linaro.org>
6303 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6306 2018-05-07 Luis Machado <luis.machado@linaro.org>
6308 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6309 <prefetch_dynamic_strides>: New const bool field.
6310 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6311 prefetch_dynamic_strides.
6312 (exynosm1_prefetch_tune): Likewise.
6313 (thunderxt88_prefetch_tune): Likewise.
6314 (thunderx_prefetch_tune): Likewise.
6315 (thunderx2t99_prefetch_tune): Likewise.
6316 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6318 (aarch64_override_options_internal): Update to set
6319 PARAM_PREFETCH_DYNAMIC_STRIDES.
6320 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6321 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6322 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6323 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6324 prefetch-dynamic-strides setting.
6326 2018-05-07 Luis Machado <luis.machado@linaro.org>
6328 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6329 <minimum_stride>: New const int field.
6330 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6331 minimum_stride field.
6332 (exynosm1_prefetch_tune): Likewise.
6333 (thunderxt88_prefetch_tune): Likewise.
6334 (thunderx_prefetch_tune): Likewise.
6335 (thunderx2t99_prefetch_tune): Likewise.
6336 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6337 (aarch64_override_options_internal): Update to set
6338 PARAM_PREFETCH_MINIMUM_STRIDE.
6339 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6340 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6341 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6342 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6343 stride is constant and is below the minimum stride threshold.
6345 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6348 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6349 the type is addressable. Don't force op into register if it has
6352 2018-05-05 Roland McGrath <mcgrathr@google.com>
6355 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6356 any section for which we don't know a specific type it should have,
6357 regardless of name. Previously this was done only for the exact
6358 names ".init_array", ".fini_array", and ".preinit_array".
6359 (default_elf_asm_named_section): Add comment about
6360 relationship with default_section_type_flags and SECTION_NOTYPE.
6361 (get_section): Don't consider it a type conflict if one side has
6362 SECTION_NOTYPE and the other doesn't, as long as neither has the
6363 SECTION_BSS et al used in the default_section_type_flags logic.
6365 2018-05-05 Tom de Vries <tom@codesourcery.com>
6368 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6369 (workaround_barsyncs): New function.
6370 (nvptx_reorg): Use workaround_barsyncs.
6371 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6372 (define_expand "nvptx_membar_cta"): New define_expand.
6373 (define_insn "*nvptx_membar_cta"): New insn.
6375 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6377 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6378 To improve optimization opportunities.
6379 * builtin-types.def: The new needed builtin types for the above.
6381 2018-05-04 Richard Biener <rguenther@suse.de>
6383 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6384 * gimple-ssa-store-merging.c
6385 (imm_store_chain_info::output_merged_store): Remove redundant create,
6386 release split_store vector contents on failure.
6387 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6388 scalar stmt vector on cache hit.
6390 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6392 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6394 * config.gcc (powerpc-xilinx-eabi*): Remove.
6395 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6397 (fusion_addis_mem_combo_load): Ditto.
6398 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6400 (rs6000_cpu_cpp_builtins): Ditto.
6401 * config/rs6000/rs6000-linux.c
6402 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6403 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6404 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6406 (rs6000_setup_reg_addr_masks): Ditto.
6407 (rs6000_init_hard_regno_mode_ok): Ditto.
6408 (rs6000_option_override_internal): Ditto.
6409 (legitimate_lo_sum_address_p): Ditto.
6410 (rs6000_legitimize_address): Ditto.
6411 (rs6000_legitimize_reload_address): Ditto.
6412 (rs6000_legitimate_address_p): Ditto.
6413 (abi_v4_pass_in_fpr): Ditto.
6414 (setup_incoming_varargs): Ditto.
6415 (rs6000_gimplify_va_arg): Ditto.
6416 (rs6000_split_multireg_move): Ditto.
6417 (rs6000_savres_strategy): Ditto.
6418 (rs6000_emit_prologue_components): Ditto.
6419 (rs6000_emit_epilogue_components): Ditto.
6420 (rs6000_emit_prologue): Ditto.
6421 (rs6000_emit_epilogue): Ditto.
6422 (rs6000_elf_file_end): Ditto.
6423 (rs6000_function_value): Ditto.
6424 (rs6000_libcall_value): Ditto.
6425 * config/rs6000/rs6000.h: Ditto.
6426 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6427 (TARGET_MINMAX): ... this. New.
6428 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6429 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6430 (*movsi_internal1_single): Delete.
6431 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6432 mfpu=, mxilinx-fpu): Delete.
6433 * config/rs6000/singlefp.h: Delete.
6434 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6435 * config/rs6000/t-rs6000: Ditto.
6436 * config/rs6000/t-xilinx: Delete.
6437 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6438 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6439 (VStype_simple): Delete.
6440 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6441 * config/rs6000/xfpu.h: Delete.
6442 * config/rs6000/xfpu.md: Delete.
6443 * config/rs6000/xilinx.h: Delete.
6444 * config/rs6000/xilinx.opt: Delete.
6445 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6446 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6448 2018-05-04 Tom de Vries <tom@codesourcery.com>
6451 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6454 2018-05-04 Richard Biener <rguenther@suse.de>
6457 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6458 (expand_complex_div_wide): Likewise.
6459 (expand_complex_operations_1): Likewise.
6460 (expand_complex_libcall): Preserve EH info of the original stmt.
6461 (tree_lower_complex): Handle removed blocks.
6462 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6463 on complex multiplication and division libcall builtins.
6465 2018-05-04 Richard Biener <rguenther@suse.de>
6468 * fold-const.c (negate_expr_p): Restrict negation of operand
6469 zero of a division to when we know that can happen without
6471 (fold_negate_expr_1): Likewise.
6473 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6476 * real.h (real_nextafter): Declare.
6477 * real.c (real_nextafter): New function.
6478 * fold-const-call.c (fold_const_nextafter): New function.
6479 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6480 CASE_CFN_NEXTTOWARD.
6481 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6482 even when arg1_mode is different from arg0_mode.
6484 2018-05-03 Nathan Sidwell <nathan@acm.org>
6486 * doc/extend.texi (Deprecated Features): Remove
6488 (Backwards Compatibility): Likewise.
6489 * doc/invoke.texi (C++ Language Options): Likewise.
6490 (C++ Dialect Options): Likewise.
6492 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6495 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6496 _mm512_mask_mullox_epi64): New intrinsics.
6498 2018-05-03 Tom de Vries <tom@codesourcery.com>
6501 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6502 dump files): Add offload-tree.
6504 2018-05-03 Richard Biener <rguenther@suse.de>
6506 PR tree-optimization/85615
6507 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6508 to loops not nested in BBs loop father to avoid creating multi-entry
6511 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6513 PR tree-optimization/70291
6514 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6515 arguments. Change return type to tree. Emit libcall as a new
6516 statement rather than replacing existing one when inplace_p is true.
6517 (expand_complex_multiplication_components): New function.
6518 (expand_complex_multiplication): Expand floating-point complex
6519 multiplication using the above.
6520 (expand_complex_division): Rename inner_type parameter to type.
6521 Update expand_complex_libcall call-site.
6522 (expand_complex_operations_1): Update expand_complex_multiplication
6523 and expand_complex_division call-sites.
6525 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6528 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6529 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6530 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6531 the highest significant bit of the shift count mask is clear. In
6532 check whether and[sq]i3 is needed verify that all significant bits
6533 of the shift count other than the highest are set.
6535 2018-05-02 Tom de Vries <tom@codesourcery.com>
6538 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6539 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6540 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6541 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6542 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6543 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6544 __builtin_goacc_parlevel_size.
6546 2018-05-02 Richard Biener <rguenther@suse.de>
6548 PR tree-optimization/85597
6549 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6550 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6553 2018-05-02 Tom de Vries <tom@codesourcery.com>
6556 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6557 dump files): Add ltrans-tree.
6559 2018-05-02 Tom de Vries <tom@codesourcery.com>
6562 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6563 dump files): Add wpa-ipa.
6565 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6567 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6568 powerpc*-*-linux*paired* target.
6569 * config/rs6000/750cl.h: Delete.
6570 * config/rs6000/paired.h: Delete.
6571 * config/rs6000/paired.md: Delete.
6572 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6574 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6575 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6576 comment. Remove paired float support.
6577 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6578 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6580 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6581 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6583 * config/rs6000/rs6000.c: Remove paired float support.
6584 (paired_expand_vector_init, paired_expand_vector_move,
6585 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6586 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6587 paired_expand_builtin, paired_expand_predicate_builtin,
6588 paired_init_builtins): Delete.
6589 * config/rs6000/rs6000.h: Remove paired float support.
6590 * config/rs6000/rs6000.md: Remove paired float support.
6591 (move_from_CR_ov_bit): Delete.
6592 * config/rs6000/rs6000.opt (mpaired): Delete.
6593 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6594 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6596 2018-05-02 Richard Biener <rguenther@suse.de>
6599 * gimplify.c (gimplify_save_expr): When in SSA form allow
6600 SAVE_EXPRs to compute to SSA vars.
6602 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6605 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6606 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6607 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6608 clobber operands[2], instead use a new pseudo. Formatting fixes.
6610 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6612 PR tree-optimization/85586
6613 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6614 exit early for statements in the same group if the accesses are
6617 2018-05-02 Tom de Vries <tom@codesourcery.com>
6620 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6623 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6625 PR tree-optimization/85143
6626 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6628 2018-05-01 Tom de Vries <tom@codesourcery.com>
6631 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6632 not found" error message.
6634 2018-05-01 Tom de Vries <tom@codesourcery.com>
6637 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6638 * vec.c (test_ordered_remove_if): New function.
6639 (vec_c_tests): Call test_ordered_remove_if.
6640 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6641 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6642 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6643 VEC_ORDERED_REMOVE_IF.
6645 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6647 PR tree-optimization/82665
6648 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6649 pointer subtraction where arguments come from a memchr call.
6651 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6653 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6654 --push-state --as-needed and --pop-state instead of --as-needed and
6655 --no-as-needed if ld supports it.
6656 * configure: Regenerated.
6659 * doc/install.texi2html: Replace _002d with - and _002a with * in
6660 generated html files using sed.
6662 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6665 * gcc-rich-location.c (blank_line_before_p): New function.
6666 (use_new_line): New function.
6667 (gcc_rich_location::add_fixit_insert_formatted): New function.
6668 * gcc-rich-location.h
6669 (gcc_rich_location::add_fixit_insert_formatted): New function.
6671 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6673 * selftest.c (assert_streq): Rename "expected" and "actual" to
6674 "val1" and "val2". Extend NULL-handling to cover both inputs
6675 symmetrically, while still requiring both to be non-NULL for a pass.
6676 * selftest.h (assert_streq): Rename "expected" and "actual" to
6678 (ASSERT_EQ): Likewise.
6679 (ASSERT_EQ_AT): Likewise.
6680 (ASSERT_KNOWN_EQ): Likewise.
6681 (ASSERT_KNOWN_EQ_AT): Likewise.
6682 (ASSERT_NE): Likewise.
6683 (ASSERT_MAYBE_NE): Likewise.
6684 (ASSERT_MAYBE_NE_AT): Likewise.
6685 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6686 the assertion to pass.
6687 (ASSERT_STREQ_AT): Likewise.
6689 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6691 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6692 interaction with -pie.
6694 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6696 * selftest.h: Fix alphabetization of per-source-file selftest
6699 2018-04-30 Jason Merrill <jason@redhat.com>
6701 PR c++/61982 - dead stores to destroyed objects.
6702 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6705 2018-04-30 Jason Merrill <jason@redhat.com>
6707 * tree.c (build_clobber): New.
6708 * tree.h: Declare it.
6709 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6711 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6713 * diagnostic-show-locus.c (layout::layout): Update for
6714 location_get_source_line returning a char_span.
6715 (struct char_span): Move to input.h.
6716 (struct correction): Update for fields in char_span becoming
6718 (struct source_line): Update for location_get_source_line
6719 returning a char_span.
6720 (layout::print_line): Likewise.
6721 * edit-context.c (edited_file::print_content): Likewise.
6722 (edited_file::print_diff_hunk): Likewise.
6723 (edited_file::print_run_of_changed_lines): Likewise.
6724 (edited_file::get_num_lines): Likewise.
6725 (edited_line::edited_line): Likewise.
6726 * final.c (asm_show_source): Likewise.
6727 * input.c (location_get_source_line): Convert return type
6728 from const char * to char_span, losing the final "line_len"
6730 (dump_location_info): Update for the above.
6731 (get_substring_ranges_for_loc): Likewise. Use a char_span
6732 when handling the literal within the line.
6733 (test_reading_source_line): Update for location_get_source_line
6734 returning a char_span.
6735 * input.h (class char_span): Move here from
6736 diagnostic-show-locus.c, converting from a struct to a class.
6737 Make data members private.
6738 (char_span::operator bool): New.
6739 (char_span::length): New.
6740 (char_span::get_buffer): New.
6741 (char_span::operator[]): New.
6742 (char_span::subspan): Make const.
6743 (char_span::xstrdup): New.
6744 (location_get_source_line): Convert return type from const char *
6745 to char_span, losing the final "line_size" param.
6747 2018-04-30 Jan Hubicka <jh@suse.cz>
6749 * lto-wrapper.c (ltrans_priorities): New static var.
6750 (cmp_priority): New.
6751 (run_gcc): Read priorities and if doing parallel build order
6752 the Makefile by them.
6754 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6756 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6758 2018-04-30 Richard Biener <rguenther@suse.de>
6760 * tree-cfg.c (verify_address): Remove base argument, add
6761 flag whether to check TREE_ADDRESSABLE and do that.
6762 (verify_expr): Remove.
6763 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6764 (verify_gimple_assign_single): Likewise.
6765 (verify_gimple_switch): Likewise.
6766 (verify_expr_location_1): Dereference tp once. Add (disabled)
6767 piece from verify_expr.
6768 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6770 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6772 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6774 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6776 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6777 (small_data_pattern): Likewise.
6778 (arc_rewrite_small_data): Likewise.
6779 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6780 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6781 (get_symbol_alignment): New function.
6782 (legitimate_small_data_address_p): Likewise.
6783 (legitimate_scaled_address): Update, call
6784 legitimate_small_data_address_p.
6785 (output_sdata): New static variable.
6786 (arc_print_operand): Update how we handle small data operands.
6787 (arc_print_operand_address): Likewise.
6788 (arc_legitimate_address_p): Update, use
6789 legitimate_small_data_address_p.
6790 (arc_rewrite_small_data_p): Remove.
6791 (arc_rewrite_small_data_1): Likewise.
6792 (arc_rewrite_small_data): Likewise.
6793 (small_data_pattern): Likewise.
6794 (compact_sda_memory_operand): Update to use
6795 legitimate_small_data_address_p and get_symbol_alignment.
6796 (prepare_move_operands): Don't rewite sdata pattern.
6797 (prepare_extend_operands): Remove.
6798 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6800 (zero_extendqisi2): Likewise.
6801 (zero_extendhisi2): Likewise.
6802 (extendqihi2): Likewise.
6803 (extendqisi2): Likewise.
6804 (extendhisi2): Likewise.
6808 * config/arc/constraints.md (Usd): Change it to memory constraint.
6810 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6812 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6813 as source of std instructions.
6814 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6815 allow 6-bit constants as source for store instructions.
6816 (movdi_insn): Update instruction pattern to allow 6-bit constants
6817 as source for store instructions.
6819 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6821 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6823 2018-04-30 Nathan Sidwell <nathan@acm.org>
6824 Sandra Loosemore <sandra@codesourcery.com>
6826 * dumpfile.c (dump_open): Allow '-' for stdout.
6827 * doc/invoke.texi (Developer Options): Document dump filename
6828 determination early. Document stdin/stdout selection.
6830 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6832 Microblaze Target: PIC data text relative
6834 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6835 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6837 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6838 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6839 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6840 New addressing mode for data-text relative position indepenedent code.
6841 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6842 'ADDRESS_SYMBOLIC_TXT_REL'.
6843 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6844 (microblaze_legitimate_pic_operand): Exclude function calls from
6845 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6846 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6848 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6849 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6850 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6851 for 'address + offset'.
6852 (microblaze_expand_prologue): Add new function prologue call for
6854 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6855 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6856 table in case of TARGET_PIC_DATA_TEXT_REL.
6857 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6858 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6859 Add new macros 'UNSPEC_TEXT',
6860 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6861 + exclude function calls from 'UNSPEC_PLT' in case of data text
6863 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6864 new target hook for generating address diff vector tables in case of
6866 * doc/tm.texi : Regenerate.
6867 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6868 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6869 of addr diff vector generation.
6870 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6871 target hook definition.
6872 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6873 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6874 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6875 option for data text relative.
6877 2018-04-30 Richard Biener <rguenther@suse.de>
6879 * tree-chrec.h (evolution_function_is_constant_p): Remove
6881 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6883 2018-04-30 Richard Biener <rguenther@suse.de>
6886 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6888 2018-04-30 Richard Biener <rguenther@suse.de>
6890 PR tree-optimization/28364
6891 PR tree-optimization/85275
6892 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6893 copying first exit test.
6895 2018-04-28 Mark Wielaard <mark@klomp.org>
6897 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6899 (dwarf_AT): Handle DW_AT_addr_base.
6900 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6902 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6905 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6906 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6907 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6908 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6910 2018-04-28 Richard Biener <rguenther@suse.de>
6912 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6913 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6914 to reflect use. Only add interesting stmts.
6916 2018-04-27 Martin Jambor <mjambor@suse.cz>
6919 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6920 the jump function allows for passing through aggregate values.
6922 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6924 * input.h (in_system_header_at): Convert from macro to inline
6926 (from_macro_expansion_at): Likewise.
6927 (from_macro_definition_at): Likewise.
6929 2018-04-27 Jeff Law <law@redhat.com>
6931 * config.gcc: Mark tile* targets as deprecated/obsolete.
6933 2018-04-27 Richard Biener <rguenther@suse.de>
6935 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6938 2018-04-27 Richard Biener <rguenther@suse.de>
6940 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6942 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6944 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6945 with Yd constraint. Set "preferred_for_speed" attribute from
6946 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6948 (*movdi_internal): Ditto.
6949 (movti_interunit splitters): Remove
6950 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6951 (movdi_interunit splitters): Ditto.
6952 * config/i386/constraints.md (Ye): Remove.
6953 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6955 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6958 * config/aarch64/constraints.md (Usg): Limit to 31.
6961 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6963 PR tree-optimization/85529
6964 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6965 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6966 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6967 zero extension or masking of the MSB bit.
6968 (optimize_range_tests): Add FIRST_BB argument, pass it through
6969 to optimize_range_tests_var_bound.
6970 (maybe_optimize_range_tests, reassociate_bb): Adjust
6971 optimize_range_tests callers.
6973 2018-04-26 Richard Biener <rguenther@suse.de>
6974 Jakub Jelinek <jakub@redhat.com>
6976 * cgraph.h (symbol_table): Just declare debug method here.
6977 * symtab.c (symbol_table::debug): Define.
6979 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6981 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6983 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6985 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6986 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6987 (*movdi_internal): Substitute Yi and Yj constraint with x
6988 and Ym and Yn constraint with y constraint. Update "isa"
6989 attribute and set "preferred_for_speed" attribute from
6990 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6991 (*movsi_internal): Ditto.
6992 (*movdf_internal): Ditto.
6993 (*movsf_internal): Ditto.
6994 (*zero_extendsidi2): Ditto.
6995 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6996 (sse2_loadld): Ditto.
6997 (*vec_extract<ssevecmodelower>_0): Ditto.
6998 (*vec_extractv4si_0_zext_sse4): Ditto.
6999 (vec_concatv2di): Ditto.
7000 (*vec_dup<mode>): Ditto.
7001 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7002 * config/i386/constraints.md (Yi): Remove.
7007 2018-04-26 Nathan Sidwell <nathan@acm.org>
7009 * dumpfile.c (dump_open): New.
7010 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7011 (dump_finish): Detect stdio/stderr by value not name.
7013 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7015 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7017 2018-04-26 Tom de Vries <tom@codesourcery.com>
7020 * config/nvptx/nvptx.c (verify_neutering_jumps)
7021 (verify_neutering_labels): New function
7022 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7024 2018-04-26 Tom de Vries <tom@codesourcery.com>
7027 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7028 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7031 2018-04-26 Richard Biener <rguenther@suse.de>
7032 Tom de Vries <tom@codesourcery.com>
7035 * lto-streamer-out.c (output_function): Fixup loops if required to match
7036 discovery done in the reader.
7038 2018-04-26 Richard Biener <rguenther@suse.de>
7040 PR tree-optimization/85116
7041 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7042 have a loop exit from the single latch predecessor. Remove
7043 case of header with just condition.
7044 (ch_base::copy_headers): Exclude infinite loops from any
7046 (pass_ch::execute): Record exits.
7048 2018-04-26 Richard Biener <rguenther@suse.de>
7050 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7051 prologue cost vector and pass it to vect_get_load_cost.
7052 (vect_get_peeling_costs_all_drs): Likewise.
7053 (vect_peeling_hash_get_lowest_cost): Likewise.
7054 (vect_enhance_data_refs_alignment): Likewise.
7056 2018-04-26 Richard Biener <rguenther@suse.de>
7059 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7060 checking of integer<->pointer conversions.
7061 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7062 sign-/zero-extending pointer types.
7063 (expand_omp_for_static_chunk): Likewise.
7065 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7066 Jean Lee <xiaoyur347@gmail.com>
7068 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7069 (TARGET_ASAN_SHADOW_OFFSET): Define.
7070 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7071 true for -fsanitize=address.
7073 2018-04-25 Mark Wielaard <mark@klomp.org>
7075 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7078 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7080 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7081 than "alu", remove explicit "memory" and "imm_disp" attributes.
7082 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7085 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7086 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7087 gen_lowpart_no_emit.
7089 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7092 * config/i386/i386.c (ix86_expand_builtin): Change memory
7093 operand to XI, extend p0 to Pmode.
7094 * config/i386/i386.md: Change unspec volatile and operand
7095 1 mode to XI, change operand 0 mode to P.
7097 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7099 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7100 GET_MODE_MASK before any checking.
7101 (nds32_can_use_bset_p): Likewise.
7102 (nds32_can_use_btgl_p): Likewise.
7104 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7106 * config/nds32/nds32-doubleword.md: New define_split pattern for
7107 illegal register number.
7109 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7111 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7113 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7115 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7117 2018-04-25 Richard Biener <rguenther@suse.de>
7119 * lto-streamer.h (LTO_major_version): Bump to 8.
7121 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7123 * BASE-VER: Set to 9.0.0.
7125 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7127 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7128 in __abskf2 and __powikf2.
7130 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7133 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7134 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7135 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7136 Use the above on operand 2. Reindent.
7137 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7139 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7142 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7144 * config/i386/i386.opt (mcet): Removed.
7145 * doc/install.texi: Remove -mcet documentation.
7146 * doc/invoke.texi: Likewise.
7148 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7151 * doc/install.texi: Remove -mcet from bootstrap-cet.
7153 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7156 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7157 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7161 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7162 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7163 containing a CONST_VECTOR.
7165 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7167 * doc/install.texi: Update newlib dependency for nvptx.
7169 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7172 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7173 instead of INTVAL when shifting x left.
7175 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7177 PR tree-optimization/85478
7178 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7179 vect_grouped_store_supported for single element vectors.
7181 2018-04-24 Richard Biener <rguenther@suse.de>
7184 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7185 load cost increase to the case of non-constant step.
7187 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7190 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7191 destination if any_malformed_asm.
7193 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7196 * expr.c (store_field): In the bitfield case, if the value comes from
7197 a function call and is returned in registers by means of a PARALLEL,
7198 do not change the mode of the temporary unless BLKmode and VOIDmode.
7200 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7202 PR rtl-optimization/85423
7203 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7204 dependencies to debug insns when the previous insn is non-debug.
7206 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7208 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7209 enums into a single definition.
7210 (fls): Fix predicates and printing.
7213 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7215 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7216 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7217 and short u6 immediate.
7218 (check_if_valid_sleep_operand): Remove.
7219 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7221 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7223 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7224 flag_always_save_lp condition.
7225 * config/nds32/nds32.opt (malways-save-lp): New option.
7227 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7229 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7230 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7231 * config/nds32/nds32.h
7232 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7233 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7235 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7237 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7238 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7240 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7241 Chung-Ju Wu <jasonwucj@gmail.com>
7243 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7245 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7246 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7248 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7250 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7252 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7254 * config/nds32/nds32-protos.h (nds32_data_alignment,
7255 nds32_local_alignment): Declare.
7256 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7257 nds32_local_alignment): New functions.
7258 (TARGET_CONSTANT_ALIGNMENT): Define.
7259 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7261 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7263 * config/nds32/nds32.c
7264 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7265 (TARGET_MODES_TIEABLE_P): Likewise.
7267 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7272 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7273 Chung-Ju Wu <jasonwucj@gmail.com>
7275 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7276 for unaligned access.
7277 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7279 * config/nds32/nds32-intrinsic.md: Likewise.
7280 * config/nds32/nds32_intrinsic.h: Likewise.
7281 * config/nds32/nds32.h (nds32_builtins): Likewise.
7282 * config/nds32/nds32.opt (munaligned-access): New option.
7283 * config/nds32/nds32.c (nds32_asm_file_start): Display
7284 flag_unaligned_access status.
7286 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7288 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7289 -mno-relax is present.
7290 * config/riscv/linux.h (LINK_SPEC): Ditto.
7292 2018-04-20 Martin Sebor <msebor@redhat.com>
7295 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7297 (gimple_fold_builtin_stxcpy_chk): Same.
7298 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7300 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7303 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7304 __powikf2 when long double is IEEE 128-bit.
7306 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7308 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7309 step to make sure stack always aligned.
7311 2018-04-20 Carl Love <cel@us.ibm.com>
7314 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7315 size check for arg0.
7317 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7318 Tom de Vries <tom@codesourcery.com>
7321 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7322 Emit insns for calls too.
7323 (nvptx_find_par): Always look for worker-level predecessor insn.
7324 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7326 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7327 (nvptx_process_pars): Propagate frames for calls.
7329 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7332 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7334 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7335 (ix86_handle_option): Don't handle OPT_mibt.
7336 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7338 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7340 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7341 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7342 (ix86_target_macros): Define __CET__ with flag_cf_protection
7343 for -fcf-protection.
7344 * config/i386/i386.c (isa2_opts): Remove -mibt.
7345 * config/i386/i386.h (TARGET_IBT): Removed.
7346 (TARGET_IBT_P): Likewise.
7347 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7348 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7349 * config/i386/i386.opt (mcet): Update help message.
7352 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7353 -mcet as an alias for -mshstk.
7355 2018-04-20 Richard Biener <rguenther@suse.de>
7358 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7359 complexity by forcing a single use of the multiply operand.
7361 2018-04-20 Martin Jambor <mjambor@suse.cz>
7364 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7365 recursion dependency to only apply to non-clones.
7367 2018-04-20 Martin Jambor <mjambor@suse.cz>
7370 * ipa-cp.c (create_specialized_node): Check that clones of
7371 self-recursive edges exist during IPA-CP.
7373 2018-04-19 Toon Moene <toon@moene.org>
7375 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7378 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7380 PR tree-optimization/85467
7381 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7382 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7383 VECTOR_CST element to type.
7385 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7388 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7389 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7390 (builtin_longjmp): Likewise.
7391 (save_stack_nonlocal): New pattern.
7392 (restore_stack_nonlocal): Likewise.
7394 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7397 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7398 Replace ASM_OUTPUT_LABEL with fprintf.
7400 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7403 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7404 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7405 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7406 define __IBT__ and __SHSTK__ for -fcf-protection.
7407 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7409 (ix86_trampoline_init): Likewise.
7410 (x86_output_mi_thunk): Likewise.
7411 (ix86_notrack_prefixed_insn_p): Likewise.
7412 (ix86_option_override_internal): Don't disallow -fcf-protection.
7413 * config/i386/i386.md (rdssp<mode>): Also enable for
7415 (incssp<mode>): Likewise.
7416 (nop_endbr): Likewise.
7417 * config/i386/i386.opt (mcet): Change help message to built-in
7421 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7422 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7423 enable CET built-in functions.
7425 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7427 * common/config/i386/i386-common.c
7428 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7429 OPTION_MASK_ISA_MOVDIRI_UNSET,
7430 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7431 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7432 * config.gcc (movdirintrin.h): New header.
7433 * config/i386/cpuid.h (bit_MOVDIRI,
7434 bit_MOVDIR64B): New bits.
7435 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7437 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7438 (VOID, PVOID, PCVOID)): New function types.
7439 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7440 __builtin_ia32_directstoreu_u64,
7441 __builtin_ia32_movdir64b): New builtins.
7442 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7443 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7445 (ix86_valid_target_attribute_inner_p): Ditto.
7446 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7447 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7448 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7449 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7450 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7451 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7452 (movdiri<mode>, movdir64b_<mode>): New.
7453 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7454 * config/i386/immintrin.h: Include movdirintrin.h.
7455 * config/i386/movdirintrin.h: New file.
7456 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7458 2018-04-19 Richard Biener <rguenther@suse.de>
7461 * cfg.c (clear_bb_flags): When loop state says we have
7462 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7464 2018-04-19 Richard Biener <rguenther@suse.de>
7466 PR tree-optimization/84737
7467 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7468 copying restrict info.
7469 (vect_setup_realignment): Use it.
7470 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7471 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7472 the first DR to all generated stores.
7473 (vectorizable_load): Likewise for loads.
7475 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7477 PR tree-optimization/85446
7478 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7479 the integral and pointer types to have the same precision.
7481 * doc/install.texi: Document --disable-cet being the default and
7484 2018-04-18 Martin Liska <mliska@suse.cz>
7486 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7489 2018-04-18 Martin Liska <mliska@suse.cz>
7492 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7495 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7496 arguments if they are comparable.
7498 2018-04-18 Martin Liska <mliska@suse.cz>
7501 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7504 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7507 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7510 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7511 ENDBR after calling __morestack.
7513 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7516 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7517 by using gcc_base_ver to generate a gcc_driver_version, and use
7518 it when generating GCC_DRIVER_NAME.
7519 * configure: Regenerate.
7521 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7524 * config.gcc: Obsolete powerpc*-*-*spe*.
7526 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7529 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7530 (stabstr_D): Change type of unum from unsigned int to
7531 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7534 2018-04-17 Jim Wilson <jimw@sifive.com>
7537 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7538 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7539 Set arg_pointer_offset after using pretend_args_size.
7541 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7543 PR rtl-optimization/85431
7544 * dse.c (record_store): Ignore zero width stores.
7547 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7548 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7549 __builtin_stack_restore rather than after it.
7550 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7551 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7552 argument instead of virtual_dynamic_stack_rtx.
7554 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7556 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7558 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7559 Add note to error message to explain internal mapping of overloaded
7560 built-in function name to non-overloaded built-in function name.
7561 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7564 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7567 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7568 where the inputs overlap with the output.
7570 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7572 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7573 (=v, v) alternative and explicit "memory" attribute.
7574 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7575 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7577 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7578 "sselog1" type instead of "sselog".
7579 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7580 "sselog". Remove explicit "memory" attribute.
7581 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7582 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7584 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7585 "sselog1" type instead of "sselog". Remove explicit "memory"
7587 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7588 use "sselog1" type instead of "sselog". Remove explicit "memory"
7590 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7591 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7593 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7594 "sselog1" type instead of "sselog". Remove explicit "memory"
7596 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7597 use "sselog1" type instead of "sselog". Remove explicit "memory"
7601 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7604 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7607 2018-04-17 Martin Jambor <mjambor@suse.cz>
7610 * ipa-cp.c (create_specialized_node): Call
7611 expand_all_artificial_thunks if necessary.
7613 2018-04-17 Martin Liska <mliska@suse.cz>
7616 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7617 in message, remote space in between '_G' and '('.
7619 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7622 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7623 avx512f_vmcmp<mode>3<round_saeonly_name>,
7624 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7625 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7626 avx512f_rndscale<mode><round_saeonly_name>,
7627 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7628 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7629 Use %<iptr>2 instead of %2 for -masm=intel.
7630 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7631 avx512f_vcvttss2usi<round_saeonly_name>,
7632 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7634 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7635 avx512f_vcvttsd2usi<round_saeonly_name>,
7636 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7637 Use %q1 instead of %1 for -masm=intel.
7638 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7639 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7640 of %3 for -masm=intel.
7641 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7643 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7645 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7647 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7648 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7650 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7652 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7653 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7654 %g1 and one with %0 and %1.
7655 (avx512er_vmrcp28<mode><round_saeonly_name>,
7656 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7658 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7659 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7660 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7661 of %0 and %{%4%} for -masm=intel.
7662 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7663 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7664 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7665 order of %0 and %{%5%}%{z%} for -masm=intel.
7667 2018-04-17 Jan Hubicka <jh@suse.cz>
7670 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7672 2018-04-17 Martin Liska <mliska@suse.cz>
7675 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7676 for target_clone error message. Make default implementation
7677 clone to be a local declaration.
7678 (separate_attrs): Add new argument and check for an empty
7680 (expand_target_clones): Handle it.
7681 (ipa_target_clone): Make redirection just for target_clones
7684 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7685 Tom de Vries <tom@codesourcery.com>
7688 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7689 tiled basic blocks without omp continue statements.
7691 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7694 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7695 vec_extract expression as having side effects to make sure it gets
7698 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7701 * config/i386/i386.c (get_builtin_code_for_version): Check
7704 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7707 * gcc/config.gcc: Support "skylake".
7708 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7710 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7711 (processor_target_table): Add "skylake".
7712 (ix86_option_override_internal): Add "skylake".
7713 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7714 PROCESSOR_CANNONLAKE.
7715 (get_builtin_code_for_version): Fix priority for
7716 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7717 PROCESSOR_SKYLAKE-AVX512.
7718 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7719 (processor_type): Add PROCESSOR_SKYLAKE.
7721 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7722 Jason Merrill <jason@redhat.com>
7725 * convert.c (convert_to_integer_1): Use direct recursion for
7726 enumeral types and types with a precision less than the number
7727 of bits in their mode.
7729 2018-04-16 Julia Koval <julia.koval@intel.com>
7732 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7733 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7735 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7738 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7739 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7740 and -mno-direct-move.
7742 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7745 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7746 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7747 Ensure negative shifts result in {0}.
7749 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7751 PR rtl-optimization/79916
7752 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7753 regs (if any) to define how to gnerate SD moves when LRA is in
7756 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7758 PR rtl-optimization/85393
7759 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7760 * except.c (expand_dw2_landing_pad_for_region): Make static.
7761 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7762 a label and unconditional jump to old_bb, rather than
7763 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7766 PR rtl-optimization/85376
7767 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7768 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7769 instead of a specific value.
7771 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7772 Bin Cheng <bin.cheng@arm.com>
7774 PR tree-optimization/82965
7775 PR tree-optimization/83991
7776 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7777 by_profile_only parameter.
7778 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7779 information if the loop was predicted to iterate too many times.
7780 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7782 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7785 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7788 2018-04-13 Martin Liska <mliska@suse.cz>
7789 Jakub Jelinek <jakub@redhat.com>
7792 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7793 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7794 * builtins.c (expand_builtin_memory_copy_args): Use
7795 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7796 handle dest_addr == pc_rtx.
7798 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7801 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7802 asked to not generate direct moves.
7803 (fix_trunc<mode>si2_stfiwx): Similar.
7804 (fix_trunc<mode>si2_internal): Similar.
7806 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7809 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7810 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7811 lookup if dest in some wider mode is known to be const0_rtx and
7812 if so, record permanent equivalence for it to be ZERO_EXTEND of
7813 the narrower mode destination.
7815 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7817 * lto-streamer-out.c (output_function): Revert 259346.
7818 * omp-expand.c (expand_oacc_for): Likewise.
7820 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7822 PR rtl-optimization/85354
7823 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7824 * sel-sched.c (sel_global_init): ... here.
7826 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7829 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7830 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7831 mode for PE-COFF targets.
7832 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7833 (i386_pe_asm_lto_end): Likewise.
7834 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7835 (TARGET_ASM_LTO_END): Likewise.
7836 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7837 (i386_pe_asm_lto_start): New function.
7838 (i386_pe_asm_lto_end): Likewise.
7840 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7841 Richard Biener <rguenther@suse.de>
7844 * lto-streamer-out.c (output_function): Fix CFG loop state before
7846 * omp-expand.c (expand_oacc_for): Handle calls to internal
7847 functions like regular functions.
7849 2018-04-12 Richard Biener <rguenther@suse.de>
7852 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7853 for the early LTO debug to properly generate references to it
7854 during DIE emission. Do not re-use that for the skeleton for
7856 (dwarf2out_early_finish): Likewise.
7858 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7861 * config/i386/sse.md
7862 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7863 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7864 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7865 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7866 and output is a reg, avoid creating invalid lowpart subreg, but
7867 instead split into a 512-bit move. Don't split if not AVX512VL,
7868 input is xmm16+ reg and output is a mem.
7869 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7870 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7871 xmm16+ reg and output is a mem.
7873 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7875 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7876 also for flag_dwarf2_cfi_asm.
7878 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7880 PR rtl-optimization/85342
7881 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7882 a bool scalar var inside of the loop instead. Don't try to update
7883 recog_data.operand after failed apply_change_group.
7885 2018-04-12 Tom de Vries <tom@codesourcery.com>
7888 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7889 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7890 array with flexible array member as array without given dimension.
7891 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7892 argument for undefined param to true.
7894 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7897 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7898 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7899 from PowerPC section.
7900 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7901 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7902 help text that is too long.
7903 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7904 help text that is too long.
7905 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7906 help text that is too long.
7908 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7910 * config/alpha/alpha.md (stack_probe_internal): Rename
7911 from "probe_stack". Update all callers.
7913 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7915 PR rtl-optimization/84566
7916 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7917 sched_macro_fuse_insns.
7919 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7922 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7923 (compute_block_dependences): ... from here.
7925 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7927 PR tree-optimization/85331
7928 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7929 from int to HOST_WIDE_INT.
7931 2018-04-11 Martin Jambor <mjambor@suse.cz>
7934 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7935 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7936 not the same as the source val.
7937 (cgraph_edge_brings_value_p): New parameter.
7938 (gather_edges_for_value): Pass destination value to
7939 cgraph_edge_brings_value_p.
7940 (perhaps_add_new_callers): Likewise.
7941 (get_info_about_necessary_edges): Likewise and exclude values brought
7942 only by self-recursive edges.
7943 (create_specialized_node): Redirect only clones of self-calling edges.
7944 (+self_recursive_pass_through_p): New function.
7945 (find_more_scalar_values_for_callers_subset): Use it.
7946 (find_aggregate_values_for_callers_subset): Likewise.
7947 (known_aggs_to_agg_replacement_list): Removed.
7948 (decide_whether_version_node): Re-calculate known constants for all
7949 remaining context clones.
7951 2018-04-11 Richard Biener <rguenther@suse.de>
7954 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7955 from early DWARF output.
7956 (dwarf2out_early_finish): Output line info unconditionally into
7957 early DWARF and add reference to it.
7959 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7962 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7963 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7964 other than V2DFmode using iptr mode attribute.
7965 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7967 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7969 PR rtl-optimization/84659
7970 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7972 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7975 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7977 (output_loc_list): Pass address of a dummy size variable even in the
7978 locview handling loop.
7979 (index_location_lists): Add comment on why skip_loc_list_entry can't
7982 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7985 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7988 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7991 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7992 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7993 and -mstring-compare-inline-limit.
7995 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7998 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7999 for stack clash protection in a register whenever we need it to be in
8002 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8004 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8005 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8007 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8010 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8012 (mlong-double-): Ditto.
8013 * config/rs6000/sysv4.opt (msdata=): Ditto.
8014 (mtls-size=): Ditto.
8016 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8018 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8019 erroneous entries for
8020 "vector int vec_ldl (int, long int *)", and
8021 "vector unsigned int vec_ldl (int, unsigned long int *)".
8022 Add comments and entries for
8023 "vector bool char vec_ldl (int, bool char *)",
8024 "vector bool short vec_ldl (int, bool short *)",
8025 "vector bool int vec_ldl (int, bool int *)",
8026 "vector bool long long vec_ldl (int, bool long long *)",
8027 "vector pixel vec_ldl (int, pixel *)",
8028 "vector long long vec_ldl (int, long long *)",
8029 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8030 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8031 type tree bool_long_long_type_node and correct definition of
8032 bool_V2DI_type_node to make reference to this new type tree.
8033 (rs6000_mangle_type): Replace erroneous reference to
8034 bool_long_type_node with bool_long_long_type_node.
8035 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8036 comments to emphasize sign distinctions for char and int types and
8037 replace RS6000_BTI_bool_long constant with
8038 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8039 use of RS6000_BTI_pixel.
8040 (bool_long_type_node): Remove this macro definition.
8041 (bool_long_long_type_node): New macro definition
8043 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8045 PR rtl-optimization/85300
8046 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8047 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8048 simplify_unary_operation fails.
8050 2018-04-10 Martin Liska <mliska@suse.cz>
8052 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8053 cgraph_edge and ipa_ref.
8055 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8059 * config/i386/sse.md
8060 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8061 computation of the VEC_MERGE selector from mask.
8062 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8063 Fix decoding of the VEC_MERGE selector into mask.
8065 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8067 PR tree-optimization/85286
8068 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8070 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8072 * final.c (final_1): Set insn_last_address as well as
8073 insn_current_address.
8075 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8078 * explow.c (emit_stack_probe): Call validize_mem on memory location
8079 before passing it to gen_probe_stack. Create address operand and
8080 legitimize it for the probe_stack_address case.
8082 2018-04-09 Jan Hubicka <jh@suse.cz>
8085 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8086 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8087 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8089 (free_lang_data): Rebuild type inheritance graph.
8091 2018-04-09 Martin Sebor <msebor@redhat.com>
8093 * invoke.texi (-finline-small-functions): Mention other optimization
8095 (-findirect-inlining, -fpartial-inlining): Same.
8096 (-finline-functions-called-once): Same.
8097 (-freorder-blocks-and-partition): Same.
8099 2018-04-09 Jan Hubicka <jh@suse.cz>
8102 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8103 jumps; choose last target that matches the criteria (i.e.
8104 no partition changes for non-crossing jumps).
8105 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8106 support for redirecting crossing jumps to non-crossing.
8108 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8110 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8111 also for naked functions.
8113 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8115 * config/arc/arc.md (add_shift): New pattern.
8116 (add_shift2): Likewise.
8117 (sub_shift): Likewise.
8118 (sub_shift_cmp0_noout): Likewise.
8119 (compare_si_ashiftsi): Likewise.
8120 (xbfu_cmp0_noout): New combine pattern.
8121 (xbfu_cmp0"): Likewise.
8122 (movsi_set_cc_insn): Place the predicable variant first.
8123 (commutative_binary_cmp0_noout): Remove clobber.
8124 (commutative_binary_cmp0): New pattern.
8125 (noncommutative_binary_cmp0): Likewise.
8126 (noncommutative_binary_cmp0_noout): Likewise.
8127 (noncommutative_binary_comparison_result_used): Removed.
8128 (rsub_cmp0): New pattern.
8129 (rsub_cmp0_noout): Likewise.
8130 (extzvsi): Changed, keep only meaningful variants.
8131 (SQH, SEZ): New iterators.
8132 (SQH_postfix): New mode attribute.
8133 (SEZ_prefix): New code attribute.
8134 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8135 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8136 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8138 (noncommutative_operator): Check the availability of barrel
8141 2018-04-09 Richard Biener <rguenther@suse.de>
8143 PR tree-optimization/85284
8144 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8145 Only use the niter constraining form of simple_iv when the exit
8148 2018-04-09 Tom de Vries <tom@codesourcery.com>
8151 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8152 (define_expand "*memory_barrier"): New define_expand.
8153 (define_insn "memory_barrier"): New insn.
8155 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8157 PR rtl-optimization/80463
8158 PR rtl-optimization/83972
8159 PR rtl-optimization/83480
8161 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8162 correct producer for the insn.
8163 (tidy_control_flow): Fixup seqnos in case of debug insns.
8165 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8167 PR rtl-optimization/83913
8169 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8170 different sched-times when merging exprs.
8172 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8174 PR rtl-optimization/83962
8176 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8177 tidy_fallthru_edge and tidy_control_flow.
8179 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8181 PR rtl-optimization/83530
8183 * sel-sched.c (force_next_insn): New global variable.
8184 (remove_insn_for_debug): When force_next_insn is true, also leave only
8185 next insn in the ready list.
8186 (sel_sched_region): When the region wasn't scheduled, make another pass
8187 over it with force_next_insn set to 1.
8189 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8191 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8193 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8194 for interrupt control.
8195 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8196 functions for interrupt control.
8197 * config/nds32/nds32-intrinsic.md: Likewise.
8198 * config/nds32/nds32_intrinsic.h: Likewise.
8199 * config/nds32/nds32.h (nds32_builtins): Likewise.
8201 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8203 * config/nds32/nds32.c (nds32_init_machine_status,
8204 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8205 strict_aligned_p field.
8206 (nds32_expand_to_rtl_hook): New function.
8207 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8208 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8210 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8211 Chung-Ju Wu <jasonwucj@gmail.com>
8213 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8214 * config/nds32/nds32-n7.md: New file.
8215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8218 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8219 * config/nds32/nds32.md (pipeline_model): Add n7.
8220 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8221 * config/nds32/pipelines.md: Include n7 settings.
8223 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8224 Chung-Ju Wu <jasonwucj@gmail.com>
8226 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8227 * config/nds32/nds32-e8.md: New file.
8228 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8231 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8232 * config/nds32/nds32.md (pipeline_model): Add e8.
8233 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8234 * config/nds32/pipelines.md: Include e8 settings.
8236 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8237 Chung-Ju Wu <jasonwucj@gmail.com>
8239 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8240 * config/nds32/nds32-n8.md: New file.
8241 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8242 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8244 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8245 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8246 * config/nds32/nds32.md (pipeline_model): Add n8.
8247 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8248 * config/nds32/pipelines.md: Include n8 settings.
8250 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8251 Chung-Ju Wu <jasonwucj@gmail.com>
8253 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8254 * config/nds32/nds32-n9-2r1w.md: New file.
8255 * config/nds32/nds32-n9-3r2w.md: New file.
8256 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8257 nds32_register_ports): New or modify for cpu n9.
8258 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8260 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8261 * config/nds32/nds32-utils.c: New file.
8262 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8263 TARGET_MUL_SLOW): Define.
8264 * config/nds32/nds32.md (pipeline_model): New attribute.
8265 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8266 New options that support cpu n9.
8267 * config/nds32/pipelines.md: Include n9 settings.
8268 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8270 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8272 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8273 information if necessary.
8274 (output_cond_branch_compare_zero): Likewise.
8275 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8276 (nds32_target_alignment): Refine for alignment.
8277 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8278 (FUNCTION_BOUNDARY): Modify.
8279 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8281 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8283 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8285 * config/nds32/constants.md (unspec_volatile_element): Add values for
8286 TLB operation and data prefetch.
8287 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8288 functions for TLB operation and data prefetch.
8289 * config/nds32/nds32-intrinsic.md: Likewise.
8290 * config/nds32/nds32_intrinsic.h: Likewise.
8291 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8292 (nds32_print_operand): Likewise.
8293 * config/nds32/nds32.h (nds32_builtins): Likewise.
8295 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8296 Andrew Pinski <pinsika@gcc.gnu.org>
8299 * match.pd: Use constant_boolean_node of correct type instead of
8300 boolean_true_node or boolean_false_node for simplifying
8301 pointer comparisons to zero.
8303 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8305 PR tree-optimization/80021
8306 * tree.c (verify_type_variant): Make error call in verify_variant_match
8307 translatable and remove final full stop.
8309 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8311 * config/nds32/constants.md (unspec_volatile_element): Add
8312 UNSPEC_VOLATILE_EH_RETURN.
8313 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8314 nds32_output_stack_pop): Support dwarf exception handling process.
8315 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8316 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8317 exception handling process.
8318 (nds32_compute_stack_frame): Likewise.
8319 (nds32_return_addr_rtx): Likewise.
8320 (nds32_initial_elimination_offset): Likewise.
8321 (nds32_expand_prologue): Likewise.
8322 (nds32_expand_epilogue): Likewise.
8323 (nds32_dynamic_chain_address): New function.
8324 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8326 (DYNAMIC_CHAIN_ADDRESS): Define.
8327 (EH_RETURN_DATA_REGNO): Define.
8328 (EH_RETURN_STACKADJ_RTX): Define.
8329 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8330 patterns for dwarf exception handling.
8332 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8334 * config/nds32/nds32.h: Clean up obsolete macros.
8336 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8338 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8339 Add enum values for particular instructions.
8340 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8341 particular intrinsic functions.
8342 * config/nds32/nds32-intrinsic.md: Likewise.
8343 * config/nds32/nds32_intrinsic.h: Likewise.
8344 * config/nds32/nds32.h (nds32_builtins): Likewise.
8345 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8346 (btst, ave): New patterns for particular instructions.
8348 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8350 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8351 Add enum values for atomic load/store and memory sync.
8352 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8354 * config/nds32/nds32-intrinsic.md: Likewise.
8355 * config/nds32/nds32_intrinsic.h: Likewise.
8356 * config/nds32/nds32.h (nds32_builtins): Likewise.
8358 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8360 PR tree-optimization/85257
8361 * fold-const.c (native_encode_vector): If not all elts could fit
8362 and off is -1, return 0 rather than offset.
8363 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8364 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8365 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8366 adjust buffer in native_interpret_expr call.
8368 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8370 * config/nds32/constants.md (unspec_volatile_element): Add cache
8371 control enum values.
8372 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8373 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8374 * config/nds32/nds32.c (nds32_cctl_names): New.
8375 (nds32_print_operand): Handle cache control register names.
8376 * config/nds32/nds32.h (nds32_builtins): New enum values.
8377 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8379 * config/nds32/nds32.md (type): Add mmu.
8380 * config/nds32/pipelines.md (simple_insn): Add mmu.
8382 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8384 * config/nds32/nds32.md (type): Remove call.
8385 * config/nds32/pipelines.md (simple_insn): Likewise.
8387 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8389 * config/nds32/constants.md (unspec_volatile_element): Add
8390 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8391 UNSPEC_VOLATILE_FMFCFG.
8392 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8393 description for fmfcfg and fmfcsr.
8394 (bdesc_1arg): Add fmtcsr.
8395 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8396 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8397 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8398 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8399 unspec_fmfcfg): New patterns.
8400 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8401 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8402 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8403 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8404 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8405 __nds32__fmfcfg): Define.
8407 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8409 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8410 intrinsic register names.
8411 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8412 intrinsic register enum values and macros.
8414 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8416 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8417 for load/store addressing form.
8418 (nds32_print_operand_address): Likewise.
8420 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8423 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8424 based on LABEL_REF. Remove useless assertion.
8425 (pic_address_needs_scratch): Fix formatting.
8426 (sparc_legitimize_pic_address): Minor tweaks.
8427 (sparc_delegitimize_address): Adjust assertion accordingly.
8428 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8429 into symbolic_operand.
8430 (movsi_high_pic_label_ref): Likewise.
8431 (movsi_lo_sum_pic_label_ref): Likewise.
8432 (movdi_pic_label_ref): Likewise.
8433 (movdi_high_pic_label_ref): Likewise.
8434 (movdi_lo_sum_pic_label_ref): Likewise.
8436 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8438 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8439 custom LIB_SPEC setup.
8441 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8442 Kito Cheng <kito.cheng@gmail.com>
8444 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8445 * config/riscv/freebsd.h: New.
8447 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8449 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8450 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8453 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8454 Kito Cheng <kito.cheng@gmail.com>
8456 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8457 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8458 * config/nds32/nds32-protos.h (nds32_output_call,
8459 nds32_output_return): Declare.
8460 * config/nds32/nds32.md: Refine all the call and return patterns.
8462 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8465 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8466 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8468 PR rtl-optimization/84872
8469 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8470 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8473 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8475 * expr.c (copy_blkmode_to_reg): Revert 254862.
8476 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8478 2018-04-06 Richard Biener <rguenther@suse.de>
8481 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8482 after seeing a component reference with an adjacent field. Treat
8483 refs to arrays at struct end of external decls similar to
8484 refs to unconstrained commons.
8486 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8489 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8490 look through SAVE_EXPRs with non-side-effects argument. Adjust
8492 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8495 2018-04-06 Richard Biener <rguenther@suse.de>
8498 * alias.c (find_base_term): New wrapper around find_base_term
8499 unwinding CSELIB_VAL_PTR changes.
8500 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8503 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8505 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8507 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8508 constant definitions.
8509 ("nop"): lr 0,0 -> nopr r0
8510 ("nop_lr0", "nop_lr1"): New insn definitions.
8512 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8514 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8515 NDS32_V3PUSH_AVAILABLE_P macro.
8517 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8518 Chung-Ju Wu <jasonwucj@gmail.com>
8520 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8521 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8522 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8523 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8524 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8525 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8526 * config/nds32/constraints.md: New constraints and checking for hard
8527 float configuration.
8528 * config/nds32/iterators.md: New mode iterator and attribute for hard
8529 float configuration.
8530 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8532 * config/nds32/nds32-fpu.md: New file.
8533 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8534 deal with hard float code generation.
8535 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8537 (abi_type, float_reg_number): New enum type.
8538 * config/nds32/nds32-predicates.c: New predicates for hard float.
8539 * config/nds32/nds32-protos.h: Declare functions for hard float.
8540 * config/nds32/nds32.c: Implementation for hard float configuration.
8541 * config/nds32/nds32.h: Definitions for hard float configuration.
8542 * config/nds32/nds32.md: Include hard float machine description and
8543 modify patterns for hard float configuration.
8544 * config/nds32/nds32.opt: New options for hard float configuration.
8545 * config/nds32/predicates.md: New predicates for hard float
8548 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8550 * common/config/nds32/nds32-common.c
8551 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8553 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8556 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8557 CONSTRUCTOR_ELT (ctor, ...)->value.
8559 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8562 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8564 2018-04-05 Tom de Vries <tom@codesourcery.com>
8567 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8570 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8571 Kito Cheng <kito.cheng@gmail.com>
8573 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8574 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8575 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8576 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8578 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8579 Kito Cheng <kito.cheng@gmail.com>
8581 * config/nds32/constraints.md (Ufe): New memory constraint.
8582 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8583 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8584 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8586 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8587 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8589 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8591 * config/nds32/nds32.md: Use optimize_size in the condition for
8592 alu-shift instructions.
8594 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8596 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8598 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8600 * config/nds32/nds32.md (negsi2): Refine pattern.
8602 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8603 Chung-Ju Wu <jasonwucj@gmail.com>
8605 * config/nds32/iterators.md (shift_rotate): New code iterator.
8606 (shift): New code attribute.
8607 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8608 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8609 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8610 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8611 bit-wise operations.
8612 (andsi3, *andsi3): Ditto.
8613 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8614 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8615 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8616 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8617 nds32_ior_operand, nds32_xor_operand): New predicates.
8619 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8621 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8622 (addsi3, subsi3): ... this.
8624 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8626 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8628 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8630 * config/nds32/nds32.md: Adjust indention.
8632 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8634 * config/nds32/nds32.md (feature): New attribute.
8636 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8638 * config/nds32/nds32.md (subtype): New attribute.
8640 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8643 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8644 expansion to perform a bitwise AND of the argument followed by a
8645 boolean negation of the result.
8647 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8649 PR rtl-optimization/84878
8650 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8651 the basic block. Assert the use reference is not artificial and that
8652 it has an associated insn.
8654 2018-04-04 Michael Matz <matz@suse.de>
8656 * builtins.c (compute_objsize): Pass correct operand
8657 to array_at_struct_end_p.
8659 2018-04-04 Richard Biener <rguenther@suse.de>
8662 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8663 from contexts for DINFO_LEVEL_TERSE and below.
8665 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8667 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8668 resiter_operand condition.
8669 * config/nds32/nds32.md (*move<mode>): Ditto.
8671 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8672 Monk Chiang <sh.chiang04@gmail.com>
8674 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8676 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8678 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8680 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8681 Kito Cheng <kito.cheng@gmail.com>
8683 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8684 nds32_cond_code_str, output_cond_branch,
8685 output_cond_branch_compare_zero, nds32_expand_cbranch,
8686 nds32_expand_cstore, nds32_expand_movcc,
8687 nds32_output_cbranchsi4_equality_zero,
8688 nds32_output_cbranchsi4_equality_reg,
8689 nds32_output_cbranchsi4_equality_reg_or_const_int,
8690 nds32_output_cbranchsi4_greater_less_zero: New functions.
8691 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8692 nds32_expand_cstore, nds32_expand_movcc,
8693 nds32_output_cbranchsi4_equality_zero,
8694 nds32_output_cbranchsi4_equality_reg,
8695 nds32_output_cbranchsi4_equality_reg_or_const_int,
8696 nds32_output_cbranchsi4_greater_less_zero): Declare.
8697 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8698 nds32_rimm11s_operand): New predicates.
8699 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8700 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8703 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8705 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8706 * config/nds32/nds32.md: Ditto.
8707 * config/nds32/pipelines.md: Ditto.
8709 2018-04-04 Richard Biener <rguenther@suse.de>
8711 PR tree-optimization/85168
8712 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8713 propagating abnormals.
8715 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8717 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8719 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8720 Kito Cheng <kito.cheng@gmail.com>
8722 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8723 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8724 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8725 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8726 * config/nds32/nds32.md (sibcall_internal): New.
8727 (sibcall_register): Remove.
8728 (sibcall_immediate): Remove.
8729 (sibcall_value_internal): New.
8730 (sibcall_value_register): Remove.
8731 (sibcall_value_immediate): Remove.
8732 * config/nds32/predicates.md (nds32_general_register_operand): New.
8733 (nds32_call_address_operand): New.
8735 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8737 PR rtl-optimization/85167
8738 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8739 bb_defs if *split_p, instead preinitialize it to NULL.
8741 PR tree-optimization/85156
8742 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8743 evaluating the argument multiple times.
8745 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8747 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8749 (_mm_cvtpd_ps): Likewise.
8750 (_mm_cvttpd_epi32): Likewise.
8751 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8752 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8753 vector, pixel, and bool following altivec.h include.
8755 2018-04-03 Martin Sebor <msebor@redhat.com>
8757 * doc/extend.texi (Common Function Attributes): Clarify.
8758 (const attribute): Likewise.
8759 (pure attribute): Likewise.
8761 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8764 * config/i386/i386.c (ix86_expand_vector_set): Use
8765 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8767 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8769 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8770 instructions when changing rounding bits to preserve precision bits
8771 in the x87 control word.
8773 2018-04-03 Martin Liska <mliska@suse.cz>
8775 PR tree-optimization/82491
8776 * rtl.h (strip_offset_and_add): Replace += suboffset with
8777 poly_uint64 () + suboffset.
8779 2018-03-29 Martin Liska <mliska@suse.cz>
8780 Martin Jambor <mjambor@suse.cz>
8783 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8784 param_type is not an integral or pointer type.
8786 2018-04-03 Richard Biener <rguenther@suse.de>
8788 * sese.h (recompute_all_dominators): Remove.
8790 2018-04-02 Martin Sebor <msebor@redhat.com>
8792 * doc/invoke.texi (-Wrestrict): Fix typos.
8794 2018-04-02 Jim Wilson <jimw@sifive.com>
8796 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8797 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8798 (<optab>di3, <optab>si3_extend): Likewise.
8799 (<optab>si3_mask, <optab>si3_mask_1): New.
8800 (<optab>di3_mask, <optab>di3_mask_1): New.
8801 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8802 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8803 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8805 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8807 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8810 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8812 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8813 (nds32_canonicalize_comparison): New function.
8815 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8816 Kito Cheng <kito.cheng@gmail.com>
8817 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8819 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8820 * config/nds32/constants.md (unspec_volatile_element): Add
8821 UNSPEC_VOLATILE_RELAX_GROUP.
8822 * config/nds32/nds32-relax-opt.c: New file.
8823 * config/nds32/nds32-predicates.c
8824 (nds32_symbol_load_store_p): New function.
8825 * config/nds32/nds32-protos.h
8826 (nds32_symbol_load_store_p): Declare function.
8827 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8828 * config/nds32/nds32.c
8829 (nds32_register_pass): New function to register pass.
8830 (nds32_register_passes): New function to register passes.
8831 * config/nds32/nds32.md (relax_group): New pattern.
8832 * config/nds32/nds32.opt (mrelax-hint): New option.
8833 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8835 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8837 * config/nds32/t-nds32: Modify files dependency.
8839 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8841 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8842 (PROFILE_HOOK): Define its implementation.
8844 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8846 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8847 type and 32-bit size.
8849 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8852 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8853 (V_128_256): New mode iterator.
8854 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8855 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8856 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8858 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8859 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8861 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8864 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8865 NaN inputs correctly.
8867 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8870 * config/rs6000/vsx.md (??r): New mode attribute.
8871 (*vsx_mov<mode>_64bit): Use it.
8872 (*vsx_mov<mode>_32bit): Likewise.
8874 2018-03-30 Martin Sebor <msebor@redhat.com>
8876 PR tree-optimization/84818
8877 * builtins.c (check_access): Use warning_n.
8879 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8882 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8884 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8887 2018-03-30 Julia Koval <julia.koval@intel.com>
8890 * x86-tune.def (movx, partial_reg_dependency): Enable for
8893 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8896 * lra-constraints.c (process_alt_operands): Move setting
8897 this_alternative_matches below.
8899 2018-03-29 Martin Liska <mliska@suse.cz>
8902 * doc/invoke.texi: Document how LTO works with debug info.
8903 Describe auto-load support of binutils. Mention 'x86-64'
8904 as valid option value of -march option.
8906 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8908 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8911 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8912 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8913 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8916 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8919 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8920 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8921 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8922 for RS6000_BTM_POWERPC64.
8923 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8924 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8925 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8930 2018-03-28 Carl Love <cel@us.ibm.com>
8933 2017-09-27 Carl Love <cel@us.ibm.com>
8935 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8936 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8937 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8940 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8942 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8943 instead of __vector bool.
8944 (_mm_max_pu8): Likewise.
8945 (_mm_min_pi16): Likewise.
8947 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8950 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8951 (DIVWEUO): Likewise.
8953 (DIVDEUO): Likewise.
8954 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8955 DIVWEUO and DIVDEUO.
8956 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8957 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8958 (div_extend): Likewise.
8959 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8961 (__builtin_divweuo): Likewise.
8962 (__builtin_divdeo): Likewise.
8963 (__builtin_divdeuo): Likewise.
8965 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8968 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8969 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8971 PR tree-optimization/82004
8972 * gimple-match-head.c (optimize_pow_to_exp): New function.
8973 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8974 Don't fold to exp if optimize_pow_to_exp is false.
8976 2018-03-28 Martin Liska <mliska@suse.cz>
8979 * calls.c (initialize_argument_information): Fix trailing space.
8980 * common.opt: Fix typo and provide better explanation for
8981 -fsanitize-coverage option.
8982 * config/i386/i386.opt: Fix typo.
8984 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8985 Martin Liska <mliska@suse.cz>
8988 * gimplify.c (asan_poison_variable): Don't do the check for
8989 gimplify_omp_ctxp here.
8990 (gimplify_decl_expr): Do it here.
8991 (gimplify_target_expr): Likewise.
8993 2018-03-28 Martin Liska <mliska@suse.cz>
8996 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8997 chkp_type_bounds_count if MPX is not enabled.
8999 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9001 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9003 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9006 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9007 function to create the function decl for complex long double
9008 multiply and divide for -mabi=ieeelongdouble.
9009 (init_float128_ieee): Call it.
9011 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9014 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9015 -fcf-protection=branch -mibt.
9016 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9018 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9021 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9022 arm_word_relocations.
9024 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9027 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9028 extern array declarations.
9030 2018-03-27 Richard Biener <rguenther@suse.de>
9033 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9034 explicit single_use checks.
9036 2018-03-27 Richard Biener <rguenther@suse.de>
9038 PR tree-optimization/85082
9039 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9042 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9044 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9045 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9046 Turn on fasynchronous-unwind-tables and funwind-tables.
9048 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9051 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9052 (*bmi_blsr_<mode>_ccz): Ditto.
9054 2018-03-26 Tom de Vries <tom@codesourcery.com>
9056 PR tree-optimization/85063
9057 * omp-general.c (offloading_function_p): New function. Factor out
9059 * omp-offload.c (pass_omp_target_link::gate): ... here.
9060 * omp-general.h (offloading_function_p): Declare.
9061 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9062 with attribute omp declare target for offloading functions.
9064 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9066 PR tree-optimization/84005
9067 * tree-data-ref.h (get_base_for_alignment): Declare.
9068 * tree-data-ref.c (get_base_for_alignment_1): New function.
9069 (get_base_for_alignment): Likewise.
9070 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9071 get_base_for_alignment to find a suitable base object, instead
9072 of always using drb->base_address.
9074 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9077 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9078 known size by default.
9080 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9083 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9084 and non BLKmode operands.
9086 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9089 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9090 Clean up attributes.
9092 2018-03-23 Richard Biener <rguenther@suse.de>
9095 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9096 we are going to emit early debug for LTO.
9098 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9101 * function.c (match_asm_constraints_1): Don't optimize if input
9102 doesn't satisfy general_operand predicate for output's mode.
9105 * alias.c (write_dependence_p): Don't require for x_canonicalized
9106 non-VOIDmode if x has VOIDmode.
9109 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9110 just don't try to optimize it rather than assert it never happens.
9112 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9114 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9115 macro expansions for definition of ST_INTERNAL_<mode> and
9116 LD_INTERNAL_<mode> builtins.
9117 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9119 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9121 (altivec_expand_st_builtin): Likewise.
9122 (altivec_expand_builtin): Remove calls to deleted functions.
9123 (rs6000_address_for_altivec): Delete this function.
9124 * config/rs6000/vector.md: Remove expands for
9125 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9127 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9130 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9131 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9132 re-computing once computed.
9133 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9134 (arm_init_machine_status): Initialize
9135 machine->static_chain_stack_bytes.
9137 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9140 * doc/extend.texi: Add four new prototypes for vec_ld.
9141 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9142 definitions for more logical presentation.
9143 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9144 entries for V1TI variants of __builtin_altivec_ld builtin.
9145 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9146 handling of V1TI variant of LVX icode pattern.
9147 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9148 (rs6000_gimple_fold_builtin): Likewise.
9149 (altivec_init_builtins): Add code to define
9150 __builtin_altivec_lvx_v1ti function.
9152 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9155 * function.c (match_asm_constraints_1): Don't do the optimization
9156 if input isn't a REG, SUBREG, MEM or constant.
9158 2018-03-22 Tom de Vries <tom@codesourcery.com>
9160 PR tree-optimization/84956
9161 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9162 bb_has_abnormal_pred.
9164 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9167 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9168 DECL_INITIAL (decl) to decl at the end.
9169 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9172 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9174 * doc/extend.texi (__builtin_tgmath): Document when complex
9175 integer types are treated as _Complex _Float64.
9177 2018-03-21 Tom de Vries <tom@codesourcery.com>
9179 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9181 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9183 PR tree-optimization/84960
9184 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9185 if it is ENTRY block, move them into single succ of ENTRY in that case.
9187 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9189 PR tree-optimization/84811
9190 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9191 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9192 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9193 (known_subrange_p): Update accordingly. Cast each value involved
9194 in the size comparison, rather than casting the result of the
9197 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9199 PR tree-optimization/84982
9200 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9201 by flipping the least significant bit rather than all bits from
9202 bitpos to bitpos + bitsize - 1.
9204 2018-03-21 Nathan Sidwell <nathan@acm.org>
9206 * doc/extend.texi (Deprecated Features): Remove mention of
9207 long-deleted deprecations.
9209 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9212 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9213 * configure: Regenerate.
9215 2018-03-21 Tom de Vries <tom@codesourcery.com>
9217 PR tree-optimization/83126
9218 * tree-parloops.c (num_phis): New function.
9219 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9221 2018-03-21 Nathan Sidwell <nathan@acm.org>
9223 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9224 mention anon-struct/union members and trailing attributes.
9226 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9228 PR tree-optimization/84969
9229 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9230 builtin memset partitions if they set different rhs values.
9232 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9234 PR rtl-optimization/84989
9235 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9236 VEC_DUPLICATE with scalar result mode.
9238 2018-03-21 Martin Liska <mliska@suse.cz>
9241 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9242 not intended return statement.
9244 2018-03-21 Martin Liska <mliska@suse.cz>
9247 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9248 (chkp_find_bound_slots_1): Limit number of iterations.
9250 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9253 * Minor grammar fixes for x86 options.
9255 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9258 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9259 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9261 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9264 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9265 (altivec_lvx_<mode>_1op): Likewise.
9266 (altivec_stvx_<mode>_2op): Likewise.
9267 (altivec_stvx_<mode>_1op): Likewise.
9268 (altivec_lvx_<VM2:mode>): New define_expand.
9269 (altivec_stvx_<VM2:mode>): Likewise.
9270 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9271 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9272 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9273 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9274 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9275 (rs6000_gen_lvx): Likewise.
9276 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9277 (altivec_expand_stv_builtin): Likewise.
9278 (altivec_expand_builtin): Likewise.
9279 * config/rs6000/vector.md: Likewise.
9281 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9284 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9287 2018-03-20 Richard Biener <rguenther@suse.de>
9290 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9291 sign-conversions as zero, fall back to standard scalar_stmt
9294 2018-03-20 Martin Liska <mliska@suse.cz>
9297 * predict.c (rebuild_frequencies): Handle case when we have
9298 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9300 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9303 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9304 flag_section_anchors.
9305 * varasm.c (use_blocks_for_decl_p): Remove hack for
9306 dw2_force_const_mem.
9309 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9311 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9312 be created, use lowpart_subreg of operands[0] rather than operands[0]
9314 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9315 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9316 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9317 and n constraint instead of aarch64_shift_imm_di and Usd.
9318 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9319 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9321 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9324 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9325 to favor GPR over NEON registers.
9326 (<shift>di3_neon): Likewise.
9328 2018-03-20 Tom de Vries <tom@codesourcery.com>
9331 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9332 (nvptx_process_pars): Emit bar.sync asap and alap.
9334 2018-03-20 Tom de Vries <tom@codesourcery.com>
9337 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9338 seen_label if seen_label is already set.
9340 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9343 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9344 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9345 Use 1U instead of 1. Formatting fixes.
9348 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9349 instead of TREE_TYPE (s1) for the return value.
9351 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9353 PR tree-optimization/84946
9354 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9355 bitsize + bitsize in poly_uint64 rather than poly_int64.
9358 * dwarf2asm.c: Include fold-const.c.
9359 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9360 of decl rather than decl itself.
9362 PR rtl-optimization/84643
9363 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9365 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9368 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9369 calling assemble_variable.
9371 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9374 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9375 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9377 2018-03-19 Jim Wilson <jimw@sifive.com>
9380 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9381 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9382 (riscv_first_stack_step): Likewise.
9383 (riscv_option_override): Use STACK_BOUNDARY instead of
9385 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9387 (BIGGEST_ALIGNMENT): Set to 128.
9388 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9389 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9392 2018-03-19 Richard Biener <rguenther@suse.de>
9394 PR tree-optimization/84933
9395 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9396 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9398 2018-03-19 Richard Biener <rguenther@suse.de>
9400 PR tree-optimization/84859
9401 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9402 (cond_if_else_store_replacement): Perform sinking operation on
9403 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9404 Generalize what a BB with a single eligible store is.
9406 2018-03-19 Richard Biener <rguenther@suse.de>
9408 PR tree-optimization/84929
9409 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9410 chrec_is_positive against non-chrec arg.
9412 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9415 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9417 2018-03-18 Martin Liska <mliska@suse.cz>
9419 PR rtl-optimization/84635
9420 * regrename.c (build_def_use): Use matches_mode only when
9423 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9425 PR tree-optimization/84913
9426 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9427 vectorize chains of COND_EXPRs.
9429 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9431 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9433 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9435 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9437 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9439 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9441 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9442 Kito Cheng <kito.cheng@gmail.com>
9444 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9445 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9446 (nds32_adjust_reg_alloc_order): New function.
9447 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9449 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9451 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9452 nds32_print_operand, nds32_print_operand_address): Use
9453 HOST_WIDE_INT_PRINT_DEC instead.
9455 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9457 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9459 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9462 * config/i386/i386.c (initial_ix86_tune_features,
9463 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9465 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9466 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9467 rather than 1u << ix86_tune. Formatting fix.
9468 (ix86_option_override_internal): Change ix86_arch_mask from
9469 unsigned int to unsigned HOST_WIDE_INT, initialize to
9470 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9471 (ix86_function_specific_restore): Likewise.
9473 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9476 * postreload.c (reload_combine_recognize_pattern): Perform
9477 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9478 truncate_int_for_mode the result for the destination's mode.
9481 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9482 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9484 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9487 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9490 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9491 regno_allocno_class_array and sorted_pseudos.
9492 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9493 insns where regno is used.
9495 2018-03-16 Martin Liska <mliska@suse.cz>
9498 * multiple_target.c (create_dispatcher_calls): Redirect
9499 reference in the symbol table.
9501 2018-03-16 Martin Liska <mliska@suse.cz>
9504 * multiple_target.c (create_dispatcher_calls): Redirect also
9507 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9511 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9512 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9513 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9515 2018-03-16 Julia Koval <julia.koval@intel.com>
9517 * doc/invoke.texi (Skylake Server): Add CLWB.
9518 Cannonlake): Remove CLWB.
9520 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9522 PR tree-optimization/84841
9523 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9525 (FLOAT_ONE_CONST_TYPE): Define.
9526 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9527 (sort_by_operand_rank): Put entries with higher constant_type last
9528 rather than first to match comments.
9530 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9532 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9535 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9538 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9539 check and use address of target always.
9541 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9544 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9545 (indirect_thunk_bnd_needed): Likewise.
9546 (indirect_thunks_used): Likewise.
9547 (indirect_thunks_bnd_used): Likewise.
9548 (indirect_return_needed): New.
9549 (indirect_return_bnd_needed): Likewise.
9550 (output_indirect_thunk_function): Add a bool argument for
9552 (output_indirect_thunk_function): Don't generate alias for
9553 function return thunk.
9554 (ix86_code_end): Call output_indirect_thunk_function to generate
9555 function return thunks.
9556 (ix86_output_function_return): Set indirect_return_bnd_needed
9557 and indirect_return_needed instead of indirect_thunk_bnd_needed
9558 and indirect_thunk_needed.
9560 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9562 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9563 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9564 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9566 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9567 Paul Hua <paul.hua.gm@gmail.com>
9570 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9572 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9574 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9575 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9578 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9581 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9582 instead of GET_MODE_SIZE when comparing Units.
9584 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9587 * varasm.c (hash_section): Return an unchangeble hash value
9588 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9589 Return !aarch64_can_use_per_function_literal_pools_p ().
9591 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9594 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9595 rather than address of cmode as last argument to prepare_cmp_insn.
9597 2018-03-15 Julia Koval <julia.koval@intel.com>
9599 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9600 F_AVX512VNNI, F_AVX512BITALG): New.
9602 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9605 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9606 insn for floating-point loads and stores.
9608 2018-03-14 Carl Love <cel@us.ibm.com>
9610 * config/rs6000/rs6000-c.c: Add macro definitions for
9611 ALTIVEC_BUILTIN_VEC_PERMXOR.
9612 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9613 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9614 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9615 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9617 * config/doc/extend.texi: Add prototypes for vec_permxor.
9619 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9622 * diagnostic-show-locus.c (class layout_point): Convert m_line
9623 from int to linenum_type.
9624 (line_span::comparator): Use linenum "compare" function when
9625 comparing line numbers.
9626 (test_line_span): New function.
9627 (layout_range::contains_point): Convert param "row" from int to
9629 (layout_range::intersects_line_p): Likewise.
9630 (layout::will_show_line_p): Likewise.
9631 (layout::print_source_line): Likewise.
9632 (layout::should_print_annotation_line_p): Likewise.
9633 (layout::print_annotation_line): Likewise.
9634 (layout::print_leading_fixits): Likewise.
9635 (layout::annotation_line_showed_range_p): Likewise.
9636 (struct line_corrections): Likewise for field m_row.
9637 (line_corrections::line_corrections): Likewise for param "row".
9638 (layout::print_trailing_fixits): Likewise.
9639 (layout::get_state_at_point): Likewise.
9640 (layout::get_x_bound_for_row): Likewise.
9641 (layout::print_line): Likewise.
9642 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9643 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9644 * input.c (selftest::test_linenum_comparisons): New function.
9645 (selftest::input_c_tests): Call it.
9646 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9647 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9648 * selftest.h (ASSERT_GT): New macro.
9649 (ASSERT_GT_AT): New macro.
9650 (ASSERT_LT): New macro.
9651 (ASSERT_LT_AT): New macro.
9653 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9655 PR rtl-optimization/84780
9656 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9658 2018-03-14 Martin Liska <mliska@suse.cz>
9660 * tree.c (record_node_allocation_statistics): Use
9661 get_stats_node_kind.
9662 (get_stats_node_kind): New function extracted from
9663 record_node_allocation_statistics.
9664 (free_node): Use get_stats_node_kind.
9666 2018-03-14 Richard Biener <rguenther@suse.de>
9668 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9669 that the value-set of ANTIC_IN doesn't grow.
9672 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9674 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9675 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9677 2018-03-14 Julia Koval <julia.koval@intel.com>
9679 * config.gcc (icelake-client, icelake-server): New.
9681 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9682 (initial_ix86_arch_features): Ditto.
9683 (PTA_SKYLAKE): Add SGX.
9684 (PTA_ICELAKE): Remove.
9685 (PTA_ICELAKE_CLIENT): New.
9686 (PTA_ICELAKE_SERVER): New.
9687 (ix86_option_override_internal): Split up icelake on icelake client and
9689 (get_builtin_code_for_version): Ditto.
9690 (fold_builtin_cpu): Ditto.
9691 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9692 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9693 * config/i386/i386.h (processor_type): Ditto.
9694 * doc/invoke.texi: Ditto.
9696 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9699 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9700 INTEGER_CST offset, add it together with bitpos / 8 and
9701 sign extend based on POINTER_SIZE.
9705 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9708 * config/i386/constraints.md (Yc): New register constraint.
9709 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9710 Use Yc constraint for alternative 2 of operand 0. Remove
9711 preferred_for_speed attribute.
9713 2018-03-14 Richard Biener <rguenther@suse.de>
9715 PR tree-optimization/84830
9716 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9717 with the old one to avoid oscillations.
9719 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9722 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9724 (assign_by_spills): Return a flag of reload assignment failure.
9725 Do not process the reload assignment failures. Do not spill other
9726 reload pseudos if they has the same reg class. Update n if
9728 (lra_assign): Add a return arg. Set up from the result of
9729 assign_by_spills call.
9730 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9731 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9732 usage_insns if it is not NULL.
9733 (spill_hard_reg_in_range): New function.
9734 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9735 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9736 function prototypes.
9737 (lra_assign): Change prototype.
9738 * lra.c (lra): Add code to deal with fails by splitting hard reg
9741 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9743 * config/riscv/riscv.opt (mrelax): New option.
9744 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9745 "norelax" when riscv_mrelax is disabled.
9746 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9748 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9751 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9752 reassociation for int modes.
9754 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9756 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9757 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9759 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9760 * config/aarch64/aarch64-sve.md
9761 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9762 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9763 (*extend<mode><Vwide>2): Rename to...
9764 (aarch64_sve_extend<mode><Vwide>2): ...this.
9765 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9766 renaming the old pattern to...
9767 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9769 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9770 define_expand, renaming the old pattern to...
9771 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9772 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9773 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9774 account when deciding which SVE instruction the optab should use.
9775 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9777 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9779 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9780 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9781 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9782 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9783 (tlsdesc_small_<mode>): Turn a define_expand and use
9784 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9785 (tlsdesc_small_advsimd_<mode>): ...this.
9786 (tlsdesc_small_sve_<mode>): New pattern.
9788 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9790 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9791 (UNSPEC_UMUL_HIGHPART): New constants.
9792 (MUL_HIGHPART): New int iteraor.
9793 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9794 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9796 (*<su>mul<mode>3_highpart): New define_insn.
9798 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9801 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9804 2018-03-13 Martin Liska <mliska@suse.cz>
9807 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9809 (sem_item_optimizer::~sem_item_optimizer): Release it.
9810 (sem_item_optimizer::merge_classes): Register variable aliases.
9811 (sem_item_optimizer::fixup_pt_set): New function.
9812 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9813 * ipa-icf.h: Declare new variables and functions.
9815 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9818 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9819 integer_pow2p@2 and test integer_pow2p in condition.
9820 (A < 0 ? C : 0): Similarly for @1.
9823 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9824 characters starting at p contain '\0' character, don't look beyond
9828 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9829 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9832 * reg-stack.c (change_stack): Change update_end var from int to
9833 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9834 also call set_block_for_insn on the newly added insns and rescan.
9837 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9838 on the last operand.
9841 * tree.c (stabilize_reference_1): Return save_expr (e) for
9842 STATEMENT_LIST even if it doesn't have side-effects.
9844 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9846 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9848 2018-03-12 Renlin Li <renlin.li@arm.com>
9850 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9851 aarch64_output_scalar_simd_mov_immediate.
9853 2018-03-12 Martin Sebor <msebor@redhat.com>
9855 PR tree-optimization/83456
9856 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9857 for perfectly overlapping calls to memcpy.
9858 (gimple_fold_builtin_memory_chk): Same.
9859 (gimple_fold_builtin_strcpy): Handle no-warning.
9860 (gimple_fold_builtin_stxcpy_chk): Same.
9861 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9863 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9865 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9866 parameter. Use it for SFmode.
9867 (rs6000_function_arg_advance_1): Adjust.
9868 (rs6000_function_arg): Adjust.
9869 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9871 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9873 PR rtl-optimization/84169
9874 PR rtl-optimization/84780
9875 * combine.c (can_combine_p): Check for a 2-insn combination whether
9876 the destination register is used between the two insns, too.
9878 2018-03-12 Richard Biener <rguenther@suse.de>
9880 PR tree-optimization/84803
9881 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9882 for refs DR analysis didn't process.
9884 2018-03-12 Richard Biener <rguenther@suse.de>
9886 PR tree-optimization/84777
9887 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9888 force-vectorize loops ignore whether we are optimizing for size.
9890 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9892 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9893 (TARGET_MD_ASM_ADJUST): Define.
9895 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9896 Kito Cheng <kito.cheng@gmail.com>
9897 Chung-Ju Wu <jasonwucj@gmail.com>
9899 * config/nds32/nds32.c (nds32_compute_stack_frame,
9900 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9901 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9902 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9903 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9904 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9905 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9906 * config/nds32/nds32.md (prologue, epilogue): Use macro
9907 NDS32_V3PUSH_AVAILABLE_P to do checking.
9909 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9912 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9913 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9914 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9915 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9916 addition of most attributes on !orig_type_die or the attribute not
9917 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9919 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9920 Chung-Ju Wu <jasonwucj@gmail.com>
9922 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9924 * config/nds32/nds32.opt (mvh): New option.
9926 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9927 Chung-Ju Wu <jasonwucj@gmail.com>
9929 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9931 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9932 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9935 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9936 Chung-Ju Wu <jasonwucj@gmail.com>
9938 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9940 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9941 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9943 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9944 Kito Cheng <kito.cheng@gmail.com>
9945 Chung-Ju Wu <jasonwucj@gmail.com>
9947 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9948 UNSPEC_FFMISM and UNSPEC_FLMISM.
9949 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9950 for ffb, ffmism and flmism.
9951 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9952 (unspec_ffmism): Ditto.
9953 (unspec_flmism): Ditto.
9954 (nds32_expand_builtin_impl): Check if string extension is available.
9955 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9956 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9958 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9961 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9964 * lra-assigns.c (assign_by_spills): Return a flag of reload
9965 assignment failure. Do not process the reload assignment
9966 failures. Do not spill other reload pseudos if they has the same
9968 (lra_assign): Add a return arg. Set up from the result of
9969 assign_by_spills call.
9970 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9971 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9972 usage_insns if it is not NULL.
9973 (spill_hard_reg_in_range): New function.
9974 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9975 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9976 function prototypes.
9977 (lra_assign): Change prototype.
9978 * lra.c (lra): Add code to deal with fails by splitting hard reg
9981 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9984 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9986 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9989 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9990 (dw_val_node): Add val_symbolic_view.
9991 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9992 (symview_upper_bound): New.
9993 (new_line_info_table): Initialize symviews_since_reset.
9994 (dwarf2out_source_line): Count symviews_since_reset and set
9995 symview_upper_bound.
9996 (dw_val_equal_p): Handle symview.
9997 (add_AT_symview): New.
9998 (print_dw_val): Handle symview.
9999 (attr_checksum, attr_checksum_ordered): Likewise.
10000 (same_dw_val_p, size_of_die): Likewise.
10001 (value_format, output_die): Likewise.
10002 (add_high_low_attributes): Use add_AT_symview for entry_view.
10003 (dwarf2out_finish): Reset symview_upper_bound, clear
10006 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10009 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10010 Add strict argument and use it.
10011 (rs6000_split_multireg_move): Update for new strict argument.
10012 (mem_operand_gpr): Disallow all non-offsettable addresses.
10013 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10015 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10018 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10019 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10020 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10023 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10024 decl, use remap_type if we want to use the type.
10026 2018-03-09 Martin Sebor <msebor@redhat.com>
10028 PR tree-optimization/84526
10029 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10031 (builtin_access::generic_overlap): Be prepared to handle non-array
10034 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10036 PR rtl-optimization/84682
10037 * lra-constraints.c (process_address_1): Check is_address flag
10038 for address constraints.
10039 (process_alt_operands): Likewise.
10040 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10041 preprocess_constraints.
10042 * recog.h (preprocess_constraints): Add oploc parameter.
10044 * recog.c (preprocess_constraints): Test address_operand for
10045 CT_ADDRESS constraints.
10047 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10050 * lra-assigns.c (assign_by_spills): Return a flag of reload
10051 assignment failure. Do not process the reload assignment
10052 failures. Do not spill other reload pseudos if they has the same
10054 (lra_assign): Add a return arg. Set up from the result of
10055 assign_by_spills call.
10056 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10057 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10058 usage_insns if it is not NULL.
10059 (spill_hard_reg_in_range): New function.
10060 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10061 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10062 function prototypes.
10063 (lra_assign): Change prototype.
10064 * lra.c (lra): Add code to deal with fails by splitting hard reg
10067 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10070 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10071 Accept complain bool parameter. Only emit errors if it is true.
10072 (arm_parse_cpu_option_name): Likewise.
10073 (arm_target_thumb_only): Adjust callers of the above.
10074 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10075 prototype to take a default true bool parameter.
10076 (arm_parse_arch_option_name): Likewise.
10078 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10079 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10083 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10084 * configure: Regenerate.
10085 * configure.ac ("linker --version-script option"): New.
10086 ("linker soname option"): New.
10088 2018-03-09 Richard Biener <rguenther@suse.de>
10090 PR tree-optimization/84775
10091 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10092 immediate uses of predicate stmts and mark them modified.
10095 PR tree-optimization/84178
10096 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10098 (version_loop_for_if_conversion): Delay update_ssa call.
10099 (tree_if_conversion): Delay update_ssa until after predicate
10102 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10105 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10106 when the function accesses prior frames.
10108 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10111 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10112 gen_llsym, otherwise call maybe_gen_llsym.
10114 PR inline-asm/84742
10115 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10116 has ',' character inside of it.
10118 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10121 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10122 as clobbering CC_REGNUM.
10124 2018-03-08 Richard Biener <rguenther@suse.de>
10126 PR middle-end/84552
10127 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10128 (follow_copies_to_constant): Do not follow SSA names registered
10131 2018-03-08 Richard Biener <rguenther@suse.de>
10133 PR tree-optimization/84178
10134 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10136 (version_loop_for_if_conversion): Delay update_ssa call.
10137 (tree_if_conversion): Delay update_ssa until after predicate
10140 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10142 PR tree-optimization/84178
10143 * tree-if-conv.c (release_bb_predicate): Remove the
10144 the assertion that the stmts have NULL use_ops.
10145 Discard the statements, asserting that they haven't
10146 yet been added to a BB.
10148 2018-03-08 Richard Biener <rguenther@suse.de>
10150 PR tree-optimization/84746
10151 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10152 (phi_translate): Pass in destination ANTIC_OUT set.
10153 (phi_translate_1): Likewise. For a simplified result lookup
10154 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10155 (phi_translate_set): Adjust.
10156 (do_pre_regular_insertion): Likewise.
10157 (do_pre_partial_partial_insertion): Likewise.
10159 2018-03-08 Martin Liska <mliska@suse.cz>
10161 PR gcov-profile/84735
10162 * doc/gcov.texi: Document usage of profile files.
10163 * gcov-io.h: Document changes in the format.
10165 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10169 * dwarf2out.c (struct dw_line_info_table): Update comments for
10171 (FORCE_RESET_NEXT_VIEW): New.
10172 (FORCE_RESETTING_VIEW_P): New.
10173 (RESETTING_VIEW_P): Check for -1 too.
10174 (ZERO_VIEW_P): Likewise.
10175 (new_line_info_table): Force-reset next view.
10176 (dwarf2out_begin_function): Likewise.
10177 (dwarf2out_source_line): Simplify zero_view_p initialization.
10178 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10179 view directly. Omit view when omitting .loc at line 0.
10181 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10183 PR tree-optimization/84740
10184 * tree-switch-conversion.c (process_switch): Call build_constructors
10185 only if info.phi_count is non-zero.
10187 PR tree-optimization/84739
10188 * tree-tailcall.c (find_tail_calls): Check call arguments against
10189 DECL_ARGUMENTS (current_function_decl) rather than
10190 DECL_ARGUMENTS (func) when checking for tail recursion.
10192 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10194 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10195 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10196 Volker Reichelt's entry and add entries for people that perform
10197 GCC fuzzy testing and report numerous bugs.
10199 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10202 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10203 readonly data in sdata, if that is disabled.
10204 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10205 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10206 -mreadonly-in-sdata option.
10208 2018-03-07 Martin Sebor <msebor@redhat.com>
10210 PR tree-optimization/84468
10211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10212 basic block when looking for nul assignment.
10214 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10217 * except.h (output_function_exception_table): Adjust prototype.
10218 * except.c (output_function_exception_table): Remove FNNAME parameter
10219 and add SECTION parameter. Ouput one part of the table at a time.
10220 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10221 the first part of the exception table and emit unwind directives.
10222 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10223 (i386_pe_seh_cold_init): Likewise.
10224 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10225 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10226 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10227 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10228 * config/i386/winnt.c: Include except.h.
10229 (struct seh_frame_state): Add reg_offset, after_prologue and
10230 in_cold_section fields.
10231 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10232 (i386_pe_seh_cold_init): New function.
10233 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10234 to seh->in_cold_section.
10235 (seh_emit_push): Record the offset of the push.
10236 (seh_emit_save): Record the offet of the save.
10237 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10238 Test seh->after_prologue to disregard the epilogue.
10239 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10240 (i386_pe_end_cold_function): New function.
10242 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10245 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10246 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10249 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10251 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10252 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10254 PR middle-end/84723
10255 * multiple_target.c: Include tree-inline.h and intl.h.
10256 (expand_target_clones): Diagnose and fail if node->definition and
10257 !tree_versionable_function_p (node->decl).
10259 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10261 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10263 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10264 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10265 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10267 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10270 * combine.c (try_combine): Use reg_or_subregno instead of handling
10271 just paradoxical SUBREGs and REGs.
10273 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10275 * config/arc/arc.c (arc_finalize_pic): Remove function.
10276 (arc_must_save_register): We use single base PIC register, remove
10277 checks to save/restore the PIC register.
10278 (arc_expand_prologue): Likewise.
10279 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10281 (arc_verify_short): Likewise.
10282 (arc_attr_type): Likewise.
10283 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10284 (walk_stores): Likewise.
10285 (arc_address_cost): Make it static.
10286 (arc_verify_short): Likewise.
10287 (branch_dest): Likewise.
10288 (arc_attr_type): Likewise.
10289 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10290 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10291 (arc_final_prescan_insn): Remove inserting the nops due to
10292 hardware hazards. It is done in reorg step.
10293 (insn_length_variant_t): Remove.
10294 (insn_length_parameters_t): Likewise.
10295 (arc_insn_length_parameters): Likewise.
10296 (arc_get_insn_variants): Likewise.
10297 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10299 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10301 PR inline-asm/84683
10302 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10305 PR tree-optimization/84687
10306 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10308 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10310 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10312 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10313 Rename to ppc_speculation_barrier.
10314 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10315 __builtin_ppc_speculation_barrier.
10317 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10320 * combine.c (combine_simplify_rtx): Don't try to simplify if
10321 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10324 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10326 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10327 to 32 bytes when compiling for POWER9.
10329 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10332 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10333 regparm >= 3 with no arg reg available also for calls with
10334 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10337 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10339 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10341 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10344 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10346 2018-03-05 Richard Biener <rguenther@suse.de>
10348 PR tree-optimization/84486
10349 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10350 When inserting a __builtin_assume_aligned call set the LHS
10351 SSA name alignment info accordingly.
10353 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10355 PR tree-optimization/84114
10356 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10357 Avoid reassociation of FLOAT_MODE addition.
10359 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10361 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10362 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10363 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10364 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10365 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10366 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10367 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10369 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10370 __builtin_ia32_wbinvd): New builtins.
10371 (SPECIAL_ARGS2): New.
10372 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10373 (SPECIAL_ARGS2): New.
10374 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10375 (ix86_valid_target_attribute_inner_p): Ditto.
10376 (ix86_init_mmx_sse_builtins): Add special_args2.
10377 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10378 TARGET_WBNOINVD_P): New.
10379 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10380 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10381 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10382 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10383 * config/i386/pconfigintrin.h: New file.
10384 * config/i386/wbnoinvdintrin.h: Ditto.
10385 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10387 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10389 2018-03-05 Richard Biener <rguenther@suse.de>
10391 PR tree-optimization/84670
10392 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10394 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10395 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10396 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10397 doesn't grow if all successors (recursively) were visited at least
10400 2018-03-05 Richard Biener <rguenther@suse.de>
10402 PR tree-optimization/84650
10403 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10404 if executed in the loop pipeline.
10406 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10408 * doc/configfiles.texi (Configuration Files): Move info about
10409 conditionalizing $target-protos.h to...
10410 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10411 differs from $target-protos.h.
10413 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10414 Chung-Ju Wu <jasonwucj@gmail.com>
10416 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10417 * config/nds32/nds32-multiple.md (setmemsi): Define.
10418 * config/nds32/nds32-memory-manipulation.c
10419 (nds32_gen_dup_4_byte_to_word_value): New.
10420 (emit_setmem_word_loop): New.
10421 (emit_setmem_byte_loop): New.
10422 (nds32_expand_setmem_loop): New.
10423 (nds32_expand_setmem_loop_v3m): New.
10424 (nds32_expand_setmem_unroll): New.
10425 (nds32_expand_setmem): New.
10427 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10428 Chung-Ju Wu <jasonwucj@gmail.com>
10430 * config/nds32/nds32-memory-manipulation.c
10431 (nds32_emit_load_store): New.
10432 (nds32_emit_post_inc_load_store): New.
10433 (nds32_emit_mem_move): New.
10434 (nds32_emit_mem_move_block): New.
10435 (nds32_expand_movmemsi_loop_unknown_size): New.
10436 (nds32_expand_movmemsi_loop_known_size): New.
10437 (nds32_expand_movmemsi_loop): New.
10438 (nds32_expand_movmemsi_unroll): New.
10439 (nds32_expand_movmemqi): Rename ...
10440 (nds32_expand_movmemsi): ... to this.
10441 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10442 (movmemsi): ... to this.
10443 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10444 (nds32_expand_movmemsi): ... to this.
10446 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10447 Monk Chiang <sh.chiang04@gmail.com>
10448 Chung-Ju Wu <jasonwucj@gmail.com>
10450 * config/nds32/nds32-protos.h
10451 (nds32_expand_load_multiple): New arguments.
10452 (nds32_expand_store_multiple): Ditto.
10453 (nds32_valid_multiple_load_store): Rename ...
10454 (nds32_valid_multiple_load_store_p): ... to this.
10455 * config/nds32/nds32-memory-manipulation.c
10456 (nds32_expand_load_multiple): Refine implementation.
10457 (nds32_expand_store_multiple): Ditto.
10458 * config/nds32/nds32-multiple.md
10459 (load_multiple): Update nds32_expand_load_multiple interface.
10460 (store_multiple): Update nds32_expand_store_multiple interface.
10461 * config/nds32/nds32-predicates.c
10462 (nds32_valid_multiple_load_store): Rename ...
10463 (nds32_valid_multiple_load_store_p): ... to this and refine
10465 * config/nds32/predicates.md
10466 (nds32_load_multiple_and_update_address_operation): New predicate.
10467 (nds32_store_multiple_and_update_address_operation): New predicate.
10469 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10470 Chung-Ju Wu <jasonwucj@gmail.com>
10472 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10473 (combo): New attribute.
10474 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10476 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10478 * config/nds32/nds32.opt: Change -mcmodel= default value.
10480 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10481 Monk Chiang <sh.chiang04@gmail.com>
10482 Chung-Ju Wu <jasonwucj@gmail.com>
10484 * config/nds32/constants.md (unspec_element): New enum.
10485 * config/nds32/constraints.md (Umw): New constraint.
10486 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10487 * config/nds32/nds32-intrinsic.md: Likewise.
10488 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10489 (nds32_valid_smw_lwm_base_p): New.
10490 (nds32_output_smw_single_word): New.
10491 (nds32_output_lmw_single_word): New.
10492 (nds32_expand_unaligned_load): New.
10493 (nds32_expand_unaligned_store): New.
10494 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10495 (nds32_output_smw_single_word): Declare.
10496 (nds32_output_lmw_single_word): Declare.
10497 (nds32_expand_unaligned_load): Declare.
10498 (nds32_expand_unaligned_store): Declare.
10499 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10500 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10501 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10502 NDS32_BUILTIN_UASTORE_DW.
10503 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10506 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10507 Kito Cheng <kito.cheng@gmail.com>
10508 Chung-Ju Wu <jasonwucj@gmail.com>
10510 * config/nds32/nds32-intrinsic.c
10511 (nds32_expand_builtin_null_ftype_reg): Delete.
10512 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10513 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10514 (nds32_read_argument): New.
10515 (nds32_legitimize_target): Ditto.
10516 (nds32_legitimize_argument): Ditto.
10517 (nds32_check_constant_argument): Ditto.
10518 (nds32_expand_unop_builtin): Ditto.
10519 (nds32_expand_unopimm_builtin): Ditto.
10520 (nds32_expand_binop_builtin): Ditto.
10521 (nds32_builtin_decl_impl): Ditto.
10522 (builtin_description): Ditto.
10523 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10524 (nds32_init_builtins_impl): Ditto.
10525 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10526 (nds32_builtin_decl): New.
10527 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10528 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10530 2018-03-02 Jeff Law <law@redhat.com>
10532 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10533 (redundant_insn, fill_simple_delay_slots): Likewise.
10534 (fill_slots_from_thread): Likewise.
10535 * resource.c (mark_referenced_resources): Likewise.
10536 (mark_set_resources, find_dead_or_set_registers): Likewise.
10538 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10540 * substring-locations.h (format_warning_va): Formatting fix for
10541 ATTRIBUTE_GCC_DIAG.
10542 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10544 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10545 * substring-locations.c: Include intl.h.
10546 (format_warning_va): Turned into small wrapper around
10547 format_warning_n_va, renamed to ...
10548 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10549 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10551 (format_warning_at_substring_n): New function.
10552 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10553 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10554 format_warning_at_substring with just a shorter name instead of
10555 const function pointer.
10556 (fmtwarn_n): New function.
10557 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10558 appropriate, get rid of all the fmtstr temporaries, move conditionals
10559 with G_() wrapped string literals directly into fmtwarn arguments,
10560 cast dir.len to (int), formatting fixes.
10562 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10564 * doc/invoke.texi: Remove "Cilk Plus" references.
10566 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10567 Richard Biener <rguenther@suse.de>
10570 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10571 for error or warning attributes if CALL_FROM_THUNK_P is set.
10574 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10577 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10578 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10581 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10582 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10584 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10585 instead of -1U in last predictors element's probability member.
10587 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10590 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10591 arguments if they are comparable.
10593 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10595 PR tree-optimization/84634
10596 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10597 masks and masked_loop_p with a single loop_masks, making sure it's
10598 null for bb vectorization.
10600 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10602 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10603 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10604 loop->force_vectorize to check whether there is no alias.
10606 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10609 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10611 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10613 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10614 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10615 instead of a loop around prev_real_insn.
10616 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10619 PR inline-asm/84625
10620 * config/i386/i386.c (ix86_print_operand): Use conditional
10621 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10624 2018-03-02 Richard Biener <rguenther@suse.de>
10626 PR tree-optimization/84427
10627 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10628 (bitmap_set_subtract_values): Rewrite to handle multiple
10631 (prune_clobbered_mems): Likewise.
10632 (phi_translate): Take edge instead of pred/phiblock.
10633 (phi_translate_1): Likewise.
10634 (phi_translate_set): Likewise. Insert all translated
10635 exprs for a value into the set, keeping possibly multiple
10636 expressions per value.
10637 (compute_antic_aux): Adjust for phi_translate changes.
10638 When intersecting union the expressions and prune those
10639 not in the final value set, keeping possibly multiple
10640 expressions per value. Do not use value-insertion
10641 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10642 all expressions. Add verification that the value-sets
10643 only shrink during iteration.
10644 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10645 (do_pre_regular_insertion): Likewise.
10646 (do_pre_partial_partial_insertion): Likewise.
10648 2018-03-02 Richard Biener <rguenther@suse.de>
10651 * config/darwin.c (saved_debug_info_level): New static global.
10652 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10653 (darwin_asm_lto_end): Restore debug info generation settings.
10655 2018-03-01 Martin Liska <mliska@suse.cz>
10658 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10659 volatile arguments.
10661 2018-03-01 Richard Biener <rguenther@suse.de>
10664 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10665 type annotation with LTO when debug was disabled at compile-time.
10667 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10669 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10671 (mips_final_postscan_insn): Likewise.
10673 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10675 PR rtl-optimization/84528
10676 * alias.c (init_alias_target): Add commentary.
10677 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10678 a unique base value if the frame pointer is not eliminated
10679 to the stack pointer.
10681 2018-03-01 Tom de Vries <tom@codesourcery.com>
10683 PR rtl-optimization/83327
10684 * lra-int.h (hard_regs_spilled_into): Declare.
10685 * lra.c (hard_regs_spilled_into): Define.
10686 (init_reg_info): Init hard_regs_spilled_into.
10687 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10688 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10689 (process_bb_lives): Handle hard_regs_spilled_into.
10690 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10691 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10693 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10695 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10696 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10697 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10698 * config/rs6000/aix72.h: New file.
10700 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10702 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10703 instead of warning_at with conditional singular and plural messages
10707 * stor-layout.c (update_alignment_for_field): For
10708 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10709 && !DECL_PACKED (field), do the alignment update, just use
10710 only desired_align instead of MAX (type_align, desired_align)
10712 (place_field): Don't do known_align < desired_align handling
10713 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10714 is non-NULL, instead do it after rli->prev_field handling and
10715 only if not within a bitfield word. For DECL_PACKED (field)
10716 use type_align of BITS_PER_UNIT.
10718 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10720 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10721 superfluous parentheses and trailing spaces.
10723 2018-02-28 Richard Biener <rguenther@suse.de>
10725 PR tree-optimization/84584
10726 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10727 SCoPs with fake exit edge.
10729 2018-02-28 Martin Liska <mliska@suse.cz>
10732 * timevar.c (timer::print): Fix format to properly print 100%
10735 2018-02-28 Richard Biener <rguenther@suse.de>
10737 PR middle-end/84607
10738 * genmatch.c (capture_info::walk_match): Do not mark
10739 captured expressions without operands as expr_p given
10740 they act more like predicates and should be subject to
10741 "lost tail" side-effect preserving.
10743 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10745 PR rtl-optimization/81611
10746 * auto-inc-dec.c (attempt_change): Move dead note from
10747 mem_insn if it's the next use of regno
10748 (find_address): Take address use of reg holding
10749 non-incremented value. Add parm to limit search to the named
10751 (merge_in_block): Attempt to use a mem insn that is the next
10752 use of the original regno.
10754 2018-02-27 Martin Sebor <msebor@redhat.com>
10757 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10758 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10760 2018-02-27 Martin Sebor <msebor@redhat.com>
10762 PR translation/84207
10763 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10764 n argument to unsigned HOST_WIDE_INT.
10765 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10766 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10767 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10768 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10770 2018-02-27 Richard Biener <rguenther@suse.de>
10772 PR tree-optimization/84512
10773 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10774 Do not use the estimate returned from record_stmt_cost for
10775 the scalar iteration cost but sum properly using add_stmt_cost.
10777 2018-02-27 Richard Biener <rguenther@suse.de>
10779 PR tree-optimization/84466
10780 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10781 Adjust last change to less strictly validate use operands.
10783 2018-02-27 Martin Liska <mliska@suse.cz>
10785 PR gcov-profile/84548
10786 * gcov.c (process_file): Allow partial overlap and consider it
10787 also as group functions.
10788 (output_lines): Properly calculate range of lines for a group.
10790 2018-02-27 Martin Liska <mliska@suse.cz>
10792 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10793 'ggc' suffixes. Change first column width.
10794 (timer::print): Fix formatting of the column.
10796 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10798 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10799 preserve inline entry blocks for the sake of debug inline
10800 entry point markers alone.
10801 (remove_unused_locals): Suggest in comments a better place to
10802 force the preservation of inline entry blocks that are
10803 otherwise unused, but do not preserve them.
10805 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10807 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10809 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10812 * config/i386/constraints.md (Bs): Replace
10813 ix86_indirect_branch_register with
10814 TARGET_INDIRECT_BRANCH_REGISTER.
10816 * config/i386/i386.md (indirect_jump): Likewise.
10817 (tablejump): Likewise.
10818 (*sibcall_memory): Likewise.
10819 (*sibcall_value_memory): Likewise.
10820 Peepholes of indirect call and jump via memory: Likewise.
10821 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10822 (*sibcall_value_GOT_32): Likewise.
10823 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10824 (GOT_memory_operand): Likewise.
10825 (call_insn_operand): Likewise.
10826 (sibcall_insn_operand): Likewise.
10827 (GOT32_symbol_operand): Likewise.
10828 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10830 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10832 PR rtl-optimization/83496
10833 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10834 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10835 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10836 redundant insn, if any.
10837 (relax_delay_slots): Likewise.
10838 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10840 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10842 PR tree-optimization/83965
10843 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10844 that grouped statements are part of a reduction chain. Return
10845 true if the statement is not marked as a reduction itself but
10846 is part of a group.
10847 (vect_recog_dot_prod_pattern): Don't check whether the statement
10848 is part of a group here.
10849 (vect_recog_sad_pattern): Likewise.
10850 (vect_recog_widen_sum_pattern): Likewise.
10852 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10855 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10857 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10860 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10862 (ix86_output_indirect_function_return): New prototype.
10863 (ix86_split_simple_return_pop_internal): Likewise.
10864 * config/i386/i386.c (indirect_return_via_cx): New.
10865 (indirect_return_via_cx_bnd): Likewise.
10866 (indirect_thunk_name): Handle return va CX_REG.
10867 (output_indirect_thunk_function): Create alias for
10868 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10869 (ix86_output_indirect_jmp): Remove the bool argument.
10870 (ix86_output_indirect_function_return): New function.
10871 (ix86_split_simple_return_pop_internal): Likewise.
10872 * config/i386/i386.md (*indirect_jump): Don't pass false
10873 to ix86_output_indirect_jmp.
10874 (*tablejump_1): Likewise.
10875 (simple_return_pop_internal): Change it to define_insn_and_split.
10876 Call ix86_split_simple_return_pop_internal to split it for
10877 -mfunction-return=.
10878 (simple_return_indirect_internal): Call
10879 ix86_output_indirect_function_return instead of
10880 ix86_output_indirect_jmp.
10882 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10885 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10886 memset and value initialization afterwards.
10888 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10890 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10892 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10895 * common/config/aarch64/aarch64-common.c
10896 (aarch_option_optimization_table[]): Switch
10897 off fomit-frame-pointer
10899 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10900 Chung-Ju Wu <jasonwucj@gmail.com>
10902 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10904 (store_multiple): Ditto.
10906 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10908 * config.gcc: Add --with-cpu support for nds32 target.
10909 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10910 * config/nds32/nds32.opt: Add -mcpu= option.
10912 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10914 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10915 isel=yes): Warn for these deprecated options.
10917 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10919 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10920 ISA_2_5_MASKS_EMBEDDED.
10922 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10924 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10925 p->max as pointers rather than using iterative_hash_expr.
10927 2018-02-23 Carl Love <cel@us.ibm.com>
10929 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10930 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10932 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10933 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10934 P8V_BUILTIN_VEC_VUNSIGNED2.
10936 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10939 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10940 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10942 * lra-constraints.c (curr_insn_transform): Set up
10943 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10945 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10946 * lra-eliminations.c (spill_pseudos): Ditto.
10947 (process_insn_for_elimination): Ditto.
10948 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10949 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10950 LRA_NON_CLOBBERED_ALT.
10952 2018-02-22 Martin Sebor <msebor@redhat.com>
10954 PR tree-optimization/84480
10955 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10956 to maybe_diag_stxncpy_trunc. Call it.
10957 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10958 from gimple_fold_builtin_strcpy. Print inlining stack.
10959 (handle_builtin_stxncpy): Print inlining stack.
10960 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10962 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10965 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10966 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10967 and -fcheck-pointer-bounds are used together.
10968 (indirect_thunk_prefix): New enum.
10969 (indirect_thunk_need_prefix): New function.
10970 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10971 "_nt" instead of "_bnd" for NOTRACK prefix.
10972 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10973 (output_indirect_thunk_function): Likewise.
10975 (ix86_code_end): Update output_indirect_thunk_function calls.
10976 (ix86_output_indirect_branch_via_reg): Replace
10977 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10978 (ix86_output_indirect_branch_via_push): Likewise.
10979 (ix86_output_function_return): Likewise.
10980 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10981 incompatible with -fcf-protection=branch and
10982 -fcheck-pointer-bounds.
10984 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10987 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10988 Change gcc_assert call to output_operand_lossage.
10990 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10992 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10994 2018-02-22 DJ Delorie <dj@redhat.com>
10995 Sebastian Perta <sebastian.perta@renesas.com>
10996 Oleg Endo <olegendo@gcc.gnu.org>
10998 * config/rx/rx.c (rx_rtx_costs): New function.
10999 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11001 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11003 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11005 2018-02-22 Martin Liska <mliska@suse.cz>
11008 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11009 Add "native" as a possible value.
11011 2018-02-22 Martin Liska <mliska@suse.cz>
11014 * config/i386/i386.c (ix86_option_override_internal):
11015 Add "native" as a possible value for -march and -mtune.
11017 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11020 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11021 to all type variants.
11023 PR tree-optimization/84503
11024 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11025 width as info->bitpos + info->bitsize - start.
11026 (merged_store_group::merge_overlapping): Simplify width computation.
11027 (check_no_overlap): New function.
11028 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11029 start + width and last_order of the group, fail if check_no_overlap
11031 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11032 to group if check_no_overlap fails.
11034 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11036 * config/rs6000/altivec.md: Delete contraint arguments to
11037 define_expand, define_split, and define_peephole2, and in
11038 define_insn_and_split if always unused.
11039 * config/rs6000/darwin.md: Ditto.
11040 * config/rs6000/dfp.md: Ditto.
11041 * config/rs6000/rs6000.md: Ditto.
11042 * config/rs6000/sync.md: Ditto.
11043 * config/rs6000/vector.md: Ditto.
11044 * config/rs6000/vsx.md: Ditto.
11046 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11048 * config/rs6000/altivec.md: Write output control strings as braced
11049 blocks instead of double-quoted strings.
11050 * config/rs6000/darwin.md: Ditto.
11051 * config/rs6000/rs6000.md: Ditto.
11052 * config/rs6000/vector.md: Ditto.
11053 * config/rs6000/vsx.md: Ditto.
11055 2018-02-21 Jason Merrill <jason@redhat.com>
11057 PR c++/84314 - ICE with templates and fastcall attribute.
11058 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11060 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11062 * ipa-cp.c (determine_versionability): Fix comment typos.
11064 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11067 * ipa-cp.c (determine_versionability): Do not version functions caling
11070 2018-02-21 Martin Liska <mliska@suse.cz>
11073 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11074 Add "native" as a possible value.
11075 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11076 the macro when native cpu detection is available.
11078 2018-02-21 Martin Liska <mliska@suse.cz>
11081 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11082 Add "native" as a possible value.
11083 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11084 when native cpu detection is available.
11086 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11087 Martin Sebor <msebor@redhat.com>
11089 PR tree-optimization/84478
11090 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11092 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11093 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11094 support which is conservatively correct, for 2 only stay conservative
11095 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11096 argument to the 2 argument get_range_strlen, adjust 6 arg
11097 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11099 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11100 (gimple_fold_builtin_strlen): Pass true as last argument to
11103 2018-02-20 Martin Sebor <msebor@redhat.com>
11105 PR middle-end/84095
11106 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11107 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11108 (builtin_memref::builtin_memref): Factor out parts into
11109 set_base_and_offset and call it.
11111 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11113 PR middle-end/84406
11114 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11115 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11116 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11117 search at the associated MODE_INT.
11119 2018-02-20 Jeff Law <law@redhat.com>
11121 PR middle-end/82123
11122 PR tree-optimization/81592
11123 PR middle-end/79257
11124 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11125 for range data rather than using global data.
11126 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11127 range data rather than using global data.
11128 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11129 pass it to children as needed.
11130 (struct directive::fmtresult): Similarly.
11131 (struct directive::set_width): Similarly.
11132 (struct directive::set_precision): Similarly.
11133 (format_integer, format_directive, parse_directive): Similarly.
11134 (format_none): Accept unnamed vr_values parameter.
11135 (format_percent, format_floating, format_character): Similarly.
11136 (format_string, format_plain): Similarly.
11137 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11138 the EVRP range analyzer for range data rather than using global data.
11139 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11140 gimple-ssa-evrp-analyze.h
11141 (class sprintf_dom_walker): Add after_dom_children member function.
11142 Add evrp_range_analyzer member.
11143 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11144 range analyzer as needed.
11145 (sprintf_dom_walker::after_dom_children): New member function.
11146 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11148 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11149 (evrp_range_analyzer::pop_to_marker): Likewise.
11151 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11153 PR tree-optimization/84419
11154 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11155 with the required type if its current type is compatible but
11158 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11160 PR middle-end/82004
11161 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11162 after vectorization.
11164 2018-02-20 Martin Liska <mliska@suse.cz>
11167 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11168 possible values if we don't have a hint.
11170 2018-02-20 Martin Liska <mliska@suse.cz>
11174 * final.c (shorten_branches): Build align_tab array with one
11176 * opts.c (finish_options): Add alignment option limit check.
11177 (MAX_CODE_ALIGN): Likewise.
11178 (MAX_CODE_ALIGN_VALUE): Likewise.
11179 * doc/invoke.texi: Document maximum allowed option value for
11180 all -falign-* options.
11182 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11185 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11186 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11187 * var-tracking.c (emit_note_insn_var_location): Remove all references
11188 to NOTE_INSN_CALL_ARG_LOCATION.
11189 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11190 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11191 Use copy_rtx_if_shared.
11192 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11193 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11194 (dwarf2out_var_location): Remove handling of
11195 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11197 * final.c (final_scan_insn): Remove all references to
11198 NOTE_INSN_CALL_ARG_LOCATION.
11199 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11200 before dumping final insns.
11201 * except.c (emit_note_eh_region_end): Remove all references to
11202 NOTE_INSN_CALL_ARG_LOCATION.
11203 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11204 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11205 * config/arc/arc.c (hwloop_optimize): Likewise.
11206 * config/arm/arm.c (create_fix_barrier): Likewise.
11207 * config/s390/s390.c (s390_chunkify_start): Likewise.
11208 * config/sh/sh.c (find_barrier): Likewise.
11209 * config/i386/i386.c (rest_of_insert_endbranch,
11210 ix86_seh_fixup_eh_fallthru): Likewise.
11211 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11212 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11213 * config/frv/frv.c (frv_function_prologue): Likewise.
11214 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11216 (note_outside_basic_block_p): Remove all references to
11217 NOTE_INSN_CALL_ARG_LOCATION.
11218 * gengtype.c (adjust_field_rtx_def): Likewise.
11219 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11221 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11222 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11225 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11228 PR tree-optimization/84452
11229 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11230 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11233 2018-02-19 Martin Liska <mliska@suse.cz>
11236 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11238 2018-02-19 Martin Liska <mliska@suse.cz>
11239 Richard Sandiford <richard.sandiford@linaro.org>
11241 PR tree-optimization/82491
11242 * gimple-fold.c (get_base_constructor): Make earlier bail out
11245 2018-02-19 Carl Love <cel@us.ibm.com>
11247 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11248 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11250 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11251 P8V_BUILTIN_VEC_NEG.
11253 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11255 * config/rl78/rl78.md (movdf): New define expand.
11257 2018-02-19 Martin Liska <mliska@suse.cz>
11260 * doc/invoke.texi: Fix typo.
11261 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11263 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11265 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11266 handle rs6000_single_float and rs6000_double_float specially for
11269 2018-02-16 Jeff Law <law@redhat.com>
11271 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11272 (pop_regs): Use it.
11274 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11277 * ipa-inline.c (inline_small_functions): Fix a typo.
11279 2018-02-16 Nathan Sidwell <nathan@acm.org>
11281 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11283 2018-02-16 Carl Love <cel@us.ibm.com>
11285 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11286 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11287 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11288 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11289 expansion to P8V_BUILTIN_VEC_FLOAT2.
11291 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11293 PR rtl-optimization/70023
11294 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11295 src_regno into account.
11297 2018-02-16 Carl Love <cel@us.ibm.com>
11299 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11300 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11301 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11302 * config/rs6000/rs6000.c: Remove case statements for
11303 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11304 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11305 and P9V_BUILTIN_VEC_VINSERT4B.
11306 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11307 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11308 * config/rs6000/vsx.md:
11309 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11312 2018-02-16 Carl Love <cel@us.ibm.com>
11314 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11316 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11318 * config/rs6000/rs6000-c.c: Add the definitions for
11319 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11320 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11321 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11322 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11323 definition for insert4b and define insn *insert3b_internal.
11324 * doc/extend.texi: Add documentation for vec_extract4b.
11326 2018-02-16 Nathan Sidwell <nathan@acm.org>
11328 * doc/extend.texi (Backwards Compatibility): Mention friend
11329 injection. Note for-scope is deprecated.
11330 * doc/invoke.texi (-ffriend-injection): Deprecate.
11332 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11334 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11335 that moved to I2, also allow destinations that are a paradoxical
11336 subreg (instead of a normal reg).
11338 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11341 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11344 2018-02-16 Richard Biener <rguenther@suse.de>
11346 PR tree-optimization/84037
11347 PR tree-optimization/84016
11349 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11350 Adjust vec_construct for the fact we need additional higher latency
11351 128bit inserts for AVX256 and AVX512 vector builds.
11352 (ix86_add_stmt_cost): Scale vector construction cost for
11355 2018-02-16 Richard Biener <rguenther@suse.de>
11357 PR tree-optimization/84417
11358 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11359 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11360 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11362 2018-02-16 Martin Liska <mliska@suse.cz>
11365 * internal-fn.def (ASAN_CHECK): Set proper flags.
11366 (ASAN_MARK): Likewise.
11368 2018-02-16 Julia Koval <julia.koval@intel.com>
11370 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11371 from PTA_CANNONLAKE.
11373 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11376 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11377 Use ++iter rather than iter++ for std::list iterators.
11378 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11379 defer deleting them until all nodes in the forest are processed. Do
11380 free even leaf nodes. Change to_process into auto_vec.
11383 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11384 * vec.h (vec_default_construct): Use memset instead of placement new
11385 if BROKEN_VALUE_INITIALIZATION is defined.
11386 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11387 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11390 PR rtl-optimization/83723
11391 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11392 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11393 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11395 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11397 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11399 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11401 PR rtl-optimization/81443
11402 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11403 from inner REGs to paradoxical SUBREGs.
11405 2018-02-16 Richard Biener <rguenther@suse.de>
11407 PR tree-optimization/84399
11408 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11409 For operands we can analyze at their definition make sure we can
11410 analyze them at each use as well.
11412 2018-02-16 Richard Biener <rguenther@suse.de>
11414 PR tree-optimization/84190
11415 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11416 volatile accesses if the decl isn't volatile.
11418 2018-02-15 Jason Merrill <jason@redhat.com>
11420 PR c++/84314 - ICE with templates and fastcall attribute.
11421 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11422 TYPE_CANONICAL on an existing type.
11424 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11426 PR tree-optimization/84383
11427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11428 dstoff nor call operand_equal_p if dstbase is NULL.
11430 PR tree-optimization/84334
11431 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11432 also a CONSTANT_CLASS_P, punt.
11434 2018-02-14 Jim Wilson <jimw@sifive.com>
11436 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11437 first SMALL_OPERAND check. New local min_second_step. Move assert
11438 to where locals are set. Add TARGET_RVC support.
11439 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11441 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11443 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11445 2018-02-14 Martin Sebor <msebor@redhat.com>
11447 PR tree-optimization/83698
11448 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11449 arrays constrain the offset range to their bounds.
11450 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11451 (builtin_access::overlap): Avoid setting the size of overlap if it's
11453 (maybe_diag_overlap): Also consider arrays when deciding what values
11454 of offsets to include in diagnostics.
11456 2018-02-14 Martin Sebor <msebor@redhat.com>
11459 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11460 that correspond to the kind of a declaration.
11462 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11465 * config/pa/pa.md: Load address of PIC label using the linkage table
11466 if the label is nonlocal.
11468 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11470 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11471 warning message if user requests -maltivec=be.
11472 * doc/invoke.texi: Document deprecation of -maltivec=be.
11474 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11477 * config/rs6000/rs6000-c.c: Update definitions for
11478 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11479 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11481 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11484 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11485 add _get_ssp intrinsics. Remove argument from
11486 __builtin_ia32_rdssp[d|q].
11487 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11488 * config/i386/i386-builtin.def: Remove argument from
11489 __builtin_ia32_rdssp[d|q].
11490 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11491 ix86_expand_special_args_builtin for _rdssp[d|q].
11492 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11493 Clear register before usage.
11494 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11495 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11497 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11499 PR tree-optimization/84357
11500 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11501 operand 1 of an ARRAY_REF too.
11503 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11506 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11507 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11509 (set_of_reg): New struct.
11510 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11511 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11512 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11514 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11515 Split into bitclr, bitset, bitinvert patterns if appropriate.
11516 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11517 use rx_fuse_in_memory_bitop.
11518 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11519 to named insn, correct maximum insn length.
11521 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11524 * machmode.def: Define a complex mode for PARTIAL_INT.
11525 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11527 * doc/rtl.texi: Document CSPImode.
11528 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11530 (msp430_hard_regno_nregs_with_padding): Likewise.
11532 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11535 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11537 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11539 PR rtl-optimization/84169
11540 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11541 we generated a parallel as new i3 and we split that to new i2 and i3
11542 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11543 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11544 those to i2, not i1. Partially rewrite this scan code.
11546 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11549 * stor-layout.c (place_field): For variable length fields, adjust
11550 offset_align afterwards not just based on the field's alignment,
11551 but also on the size.
11553 PR middle-end/84309
11554 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11555 of exps and logs in the use_exp2 case.
11557 2018-02-13 Jeff Law <law@redhat.com>
11559 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11560 entry for "vector".
11562 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11565 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11569 * common.opt (gas-loc-support, gas-locview-support): New.
11570 (ginline-points, ginternal-reset-location-views): New.
11571 * doc/invoke.texi: Document them. Use @itemx where intended.
11572 (gvariable-location-views): Adjust.
11573 * target.def (reset_location_view): New.
11574 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11575 (TARGET_RESET_LOCATION_VIEW): New.
11576 * doc/tm.texi: Rebuilt.
11577 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11578 (dwarf2out_default_as_locview_support): New.
11579 (output_asm_line_debug_info): Use option variables.
11580 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11581 (output_loc_list): Likewise.
11582 (add_high_low_attributes): Check option variables.
11583 Don't output entry view attribute in strict mode.
11584 (gen_inlined_subroutine_die): Check option variables.
11585 (dwarf2out_inline_entry): Likewise.
11586 (init_sections_and_labels): Likewise.
11587 (dwarf2out_early_finish): Likewise.
11588 (maybe_reset_location_view): New, from...
11589 (dwarf2out_var_location): ... here. Call it.
11590 * debug.h (dwarf2out_default_as_loc_support): Declare.
11591 (dwarf2out_default_as_locview_support): Declare.
11592 * hooks.c (hook_int_rtx_insn_0): New.
11593 * hooks.h (hook_int_rtx_insn_0): Declare.
11594 * toplev.c (process_options): Take -gas-loc-support and
11595 -gas-locview-support from dwarf2out. Enable
11596 -gvariable-location-views by default only with locview
11597 assembler support. Enable -ginternal-reset-location-views by
11598 default only if the target defines the corresponding hook.
11599 Enable -ginline-points by default if location views are
11600 enabled; force it disabled if statement frontiers are
11602 * tree-inline.c (expand_call_inline): Check option variables.
11603 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11605 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11607 PR tree-optimization/84321
11608 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11609 handling. Also check whether the anti-range contains any values
11610 that satisfy the mask; switch to a VR_RANGE if not.
11612 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11615 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11617 2018-02-13 Martin Jambor <mjambor@suse.cz>
11620 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11621 of call statements, also set location of a load to a temporary.
11623 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11625 * config/rl78/rl78.c (add_vector_labels): New function.
11626 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11627 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11628 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11629 which checks that no arguments are passed.
11630 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11631 * doc/extend.texi: Documentation for the new attribute.
11633 2018-02-13 Andreas Schwab <schwab@suse.de>
11635 * config/riscv/linux.h (CPP_SPEC): Define.
11637 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11640 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11641 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11642 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11643 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11644 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11645 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11646 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11647 temporarily for AES and PCLMUL builtins.
11649 PR tree-optimization/84339
11650 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11651 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11654 PR middle-end/84309
11655 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11656 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11657 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11659 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11661 * omp-simd-clone.h: New file.
11662 * omp-simd-clone.c: Include omp-simd-clone.h.
11663 (expand_simd_clones): No longer static.
11664 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11665 cgraph.h and omp-simd-clone.h.
11666 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11667 (vect_recog_widen_shift_pattern): Formatting fix.
11668 (vect_pattern_recog_1): Don't check optab for calls.
11671 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11672 operands[2] into a REG before using gen_lowpart on it.
11674 2018-02-12 Jeff Law <law@redhat.com>
11677 * config/sh/sh.c (find_barrier): Consider a sibling call
11680 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11681 successfully back substituting a reg.
11683 2018-02-12 Richard Biener <rguenther@suse.de>
11685 PR tree-optimization/84037
11686 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11687 parameter, move visited init to caller.
11688 (vect_slp_analyze_operations): Separate cost from validity
11689 check, initialize visited once for all instances.
11690 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11692 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11694 (vect_model_promotion_demotion_cost): Likewise.
11695 (vectorizable_bswap): Guard cost modeling with !slp_node
11696 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11698 (vectorizable_call): Likewise.
11699 (vectorizable_conversion): Likewise.
11700 (vectorizable_assignment): Likewise.
11701 (vectorizable_shift): Likewise.
11702 (vectorizable_operation): Likewise.
11703 (vectorizable_store): Likewise.
11704 (vectorizable_load): Likewise.
11705 (vectorizable_condition): Likewise.
11706 (vectorizable_comparison): Likewise.
11708 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11711 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11712 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11713 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11715 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11717 PR middle-end/83665
11718 * params.def (inline-min-speedup): Increase from 8 to 15.
11719 (max-inline-insns-auto): Decrease from 40 to 30.
11720 * ipa-split.c (consider_split): Add some buffer for function to
11721 be considered inlining candidate.
11722 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11725 2018-02-12 Richard Biener <rguenther@suse.de>
11727 PR tree-optimization/84037
11728 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11729 matched stmts if we cannot swap the non-matched ones.
11731 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11733 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11734 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11735 _mm_maskz_scalef_round_ss): New intrinsics.
11736 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11737 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11738 __builtin_ia32_scalefss_round): Remove.
11739 (__builtin_ia32_scalefsd_mask_round,
11740 __builtin_ia32_scalefss_mask_round): New intrinsics.
11741 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11742 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11743 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11744 "<round_constraint>")): Changed to ...
11745 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11746 "<round_scalar_constraint>")): ... this.
11747 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11748 %0, %1, %2<round_op3>}"): Changed to ...
11749 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11750 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11751 %2<round_scalar_mask_op3>}"): ... this.
11752 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11754 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11756 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11757 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11758 (_mm_maskz_sqrt_round_ss): New intrinsics.
11759 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11760 (__builtin_ia32_sqrtsd_mask_round)
11761 (__builtin_ia32_sqrtss_mask_round): New builtins.
11762 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11763 (__builtin_ia32_sqrtss_round): Remove.
11764 (__builtin_ia32_sqrtsd_mask_round)
11765 (__builtin_ia32_sqrtss_mask_round): New builtins.
11766 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11767 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11768 ((match_operand:VF_128 1 "vector_operand"
11769 "xBm,<round_constraint>")): Changed to ...
11770 ((match_operand:VF_128 1 "vector_operand"
11771 "xBm,<round_scalar_constraint>")): ... this.
11772 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11773 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11774 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11775 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11776 %<iptr>1<round_scalar_mask_op3>}): ... this.
11777 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11778 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11780 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11783 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11784 Cast vec_cmpeq result to correct type.
11785 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11786 Cast vec_cmpgt result to correct type.
11788 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11790 * final.c (final_scan_insn_1): Renamed from...
11791 (final_scan_insn): ... this. New wrapper, to recover
11792 seen from the outermost call in recursive ones.
11793 * config/sparc/sparc.c (output_return): Drop seen from call.
11794 (output_sibcall): Likewise.
11795 * config/visium/visium.c (output_branch): Likewise.
11797 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11799 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11802 2018-02-10 Alan Modra <amodra@gmail.com>
11805 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11806 Specify LR as an input.
11808 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11811 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11812 remove_member_access_dummy_vars): New functions.
11813 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11814 lower_omp_1, execute_lower_omp): Use them.
11816 PR rtl-optimization/84308
11817 * shrink-wrap.c (spread_components): Release todo vector.
11819 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11821 PR rtl-optimization/57193
11822 * ira-color.c (struct allocno_color_data): Add member
11823 conflict_allocno_hard_prefs.
11824 (update_conflict_allocno_hard_prefs): New.
11825 (bucket_allocno_compare_func): Add a preference based on
11826 conflict_allocno_hard_prefs.
11827 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11828 (color_allocnos): Remove a dead code. Initiate
11829 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11831 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11834 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11835 constraint from =wa to wa. Avoid a subreg on the output operand,
11836 instead use a pseudo and subreg it in a move.
11837 (p9_xxbrd_<mode>): Changed to ...
11838 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11839 (p9_xxbrd_v2df): New expander.
11840 (p9_xxbrw_<mode>): Changed to ...
11841 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11842 (p9_xxbrw_v4sf): New expander.
11844 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11846 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11848 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11851 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11852 multiply in 32-bit mode.
11853 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11854 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11857 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11859 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11860 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11861 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11862 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11864 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11867 * dwarf2out.c (is_trivial_indirect_ref): New function.
11868 (dwarf2out_late_global_decl): Do not generate a location
11869 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11870 and that are not defined in the current unit.
11872 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11874 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11875 instead of a libcall for UNORDERED.
11877 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11880 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11881 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11883 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11886 * config/s390/s390.c (s390_set_current_function): Invoke
11887 s390_indirect_branch_settings also if fndecl didn't change.
11889 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11891 * config/rs6000/rs6000.md (blockage): Set length to zero.
11893 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11895 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11897 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11900 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11901 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11905 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11906 PARALLEL incoming that failed vt_get_decl_and_offset check.
11908 PR middle-end/84237
11909 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11910 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11912 (get_variable_section): For decls in named .bss* sections pass true as
11913 second argument to bss_initializer_p.
11915 2018-02-09 Marek Polacek <polacek@redhat.com>
11916 Jakub Jelinek <jakub@redhat.com>
11919 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11920 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11921 Sync some changes from cxx_fold_indirect_ref.
11923 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11925 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11927 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11928 (BLOCK_INLINE_ENTRY_LABEL): New.
11929 (dwarf2out_var_location): Disregard inline entry markers.
11930 (inline_entry_data): New struct.
11931 (inline_entry_data_hasher): New hashtable type.
11932 (inline_entry_data_hasher::hash): New.
11933 (inline_entry_data_hasher::equal): New.
11934 (inline_entry_data_table): New variable.
11935 (add_high_low_attributes): Add DW_AT_entry_pc and
11936 DW_AT_GNU_entry_view attributes if a pending entry is found
11937 in inline_entry_data_table. Add old entry_pc attribute only
11938 if debug nonbinding markers are disabled.
11939 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11940 markers are enabled.
11941 (block_within_block_p, dwarf2out_inline_entry): New.
11942 (dwarf2out_finish): Check that no entries remained in
11943 inline_entry_data_table.
11944 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11945 (final_scan_insn, notice_source_line): Likewise.
11946 (rest_of_clean_state): Skip inline entry markers.
11947 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11949 * gimple.c (gimple_build_debug_inline_entry): New.
11950 * gimple.h (enum gimple_debug_subcode): Add
11951 GIMPLE_DEBUG_INLINE_ENTRY.
11952 (gimple_build_debug_inline_entry): Declare.
11953 (gimple_debug_inline_entry_p): New.
11954 (gimple_debug_nonbind_marker_p): Adjust.
11955 * insn-notes.def (INLINE_ENTRY): New.
11956 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11957 inline entry marker notes.
11958 (print_insn): Likewise.
11959 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11960 (INSN_DEBUG_MARKER_KIND): Likewise.
11961 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11962 * tree-inline.c (expand_call_inline): Build and insert
11963 debug_inline_entry stmt.
11964 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11965 inline entry blocks early, if nonbind markers are enabled.
11966 (dump_scope_block): Dump fragment info.
11967 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11968 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11969 (gimple_build_debug_inline_entry): New.
11970 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11971 Enable/disable inline entry points too.
11972 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11973 (DEBUG_INSN): Describe inline entry markers.
11975 * common.opt (gvariable-location-views): New.
11976 (gvariable-location-views=incompat5): New.
11977 * config.in: Rebuilt.
11978 * configure: Rebuilt.
11979 * configure.ac: Test assembler for view support.
11980 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11981 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11982 * dwarf2out.c (var_loc_view): New typedef.
11983 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11984 (dwarf2out_locviews_in_attribute): New.
11985 (dwarf2out_locviews_in_loclist): New.
11986 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11987 (enum dw_line_info_opcode): Add LI_adv_address.
11988 (struct dw_line_info_table): Add view.
11989 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11990 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11991 (zero_view_p): New variable.
11992 (ZERO_VIEW_P): New macro.
11993 (output_asm_line_debug_info): New.
11994 (struct var_loc_node): Add view.
11995 (add_AT_view_list, AT_loc_list): New.
11996 (add_var_loc_to_decl): Add view param. Test it against last.
11997 (new_loc_list): Add view params. Record them.
11998 (AT_loc_list_ptr): Handle loc and view lists.
11999 (view_list_to_loc_list_val_node): New.
12000 (print_dw_val): Handle dw_val_class_view_list.
12001 (size_of_die): Likewise.
12002 (value_format): Likewise.
12003 (loc_list_has_views): New.
12004 (gen_llsym): Set vl_symbol too.
12005 (maybe_gen_llsym, skip_loc_list_entry): New.
12006 (dwarf2out_maybe_output_loclist_view_pair): New.
12007 (output_loc_list): Output view list or entries too.
12008 (output_view_list_offset): New.
12009 (output_die): Handle dw_val_class_view_list.
12010 (output_dwarf_version): New.
12011 (output_compilation_unit_header): Use it.
12012 (output_skeleton_debug_sections): Likewise.
12013 (output_rnglists, output_line_info): Likewise.
12014 (output_pubnames, output_aranges): Update version comments.
12015 (output_one_line_info_table): Output view numbers in asm comments.
12016 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12017 Call maybe_gen_llsym.
12018 (loc_list_from_tree_1): Adjust.
12019 (add_AT_location_description): Create view list attribute if
12020 needed, check it's absent otherwise.
12021 (convert_cfa_to_fb_loc_list): Adjust.
12022 (maybe_emit_file): Call output_asm_line_debug_info for test.
12023 (dwarf2out_var_location): Reset views as needed. Precompute
12024 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12025 attribute. Set view.
12026 (new_line_info_table): Reset next view.
12027 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12028 (dwarf2out_source_line): Likewise. Output view resets and labels to
12029 the assembler, or select appropriate line info opcodes.
12030 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12031 (optimize_string_length): Catch it. Adjust.
12032 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12033 dw_val_class_view_list, and remove it if no longer needed.
12034 (hash_loc_list): Hash view numbers.
12035 (loc_list_hasher::equal): Compare them.
12036 (optimize_location_lists): Check whether a view list symbol is
12037 needed, and whether the locview attribute is present, and
12038 whether they match. Remove the locview attribute if no longer
12040 (index_location_lists): Call skip_loc_list_entry for test.
12041 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12042 Use output_dwarf_version.
12043 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12044 (struct dw_val_node): Add val_view_list.
12045 * final.c (SEEN_NEXT_VIEW): New.
12046 (set_next_view_needed): New.
12047 (clear_next_view_needed): New.
12048 (maybe_output_next_view): New.
12049 (final_start_function): Rename to...
12050 (final_start_function_1): ... this. Take pointer to FIRST,
12051 add SEEN parameter. Emit param bindings in the initial view.
12052 (final_start_function): Reintroduce SEEN-less interface.
12053 (final): Rename to...
12054 (final_1): ... this. Take SEEN parameter. Output final pending
12055 next view at the end.
12056 (final): Reintroduce seen-less interface.
12057 (final_scan_insn): Output pending next view before switching
12058 sections or ending a block. Mark the next view as needed when
12059 outputting variable locations. Notify debug backend of section
12060 changes, and of location view changes.
12061 (rest_of_handle_final): Adjust.
12062 * toplev.c (process_options): Autodetect value for debug variable
12063 location views option. Warn on incompat5 without -gdwarf-5.
12064 * doc/invoke.texi (gvariable-location-views): New.
12065 (gvariable-location-views=incompat5): New.
12066 (gno-variable-location-views): New.
12068 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12070 PR tree-optimization/84136
12071 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12072 that the result of find_edge is non-NULL.
12074 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12077 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12078 storing integer register in SImode. Fix cost of 256 and 512
12079 byte aligned SSE register store.
12081 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12083 * config/i386/i386.c (ix86_multiplication_cost): Fix
12084 multiplication cost for TARGET_AVX512DQ.
12086 2018-02-08 Marek Polacek <polacek@redhat.com>
12088 PR tree-optimization/84238
12089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12092 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12094 PR tree-optimization/84265
12095 * tree-vect-stmts.c (vectorizable_store): Don't treat
12096 VMAT_CONTIGUOUS accesses as grouped.
12097 (vectorizable_load): Likewise.
12099 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12101 PR tree-optimization/81635
12102 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12103 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12104 (test_round_for_mask): New functions.
12105 (wide_int_cc_tests): Call test_round_for_mask.
12106 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12107 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12108 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12109 range returned by get_range_info.
12111 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12114 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12115 * symtab.c: Include builtins.h
12116 (symtab_node::output_to_lto_symbol_table_p): Move here
12117 from lto-streamer-out.c:output_symbol_p.
12118 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12119 (output_symbol_p): Move all logic to symtab.c
12120 (produce_symtab): Update.
12122 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12124 * config/s390/s390-opts.h (enum indirect_branch): Define.
12125 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12126 (s390_indirect_branch_via_thunk)
12127 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12128 (enum s390_indirect_branch_type): Define.
12129 * config/s390/s390.c (struct s390_frame_layout, struct
12130 machine_function): Remove.
12131 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12132 (indirect_branch_table_label_no, indirect_branch_table_name):
12134 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12135 (enum s390_indirect_branch_option): Define.
12136 (s390_return_addr_from_memory): New function.
12137 (s390_handle_string_attribute): New function.
12138 (s390_attribute_table): Add new attribute handler.
12139 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12140 (s390_indirect_branch_via_thunk): New function.
12141 (s390_indirect_branch_via_inline_thunk): New function.
12142 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12143 sibling call optimization for non z10 compiles.
12144 (s390_emit_call): Force indirect branch target to be a single
12145 register. Add r1 clobber for non-z10 compiles.
12146 (s390_emit_epilogue): Emit return jump via return_use expander.
12147 (s390_reorg): Handle JUMP_INSNs as execute targets.
12148 (s390_option_override_internal): Perform validity checks for the
12149 new command line options.
12150 (s390_indirect_branch_attrvalue): New function.
12151 (s390_indirect_branch_settings): New function.
12152 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12153 (s390_output_indirect_thunk_function): New function.
12154 (s390_code_end): Implement target hook.
12155 (s390_case_values_threshold): Implement target hook.
12156 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12158 * config/s390/s390.h (struct s390_frame_layout)
12159 (struct machine_function): Move here from s390.c.
12160 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12161 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12162 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12163 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12164 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12165 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12166 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12167 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12168 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12169 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12170 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12171 (mnemonic attribute): Add values which aren't recognized
12173 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12174 pattern for branch conversion. Fix mnemonic attribute.
12175 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12176 indirect branch via thunk if requested.
12177 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12178 ("*indirect_jump"): Disable for branch conversion using out of
12180 ("indirect_jump_via_thunk<mode>_z10")
12181 ("indirect_jump_via_thunk<mode>")
12182 ("indirect_jump_via_inlinethunk<mode>_z10")
12183 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12184 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12185 ("casesi_jump_via_inlinethunk<mode>_z10")
12186 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12187 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12188 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12189 ("*indirect2_jump"): Disable for branch conversion.
12190 ("casesi_jump"): Turn into expander and expand patterns for branch
12192 ("return_use"): New expander.
12193 ("*return"): Emit return via thunk and rename it to ...
12194 ("*return<mode>"): ... this one.
12195 * config/s390/s390.opt: Add new options and and enum for the
12198 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12200 * lra-constraints.c (match_reload): Unconditionally use
12201 gen_lowpart_SUBREG, rather than selecting between that
12202 and equivalent gen_rtx_SUBREG code.
12204 2018-02-08 Richard Biener <rguenther@suse.de>
12206 PR tree-optimization/84233
12207 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12208 changed flag instead of boguously re-using phi_inserted.
12210 2018-02-08 Martin Jambor <mjambor@suse.cz>
12212 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12213 static local variables.
12215 2018-02-08 Richard Biener <rguenther@suse.de>
12217 PR tree-optimization/84278
12218 * tree-vect-stmts.c (vectorizable_store): When looking for
12219 smaller vector types to perform grouped strided loads/stores
12220 make sure the mode is supported by the target.
12221 (vectorizable_load): Likewise.
12223 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12225 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12226 Increase LDP/STP opportunities by adding adjacent callee-saves.
12228 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12230 PR rtl-optimization/84068
12231 PR rtl-optimization/83459
12232 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12234 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12236 PR tree-optimization/84224
12237 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12238 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12239 non-zero arguments.
12241 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12244 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12245 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12246 count, remove one USE.
12248 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12250 * doc/install.texi (Configuration): Document the
12251 --with-long-double-format={ibm,ieee} PowerPC configuration
12255 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12256 Convert from define_expand to be define_insn_and_split. Rework
12257 float/double/_Float128 conversions to QI/HI/SImode to work with
12258 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12259 conversions to QI/HImode types did a store and then a load to
12260 truncate the value. For conversions to VSX registers, don't split
12261 the insn, instead emit the code directly. Use the code iterator
12262 any_fix to combine signed and unsigned conversions.
12263 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12264 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12265 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12266 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12267 (fix_<mode>di2_hw): Likewise.
12268 (fixuns_<mode>di2_hw): Likewise.
12269 (fix_<mode>si2_hw): Likewise.
12270 (fixuns_<mode>si2_hw): Likewise.
12271 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12272 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12273 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12274 fix<uns>_trunc<SFDF:mode>si2_p8.
12275 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12276 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12277 (fix<uns>_<mode>_mem): Likewise.
12278 (fctiw<u>z_<mode>_mem): Likewise.
12279 (fix<uns>_<mode>_mem): Likewise.
12280 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12281 the register allocator from doing a direct move to the GPRs to do
12282 a store, and instead use the ISA 3.0 store byte/half-word from
12283 vector register instruction. For IEEE 128-bit floating point,
12284 also optimize stores of 32-bit ints.
12285 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12287 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12289 * genextract.c (push_pathstr_operand): New function to support
12291 (walk_rtx): Call push_pathstr_operand.
12292 (print_path): Support [a-zA-Z].
12294 2018-02-07 Richard Biener <rguenther@suse.de>
12296 PR tree-optimization/84037
12297 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12298 (cse_and_gimplify_to_preheader): Declare.
12299 (vect_get_place_in_interleaving_chain): Likewise.
12300 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12302 (_loop_vec_info::~_loop_vec_info): Delete it.
12303 (cse_and_gimplify_to_preheader): New function.
12304 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12305 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12306 (vectorizable_load): Likewise. For grouped stores always base
12307 the IV on the first element.
12308 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12309 condition before gimplifying.
12311 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12313 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12314 *DIV_EXPR and *MOD_EXPR.
12316 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12319 * config/i386/i386.c (ix86_option_override_internal): Mask out
12320 the CF_SET bit when checking -fcf-protection.
12322 2018-02-07 Tom de Vries <tom@codesourcery.com>
12325 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12328 2018-02-07 Richard Biener <rguenther@suse.de>
12330 PR tree-optimization/84204
12331 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12334 PR tree-optimization/84205
12335 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12336 special-case isl_ast_op_zdiv_r.
12338 PR tree-optimization/84223
12339 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12340 Only add conditions from within the region.
12341 (gather_bbs::after_dom_children): Adjust.
12343 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12346 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12347 * config/avr/avr.md: Only post-reload split REG-REG moves if
12348 either register is GENERAL_REG_P.
12350 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12352 PR tree-optimization/84235
12353 * tree-ssa-scopedtables.c
12354 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12355 if the subtraction is performed in floating point type where NaNs are
12356 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12357 build 1. Formatting fix.
12359 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12362 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12363 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12364 and skip it regardless of bb boundaries. Use CALL_P macro,
12365 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12366 unnecessarily, formatting fix.
12368 2018-02-06 Michael Collison <michael.collison@arm.com>
12370 * config/arm/thumb2.md:
12371 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12372 (*thumb_mov_notscc): Ditto.
12374 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12377 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12378 unsigned_fix, not "s".
12380 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12382 * configure.ac (gcc_fn_eh_frame_ro): New function.
12383 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12384 correct .eh_frame permissions.
12385 * configure: Regenerate.
12387 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12389 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12390 irrelevant options.
12392 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12394 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12395 Display warning message for -mno-speculate-indirect-jumps.
12397 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12399 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12401 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12403 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12405 PR tree-optimization/84225
12406 * tree-eh.c (find_trapping_overflow): Only call
12407 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12409 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12412 * config/i386/i386.c: Reimplement the check of possible options
12413 -mibt/-mshstk conbination. Change error messages.
12414 * doc/invoke.texi: Fix a typo: remove extra '='.
12416 2018-02-06 Marek Polacek <polacek@redhat.com>
12418 PR tree-optimization/84228
12419 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12421 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12424 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12425 emitted arch directives.
12426 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12427 __ARM_FEATURE_COPROC before changing architectures.
12429 2018-02-06 Richard Biener <rguenther@suse.de>
12431 * config/i386/i386.c (print_reg): Fix typo.
12432 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12434 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12436 * configure: Regenerate.
12438 2018-02-05 Martin Sebor <msebor@redhat.com>
12440 PR tree-optimization/83369
12441 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12444 2018-02-05 Martin Liska <mliska@suse.cz>
12446 * doc/invoke.texi: Cherry-pick upstream r323995.
12448 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12450 * ira.c (ira_init_register_move_cost): Adjust comment.
12452 2018-02-05 Martin Liska <mliska@suse.cz>
12454 PR gcov-profile/84137
12455 * doc/gcov.texi: Fix typo in documentation.
12457 2018-02-05 Martin Liska <mliska@suse.cz>
12459 PR gcov-profile/83879
12460 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12461 using dlopen functionality.
12463 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12465 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12466 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12467 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12468 _mm_maskz_range_round_ss): New intrinsics.
12469 (__builtin_ia32_rangesd128_round)
12470 (__builtin_ia32_rangess128_round): Remove.
12471 (__builtin_ia32_rangesd128_mask_round,
12472 __builtin_ia32_rangess128_mask_round): New builtins.
12473 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12474 __builtin_ia32_rangess128_round): Remove.
12475 (__builtin_ia32_rangesd128_mask_round,
12476 __builtin_ia32_rangess128_mask_round): New builtins.
12477 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12478 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12479 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12480 "<round_saeonly_constraint>")): Changed to ...
12481 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12482 "<round_saeonly_scalar_constraint>")): ... this.
12483 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12484 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12485 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12486 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12487 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12489 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12491 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12493 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12494 Remove all values except native, 8540 and 8548.
12496 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12498 * config/i386/i386.c (ix86_output_function_return): Pass
12499 INVALID_REGNUM, instead of -1, as invalid register number to
12500 indirect_thunk_name and output_indirect_thunk.
12502 2018-02-02 Julia Koval <julia.koval@intel.com>
12504 * config.gcc: Add -march=icelake.
12505 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12506 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12507 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12508 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12509 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12510 (processor_target_table): Add icelake.
12511 (ix86_option_override_internal): Handle new PTAs.
12512 (get_builtin_code_for_version): Handle icelake.
12513 (M_INTEL_COREI7_ICELAKE): New.
12514 (fold_builtin_cpu): Handle icelake.
12515 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12516 * doc/invoke.texi: Add -march=icelake.
12518 2018-02-02 Julia Koval <julia.koval@intel.com>
12520 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12521 to wide_int_bitmask.
12522 * wide-int-bitmask.h: New.
12524 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12527 * config/i386/i386.md: Replace Pmode with word_mode in
12528 builtin_setjmp_setup and builtin_longjmp to support x32.
12530 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12534 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12536 (rs6000_supported_cpu_names): New static variable.
12537 (linux_cpu_translation_table): Likewise.
12538 (elf_platform) <cpu>: Define new static variable and use it.
12539 Translate kernel AT_PLATFORM name to canonical name if needed.
12540 Error if platform name is unknown.
12542 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12545 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12547 2018-02-01 Jeff Law <law@redhat.com>
12550 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12551 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12552 the scratch if RELEASE_VIA_POP is false.
12553 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12554 If we have to save a temporary register, decrement SIZE appropriately.
12555 Pass new arguments to release_scratch_register_on_entry.
12556 (ix86_adjust_stack_and_probe): Likewise.
12557 (ix86_emit_probe_stack_range): Pass new arguments to
12558 release_scratch_register_on_entry.
12560 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12562 PR rtl-optimization/84157
12563 * combine.c (change_zero_ext): Use REG_P predicate in
12564 front of HARD_REGISTER_P predicate.
12566 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12568 * config/avr/avr.c (avr_option_override): Move disabling of
12569 -fdelete-null-pointer-checks to...
12570 * common/config/avr/avr-common.c (avr_option_optimization_table):
12573 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12575 PR tree-optimization/81635
12576 * tree-data-ref.c (split_constant_offset_1): For types that
12577 wrap on overflow, try to use range info to prove that wrapping
12580 2018-02-01 Renlin Li <renlin.li@arm.com>
12583 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12584 TAILCALL_ADDR_REGS.
12585 (aarch64_register_move_cost): Likewise.
12586 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12587 TAILCALL_ADDR_REGS.
12588 (REG_CLASS_NAMES): Likewise.
12589 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12590 TAILCALL_ADDR_REGS. Remove IP registers.
12591 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12593 2018-02-01 Richard Biener <rguenther@suse.de>
12595 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12596 for specifying RPO order and allow NULL for that.
12597 * domwalk.c (dom_walker::dom_walker): Likewise.
12598 (dom_walker::walk): Handle NULL RPO order.
12599 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12601 (rewrite_update_dom_walker): Likewise.
12602 (mark_def_dom_walker): Likewise.
12604 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12606 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12607 (aarch64_maybe_expand_sve_subreg_move): Declare.
12608 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12609 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12611 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12612 that are semantically a reverse operation.
12613 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12614 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12615 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12617 (aarch64_can_change_mode_class): For big-endian, forbid changes
12618 between two SVE modes if they have different element sizes.
12620 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12622 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12623 the TImode handling for big-endian targets.
12625 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12627 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12628 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12630 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12631 Remove BSWAP handing for big-endian targets and use the form of
12632 LD1RQ appropariate for the mode.
12634 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12636 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12637 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12638 duplicated element.
12640 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12643 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12644 check for operands that need to go through aarch64_sve_reload_be.
12646 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12648 PR tree-optimization/81661
12649 PR tree-optimization/84117
12650 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12651 * tree-eh.c: Include gimplify.h.
12652 (find_trapping_overflow, replace_trapping_overflow,
12653 rewrite_to_non_trapping_overflow): New functions.
12654 * tree-vect-loop.c: Include tree-eh.h.
12655 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12656 * tree-data-ref.c: Include tree-eh.h.
12657 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12659 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12661 PR rtl-optimization/84123
12662 * combine.c (change_zero_ext): Check if hard register satisfies
12663 can_change_dest_mode before calling gen_lowpart_SUBREG.
12665 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12668 * ira.c (ira_init_register_move_cost): Remove assert.
12670 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12672 PR rtl-optimization/84071
12673 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12674 * doc/tm.texi: Regenerate.
12676 2018-01-31 Richard Biener <rguenther@suse.de>
12678 PR tree-optimization/84132
12679 * tree-data-ref.c (analyze_miv_subscript): Properly
12680 check whether evolution_function_is_affine_multivariate_p
12681 before calling gcd_of_steps_may_divide_p.
12683 2018-01-31 Julia Koval <julia.koval@intel.com>
12686 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12687 * config/i386/i386.md (rdpid_rex64) New.
12688 (rdpid): Make 32bit only.
12690 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12693 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12694 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12696 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12699 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12701 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12703 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12705 PR rtl-optimization/84071
12706 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12707 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12709 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12711 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12712 (arc_attribute_table): Add 'aux' attribute.
12713 (arc_in_small_data_p): Consider aux like variables.
12714 (arc_is_aux_reg_p): New function.
12715 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12716 (arc_get_aux_arg): New function.
12717 (prepare_move_operands): Handle aux-register access.
12718 (arc_handle_aux_attribute): New function.
12719 * doc/extend.texi (ARC Variable attributes): Add subsection.
12721 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12723 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12724 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12725 (arc_attribute_table): Add 'uncached' attribute.
12726 (arc_print_operand): Print '.di' flag for uncached memory
12728 (arc_in_small_data_p): Do not consider for small data the uncached
12730 (arc_is_uncached_mem_p): New function.
12731 * config/arc/predicates.md (compact_store_memory_operand): Check
12732 for uncached memory accesses.
12733 (nonvol_nonimm_operand): Likewise.
12734 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12736 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12739 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12740 falign-loops=): Add Optimization flag.
12742 2018-01-30 Jeff Law <law@redhat.com>
12745 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12746 INT_REGISTERS_SAVED. Check it prior to calling
12747 get_scratch_register_on_entry.
12748 (ix86_adjust_stack_and_probe): Similarly.
12749 (ix86_emit_probe_stack_range): Similarly.
12750 (ix86_expand_prologue): Corresponding changes.
12752 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12755 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12756 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12758 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12761 * lra-constraints.c (curr_insn_transform): Process AND in the
12764 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12766 PR rtl-optimization/83986
12767 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12768 dependence against last_pending_memory_flush in addition to
12769 pending_jump_insns.
12771 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12773 PR tree-optimization/81611
12774 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12777 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12780 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12783 2018-01-30 Richard Biener <rguenther@suse.de>
12784 Jakub Jelinek <jakub@redhat.com>
12786 PR tree-optimization/84111
12787 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12788 inner loops added during recursion, as they don't have up-to-date
12791 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12794 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12795 (can_inline_edge_by_limits_p): ... here.
12796 (can_early_inline_edge_p, check_callers,
12797 update_caller_keys, update_callee_keys, recursive_inlining,
12798 add_new_edges_to_heap, speculation_useful_p,
12799 inline_small_functions,
12800 inline_small_functions, flatten_function,
12801 inline_to_all_callers_1): Update.
12803 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12805 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12808 2018-01-30 Richard Biener <rguenther@suse.de>
12810 PR tree-optimization/83008
12811 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12812 invariant and constant vector uses in stmts when they need
12813 more than one stmt.
12815 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12818 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12819 * configure: Regenerate.
12821 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12823 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12825 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12826 Use gen_rtx_REG rather than gen_lowpart.
12828 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12830 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12831 rather than 0 when creating partial subregs.
12833 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12835 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12838 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12841 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12842 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12843 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12844 flags. This restores the settings used before the 2017-07-24.
12845 Turning off pre increment/decrement/modify allows IVOPTS to
12846 optimize DF/SF loops where the index is an int.
12848 2018-01-29 Richard Biener <rguenther@suse.de>
12849 Kelvin Nilsen <kelvin@gcc.gnu.org>
12852 * tree-vect-stmts.c (vectorizable_call): Don't call
12853 targetm.vectorize_builtin_md_vectorized_function if callee is
12856 2018-01-22 Carl Love <cel@us.ibm.com>
12858 * doc/extend.tex: Fix typo in second arg in
12859 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12861 2018-01-29 Richard Biener <rguenther@suse.de>
12863 PR tree-optimization/84086
12864 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12865 (flush_ssaname_freelist): When SSA names were released reset
12866 the SCEV hash table.
12868 2018-01-29 Richard Biener <rguenther@suse.de>
12870 PR tree-optimization/84057
12871 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12872 removed paths when removing edges.
12874 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12876 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12877 -mfunction-return=@var{choice}.
12879 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12881 PR diagnostic/84034
12882 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12883 Handle CR like TAB.
12884 (layout::print_source_line): Likewise.
12885 (test_get_line_width_without_trailing_whitespace): Add test cases.
12887 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12889 PR middle-end/84040
12890 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12893 2018-01-26 Jim Wilson <jimw@sifive.com>
12895 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12897 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12900 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12902 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12903 and CMP + SUB-immediate -> SUBS.
12905 2018-01-26 Martin Sebor <msebor@redhat.com>
12907 PR tree-optimization/83896
12908 * tree-ssa-strlen.c (get_string_len): Rename...
12909 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12910 Avoid assuming length is constant.
12911 (handle_char_store): Use HOST_WIDE_INT for string length.
12913 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12916 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12917 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12919 2018-01-26 Richard Biener <rguenther@suse.de>
12921 PR rtl-optimization/84003
12922 * dse.c (record_store): Only record redundant stores when
12923 the earlier store aliases at least all accesses the later one does.
12925 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12927 PR rtl-optimization/83985
12928 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12929 REG_CFA_RESTORE insns.
12930 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12931 REG_CFA_RESTORE insns here.
12934 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12935 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12937 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12939 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12940 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12941 (arc_init): Likewise.
12942 (arc_override_options): Likewise.
12943 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12945 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12947 * config/arc/arc.h (TARGET_DBNZ): Define.
12948 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12949 properly set the tune attribute.
12950 (dbnz): Use TARGET_DBNZ guard.
12951 * config/arc/arc.opt (mtune): Add core3 option.
12953 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12955 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12956 recognize new pic like addresses.
12957 (arc_delegitimize_address): Clean up.
12959 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12961 * config/arc/arc-arches.def: Option mrf16 valid for all
12963 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12964 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12965 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12966 * config/arc/arc-tables.opt: Regenerate.
12967 * config/arc/arc.c (arc_conditional_register_usage): Handle
12968 reduced register file case.
12969 (arc_file_start): Set must have build attributes.
12970 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12971 mrf16 option value.
12972 * config/arc/arc.opt (mrf16): Add new option.
12973 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12974 * config/arc/genmultilib.awk: Handle new mrf16 option.
12975 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12976 * config/arc/t-multilib: Regenerate.
12977 * doc/invoke.texi (ARC Options): Document mrf16 option.
12979 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12981 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12982 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12983 (arc_attribute_table): Add 'secure_call' attribute.
12984 (arc_print_operand): Print secure call operand.
12985 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12987 (arc_is_secure_call_p): New function. * config/arc/arc.md
12988 (call_i): Add support for sjli instruction.
12989 (call_value_i): Likewise.
12990 * config/arc/constraints.md (Csc): New constraint.
12992 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12993 John Eric Martin <John.Martin@emmicro-us.com>
12995 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12996 * config/arc/arc.c (_arc_jli_section): New struct.
12997 (arc_jli_section): New type.
12998 (rc_jli_sections): New static variable.
12999 (arc_handle_jli_attribute): New function.
13000 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13001 (arc_file_end): New function.
13002 (TARGET_ASM_FILE_END): Define.
13003 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13004 (arc_add_jli_section): New function.
13005 (jli_call_scan): Likewise.
13006 (arc_reorg): Call jli_call_scan.
13007 (arc_output_addsi): Remove 'S' from printing asm operand.
13008 (arc_is_jli_call_p): New function.
13009 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13011 (movhi_insn): Likewise.
13012 (movsi_insn): Likewise.
13013 (movsi_set_cc_insn): Likewise.
13014 (loadqi_update): Likewise.
13015 (load_zeroextendqisi_update): Likewise.
13016 (load_signextendqisi_update): Likewise.
13017 (loadhi_update): Likewise.
13018 (load_zeroextendhisi_update): Likewise.
13019 (load_signextendhisi_update): Likewise.
13020 (loadsi_update): Likewise.
13021 (loadsf_update): Likewise.
13022 (movsicc_insn): Likewise.
13023 (bset_insn): Likewise.
13024 (bxor_insn): Likewise.
13025 (bclr_insn): Likewise.
13026 (bmsk_insn): Likewise.
13027 (bicsi3_insn): Likewise.
13028 (cmpsi_cc_c_insn): Likewise.
13029 (movsi_ne): Likewise.
13030 (movsi_cond_exec): Likewise.
13031 (clrsbsi2): Likewise.
13032 (norm_f): Likewise.
13042 (call_i): Remove 'S' asm letter, add jli instruction.
13043 (call_value_i): Likewise.
13044 * config/arc/arc.op (mjli-always): New option.
13045 * config/arc/constraints.md (Cji): New constraint.
13046 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13048 (subsf3_fpx): Likewise.
13049 (mulsf3_fpx): Likewise.
13050 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13052 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13053 function attrbutes.
13054 * doc/invoke.texi (ARC): Document mjli-always option.
13056 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13058 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13059 avoid addition with 0 and use incw and decw where possible.
13061 2018-01-26 Richard Biener <rguenther@suse.de>
13063 PR tree-optimization/81082
13064 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13065 association if it requires casting to unsigned.
13066 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13067 from fold_plusminus_mult_expr to catch important cases late when
13068 range info is available.
13070 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13072 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13073 * configure.ac (hidden_linkonce): New test.
13074 * configure: Regenerate.
13075 * config.in: Regenerate.
13077 2018-01-26 Julia Koval <julia.koval@intel.com>
13079 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13080 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13081 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13082 _mm_mask_bitshuffle_epi64_mask): Fix type.
13083 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13084 USI_FTYPE_V4DI_V4DI_USI): Remove.
13085 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13086 __builtin_ia32_vpshufbitqmb256_mask,
13087 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13088 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13089 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13091 2018-01-26 Alan Modra <amodra@gmail.com>
13094 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13095 UNSPEC_VBPERMQ. Sort other unspecs.
13097 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13099 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13101 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13103 PR middle-end/83055
13104 * predict.c (drop_profile): Do not push/pop cfun; update also
13106 (handle_missing_profiles): Fix logic looking for zero profiles.
13108 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13110 PR middle-end/83977
13111 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13112 on functions with #pragma omp declare simd or functions with simd
13114 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13115 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13116 Remove trailing \n from warning_at calls.
13118 2018-01-25 Tom de Vries <tom@codesourcery.com>
13121 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13122 for neutered workers.
13124 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13127 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13128 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13130 2018-01-24 Jeff Law <law@redhat.com>
13133 * i386.c (get_probe_interval): Move to earlier point.
13134 (ix86_compute_frame_layout): If -fstack-clash-protection and
13135 the frame is larger than the probe interval, then use pushes
13136 to save registers rather than reg->mem moves.
13137 (ix86_expand_prologue): Remove conditional for int_registers_saved
13140 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13143 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13144 min/max for never referenced object.
13146 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13148 PR middle-end/83977
13149 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13151 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13152 attributes from DECL_ATTRIBUTES (decl) without affecting
13153 DECL_ATTRIBUTES (current_function_decl).
13154 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13155 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13157 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13159 PR tree-optimization/83979
13160 * fold-const.c (fold_comparison): Use constant_boolean_node
13161 instead of boolean_{true,false}_node.
13163 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13165 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13168 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13170 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13171 Simplify the clause that sets the length attribute.
13172 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13173 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13174 clause that sets the length attribute.
13175 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13177 2018-01-24 Tom de Vries <tom@codesourcery.com>
13180 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13181 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13182 Add strict parameter.
13183 (prevent_branch_around_nothing): Insert dummy insn between branch to
13184 label and label with no ptx insn inbetween.
13185 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13187 2018-01-24 Tom de Vries <tom@codesourcery.com>
13190 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13191 for neutered threads in warp.
13192 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13194 2018-01-24 Richard Biener <rguenther@suse.de>
13196 PR tree-optimization/83176
13197 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13200 2018-01-24 Richard Biener <rguenther@suse.de>
13202 PR tree-optimization/82819
13203 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13204 code generating pluses that are no-ops in the target precision.
13206 2018-01-24 Richard Biener <rguenther@suse.de>
13208 PR middle-end/84000
13209 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13211 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13213 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13214 to merge probabilities.
13215 * predict.c (probably_never_executed): Also mark as cold functions
13216 with global 0 profile and guessed local profile.
13217 * profile-count.c (profile_probability::combine_with_count): New
13219 * profile-count.h (profile_probability::operator*,
13220 profile_probability::operator*=, profile_probability::operator/,
13221 profile_probability::operator/=): Reduce precision to adjusted
13222 and set value to guessed on contradictory divisions.
13223 (profile_probability::combine_with_freq): Remove.
13224 (profile_probability::combine_wiht_count): Declare.
13225 (profile_count::force_nonzero):: Set to adjusted.
13226 (profile_count::probability_in):: Set quality to adjusted.
13227 * tree-ssa-tail-merge.c (replace_block_by): Use
13228 combine_with_count.
13230 2018-01-23 Andrew Waterman <andrew@sifive.com>
13231 Jim Wilson <jimw@sifive.com>
13233 * config/riscv/riscv.c (riscv_stack_boundary): New.
13234 (riscv_option_override): Set riscv_stack_boundary. Handle
13235 riscv_preferred_stack_boundary_arg.
13236 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13237 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13238 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13239 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13240 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13241 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13243 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13246 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13247 of struct ix86_frame.
13248 (ix86_expand_epilogue): Likewise. Add a local variable for
13249 the reg_save_offset field in struct ix86_frame.
13251 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13253 PR tree-optimization/82604
13254 * tree-loop-distribution.c (enum partition_kind): New enum item
13255 PKIND_PARTIAL_MEMSET.
13256 (partition_builtin_p): Support above new enum item.
13257 (generate_code_for_partition): Ditto.
13258 (compute_access_range): Differentiate cases that equality can be
13259 proven at all loops, the innermost loops or no loops.
13260 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13261 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13262 (finalize_partitions, distribute_loop): Don't fuse partition of
13263 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13264 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13265 parloop is enabled.
13267 2018-01-23 Martin Liska <mliska@suse.cz>
13269 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13270 order to ignore the predictor.
13271 (PRED_POLYMORPHIC_CALL): Likewise.
13272 (PRED_RECURSIVE_CALL): Likewise.
13274 2018-01-23 Martin Liska <mliska@suse.cz>
13276 * tree-profile.c (tree_profiling): Print function header to
13277 aware reader which function we are working on.
13278 * value-prof.c (gimple_find_values_to_profile): Do not print
13279 not interesting value histograms.
13281 2018-01-23 Martin Liska <mliska@suse.cz>
13283 * profile-count.h (enum profile_quality): Add
13284 profile_uninitialized as the first value. Do not number values
13285 as they are zero based.
13286 (profile_count::verify): Update sanity check.
13287 (profile_probability::verify): Likewise.
13289 2018-01-23 Nathan Sidwell <nathan@acm.org>
13291 * doc/invoke.texi (ffor-scope): Deprecate.
13293 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13295 PR tree-optimization/83510
13296 * domwalk.c (set_all_edges_as_executable): New function.
13297 (dom_walker::dom_walker): Convert bool param
13298 "skip_unreachable_blocks" to enum reachability. Move setup of
13299 edge flags to set_all_edges_as_executable and only do it when
13300 reachability is REACHABLE_BLOCKS.
13301 * domwalk.h (enum dom_walker::reachability): New enum.
13302 (dom_walker::dom_walker): Convert bool param
13303 "skip_unreachable_blocks" to enum reachability.
13304 (set_all_edges_as_executable): New decl.
13305 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13306 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13308 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13309 but converting true to REACHABLE_BLOCKS.
13310 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13312 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13313 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13314 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13316 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13317 if check_all_array_refs will be called.
13319 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13321 * tree.c (selftest::test_location_wrappers): Add more test
13324 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13326 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13327 (selftest::test_bit_in_range): Likewise.
13329 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13332 * doc/sourcebuild.texi (vect_float): Say that the selector
13333 only describes the situation when -funsafe-math-optimizations is on.
13334 (vect_float_strict): Document.
13336 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13338 PR tree-optimization/83965
13339 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13340 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13341 instead of checking only for a reduction.
13342 (vect_recog_widen_sum_pattern): Likewise.
13344 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13346 * predict.c (probably_never_executed): Only use precise profile info.
13347 (compute_function_frequency): Skip after inlining hack since we now
13348 have quality checking.
13350 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13352 * profile-count.h (profile_probability::very_unlikely,
13353 profile_probability::unlikely, profile_probability::even): Set
13354 precision to guessed.
13356 2018-01-23 Richard Biener <rguenther@suse.de>
13358 PR tree-optimization/83963
13359 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13360 Properly terminate dominator walk when crossing the exit edge not
13361 when visiting its source block.
13363 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13366 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13367 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13369 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13371 PR tree-optimization/83957
13372 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13373 semicolon after for body surrounded by braces.
13375 PR tree-optimization/83081
13376 * profile-count.h (profile_probability::split): New method.
13377 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13378 Use profile_probability::split.
13379 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13380 when splitting a single conditional jump into 2.
13382 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13384 PR tree-optimization/69452
13385 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13388 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13390 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13391 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13392 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13394 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13396 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13398 * config/rl78/rl78.md (movdi): New define_expand.
13399 * config/rl78/rl78.c (rl78_split_movdi): New function.
13401 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13404 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13406 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13407 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13408 128-bit to produce an UNSPEC move to get the double word with the
13409 signbit and then a shift directly to do signbit.
13410 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13411 implementation with a new version that just does either a direct
13412 move or a regular move. Move memory interface to separate insns.
13413 Move insns so they are next to the expander.
13414 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13415 with signbit move. Split big and little endian case.
13416 (signbit<mode>2_dm_mem_le): Likewise.
13417 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13418 (signbit<mode>2_dm2): Likewise.
13420 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13422 * config/rl78/rl78.md (anddi3): New define_expand.
13424 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13426 * config/rl78/rl78.md (umindi3): New define_expand.
13428 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13430 * config/rl78/rl78.md (smindi3): New define_expand.
13432 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13434 * config/rl78/rl78.md (smaxdi3): New define_expand.
13436 2018-01-22 Carl Love <cel@us.ibm.com>
13438 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13439 LVX_V1TI): Add macro expansion.
13440 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13441 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13442 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13443 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13444 Change check to determine if the instruction is a byte reversing
13445 entry. Fix typo in comment.
13446 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13447 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13448 Add def_builtin calls for new builtins.
13449 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13450 Add define_insn expansion.
13452 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13454 * config/rl78/rl78.md (umaxdi3): New define_expand.
13456 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13458 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13459 for non-QImode registers.
13461 2018-01-22 Richard Biener <rguenther@suse.de>
13463 PR tree-optimization/83963
13464 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13465 including the loop exit block.
13466 (scop_detection::merge_sese): Likewise.
13467 (scop_detection::add_scop): Do it here instead.
13469 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13471 * doc/sourcebuild.texi (arm_softfloat): Document.
13473 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13476 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13477 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13478 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13480 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13481 David Edelsohn <dje.gcc@gmail.com>
13484 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13485 Change "crset eq" to "crset 2".
13486 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13487 (*call_indirect_aix<mode>_nospec): Likewise.
13488 (*call_value_indirect_aix<mode>_nospec): Likewise.
13489 (*call_indirect_elfv2<mode>_nospec): Likewise.
13490 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13491 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13492 change assembly output from . to $.
13493 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13494 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13495 (*tablejump<mode>_internal1_nospec): Likewise.
13497 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13500 * config/sh/sh_optimize_sett_clrt.cc:
13501 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13503 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13505 PR tree-optimization/83940
13506 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13507 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13508 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13509 use it to pass back the definition type.
13510 (vect_check_store_rhs): Likewise rhs_dt_out.
13511 (vect_build_gather_load_calls): Add a mask_dt argument and use
13512 it instead of a call to vect_is_simple_use.
13513 (vectorizable_store): Update calls to vect_check_load_store_mask
13514 and vect_check_store_rhs. Use the dt returned by the latter instead
13515 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13516 instead of calls to vect_is_simple_use. Pass the scalar rather
13517 than the vector operand to vect_is_simple_use when handling
13518 second and subsequent copies of an rhs value.
13519 (vectorizable_load): Update calls to vect_check_load_store_mask
13520 and vect_build_gather_load_calls. Use the cached mask_dt and
13521 gs_info.offset_dt instead of calls to vect_is_simple_use.
13523 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13525 PR middle-end/83945
13526 * tree-emutls.c: Include gimplify.h.
13527 (lower_emutls_2): New function.
13528 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13529 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13530 it before further processing.
13533 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13534 UINTVAL (trueop1) instead of INTVAL (op1).
13536 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13540 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13541 INCOMING_FRAME_SP_OFFSET if not defined.
13542 (scan_trace): Add ENTRY argument. If true and
13543 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13544 emit a note to adjust the CFA offset.
13545 (create_cfi_notes): Adjust scan_trace callers.
13546 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13547 INCOMING_FRAME_SP_OFFSET in the CIE.
13548 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13549 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13551 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13552 * doc/tm.texi: Regenerated.
13554 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13556 PR rtl-optimization/83147
13557 * lra-constraints.c (remove_inheritance_pseudos): Use
13558 lra_substitute_pseudo_within_insn.
13560 2018-01-19 Tom de Vries <tom@codesourcery.com>
13561 Cesar Philippidis <cesar@codesourcery.com>
13564 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13566 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13569 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13570 spaces for function labels.
13572 2018-01-19 Martin Liska <mliska@suse.cz>
13574 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13575 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13576 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13577 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13578 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13579 (PRED_CONST_RETURN): Change from 69 to 65.
13580 (PRED_NULL_RETURN): Change from 91 to 71.
13581 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13582 (PRED_LOOP_GUARD): Change from 66 to 73.
13584 2018-01-19 Martin Liska <mliska@suse.cz>
13586 * predict.c (predict_insn_def): Add new assert.
13587 (struct branch_predictor): Change type to signed integer.
13588 (test_prediction_value_range): Amend test to cover
13589 PROB_UNINITIALIZED.
13590 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13591 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13592 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13593 (PRED_LOOP_IV_COMPARE): Likewise.
13594 * predict.h (PROB_UNINITIALIZED): Define new constant.
13596 2018-01-19 Martin Liska <mliska@suse.cz>
13598 * predict.c (dump_prediction): Add new format for
13599 analyze_brprob.py script which is enabled with -details
13601 * profile-count.h (precise_p): New function.
13603 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13605 PR tree-optimization/83922
13606 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13607 there are no statements that need masking.
13608 (vect_active_double_reduction_p): New function.
13609 (vect_analyze_loop_operations): Use it when handling phis that
13610 are not in the loop header.
13612 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13614 PR tree-optimization/83914
13615 * tree-vect-loop.c (vectorizable_induction): Don't convert
13616 init_expr or apply the peeling adjustment for inductions
13617 that are nested within the vectorized loop.
13619 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13621 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13624 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13628 * function.h (gimplify_parameters): Add gimple_seq * argument.
13629 * function.c: Include gimple.h and options.h.
13630 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13631 for the added local temporaries if needed.
13632 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13633 if there are any parameter cleanups, wrap whole body into a
13634 try/finally with the cleanups.
13636 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13639 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13640 Use GET_MODE_CLASS for scalar floating point.
13642 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13646 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13647 Fix call of call_cgraph_insertion_hooks.
13649 2018-01-18 Martin Sebor <msebor@redhat.com>
13651 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13653 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13656 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13659 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13662 * common.opt: (-ffile-prefix-map): New option.
13663 * opts.c (common_handle_option): Defer it.
13664 * opts-global.c (handle_common_deferred_options): Handle it.
13665 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13666 * file-prefix-map.h: New file.
13667 (remap_debug_filename, add_debug_prefix_map): ...here.
13668 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13669 * final.c (debug_prefix_map, add_debug_prefix_map
13670 remap_debug_filename): Move to...
13671 * file-prefix-map.c: New file.
13672 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13673 generalize, get rid of alloca(), use strrchr() instead of strchr().
13674 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13675 Implement in terms of add_prefix_map().
13676 (remap_macro_filename, remap_debug_filename): Implement in term of
13678 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13679 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13680 * dbxout.c: Include file-prefix-map.h.
13681 * varasm.c: Likewise.
13682 * vmsdbgout.c: Likewise.
13683 * xcoffout.c: Likewise.
13684 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13685 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13686 * doc/invoke.texi (-ffile-prefix-map): Document.
13687 (-fdebug-prefix-map): Update description.
13689 2018-01-18 Martin Liska <mliska@suse.cz>
13691 * config/i386/i386.c (indirect_thunk_name): Document that also
13693 (output_indirect_thunk): Document why both instructions
13694 (pause and lfence) are generated.
13696 2018-01-18 Richard Biener <rguenther@suse.de>
13698 PR tree-optimization/83887
13699 * graphite-scop-detection.c
13700 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13701 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13702 (scop_detection::merge_sese): Re-implement with a flood-fill
13703 algorithm that properly finds a SESE region if it exists.
13705 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13708 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13709 pointer_diff optimizations use view_convert instead of convert.
13711 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13713 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13714 Generate different code for -mno-speculate-indirect-jumps.
13715 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13716 (*call_indirect_aix<mode>): Disable for
13717 -mno-speculate-indirect-jumps.
13718 (*call_indirect_aix<mode>_nospec): New define_insn.
13719 (*call_value_indirect_aix<mode>): Disable for
13720 -mno-speculate-indirect-jumps.
13721 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13722 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13723 -mno-speculate-indirect-jumps.
13724 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13726 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13728 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13729 long double type, set the flags for noting the default long double
13730 type, even if we don't pass or return a long double type.
13732 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13735 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13738 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13740 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13741 support for merge[hl].
13742 (fold_mergehl_helper): New helper function.
13743 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13744 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13745 (altivec_vmrglw_direct): Add xxmrglw insn.
13747 2018-01-17 Andrew Waterman <andrew@sifive.com>
13749 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13750 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13752 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13755 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13756 call the lto_location_cache before reading the
13757 DECL_SOURCE_LOCATION of the types.
13759 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13760 Richard Sandiford <richard.sandiford@linaro.org>
13762 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13763 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13764 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13765 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13766 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13768 * config/aarch64/constraints.md (aarch64_movti_operand):
13770 * config/aarch64/predicates.md (Uti): Add new constraint.
13772 2018-01-17 Carl Love <cel@us.ibm.com>
13774 * config/rs6000/vsx.md (define_expand xl_len_r,
13775 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13776 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13778 (define_expand, define_insn): Move the shift left from the
13779 define_insn to the define_expand for lxvl and stxvl instructions.
13780 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13781 and XL_LEN_R definitions to PURE.
13783 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13785 * config/i386/i386.c (indirect_thunk_name): Declare regno
13786 as unsigned int. Compare regno with INVALID_REGNUM.
13787 (output_indirect_thunk): Ditto.
13788 (output_indirect_thunk_function): Ditto.
13789 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13790 in the call to output_indirect_thunk_function.
13792 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13794 PR middle-end/83884
13795 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13796 rather than the size of inner_type to determine the stack slot size
13797 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13799 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13802 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13805 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13807 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13808 endian Linux systems to optionally enable multilibs for selecting
13809 the long double type if the user configured an explicit type.
13810 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13811 have no long double multilibs if not defined.
13812 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13813 warn if the user used -mabi={ieee,ibm}longdouble and we built
13814 multilibs for long double.
13815 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13816 appropriate multilib option.
13817 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13819 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13820 for building long double multilibs.
13821 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13823 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13825 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13828 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13830 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13833 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13836 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13839 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13841 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13842 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13844 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13846 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13847 different rtl trees depending on TARGET_64BIT.
13848 (rs6000_gen_lvx): Likewise.
13850 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13852 * config/visium/visium.md (nop): Tweak comment.
13853 (hazard_nop): Likewise.
13855 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13857 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13858 -mspeculate-indirect-jumps.
13859 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13860 for -mno-speculate-indirect-jumps.
13861 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13862 (*call_value_indirect_elfv2<mode>): Disable for
13863 -mno-speculate-indirect-jumps.
13864 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13865 (indirect_jump): Emit different RTL for
13866 -mno-speculate-indirect-jumps.
13867 (*indirect_jump<mode>): Disable for
13868 -mno-speculate-indirect-jumps.
13869 (*indirect_jump<mode>_nospec): New define_insn.
13870 (tablejump): Emit different RTL for
13871 -mno-speculate-indirect-jumps.
13872 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13873 (tablejumpsi_nospec): New define_expand.
13874 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13875 (tablejumpdi_nospec): New define_expand.
13876 (*tablejump<mode>_internal1): Disable for
13877 -mno-speculate-indirect-jumps.
13878 (*tablejump<mode>_internal1_nospec): New define_insn.
13879 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13882 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13884 * caller-save.c (insert_save): Drop unnecessary parameter. All
13887 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13888 Richard Biener <rguenth@suse.de>
13891 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13892 return early, inline manually is_gimple_sizepos. Make sure if we
13893 call gimplify_expr we don't end up with a gimple constant.
13894 * tree.c (variably_modified_type_p): Don't return true for
13895 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13896 * gimplify.h (is_gimple_sizepos): Remove.
13898 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13900 PR tree-optimization/83857
13901 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13902 vectorizable_live_operation for pure SLP statements.
13903 (vectorizable_live_operation): Handle PHIs.
13905 2018-01-16 Richard Biener <rguenther@suse.de>
13907 PR tree-optimization/83867
13908 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13909 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13911 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13914 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13915 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13916 If off is not INTEGER_CST, issue a may not be aligned warning
13917 rather than isn't aligned. Use isn%'t rather than isn't.
13918 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13920 <case MULT_EXPR>: Improve the case when bottom and one of the
13921 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13922 operand, in that case check if the other operand is multiple of
13923 bottom divided by the INTEGER_CST operand.
13925 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13928 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13929 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13930 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13931 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13932 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13933 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13934 (pa_function_arg_size): New function.
13936 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13938 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13939 in a separate statement.
13941 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13943 PR tree-optimization/83847
13944 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13945 group gathers and scatters.
13947 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13949 PR rtl-optimization/86620
13950 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13952 PR rtl-optimization/83213
13953 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13954 to last if both are JUMP_INSNs.
13956 PR tree-optimization/83843
13957 * gimple-ssa-store-merging.c
13958 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13959 store_immediate_info for bswap/nop orig_stores.
13961 2018-01-15 Andrew Waterman <andrew@sifive.com>
13963 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13965 <UDIV>: Increase cost if !TARGET_DIV.
13967 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13969 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13970 (define_attr "cr_logical_3op"): New.
13971 (cceq_ior_compare): Adjust.
13972 (cceq_ior_compare_complement): Adjust.
13973 (*cceq_rev_compare): Adjust.
13974 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13975 (is_cracked_insn): Adjust.
13976 (insn_must_be_first_in_group): Adjust.
13977 * config/rs6000/40x.md: Adjust.
13978 * config/rs6000/440.md: Adjust.
13979 * config/rs6000/476.md: Adjust.
13980 * config/rs6000/601.md: Adjust.
13981 * config/rs6000/603.md: Adjust.
13982 * config/rs6000/6xx.md: Adjust.
13983 * config/rs6000/7450.md: Adjust.
13984 * config/rs6000/7xx.md: Adjust.
13985 * config/rs6000/8540.md: Adjust.
13986 * config/rs6000/cell.md: Adjust.
13987 * config/rs6000/e300c2c3.md: Adjust.
13988 * config/rs6000/e500mc.md: Adjust.
13989 * config/rs6000/e500mc64.md: Adjust.
13990 * config/rs6000/e5500.md: Adjust.
13991 * config/rs6000/e6500.md: Adjust.
13992 * config/rs6000/mpc.md: Adjust.
13993 * config/rs6000/power4.md: Adjust.
13994 * config/rs6000/power5.md: Adjust.
13995 * config/rs6000/power6.md: Adjust.
13996 * config/rs6000/power7.md: Adjust.
13997 * config/rs6000/power8.md: Adjust.
13998 * config/rs6000/power9.md: Adjust.
13999 * config/rs6000/rs64.md: Adjust.
14000 * config/rs6000/titan.md: Adjust.
14002 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14004 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14005 ix86_indirect_branch_register logic.
14007 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14009 * config/i386/constraints.md (Bs): Update
14010 ix86_indirect_branch_register check. Don't check
14011 ix86_indirect_branch_register with GOT_memory_operand.
14013 * config/i386/predicates.md (GOT_memory_operand): Don't check
14014 ix86_indirect_branch_register here.
14015 (GOT32_symbol_operand): Likewise.
14017 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14019 * config/i386/predicates.md (constant_call_address_operand):
14020 Rewrite ix86_indirect_branch_register logic.
14021 (sibcall_insn_operand): Likewise.
14023 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14025 * config/i386/constraints.md (Bs): Replace
14026 ix86_indirect_branch_thunk_register with
14027 ix86_indirect_branch_register.
14029 * config/i386/i386.md (indirect_jump): Likewise.
14030 (tablejump): Likewise.
14031 (*sibcall_memory): Likewise.
14032 (*sibcall_value_memory): Likewise.
14033 Peepholes of indirect call and jump via memory: Likewise.
14034 * config/i386/i386.opt: Likewise.
14035 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14036 (GOT_memory_operand): Likewise.
14037 (call_insn_operand): Likewise.
14038 (sibcall_insn_operand): Likewise.
14039 (GOT32_symbol_operand): Likewise.
14041 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14043 PR middle-end/83837
14044 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14045 type rather than type addr's type points to.
14046 (expand_omp_atomic_mutex): Likewise.
14047 (expand_omp_atomic): Likewise.
14049 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14052 * config/i386/i386.c (output_indirect_thunk_function): Use
14053 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14054 for __x86_return_thunk.
14056 2018-01-15 Richard Biener <rguenther@suse.de>
14058 PR middle-end/83850
14059 * expmed.c (extract_bit_field_1): Fix typo.
14061 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14064 * config/arm/iterators.md (VF): New mode iterator.
14065 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14066 Remove integer-related logic from pattern.
14067 (neon_vabd<mode>_3): Likewise.
14069 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14071 PR middle-end/82694
14072 * common.opt (fstrict-overflow): No longer an alias.
14073 (fwrapv-pointer): New option.
14074 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14075 also for pointer types based on flag_wrapv_pointer.
14076 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14077 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14078 opts->x_flag_wrapv got set.
14079 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14080 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14081 POINTER_TYPE_OVERFLOW_UNDEFINED.
14082 * match.pd: Likewise in address comparison pattern.
14083 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14085 2018-01-15 Richard Biener <rguenther@suse.de>
14088 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14089 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14090 Reset type names to their identifier if their TYPE_DECL doesn't
14091 have linkage (and thus is used for ODR and devirt).
14092 (save_debug_info_for_decl): Remove.
14093 (save_debug_info_for_type): Likewise.
14094 (add_tree_to_fld_list): Adjust.
14095 * tree-pretty-print.c (dump_generic_node): Make dumping of
14096 type names more robust.
14098 2018-01-15 Richard Biener <rguenther@suse.de>
14100 * BASE-VER: Bump to 8.0.1.
14102 2018-01-14 Martin Sebor <msebor@redhat.com>
14105 * builtins.c (check_access): Avoid warning when the no-warning bit
14108 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14110 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14111 * ira-color (allocno_hard_regs_compare): Likewise.
14113 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14116 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14117 Use .pushsection/.popsection.
14119 2018-01-14 Martin Sebor <msebor@redhat.com>
14122 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14124 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14126 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14127 entry from extra_headers.
14128 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14129 extra_headers, make the list bitwise identical to the i?86-*-* one.
14131 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14133 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14134 -mcmodel=large with -mindirect-branch=thunk,
14135 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14136 -mfunction-return=thunk-extern.
14137 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14138 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14139 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14141 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14143 * config/i386/i386.c (print_reg): Print the name of the full
14144 integer register without '%'.
14145 (ix86_print_operand): Handle 'V'.
14146 * doc/extend.texi: Document 'V' modifier.
14148 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14150 * config/i386/constraints.md (Bs): Disallow memory operand for
14151 -mindirect-branch-register.
14153 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14154 (GOT_memory_operand): Likewise.
14155 (call_insn_operand): Likewise.
14156 (sibcall_insn_operand): Likewise.
14157 (GOT32_symbol_operand): Likewise.
14158 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14159 for -mindirect-branch-register.
14160 (tablejump): Likewise.
14161 (*sibcall_memory): Likewise.
14162 (*sibcall_value_memory): Likewise.
14163 Disallow peepholes of indirect call and jump via memory for
14164 -mindirect-branch-register.
14165 (*call_pop): Replace m with Bw.
14166 (*call_value_pop): Likewise.
14167 (*sibcall_pop_memory): Replace m with Bs.
14168 * config/i386/i386.opt (mindirect-branch-register): New option.
14169 * doc/invoke.texi: Document -mindirect-branch-register option.
14171 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14173 * config/i386/i386-protos.h (ix86_output_function_return): New.
14174 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14175 set function_return_type.
14176 (indirect_thunk_name): Add ret_p to indicate thunk for function
14178 (output_indirect_thunk_function): Pass false to
14179 indirect_thunk_name.
14180 (ix86_output_indirect_branch_via_reg): Likewise.
14181 (ix86_output_indirect_branch_via_push): Likewise.
14182 (output_indirect_thunk_function): Create alias for function
14183 return thunk if regno < 0.
14184 (ix86_output_function_return): New function.
14185 (ix86_handle_fndecl_attribute): Handle function_return.
14186 (ix86_attribute_table): Add function_return.
14187 * config/i386/i386.h (machine_function): Add
14188 function_return_type.
14189 * config/i386/i386.md (simple_return_internal): Use
14190 ix86_output_function_return.
14191 (simple_return_internal_long): Likewise.
14192 * config/i386/i386.opt (mfunction-return=): New option.
14193 (indirect_branch): Mention -mfunction-return=.
14194 * doc/extend.texi: Document function_return function attribute.
14195 * doc/invoke.texi: Document -mfunction-return= option.
14197 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14199 * config/i386/i386-opts.h (indirect_branch): New.
14200 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14201 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14202 with local indirect jump when converting indirect call and jump.
14203 (ix86_set_indirect_branch_type): New.
14204 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14205 (indirectlabelno): New.
14206 (indirect_thunk_needed): Likewise.
14207 (indirect_thunk_bnd_needed): Likewise.
14208 (indirect_thunks_used): Likewise.
14209 (indirect_thunks_bnd_used): Likewise.
14210 (INDIRECT_LABEL): Likewise.
14211 (indirect_thunk_name): Likewise.
14212 (output_indirect_thunk): Likewise.
14213 (output_indirect_thunk_function): Likewise.
14214 (ix86_output_indirect_branch_via_reg): Likewise.
14215 (ix86_output_indirect_branch_via_push): Likewise.
14216 (ix86_output_indirect_branch): Likewise.
14217 (ix86_output_indirect_jmp): Likewise.
14218 (ix86_code_end): Call output_indirect_thunk_function if needed.
14219 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14221 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14222 (ix86_attribute_table): Add indirect_branch.
14223 * config/i386/i386.h (machine_function): Add indirect_branch_type
14224 and has_local_indirect_jump.
14225 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14227 (tablejump): Likewise.
14228 (*indirect_jump): Use ix86_output_indirect_jmp.
14229 (*tablejump_1): Likewise.
14230 (simple_return_indirect_internal): Likewise.
14231 * config/i386/i386.opt (mindirect-branch=): New option.
14232 (indirect_branch): New.
14235 (thunk-inline): Likewise.
14236 (thunk-extern): Likewise.
14237 * doc/extend.texi: Document indirect_branch function attribute.
14238 * doc/invoke.texi: Document -mindirect-branch= option.
14240 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14243 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14245 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14247 * ipa-inline.c (want_inline_small_function_p): Return false if
14248 inlining has already failed with CIF_FINAL_ERROR.
14249 (update_caller_keys): Call want_inline_small_function_p before
14251 (update_callee_keys): Likewise.
14253 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14255 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14257 (rs6000_quadword_masked_address_p): Likewise.
14258 (quad_aligned_load_p): Likewise.
14259 (quad_aligned_store_p): Likewise.
14260 (const_load_sequence_p): Add comment to describe the outer-most loop.
14261 (mimic_memory_attributes_and_flags): New function.
14262 (rs6000_gen_stvx): Likewise.
14263 (replace_swapped_aligned_store): Likewise.
14264 (rs6000_gen_lvx): Likewise.
14265 (replace_swapped_aligned_load): Likewise.
14266 (replace_swapped_load_constant): Capitalize argument name in
14267 comment describing this function.
14268 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14269 and stores that access quad-word aligned addresses and replace
14270 with stvx or lvx instructions when appropriate.
14271 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14272 New function prototype.
14273 (rs6000_quadword_masked_address_p): Likewise.
14274 (rs6000_gen_lvx): Likewise.
14275 (rs6000_gen_stvx): Likewise.
14276 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14277 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14278 when memory address is aligned.
14279 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14280 this split to select lvx instruction when memory address is aligned.
14281 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14282 instruction when memory address is aligned.
14283 (*vsx_le_perm_load_v16qi): Likewise.
14284 (four unnamed splitters): Modify to select the stvx instruction
14285 when memory is aligned.
14287 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14289 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14290 which appears in the queue multiple times.
14292 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14293 Alan Hayward <alan.hayward@arm.com>
14294 David Sherwood <david.sherwood@arm.com>
14296 * tree-vectorizer.h (vec_lower_bound): New structure.
14297 (_loop_vec_info): Add check_nonzero and lower_bounds.
14298 (LOOP_VINFO_CHECK_NONZERO): New macro.
14299 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14300 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14301 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14302 fields. Make seg_len the distance travelled, not including the
14304 (dr_direction_indicator): Declare.
14305 (dr_zero_step_indicator): Likewise.
14306 (dr_known_forward_stride_p): Likewise.
14307 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14309 (runtime_alias_check_p): Allow runtime alias checks with
14311 (operator ==): Compare access_size and align.
14312 (prune_runtime_alias_test_list): Rework for new distinction between
14313 the access_size and seg_len.
14314 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14316 (get_segment_min_max): New function.
14317 (create_intersect_range_checks): Use it.
14318 (dr_step_indicator): New function.
14319 (dr_direction_indicator): Likewise.
14320 (dr_zero_step_indicator): Likewise.
14321 (dr_known_forward_stride_p): Likewise.
14322 * tree-loop-distribution.c (data_ref_segment_size): Return
14323 DR_STEP * (niters - 1).
14324 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14326 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14327 (vect_preserves_scalar_order_p): New function, split out from...
14328 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14329 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14330 (vect_vfa_access_size): New function.
14331 (vect_vfa_align): Likewise.
14332 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14333 (dump_lower_bound): New function.
14334 (vect_check_lower_bound): Likewise.
14335 (vect_small_gap_p): Likewise.
14336 (vectorizable_with_step_bound_p): Likewise.
14337 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14338 depencies if the vectorization factor is 1. Convert the checks
14339 for nonzero steps into checks on the bounds of DR_STEP. Try using
14340 a bunds check for variable steps if the minimum required step is
14341 relatively small. Update calls to the dr_with_seg_len
14342 constructor and to vect_compile_time_alias.
14343 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14345 (vect_loop_versioning): Call it.
14346 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14348 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14351 Alan Hayward <alan.hayward@arm.com>
14352 David Sherwood <david.sherwood@arm.com>
14354 * doc/sourcebuild.texi (vect_scatter_store): Document.
14355 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14357 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14359 * genopinit.c (main): Add supports_vec_scatter_store and
14360 supports_vec_scatter_store_cached to target_optabs.
14361 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14362 IFN_MASK_SCATTER_STORE.
14363 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14365 * internal-fn.h (internal_store_fn_p): Declare.
14366 (internal_fn_stored_value_index): Likewise.
14367 * internal-fn.c (scatter_store_direct): New macro.
14368 (expand_scatter_store_optab_fn): New function.
14369 (direct_scatter_store_optab_supported_p): New macro.
14370 (internal_store_fn_p): New function.
14371 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14372 IFN_MASK_SCATTER_STORE.
14373 (internal_fn_mask_index): Likewise.
14374 (internal_fn_stored_value_index): New function.
14375 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14376 for scatter stores.
14377 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14378 * optabs-query.c (supports_vec_scatter_store_p): New function.
14379 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14380 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14381 true for scatter stores.
14382 (vect_gather_scatter_fn_p): Handle scatter stores too.
14383 (vect_check_gather_scatter): Consider using scatter stores if
14384 supports_vec_scatter_store_p.
14385 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14386 scatter stores too.
14387 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14388 internal_fn_stored_value_index.
14389 (check_load_store_masking): Handle scatter stores too.
14390 (vect_get_store_rhs): Make public.
14391 (vectorizable_call): Use internal_store_fn_p.
14392 (vectorizable_store): Handle scatter store internal functions.
14393 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14394 when deciding whether the end of the group has been reached.
14395 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14396 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14397 (mask_scatter_store<mode>): New insns.
14399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14400 Alan Hayward <alan.hayward@arm.com>
14401 David Sherwood <david.sherwood@arm.com>
14403 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14404 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14405 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14407 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14408 Use vect_truncate_gather_scatter_offset if we can't treat the
14409 operation as a normal gather load or scatter store.
14410 (get_group_load_store_type): Take the gather_scatter_info
14411 as argument. Try using a gather load or scatter store for
14412 single-element groups.
14413 (get_load_store_type): Update calls to get_group_load_store_type
14414 and vect_use_strided_gather_scatters_p.
14416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14417 Alan Hayward <alan.hayward@arm.com>
14418 David Sherwood <david.sherwood@arm.com>
14420 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14421 optional tree argument.
14422 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14424 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14425 but continue to use the current value as a fallback.
14426 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14427 to compare the updates.
14428 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14429 (get_load_store_type): Use it when handling a strided access.
14430 (vect_get_strided_load_store_ops): New function.
14431 (vect_get_data_ptr_increment): Likewise.
14432 (vectorizable_load): Handle strided gather loads. Always pass
14433 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14436 Alan Hayward <alan.hayward@arm.com>
14437 David Sherwood <david.sherwood@arm.com>
14439 * doc/md.texi (gather_load@var{m}): Document.
14440 (mask_gather_load@var{m}): Likewise.
14441 * genopinit.c (main): Add supports_vec_gather_load and
14442 supports_vec_gather_load_cached to target_optabs.
14443 * optabs-tree.c (init_tree_optimization_optabs): Use
14444 ggc_cleared_alloc to allocate target_optabs.
14445 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14446 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14448 * internal-fn.h (internal_load_fn_p): Declare.
14449 (internal_gather_scatter_fn_p): Likewise.
14450 (internal_fn_mask_index): Likewise.
14451 (internal_gather_scatter_fn_supported_p): Likewise.
14452 * internal-fn.c (gather_load_direct): New macro.
14453 (expand_gather_load_optab_fn): New function.
14454 (direct_gather_load_optab_supported_p): New macro.
14455 (direct_internal_fn_optab): New function.
14456 (internal_load_fn_p): Likewise.
14457 (internal_gather_scatter_fn_p): Likewise.
14458 (internal_fn_mask_index): Likewise.
14459 (internal_gather_scatter_fn_supported_p): Likewise.
14460 * optabs-query.c (supports_at_least_one_mode_p): New function.
14461 (supports_vec_gather_load_p): Likewise.
14462 * optabs-query.h (supports_vec_gather_load_p): Declare.
14463 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14464 and memory_type field.
14465 (NUM_PATTERNS): Bump to 15.
14466 * tree-vect-data-refs.c: Include internal-fn.h.
14467 (vect_gather_scatter_fn_p): New function.
14468 (vect_describe_gather_scatter_call): Likewise.
14469 (vect_check_gather_scatter): Try using internal functions for
14470 gather loads. Recognize existing calls to a gather load function.
14471 (vect_analyze_data_refs): Consider using gather loads if
14472 supports_vec_gather_load_p.
14473 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14474 (vect_get_gather_scatter_offset_type): Likewise.
14475 (vect_convert_mask_for_vectype): Likewise.
14476 (vect_add_conversion_to_patterm): Likewise.
14477 (vect_try_gather_scatter_pattern): Likewise.
14478 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14479 (vect_vect_recog_func_ptrs): Add it.
14480 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14481 internal_fn_mask_index and internal_gather_scatter_fn_p.
14482 (check_load_store_masking): Take the gather_scatter_info as an
14483 argument and handle gather loads.
14484 (vect_get_gather_scatter_ops): New function.
14485 (vectorizable_call): Check internal_load_fn_p.
14486 (vectorizable_load): Likewise. Handle gather load internal
14488 (vectorizable_store): Update call to check_load_store_masking.
14489 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14490 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14491 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14492 (aarch64_gather_scale_operand_d): New predicates.
14493 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14494 (mask_gather_load<mode>): New insns.
14496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14497 Alan Hayward <alan.hayward@arm.com>
14498 David Sherwood <david.sherwood@arm.com>
14500 * optabs.def (fold_left_plus_optab): New optab.
14501 * doc/md.texi (fold_left_plus_@var{m}): Document.
14502 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14503 * internal-fn.c (fold_left_direct): Define.
14504 (expand_fold_left_optab_fn): Likewise.
14505 (direct_fold_left_optab_supported_p): Likewise.
14506 * fold-const-call.c (fold_const_fold_left): New function.
14507 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14508 * tree-parloops.c (valid_reduction_p): New function.
14509 (gather_scalar_reductions): Use it.
14510 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14511 (vect_finish_replace_stmt): Declare.
14512 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14513 (needs_fold_left_reduction_p): New function, split out from...
14514 (vect_is_simple_reduction): ...here. Accept reductions that
14515 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14516 (vect_force_simple_reduction): Also store the reduction type in
14517 the assignment's STMT_VINFO_REDUC_TYPE.
14518 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14519 (merge_with_identity): New function.
14520 (vect_expand_fold_left): Likewise.
14521 (vectorize_fold_left_reduction): Likewise.
14522 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14523 scalar phi in place for it. Check for target support and reject
14524 cases that would reassociate the operation. Defer the transform
14525 phase to vectorize_fold_left_reduction.
14526 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14527 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14528 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14532 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14533 call to ifc_temp_var.
14535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14536 Alan Hayward <alan.hayward@arm.com>
14537 David Sherwood <david.sherwood@arm.com>
14539 * target.def (legitimize_address_displacement): Take the original
14540 offset as a poly_int.
14541 * targhooks.h (default_legitimize_address_displacement): Update
14543 * targhooks.c (default_legitimize_address_displacement): Likewise.
14544 * doc/tm.texi: Regenerate.
14545 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14546 as an argument, moving assert of ad->disp == ad->disp_term to...
14547 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14548 Try calling targetm.legitimize_address_displacement before expanding
14549 the address rather than afterwards, and adjust for the new interface.
14550 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14551 Match the new hook interface. Handle SVE addresses.
14552 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14553 new hook interface.
14555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14557 * Makefile.in (OBJS): Add early-remat.o.
14558 * target.def (select_early_remat_modes): New hook.
14559 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14560 * doc/tm.texi: Regenerate.
14561 * targhooks.h (default_select_early_remat_modes): Declare.
14562 * targhooks.c (default_select_early_remat_modes): New function.
14563 * timevar.def (TV_EARLY_REMAT): New timevar.
14564 * passes.def (pass_early_remat): New pass.
14565 * tree-pass.h (make_pass_early_remat): Declare.
14566 * early-remat.c: New file.
14567 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14569 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14572 Alan Hayward <alan.hayward@arm.com>
14573 David Sherwood <david.sherwood@arm.com>
14575 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14576 vfm1 with a bound_epilog parameter.
14577 (vect_do_peeling): Update calls accordingly, and move the prologue
14578 call earlier in the function. Treat the base bound_epilog as 0 for
14579 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14580 this base when peeling for gaps.
14581 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14582 with fully-masked loops.
14583 (vect_estimate_min_profitable_iters): Handle the single peeled
14584 iteration in that case.
14586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14587 Alan Hayward <alan.hayward@arm.com>
14588 David Sherwood <david.sherwood@arm.com>
14590 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14591 single-element interleaving even if the size is not a power of 2.
14592 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14593 accesses for single-element interleaving if the group size is
14596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14597 Alan Hayward <alan.hayward@arm.com>
14598 David Sherwood <david.sherwood@arm.com>
14600 * doc/md.texi (fold_extract_last_@var{m}): Document.
14601 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14602 * optabs.def (fold_extract_last_optab): New optab.
14603 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14604 * internal-fn.c (fold_extract_direct): New macro.
14605 (expand_fold_extract_optab_fn): Likewise.
14606 (direct_fold_extract_optab_supported_p): Likewise.
14607 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14608 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14609 EXTRACT_LAST_REDUCTION.
14610 (get_initial_def_for_reduction): Do not create an initial vector
14611 for EXTRACT_LAST_REDUCTION reductions.
14612 (vectorizable_reduction): Leave the scalar phi in place for
14613 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14614 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14615 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14616 transform phase to vectorizable_condition.
14617 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14619 (vect_finish_stmt_generation): ...here.
14620 (vect_finish_replace_stmt): New function.
14621 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14622 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14624 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14627 Alan Hayward <alan.hayward@arm.com>
14628 David Sherwood <david.sherwood@arm.com>
14630 * doc/md.texi (extract_last_@var{m}): Document.
14631 * optabs.def (extract_last_optab): New optab.
14632 * internal-fn.def (EXTRACT_LAST): New internal function.
14633 * internal-fn.c (cond_unary_direct): New macro.
14634 (expand_cond_unary_optab_fn): Likewise.
14635 (direct_cond_unary_optab_supported_p): Likewise.
14636 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14637 loops using EXTRACT_LAST.
14638 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14639 (extract_last_<mode>): ...this optab.
14640 (vec_extract<mode><Vel>): Update accordingly.
14642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14643 Alan Hayward <alan.hayward@arm.com>
14644 David Sherwood <david.sherwood@arm.com>
14646 * target.def (empty_mask_is_expensive): New hook.
14647 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14648 * doc/tm.texi: Regenerate.
14649 * targhooks.h (default_empty_mask_is_expensive): Declare.
14650 * targhooks.c (default_empty_mask_is_expensive): New function.
14651 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14652 if the target says that empty masks are expensive.
14653 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14655 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14658 Alan Hayward <alan.hayward@arm.com>
14659 David Sherwood <david.sherwood@arm.com>
14661 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14662 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14663 (vect_use_loop_mask_for_alignment_p): New function.
14664 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14665 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14666 niters_skip argument. Make sure that the first niters_skip elements
14667 of the first iteration are inactive.
14668 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14669 Update call to vect_set_loop_masks_directly.
14670 (get_misalign_in_elems): New function, split out from...
14671 (vect_gen_prolog_loop_niters): ...here.
14672 (vect_update_init_of_dr): Take a code argument that specifies whether
14673 the adjustment should be added or subtracted.
14674 (vect_update_init_of_drs): Likewise.
14675 (vect_prepare_for_masked_peels): New function.
14676 (vect_do_peeling): Skip prologue peeling if we're using a mask
14677 instead. Update call to vect_update_inits_of_drs.
14678 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14680 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14681 alignment. Do not include the number of peeled iterations in
14682 the minimum threshold in that case.
14683 (vectorizable_induction): Adjust the start value down by
14684 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14685 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14686 Take the number of skipped iterations into account when calculating
14688 * tree-vect-stmts.c (vect_gen_while_not): New function.
14690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14691 Alan Hayward <alan.hayward@arm.com>
14692 David Sherwood <david.sherwood@arm.com>
14694 * doc/sourcebuild.texi (vect_fully_masked): Document.
14695 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14696 default value to 0.
14697 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14699 (vect_analyze_loop_2): ...here. Don't check the vectorization
14700 factor against the number of loop iterations if the loop is
14703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14704 Alan Hayward <alan.hayward@arm.com>
14705 David Sherwood <david.sherwood@arm.com>
14707 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14708 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14709 (dump_groups): Update accordingly.
14710 (iv_use::mem_type): New member variable.
14711 (address_p): New function.
14712 (record_use): Add a mem_type argument and initialize the new
14714 (record_group_use): Add a mem_type argument. Use address_p.
14715 Remove obsolete null checks of base_object. Update call to record_use.
14716 (find_interesting_uses_op): Update call to record_group_use.
14717 (find_interesting_uses_cond): Likewise.
14718 (find_interesting_uses_address): Likewise.
14719 (get_mem_type_for_internal_fn): New function.
14720 (find_address_like_use): Likewise.
14721 (find_interesting_uses_stmt): Try find_address_like_use before
14722 calling find_interesting_uses_op.
14723 (addr_offset_valid_p): Use the iv mem_type field as the type
14724 of the addressed memory.
14725 (add_autoinc_candidates): Likewise.
14726 (get_address_cost): Likewise.
14727 (split_small_address_groups_p): Use address_p.
14728 (split_address_groups): Likewise.
14729 (add_iv_candidate_for_use): Likewise.
14730 (autoinc_possible_for_pair): Likewise.
14731 (rewrite_groups): Likewise.
14732 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14733 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14734 (get_alias_ptr_type_for_ptr_address): New function.
14735 (rewrite_use_address): Rewrite address uses in calls that were
14736 identified by find_address_like_use.
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 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14744 * gimple-expr.h (is_gimple_addressable: Likewise.
14745 * gimple-expr.c (is_gimple_address): Likewise.
14746 * internal-fn.c (expand_call_mem_ref): New function.
14747 (expand_mask_load_optab_fn): Use it.
14748 (expand_mask_store_optab_fn): Likewise.
14750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14751 Alan Hayward <alan.hayward@arm.com>
14752 David Sherwood <david.sherwood@arm.com>
14754 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14755 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14756 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14757 (cond_umax@var{mode}): Document.
14758 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14759 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14760 (cond_umin_optab, cond_umax_optab): New optabs.
14761 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14762 (COND_IOR, COND_XOR): New internal functions.
14763 * internal-fn.h (get_conditional_internal_fn): Declare.
14764 * internal-fn.c (cond_binary_direct): New macro.
14765 (expand_cond_binary_optab_fn): Likewise.
14766 (direct_cond_binary_optab_supported_p): Likewise.
14767 (get_conditional_internal_fn): New function.
14768 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14769 Cope with reduction statements that are vectorized as calls rather
14771 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14772 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14773 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14774 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14775 (UNSPEC_COND_EOR): New unspecs.
14776 (optab): Add mappings for them.
14777 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14778 (sve_int_op, sve_fp_op): New int attributes.
14780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14781 Alan Hayward <alan.hayward@arm.com>
14782 David Sherwood <david.sherwood@arm.com>
14784 * optabs.def (while_ult_optab): New optab.
14785 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14786 * internal-fn.def (WHILE_ULT): New internal function.
14787 * internal-fn.h (direct_internal_fn_supported_p): New override
14788 that takes two types as argument.
14789 * internal-fn.c (while_direct): New macro.
14790 (expand_while_optab_fn): New function.
14791 (convert_optab_supported_p): Likewise.
14792 (direct_while_optab_supported_p): New macro.
14793 * wide-int.h (wi::udiv_ceil): New function.
14794 * tree-vectorizer.h (rgroup_masks): New structure.
14795 (vec_loop_masks): New typedef.
14796 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14797 and fully_masked_p.
14798 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14799 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14800 (vect_max_vf): New function.
14801 (slpeel_make_loop_iterate_ntimes): Delete.
14802 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14803 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14804 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14805 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14806 internal-fn.h, stor-layout.h and optabs-query.h.
14807 (vect_set_loop_mask): New function.
14808 (add_preheader_seq): Likewise.
14809 (add_header_seq): Likewise.
14810 (interleave_supported_p): Likewise.
14811 (vect_maybe_permute_loop_masks): Likewise.
14812 (vect_set_loop_masks_directly): Likewise.
14813 (vect_set_loop_condition_masked): Likewise.
14814 (vect_set_loop_condition_unmasked): New function, split out from
14815 slpeel_make_loop_iterate_ntimes.
14816 (slpeel_make_loop_iterate_ntimes): Rename to..
14817 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14818 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14819 (vect_do_peeling): Update call accordingly.
14820 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14822 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14823 mask_compare_type, can_fully_mask_p and fully_masked_p.
14824 (release_vec_loop_masks): New function.
14825 (_loop_vec_info): Use it to free the loop masks.
14826 (can_produce_all_loop_masks_p): New function.
14827 (vect_get_max_nscalars_per_iter): Likewise.
14828 (vect_verify_full_masking): Likewise.
14829 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14830 retries, and free the mask rgroups before retrying. Check loop-wide
14831 reasons for disallowing fully-masked loops. Make the final decision
14832 about whether use a fully-masked loop or not.
14833 (vect_estimate_min_profitable_iters): Do not assume that peeling
14834 for the number of iterations will be needed for fully-masked loops.
14835 (vectorizable_reduction): Disable fully-masked loops.
14836 (vectorizable_live_operation): Likewise.
14837 (vect_halve_mask_nunits): New function.
14838 (vect_double_mask_nunits): Likewise.
14839 (vect_record_loop_mask): Likewise.
14840 (vect_get_loop_mask): Likewise.
14841 (vect_transform_loop): Handle the case in which the final loop
14842 iteration might handle a partial vector. Call vect_set_loop_condition
14843 instead of slpeel_make_loop_iterate_ntimes.
14844 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14845 (check_load_store_masking): New function.
14846 (prepare_load_store_mask): Likewise.
14847 (vectorizable_store): Handle fully-masked loops.
14848 (vectorizable_load): Likewise.
14849 (supportable_widening_operation): Use vect_halve_mask_nunits for
14851 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14852 (vect_gen_while): New function.
14853 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14854 (aarch64_uqdec<mode>): New insn.
14856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14857 Alan Hayward <alan.hayward@arm.com>
14858 David Sherwood <david.sherwood@arm.com>
14860 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14861 (reduc_xor_scal_optab): New optabs.
14862 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14863 (reduc_xor_scal_@var{m}): Document.
14864 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14865 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14866 internal functions.
14867 * fold-const-call.c (fold_const_call): Handle them.
14868 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14869 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14870 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14871 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14872 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14873 (UNSPEC_XORV): New unspecs.
14874 (optab): Add entries for them.
14875 (BITWISEV): New int iterator.
14876 (bit_reduc_op): New int attributes.
14878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14879 Alan Hayward <alan.hayward@arm.com>
14880 David Sherwood <david.sherwood@arm.com>
14882 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14883 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14884 * optabs.def (vec_shl_insert_optab): New optab.
14885 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14886 (duplicate_and_interleave): Likewise.
14887 * tree-vect-loop.c: Include internal-fn.h.
14888 (neutral_op_for_slp_reduction): New function, split out from
14889 get_initial_defs_for_reduction.
14890 (get_initial_def_for_reduction): Handle option 2 for variable-length
14891 vectors by loading the neutral value into a vector and then shifting
14892 the initial value into element 0.
14893 (get_initial_defs_for_reduction): Replace the code argument with
14894 the neutral value calculated by neutral_op_for_slp_reduction.
14895 Use gimple_build_vector for constant-length vectors.
14896 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14897 but the first group_size elements have a neutral value.
14898 Use duplicate_and_interleave otherwise.
14899 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14900 Update call to get_initial_defs_for_reduction. Handle SLP
14901 reductions for variable-length vectors by creating one vector
14902 result for each scalar result, with the elements associated
14903 with other scalar results stubbed out with the neutral value.
14904 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14905 Require IFN_VEC_SHL_INSERT for double reductions on
14906 variable-length vectors, or SLP reductions that have
14907 a neutral value. Require can_duplicate_and_interleave_p
14908 support for variable-length unchained SLP reductions if there
14909 is no neutral value, such as for MIN/MAX reductions. Also require
14910 the number of vector elements to be a multiple of the number of
14911 SLP statements when doing variable-length unchained SLP reductions.
14912 Update call to vect_create_epilog_for_reduction.
14913 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14914 and remove initial values.
14915 (duplicate_and_interleave): Make public.
14916 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14917 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14920 Alan Hayward <alan.hayward@arm.com>
14921 David Sherwood <david.sherwood@arm.com>
14923 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14924 (can_duplicate_and_interleave_p): New function.
14925 (vect_get_and_check_slp_defs): Take the vector of statements
14926 rather than just the current one. Remove excess parentheses.
14927 Restriction rejectinon of vect_constant_def and vect_external_def
14928 for variable-length vectors to boolean types, or types for which
14929 can_duplicate_and_interleave_p is false.
14930 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14931 (duplicate_and_interleave): New function.
14932 (vect_get_constant_vectors): Use gimple_build_vector for
14933 constant-length vectors and suitable variable-length constant
14934 vectors. Use duplicate_and_interleave for other variable-length
14935 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14942 min_profitable_iters doesn't go negative.
14944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14945 Alan Hayward <alan.hayward@arm.com>
14946 David Sherwood <david.sherwood@arm.com>
14948 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14949 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14950 * optabs.def (vec_mask_load_lanes_optab): New optab.
14951 (vec_mask_store_lanes_optab): Likewise.
14952 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14953 (MASK_STORE_LANES): Likewise.
14954 * internal-fn.c (mask_load_lanes_direct): New macro.
14955 (mask_store_lanes_direct): Likewise.
14956 (expand_mask_load_optab_fn): Handle masked operations.
14957 (expand_mask_load_lanes_optab_fn): New macro.
14958 (expand_mask_store_optab_fn): Handle masked operations.
14959 (expand_mask_store_lanes_optab_fn): New macro.
14960 (direct_mask_load_lanes_optab_supported_p): Likewise.
14961 (direct_mask_store_lanes_optab_supported_p): Likewise.
14962 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14964 (vect_load_lanes_supported): Likewise.
14965 * tree-vect-data-refs.c (strip_conversion): New function.
14966 (can_group_stmts_p): Likewise.
14967 (vect_analyze_data_ref_accesses): Use it instead of checking
14968 for a pair of assignments.
14969 (vect_store_lanes_supported): Take a masked_p parameter.
14970 (vect_load_lanes_supported): Likewise.
14971 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14972 vect_store_lanes_supported and vect_load_lanes_supported.
14973 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14974 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14975 parameter. Don't allow gaps for masked accesses.
14976 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14977 and vect_load_lanes_supported.
14978 (get_load_store_type): Take a masked_p parameter and update
14979 call to get_group_load_store_type.
14980 (vectorizable_store): Update call to get_load_store_type.
14981 Handle IFN_MASK_STORE_LANES.
14982 (vectorizable_load): Update call to get_load_store_type.
14983 Handle IFN_MASK_LOAD_LANES.
14985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14986 Alan Hayward <alan.hayward@arm.com>
14987 David Sherwood <david.sherwood@arm.com>
14989 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14991 * config/aarch64/aarch64-protos.h
14992 (aarch64_sve_struct_memory_operand_p): Declare.
14993 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14994 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14995 (VPRED, vpred): Handle SVE structure modes.
14996 * config/aarch64/constraints.md (Utx): New constraint.
14997 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14998 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14999 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15000 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15001 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15002 structure modes. Split into pieces after RA.
15003 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15004 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15006 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15007 SVE structure modes.
15008 (aarch64_classify_address): Likewise.
15009 (sizetochar): Move earlier in file.
15010 (aarch64_print_operand): Handle SVE register lists.
15011 (aarch64_array_mode): New function.
15012 (aarch64_sve_struct_memory_operand_p): Likewise.
15013 (TARGET_ARRAY_MODE): Redefine.
15015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15016 Alan Hayward <alan.hayward@arm.com>
15017 David Sherwood <david.sherwood@arm.com>
15019 * target.def (array_mode): New target hook.
15020 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15021 * doc/tm.texi: Regenerate.
15022 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15023 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15024 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15025 targetm.array_mode.
15026 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15030 Alan Hayward <alan.hayward@arm.com>
15031 David Sherwood <david.sherwood@arm.com>
15033 * fold-const.c (fold_binary_loc): Check the argument types
15034 rather than the result type when testing for a vector operation.
15036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15038 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15039 * doc/tm.texi: Regenerate.
15041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15042 Alan Hayward <alan.hayward@arm.com>
15043 David Sherwood <david.sherwood@arm.com>
15045 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15046 (sve): Document new AArch64 extension.
15047 * doc/md.texi (w): Extend the description of the AArch64
15048 constraint to include SVE vectors.
15049 (Upl, Upa): Document new AArch64 predicate constraints.
15050 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15052 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15053 (msve-vector-bits=): New option.
15054 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15055 SVE when these are disabled.
15056 (sve): New extension.
15057 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15058 modes. Adjust their number of units based on aarch64_sve_vg.
15059 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15060 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15061 aarch64_addr_query_type.
15062 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15063 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15064 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15065 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15066 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15067 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15068 (aarch64_simd_imm_zero_p): Delete.
15069 (aarch64_check_zero_based_sve_index_immediate): Declare.
15070 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15071 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15072 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15073 (aarch64_sve_float_mul_immediate_p): Likewise.
15074 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15075 rather than an rtx.
15076 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15077 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15078 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15079 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15080 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15081 (aarch64_regmode_natural_size): Likewise.
15082 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15083 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15085 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15086 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15087 for VG and the SVE predicate registers.
15088 (V_ALIASES): Add a "z"-prefixed alias.
15089 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15090 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15091 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15092 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15093 (REG_CLASS_NAMES): Add entries for them.
15094 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15095 and the predicate registers.
15096 (aarch64_sve_vg): Declare.
15097 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15098 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15099 (REGMODE_NATURAL_SIZE): Define.
15100 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15102 * config/aarch64/aarch64.c: Include cfgrtl.h.
15103 (simd_immediate_info): Add a constructor for series vectors,
15104 and an associated step field.
15105 (aarch64_sve_vg): New variable.
15106 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15107 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15108 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15109 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15110 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15111 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15112 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15113 (aarch64_get_mask_mode): New functions.
15114 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15115 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15116 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15117 predicate modes and predicate registers. Explicitly restrict
15118 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15119 to store a vector mode if it is recognized by
15120 aarch64_classify_vector_mode.
15121 (aarch64_regmode_natural_size): New function.
15122 (aarch64_hard_regno_caller_save_mode): Return the original mode
15124 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15125 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15126 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15127 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15129 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15130 does not overlap dest if the function is frame-related. Handle
15132 (aarch64_split_add_offset): New function.
15133 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15134 them aarch64_add_offset.
15135 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15136 and update call to aarch64_sub_sp.
15137 (aarch64_add_cfa_expression): New function.
15138 (aarch64_expand_prologue): Pass extra temporary registers to the
15139 functions above. Handle the case in which we need to emit new
15140 DW_CFA_expressions for registers that were originally saved
15141 relative to the stack pointer, but now have to be expressed
15142 relative to the frame pointer.
15143 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15145 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15146 IP0 and IP1 values for SVE frames.
15147 (aarch64_expand_vec_series): New function.
15148 (aarch64_expand_sve_widened_duplicate): Likewise.
15149 (aarch64_expand_sve_const_vector): Likewise.
15150 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15151 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15152 into the register, rather than emitting a SET directly.
15153 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15154 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15155 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15156 (offset_9bit_signed_scaled_p): New functions.
15157 (aarch64_replicate_bitmask_imm): New function.
15158 (aarch64_bitmask_imm): Use it.
15159 (aarch64_cannot_force_const_mem): Reject expressions involving
15160 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15161 (aarch64_classify_index): Handle SVE indices, by requiring
15162 a plain register index with a scale that matches the element size.
15163 (aarch64_classify_address): Handle SVE addresses. Assert that
15164 the mode of the address is VOIDmode or an integer mode.
15165 Update call to aarch64_classify_symbol.
15166 (aarch64_classify_symbolic_expression): Update call to
15167 aarch64_classify_symbol.
15168 (aarch64_const_vec_all_in_range_p): New function.
15169 (aarch64_print_vector_float_operand): Likewise.
15170 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15171 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15172 and the FP immediates 1.0 and 0.5.
15173 (aarch64_print_address_internal): Handle SVE addresses.
15174 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15175 (aarch64_regno_regclass): Handle predicate registers.
15176 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15178 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15179 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15180 (aarch64_convert_sve_vector_bits): New function.
15181 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15182 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15183 rather than an rtx.
15184 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15185 Handle SVE vector and predicate modes. Accept VL-based constants
15186 that need only one temporary register, and VL offsets that require
15187 no temporary registers.
15188 (aarch64_conditional_register_usage): Mark the predicate registers
15189 as fixed if SVE isn't available.
15190 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15191 Return true for SVE vector and predicate modes.
15192 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15193 rather than an unsigned int. Handle SVE modes.
15194 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15196 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15198 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15199 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15200 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15201 (aarch64_sve_float_mul_immediate_p): New functions.
15202 (aarch64_sve_valid_immediate): New function.
15203 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15204 Explicitly reject structure modes. Check for INDEX constants.
15205 Handle PTRUE and PFALSE constants.
15206 (aarch64_check_zero_based_sve_index_immediate): New function.
15207 (aarch64_simd_imm_zero_p): Delete.
15208 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15209 vector modes. Accept constants in the range of CNT[BHWD].
15210 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15211 ask for an Advanced SIMD mode.
15212 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15213 (aarch64_simd_vector_alignment): Handle SVE predicates.
15214 (aarch64_vectorize_preferred_vector_alignment): New function.
15215 (aarch64_simd_vector_alignment_reachable): Use it instead of
15217 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15218 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15220 (MAX_VECT_LEN): Delete.
15221 (expand_vec_perm_d): Add a vec_flags field.
15222 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15223 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15224 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15226 (aarch64_evpc_rev): Rename to...
15227 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15228 (aarch64_evpc_rev_global): New function.
15229 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15230 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15232 (aarch64_evpc_sve_tbl): New function.
15233 (aarch64_expand_vec_perm_const_1): Update after rename of
15234 aarch64_evpc_rev. Handle SVE permutes too, trying
15235 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15236 than aarch64_evpc_tbl.
15237 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15238 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15239 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15240 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15241 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15242 (aarch64_expand_sve_vcond): New functions.
15243 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15244 of aarch64_vector_mode_p.
15245 (aarch64_dwarf_poly_indeterminate_value): New function.
15246 (aarch64_compute_pressure_classes): Likewise.
15247 (aarch64_can_change_mode_class): Likewise.
15248 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15249 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15250 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15251 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15252 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15253 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15254 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15255 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15257 (Dn, Dl, Dr): Accept const as well as const_vector.
15258 (Dz): Likewise. Compare against CONST0_RTX.
15259 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15260 of "vector" where appropriate.
15261 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15262 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15263 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15264 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15265 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15266 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15267 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15268 (v_int_equiv): Extend to SVE modes.
15269 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15271 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15272 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15273 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15274 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15275 (SVE_COND_FP_CMP): New int iterators.
15276 (perm_hilo): Handle the new unpack unspecs.
15277 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15279 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15280 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15281 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15282 (aarch64_equality_operator, aarch64_constant_vector_operand)
15283 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15284 (aarch64_sve_nonimmediate_operand): Likewise.
15285 (aarch64_sve_general_operand): Likewise.
15286 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15287 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15288 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15289 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15290 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15291 (aarch64_sve_float_arith_immediate): Likewise.
15292 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15293 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15294 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15295 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15296 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15297 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15298 (aarch64_sve_float_arith_operand): Likewise.
15299 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15300 (aarch64_sve_float_mul_operand): Likewise.
15301 (aarch64_sve_vec_perm_operand): Likewise.
15302 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15303 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15304 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15305 as well as const_vector.
15306 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15307 in file. Use CONST0_RTX and CONSTM1_RTX.
15308 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15309 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15310 Use aarch64_simd_imm_zero.
15311 * config/aarch64/aarch64-sve.md: New file.
15312 * config/aarch64/aarch64.md: Include it.
15313 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15314 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15315 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15316 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15317 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15318 (sve): New attribute.
15319 (enabled): Disable instructions with the sve attribute unless
15321 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15322 aarch64_expand_mov_immediate.
15323 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15324 CNT[BHSD] immediates.
15325 (movti): Split CONST_POLY_INT moves into two halves.
15326 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15327 Split additions that need a temporary here if the destination
15328 is the stack pointer.
15329 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15330 (*add<mode>3_poly_1): New instruction.
15331 (set_clobber_cc): New expander.
15333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15335 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15336 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15337 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15338 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15339 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15340 Change innermode from fixed_mode_size to machine_mode.
15341 (simplify_subreg): Update call accordingly. Handle a constant-sized
15342 subreg of a variable-length CONST_VECTOR.
15344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15345 Alan Hayward <alan.hayward@arm.com>
15346 David Sherwood <david.sherwood@arm.com>
15348 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15349 (add_offset_to_base): New function, split out from...
15350 (create_mem_ref): ...here. When handling a scale other than 1,
15351 check first whether the address is valid without the offset.
15352 Add it into the base if so, leaving the index and scale as-is.
15354 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15357 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15358 fold_for_warn before checking if arg2 is INTEGER_CST.
15360 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15362 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15363 (store_multiple_operation): Delete.
15364 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15365 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15366 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15367 guarded by TARGET_STRING.
15368 (rs6000_output_load_multiple): Delete.
15369 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15370 OPTION_MASK_STRING / TARGET_STRING handling.
15371 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15372 (const rs6000_opt_masks) <"string">: Change mask to 0.
15373 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15374 (MASK_STRING): Delete.
15375 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15377 (load_multiple): Delete.
15384 (store_multiple): Delete.
15391 (movmemsi_8reg): Delete.
15392 (corresponding unnamed define_insn): Delete.
15393 (movmemsi_6reg): Delete.
15394 (corresponding unnamed define_insn): Delete.
15395 (movmemsi_4reg): Delete.
15396 (corresponding unnamed define_insn): Delete.
15397 (movmemsi_2reg): Delete.
15398 (corresponding unnamed define_insn): Delete.
15399 (movmemsi_1reg): Delete.
15400 (corresponding unnamed define_insn): Delete.
15401 * config/rs6000/rs6000.opt (mno-string): New.
15402 (mstring): Replace by deprecation warning stub.
15403 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15405 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15407 * regrename.c (regrename_do_replace): If replacing the same
15408 reg multiple times, try to reuse last created gen_raw_REG.
15411 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15412 main to workaround a bug in GDB.
15414 2018-01-12 Tom de Vries <tom@codesourcery.com>
15417 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15419 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15421 PR rtl-optimization/80481
15422 * ira-color.c (get_cap_member): New function.
15423 (allocnos_conflict_by_live_ranges_p): Use it.
15424 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15425 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15427 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15430 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15431 (*saddl_se_1): Ditto.
15432 (*ssubsi_1): Ditto.
15433 (*ssubl_se_1): Ditto.
15435 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15437 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15438 rather than wi::to_widest for DR_INITs.
15439 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15440 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15441 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15443 (vect_analyze_group_access_1): Note that here.
15445 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15447 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15448 polynomial type sizes.
15450 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15452 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15453 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15454 (gimple_add_tmp_var): Likewise.
15456 2018-01-12 Martin Liska <mliska@suse.cz>
15458 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15459 (gimple_alloc_sizes): Likewise.
15460 (dump_gimple_statistics): Use PRIu64 in printf format.
15461 * gimple.h: Change uint64_t to int.
15463 2018-01-12 Martin Liska <mliska@suse.cz>
15465 * tree-core.h: Use uint64_t instead of int.
15466 * tree.c (tree_node_counts): Likewise.
15467 (tree_node_sizes): Likewise.
15468 (dump_tree_statistics): Use PRIu64 in printf format.
15470 2018-01-12 Martin Liska <mliska@suse.cz>
15472 * Makefile.in: As qsort_chk is implemented in vec.c, add
15473 vec.o to linkage of gencfn-macros.
15474 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15475 passing the info to record_node_allocation_statistics.
15476 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15478 * ggc-common.c (struct ggc_usage): Add operator== and use
15479 it in operator< and compare function.
15480 * mem-stats.h (struct mem_usage): Likewise.
15481 * vec.c (struct vec_usage): Remove operator< and compare
15482 function. Can be simply inherited.
15484 2018-01-12 Martin Jambor <mjambor@suse.cz>
15487 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15488 * tree-ssa-math-opts.c: Include domwalk.h.
15489 (convert_mult_to_fma_1): New function.
15490 (fma_transformation_info): New type.
15491 (fma_deferring_state): Likewise.
15492 (cancel_fma_deferring): New function.
15493 (result_of_phi): Likewise.
15494 (last_fma_candidate_feeds_initial_phi): Likewise.
15495 (convert_mult_to_fma): Added deferring logic, split actual
15496 transformation to convert_mult_to_fma_1.
15497 (math_opts_dom_walker): New type.
15498 (math_opts_dom_walker::after_dom_children): New method, body moved
15499 here from pass_optimize_widening_mul::execute, added deferring logic
15501 (pass_optimize_widening_mul::execute): Moved most of code to
15502 math_opts_dom_walker::after_dom_children.
15503 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15504 * config/i386/i386.c (ix86_option_override_internal): Added
15505 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15507 2018-01-12 Richard Biener <rguenther@suse.de>
15510 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15511 inline instance vars.
15513 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15516 * config/rx/rx.c (rx_is_restricted_memory_address):
15517 Handle SUBREG case.
15519 2018-01-12 Richard Biener <rguenther@suse.de>
15521 PR tree-optimization/80846
15522 * target.def (split_reduction): New target hook.
15523 * targhooks.c (default_split_reduction): New function.
15524 * targhooks.h (default_split_reduction): Declare.
15525 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15526 target requests first reduce vectors by combining low and high
15528 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15529 (get_vectype_for_scalar_type_and_size): Export.
15530 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15531 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15532 * doc/tm.texi: Regenerate.
15533 * config/i386/i386.c (ix86_split_reduction): Implement
15534 TARGET_VECTORIZE_SPLIT_REDUCTION.
15536 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15539 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15540 in PIC mode except for TARGET_VXWORKS_RTP.
15541 * config/sparc/sparc.c: Include cfgrtl.h.
15542 (TARGET_INIT_PIC_REG): Define.
15543 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15544 (sparc_pic_register_p): New predicate.
15545 (sparc_legitimate_address_p): Use it.
15546 (sparc_legitimize_pic_address): Likewise.
15547 (sparc_delegitimize_address): Likewise.
15548 (sparc_mode_dependent_address_p): Likewise.
15549 (gen_load_pcrel_sym): Remove 4th parameter.
15550 (load_got_register): Adjust call to above. Remove obsolete stuff.
15551 (sparc_expand_prologue): Do not call load_got_register here.
15552 (sparc_flat_expand_prologue): Likewise.
15553 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15554 (sparc_use_pseudo_pic_reg): New function.
15555 (sparc_init_pic_reg): Likewise.
15556 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15557 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15559 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15561 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15562 Add item for branch_cost.
15564 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15566 PR rtl-optimization/83565
15567 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15568 not extend the result to a larger mode for rotate operations.
15569 (num_sign_bit_copies1): Likewise.
15571 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15574 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15576 Use values-Xc.o for -pedantic.
15577 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15579 2018-01-12 Martin Liska <mliska@suse.cz>
15582 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15584 (possible_polymorphic_call_targets): Use it.
15585 (ipa_devirt): Likewise.
15587 2018-01-12 Martin Liska <mliska@suse.cz>
15589 * profile-count.h (enum profile_quality): Use 0 as invalid
15590 enum value of profile_quality.
15592 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15594 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15595 -mext-string options.
15597 2018-01-12 Richard Biener <rguenther@suse.de>
15599 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15600 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15601 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15603 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15605 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15607 * configure.ac (--with-long-double-format): Add support for the
15608 configuration option to change the default long double format on
15610 * config.gcc (powerpc*-linux*-*): Likewise.
15611 * configure: Regenerate.
15612 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15613 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15614 used without modification.
15616 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15618 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15619 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15620 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15621 MISC_BUILTIN_SPEC_BARRIER.
15622 (rs6000_init_builtins): Likewise.
15623 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15625 (speculation_barrier): New define_insn.
15626 * doc/extend.texi: Document __builtin_speculation_barrier.
15628 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15631 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15632 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15633 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15635 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15636 integral modes instead of "ss" and "sd".
15637 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15638 vectors with 32-bit and 64-bit elements.
15639 (vecdupssescalarmodesuffix): New mode attribute.
15640 (vec_dup<mode>): Use it.
15642 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15645 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15646 frame if argument is passed on stack.
15648 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15651 * ree.c (combine_reaching_defs): Optimize also
15652 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15653 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15655 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15657 PR middle-end/83189
15658 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15660 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15662 PR middle-end/83718
15663 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15664 after they are computed.
15666 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15668 PR tree-optimization/83695
15669 * gimple-loop-linterchange.cc
15670 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15671 reset cached scev information after interchange.
15672 (pass_linterchange::execute): Remove call to scev_reset_htab.
15674 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15676 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15677 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15678 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15679 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15680 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15681 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15682 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15683 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15684 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15685 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15686 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15687 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15688 (V_lane_reg): Likewise.
15689 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15691 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15692 (vfmal_lane_low<mode>_intrinsic,
15693 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15694 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15695 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15696 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15697 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15698 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15700 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15702 * config/arm/arm-cpus.in (fp16fml): New feature.
15703 (ALL_SIMD): Add fp16fml.
15704 (armv8.2-a): Add fp16fml as an option.
15705 (armv8.3-a): Likewise.
15706 (armv8.4-a): Add fp16fml as part of fp16.
15707 * config/arm/arm.h (TARGET_FP16FML): Define.
15708 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15710 * config/arm/arm-modes.def (V2HF): Define.
15711 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15712 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15713 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15714 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15715 vfmsl_low, vfmsl_high): New set of builtins.
15716 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15717 (vfml_op): New code attribute.
15718 (VFMLHALVES): New int iterator.
15719 (VFML, VFMLSEL): New mode attributes.
15720 (V_reg): Define mapping for V2HF.
15721 (V_hi, V_lo): New mode attributes.
15722 (VF_constraint): Likewise.
15723 (vfml_half, vfml_half_selector): New int attributes.
15724 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15726 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15727 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15729 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15730 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15731 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15732 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15734 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15735 Document new effective target and option set.
15737 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15739 * config/arm/arm-cpus.in (armv8_4): New feature.
15740 (ARMv8_4a): New fgroup.
15741 (armv8.4-a): New arch.
15742 * config/arm/arm-tables.opt: Regenerate.
15743 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15744 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15745 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15746 Add matching rules for -march=armv8.4-a and extensions.
15747 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15749 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15752 * config/rx/rx.md (BW): New mode attribute.
15753 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15755 2018-01-11 Richard Biener <rguenther@suse.de>
15757 PR tree-optimization/83435
15758 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15759 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15760 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15762 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15763 Alan Hayward <alan.hayward@arm.com>
15764 David Sherwood <david.sherwood@arm.com>
15766 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15768 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15769 (aarch64_print_address_internal): Use it to check for a zero offset.
15771 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15772 Alan Hayward <alan.hayward@arm.com>
15773 David Sherwood <david.sherwood@arm.com>
15775 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15776 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15777 Return a poly_int64 rather than a HOST_WIDE_INT.
15778 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15779 rather than a HOST_WIDE_INT.
15780 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15781 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15782 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15783 final_offset from HOST_WIDE_INT to poly_int64.
15784 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15785 to_constant when getting the number of units in an Advanced SIMD
15787 (aarch64_builtin_vectorized_function): Check for a constant number
15789 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15791 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15792 attribute instead of GET_MODE_NUNITS.
15793 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15794 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15795 GET_MODE_SIZE for fixed-size registers.
15796 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15797 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15798 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15799 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15800 (aarch64_print_operand, aarch64_print_address_internal)
15801 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15802 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15803 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15804 Handle polynomial GET_MODE_SIZE.
15805 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15806 wider than SImode without modification.
15807 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15808 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15809 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15810 passing and returning SVE modes.
15811 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15812 rather than GEN_INT.
15813 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15814 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15815 (aarch64_allocate_and_probe_stack_space): Likewise.
15816 (aarch64_layout_frame): Cope with polynomial offsets.
15817 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15818 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15819 polynomial offsets.
15820 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15821 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15822 poly_int64 rather than a HOST_WIDE_INT.
15823 (aarch64_get_separate_components, aarch64_process_components)
15824 (aarch64_expand_prologue, aarch64_expand_epilogue)
15825 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15826 (aarch64_anchor_offset): New function, split out from...
15827 (aarch64_legitimize_address): ...here.
15828 (aarch64_builtin_vectorization_cost): Handle polynomial
15829 TYPE_VECTOR_SUBPARTS.
15830 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15832 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15833 number of elements from the PARALLEL rather than the mode.
15834 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15835 rather than GET_MODE_BITSIZE.
15836 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15837 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15838 (aarch64_expand_vec_perm_const_1): Handle polynomial
15839 d->perm.length () and d->perm elements.
15840 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15841 Apply to_constant to d->perm elements.
15842 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15843 polynomial CONST_VECTOR_NUNITS.
15844 (aarch64_move_pointer): Take amount as a poly_int64 rather
15846 (aarch64_progress_pointer): Avoid temporary variable.
15847 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15848 the mode attribute instead of GET_MODE.
15850 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15851 Alan Hayward <alan.hayward@arm.com>
15852 David Sherwood <david.sherwood@arm.com>
15854 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15855 x exists before using it.
15856 (aarch64_add_constant_internal): Rename to...
15857 (aarch64_add_offset_1): ...this. Replace regnum with separate
15858 src and dest rtxes. Handle the case in which they're different,
15859 including when the offset is zero. Replace scratchreg with an rtx.
15860 Use 2 additions if there is no spare register into which we can
15861 move a 16-bit constant.
15862 (aarch64_add_constant): Delete.
15863 (aarch64_add_offset): Replace reg with separate src and dest
15864 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15865 Use aarch64_add_offset_1.
15866 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15867 an rtx rather than an int. Take the delta as a poly_int64
15868 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15869 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15870 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15871 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15872 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15873 and aarch64_add_sp.
15874 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15875 aarch64_add_constant.
15877 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15879 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15880 Use scalar_float_mode.
15882 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15884 * config/aarch64/aarch64-simd.md
15885 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15886 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15887 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15888 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15889 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15890 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15891 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15892 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15893 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15894 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15896 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15899 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15900 targ_options->x_arm_arch_string is non NULL.
15902 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15904 * config/aarch64/aarch64.h
15905 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15907 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15910 * expmed.c (emit_store_flag_force): Swap if const op0
15911 and change VOIDmode to mode of op0.
15913 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15915 PR rtl-optimization/83761
15916 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15917 than bytes to mode_for_size.
15919 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15921 PR middle-end/83189
15922 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15923 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15926 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15928 PR middle-end/83575
15929 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15930 when in layout mode.
15931 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15932 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15935 2018-01-10 Michael Collison <michael.collison@arm.com>
15937 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15938 * config/aarch64/aarch64-option-extension.def: Add
15939 AARCH64_OPT_EXTENSION of 'fp16fml'.
15940 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15941 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15942 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15943 * config/aarch64/constraints.md (Ui7): New constraint.
15944 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15945 (VFMLA_SEL_W): Ditto.
15948 (VFMLA16_LOW): New int iterator.
15949 (VFMLA16_HIGH): Ditto.
15950 (UNSPEC_FMLAL): New unspec.
15951 (UNSPEC_FMLSL): Ditto.
15952 (UNSPEC_FMLAL2): Ditto.
15953 (UNSPEC_FMLSL2): Ditto.
15954 (f16mac): New code attribute.
15955 * config/aarch64/aarch64-simd-builtins.def
15956 (aarch64_fmlal_lowv2sf): Ditto.
15957 (aarch64_fmlsl_lowv2sf): Ditto.
15958 (aarch64_fmlalq_lowv4sf): Ditto.
15959 (aarch64_fmlslq_lowv4sf): Ditto.
15960 (aarch64_fmlal_highv2sf): Ditto.
15961 (aarch64_fmlsl_highv2sf): Ditto.
15962 (aarch64_fmlalq_highv4sf): Ditto.
15963 (aarch64_fmlslq_highv4sf): Ditto.
15964 (aarch64_fmlal_lane_lowv2sf): Ditto.
15965 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15966 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15967 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15968 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15969 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15970 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15971 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15972 (aarch64_fmlal_lane_highv2sf): Ditto.
15973 (aarch64_fmlsl_lane_highv2sf): Ditto.
15974 (aarch64_fmlal_laneq_highv2sf): Ditto.
15975 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15976 (aarch64_fmlalq_lane_highv4sf): Ditto.
15977 (aarch64_fmlsl_lane_highv4sf): Ditto.
15978 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15979 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15980 * config/aarch64/aarch64-simd.md:
15981 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15982 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15983 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15984 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15985 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15986 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15987 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15988 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15989 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15990 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15991 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15992 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15993 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15994 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15995 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15996 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15997 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15998 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15999 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16000 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16001 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16002 (vfmlsl_low_u32): Ditto.
16003 (vfmlalq_low_u32): Ditto.
16004 (vfmlslq_low_u32): Ditto.
16005 (vfmlal_high_u32): Ditto.
16006 (vfmlsl_high_u32): Ditto.
16007 (vfmlalq_high_u32): Ditto.
16008 (vfmlslq_high_u32): Ditto.
16009 (vfmlal_lane_low_u32): Ditto.
16010 (vfmlsl_lane_low_u32): Ditto.
16011 (vfmlal_laneq_low_u32): Ditto.
16012 (vfmlsl_laneq_low_u32): Ditto.
16013 (vfmlalq_lane_low_u32): Ditto.
16014 (vfmlslq_lane_low_u32): Ditto.
16015 (vfmlalq_laneq_low_u32): Ditto.
16016 (vfmlslq_laneq_low_u32): Ditto.
16017 (vfmlal_lane_high_u32): Ditto.
16018 (vfmlsl_lane_high_u32): Ditto.
16019 (vfmlal_laneq_high_u32): Ditto.
16020 (vfmlsl_laneq_high_u32): Ditto.
16021 (vfmlalq_lane_high_u32): Ditto.
16022 (vfmlslq_lane_high_u32): Ditto.
16023 (vfmlalq_laneq_high_u32): Ditto.
16024 (vfmlslq_laneq_high_u32): Ditto.
16025 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16026 (AARCH64_FL_FOR_ARCH8_4): New.
16027 (AARCH64_ISA_F16FML): New ISA flag.
16028 (TARGET_F16FML): New feature flag for fp16fml.
16029 (doc/invoke.texi): Document new fp16fml option.
16031 2018-01-10 Michael Collison <michael.collison@arm.com>
16033 * config/aarch64/aarch64-builtins.c:
16034 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16035 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16036 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16037 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16038 (AARCH64_ISA_SHA3): New ISA flag.
16039 (TARGET_SHA3): New feature flag for sha3.
16040 * config/aarch64/iterators.md (sha512_op): New int attribute.
16041 (CRYPTO_SHA512): New int iterator.
16042 (UNSPEC_SHA512H): New unspec.
16043 (UNSPEC_SHA512H2): Ditto.
16044 (UNSPEC_SHA512SU0): Ditto.
16045 (UNSPEC_SHA512SU1): Ditto.
16046 * config/aarch64/aarch64-simd-builtins.def
16047 (aarch64_crypto_sha512hqv2di): New builtin.
16048 (aarch64_crypto_sha512h2qv2di): Ditto.
16049 (aarch64_crypto_sha512su0qv2di): Ditto.
16050 (aarch64_crypto_sha512su1qv2di): Ditto.
16051 (aarch64_eor3qv8hi): Ditto.
16052 (aarch64_rax1qv2di): Ditto.
16053 (aarch64_xarqv2di): Ditto.
16054 (aarch64_bcaxqv8hi): Ditto.
16055 * config/aarch64/aarch64-simd.md:
16056 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16057 (aarch64_crypto_sha512su0qv2di): Ditto.
16058 (aarch64_crypto_sha512su1qv2di): Ditto.
16059 (aarch64_eor3qv8hi): Ditto.
16060 (aarch64_rax1qv2di): Ditto.
16061 (aarch64_xarqv2di): Ditto.
16062 (aarch64_bcaxqv8hi): Ditto.
16063 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16064 (vsha512h2q_u64): Ditto.
16065 (vsha512su0q_u64): Ditto.
16066 (vsha512su1q_u64): Ditto.
16067 (veor3q_u16): Ditto.
16068 (vrax1q_u64): Ditto.
16069 (vxarq_u64): Ditto.
16070 (vbcaxq_u16): Ditto.
16071 * config/arm/types.md (crypto_sha512): New type attribute.
16072 (crypto_sha3): Ditto.
16073 (doc/invoke.texi): Document new sha3 option.
16075 2018-01-10 Michael Collison <michael.collison@arm.com>
16077 * config/aarch64/aarch64-builtins.c:
16078 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16079 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16080 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16081 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16082 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16083 (AARCH64_ISA_SM4): New ISA flag.
16084 (TARGET_SM4): New feature flag for sm4.
16085 * config/aarch64/aarch64-simd-builtins.def
16086 (aarch64_sm3ss1qv4si): Ditto.
16087 (aarch64_sm3tt1aq4si): Ditto.
16088 (aarch64_sm3tt1bq4si): Ditto.
16089 (aarch64_sm3tt2aq4si): Ditto.
16090 (aarch64_sm3tt2bq4si): Ditto.
16091 (aarch64_sm3partw1qv4si): Ditto.
16092 (aarch64_sm3partw2qv4si): Ditto.
16093 (aarch64_sm4eqv4si): Ditto.
16094 (aarch64_sm4ekeyqv4si): Ditto.
16095 * config/aarch64/aarch64-simd.md:
16096 (aarch64_sm3ss1qv4si): Ditto.
16097 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16098 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16099 (aarch64_sm4eqv4si): Ditto.
16100 (aarch64_sm4ekeyqv4si): Ditto.
16101 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16102 (sm3part_op): Ditto.
16103 (CRYPTO_SM3TT): Ditto.
16104 (CRYPTO_SM3PART): Ditto.
16105 (UNSPEC_SM3SS1): New unspec.
16106 (UNSPEC_SM3TT1A): Ditto.
16107 (UNSPEC_SM3TT1B): Ditto.
16108 (UNSPEC_SM3TT2A): Ditto.
16109 (UNSPEC_SM3TT2B): Ditto.
16110 (UNSPEC_SM3PARTW1): Ditto.
16111 (UNSPEC_SM3PARTW2): Ditto.
16112 (UNSPEC_SM4E): Ditto.
16113 (UNSPEC_SM4EKEY): Ditto.
16114 * config/aarch64/constraints.md (Ui2): New constraint.
16115 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16116 * config/arm/types.md (crypto_sm3): New type attribute.
16117 (crypto_sm4): Ditto.
16118 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16119 (vsm3tt1aq_u32): Ditto.
16120 (vsm3tt1bq_u32): Ditto.
16121 (vsm3tt2aq_u32): Ditto.
16122 (vsm3tt2bq_u32): Ditto.
16123 (vsm3partw1q_u32): Ditto.
16124 (vsm3partw2q_u32): Ditto.
16125 (vsm4eq_u32): Ditto.
16126 (vsm4ekeyq_u32): Ditto.
16127 (doc/invoke.texi): Document new sm4 option.
16129 2018-01-10 Michael Collison <michael.collison@arm.com>
16131 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16132 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16133 (AARCH64_FL_FOR_ARCH8_4): New.
16134 (AARCH64_FL_V8_4): New flag.
16135 (doc/invoke.texi): Document new armv8.4-a option.
16137 2018-01-10 Michael Collison <michael.collison@arm.com>
16139 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16140 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16141 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16142 * config/aarch64/aarch64-option-extension.def: Add
16143 AARCH64_OPT_EXTENSION of 'sha2'.
16144 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16145 (crypto): Disable sha2 and aes if crypto disabled.
16146 (crypto): Enable aes and sha2 if enabled.
16147 (simd): Disable sha2 and aes if simd disabled.
16148 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16150 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16151 (TARGET_SHA2): New feature flag for sha2.
16152 (TARGET_AES): New feature flag for aes.
16153 * config/aarch64/aarch64-simd.md:
16154 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16155 conditional on TARGET_AES.
16156 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16157 (aarch64_crypto_sha1hsi): Make pattern conditional
16159 (aarch64_crypto_sha1hv4si): Ditto.
16160 (aarch64_be_crypto_sha1hv4si): Ditto.
16161 (aarch64_crypto_sha1su1v4si): Ditto.
16162 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16163 (aarch64_crypto_sha1su0v4si): Ditto.
16164 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16165 (aarch64_crypto_sha256su0v4si): Ditto.
16166 (aarch64_crypto_sha256su1v4si): Ditto.
16167 (doc/invoke.texi): Document new aes and sha2 options.
16169 2018-01-10 Martin Sebor <msebor@redhat.com>
16171 PR tree-optimization/83781
16172 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16175 2018-01-11 Martin Sebor <msebor@gmail.com>
16176 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16178 PR tree-optimization/83501
16179 PR tree-optimization/81703
16181 * tree-ssa-strlen.c (get_string_cst): Rename...
16182 (get_string_len): ...to this. Handle global constants.
16183 (handle_char_store): Adjust.
16185 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16186 Jim Wilson <jimw@sifive.com>
16188 * config/riscv/riscv-protos.h (riscv_output_return): New.
16189 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16190 (riscv_attribute_table, riscv_output_return),
16191 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16192 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16193 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16195 (riscv_expand_prologue): Add early return for naked function.
16196 (riscv_expand_epilogue): Likewise.
16197 (riscv_function_ok_for_sibcall): Return false for naked function.
16198 (riscv_set_current_function): New.
16199 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16200 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16201 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16202 * doc/extend.texi (RISC-V Function Attributes): New.
16204 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16206 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16207 check for 128-bit long double before checking TCmode.
16208 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16209 128-bit long doubles before checking TFmode or TCmode.
16210 (FLOAT128_IBM_P): Likewise.
16212 2018-01-10 Martin Sebor <msebor@redhat.com>
16214 PR tree-optimization/83671
16215 * builtins.c (c_strlen): Unconditionally return zero for the empty
16217 Use -Warray-bounds for warnings.
16218 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16219 for non-constant array indices with COMPONENT_REF, arrays of
16220 arrays, and pointers to arrays.
16221 (gimple_fold_builtin_strlen): Determine and set length range for
16222 non-constant character arrays.
16224 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16226 PR middle-end/81897
16227 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16230 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16232 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16234 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16237 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16238 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16239 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16240 indexed_or_indirect_operand predicate.
16241 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16242 (*vsx_le_perm_load_v8hi): Likewise.
16243 (*vsx_le_perm_load_v16qi): Likewise.
16244 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16245 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16246 (*vsx_le_perm_store_v8hi): Likewise.
16247 (*vsx_le_perm_store_v16qi): Likewise.
16248 (eight unnamed splitters): Likewise.
16250 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16252 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16253 * config/rs6000/emmintrin.h: Likewise.
16254 * config/rs6000/mmintrin.h: Likewise.
16255 * config/rs6000/xmmintrin.h: Likewise.
16257 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16260 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16262 * tree.c (tree_nop_conversion): Return true for location wrapper
16264 (maybe_wrap_with_location): New function.
16265 (selftest::check_strip_nops): New function.
16266 (selftest::test_location_wrappers): New function.
16267 (selftest::tree_c_tests): Call it.
16268 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16269 (maybe_wrap_with_location): New decl.
16270 (EXPR_LOCATION_WRAPPER_P): New macro.
16271 (location_wrapper_p): New inline function.
16272 (tree_strip_any_location_wrapper): New inline function.
16274 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16277 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16278 stack_realign_offset for the largest alignment of stack slot
16280 (ix86_find_max_used_stack_alignment): New function.
16281 (ix86_finalize_stack_frame_flags): Use it. Set
16282 max_used_stack_alignment if we don't realign stack.
16283 * config/i386/i386.h (machine_function): Add
16284 max_used_stack_alignment.
16286 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16288 * config/arm/arm.opt (-mbranch-cost): New option.
16289 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16292 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16295 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16296 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16298 2018-01-10 Richard Biener <rguenther@suse.de>
16301 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16302 early out so it also covers the case where we have a non-NULL
16305 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16307 PR tree-optimization/83753
16308 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16309 for non-strided grouped accesses if the number of elements is 1.
16311 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16314 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16315 * i386.h (TARGET_USE_GATHER): Define.
16316 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16318 2018-01-10 Martin Liska <mliska@suse.cz>
16321 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16322 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16324 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16325 CLEANUP_NO_PARTITIONING is not set.
16327 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16329 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16330 for vectors, as a partial revert of r254296.
16331 * rtl.h (const_vec_p): Delete.
16332 (const_vec_duplicate_p): Don't test for vector CONSTs.
16333 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16334 * expmed.c (make_tree): Likewise.
16337 * common.md (E, F): Use CONSTANT_P instead of checking for
16339 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16340 checking for CONST_VECTOR.
16342 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16344 PR middle-end/83575
16345 * predict.c (force_edge_cold): Handle in more sane way edges
16346 with no prediction.
16348 2018-01-09 Carl Love <cel@us.ibm.com>
16350 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16352 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16353 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16354 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16355 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16356 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16357 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16358 * config/rs6000/rs6000-protos.h: Add extern defition for
16359 rs6000_generate_float2_double_code.
16360 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16362 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16363 (float2_v2df): Add define_expand.
16365 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16368 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16369 op_mode in the force_to_mode call.
16371 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16373 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16374 instead of checking each element individually.
16375 (aarch64_evpc_uzp): Likewise.
16376 (aarch64_evpc_zip): Likewise.
16377 (aarch64_evpc_ext): Likewise.
16378 (aarch64_evpc_rev): Likewise.
16379 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16380 instead of checking each element individually. Return true without
16382 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16383 whether all selected elements come from the same input, instead of
16384 checking each element individually. Remove calls to gen_rtx_REG,
16385 start_sequence and end_sequence and instead assert that no rtl is
16388 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16390 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16391 order of HIGH and CONST checks.
16393 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16395 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16396 if the destination isn't an SSA_NAME.
16398 2018-01-09 Richard Biener <rguenther@suse.de>
16400 PR tree-optimization/83668
16401 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16403 (canonicalize_loop_form): ... here, renamed from ...
16404 (canonicalize_loop_closed_ssa_form): ... this and amended to
16405 swap successor edges for loop exit blocks to make us use
16406 the RPO order we need for initial schedule generation.
16408 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16410 PR tree-optimization/64811
16411 * match.pd: When optimizing comparisons with Inf, avoid
16412 introducing or losing exceptions from comparisons with NaN.
16414 2018-01-09 Martin Liska <mliska@suse.cz>
16417 * asan.c (shadow_mem_size): Add gcc_assert.
16419 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16421 Don't save registers in main().
16424 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16425 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16426 * config/avr/avr.c (avr_set_current_function): Don't error if
16427 naked, OS_task or OS_main are specified at the same time.
16428 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16430 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16432 * common/config/avr/avr-common.c (avr_option_optimization_table):
16433 Switch on -mmain-is-OS_task for optimizing compilations.
16435 2018-01-09 Richard Biener <rguenther@suse.de>
16437 PR tree-optimization/83572
16438 * graphite.c: Include cfganal.h.
16439 (graphite_transform_loops): Connect infinite loops to exit
16440 and remove fake edges at the end.
16442 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16444 * ipa-inline.c (edge_badness): Revert accidental checkin.
16446 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16449 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16450 symbols; not inline clones.
16452 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16455 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16456 hard registers. Formatting fixes.
16458 PR preprocessor/83722
16459 * gcc.c (try_generate_repro): Pass
16460 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16461 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16464 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16465 Kito Cheng <kito.cheng@gmail.com>
16467 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16468 (riscv_leaf_function_p): Delete.
16469 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16471 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16473 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16475 (do_ifelse): New function.
16476 (do_isel): New function.
16477 (do_sub3): New function.
16478 (do_add3): New function.
16479 (do_load_mask_compare): New function.
16480 (do_overlap_load_compare): New function.
16481 (expand_compare_loop): New function.
16482 (expand_block_compare): Call expand_compare_loop() when appropriate.
16483 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16484 option description.
16485 (-mblock-compare-inline-loop-limit): New option.
16487 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16490 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16491 Reverse order of second and third operands in first alternative.
16492 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16493 of first and second elements in UNSPEC_VPERMR vector.
16494 (altivec_expand_vec_perm_le): Likewise.
16496 2018-01-08 Jeff Law <law@redhat.com>
16498 PR rtl-optimizatin/81308
16499 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16500 (process_switch): If group_case_labels makes a change, then set
16502 (pass_convert_switch::execute): If a switch is converted, then
16503 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16505 PR rtl-optimization/81308
16506 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16509 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16511 PR target/83663 - Revert r255946
16512 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16513 generation for cases where splatting a value is not useful.
16514 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16515 across a vec_duplicate and a paradoxical subreg forming a vector
16516 mode to a vec_concat.
16518 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16520 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16521 -march=armv8.3-a variants.
16522 * config/arm/t-multilib: Likewise.
16523 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16525 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16527 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16529 (cceq_ior_compare_complement): Give it a name so I can use it, and
16530 change boolean_or_operator predicate to boolean_operator so it can
16531 be used to generate a crand.
16532 (eqne): New code iterator.
16533 (bd/bd_neg): New code_attrs.
16534 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16535 a single define_insn.
16536 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16537 decrement (bdnzt/bdnzf/bdzt/bdzf).
16538 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16539 with the new names of the branch decrement patterns, and added the
16540 names of the branch decrement conditional patterns.
16542 2018-01-08 Richard Biener <rguenther@suse.de>
16544 PR tree-optimization/83563
16545 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16548 2018-01-08 Richard Biener <rguenther@suse.de>
16550 PR middle-end/83713
16551 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16553 2018-01-08 Richard Biener <rguenther@suse.de>
16555 PR tree-optimization/83685
16556 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16557 references to abnormals.
16559 2018-01-08 Richard Biener <rguenther@suse.de>
16562 * dwarf2out.c (output_indirect_strings): Handle empty
16563 skeleton_debug_str_hash.
16564 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16566 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16568 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16569 (emit_store_direct): Likewise.
16570 (arc_trampoline_adjust_address): Likewise.
16571 (arc_asm_trampoline_template): New function.
16572 (arc_initialize_trampoline): Use asm_trampoline_template.
16573 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16574 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16575 * config/arc/arc.md (flush_icache): Delete pattern.
16577 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16579 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16580 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16583 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16586 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16587 by not USED_FOR_TARGET.
16588 (make_pass_resolve_sw_modes): Likewise.
16590 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16592 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16595 2018-01-08 Richard Biener <rguenther@suse.de>
16597 PR middle-end/83580
16598 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16600 2018-01-08 Richard Biener <rguenther@suse.de>
16602 PR middle-end/83517
16603 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16605 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16607 PR middle-end/81897
16608 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16609 basic blocks with a small number of successors.
16610 (convert_control_dep_chain_into_preds): Improve handling of
16612 (dump_predicates): Split apart into...
16613 (dump_pred_chain): ...here...
16614 (dump_pred_info): ...and here.
16615 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16616 (can_chain_union_be_invalidated_p): Improve check for invalidation
16618 (uninit_uses_cannot_happen): Avoid unnecessary if
16619 convert_control_dep_chain_into_preds yielded nothing.
16621 2018-01-06 Martin Sebor <msebor@redhat.com>
16623 PR tree-optimization/83640
16624 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16625 subtracting negative offset from size.
16626 (builtin_access::overlap): Adjust offset bounds of the access to fall
16627 within the size of the object if possible.
16629 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16631 PR rtl-optimization/83699
16632 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16633 extract_bit_field_as_subreg to cases in which the extracted
16634 value is also a vector.
16636 * lra-constraints.c (process_alt_operands): Test for the equivalence
16637 substitutions when detecting a possible reload cycle.
16639 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16642 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16643 by default if flag_selective_schedling{,2}. Formatting fixes.
16645 PR rtl-optimization/83682
16646 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16647 if it has non-VECTOR_MODE element mode.
16648 (vec_duplicate_p): Likewise.
16650 PR middle-end/83694
16651 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16652 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16654 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16657 * config/i386/i386-builtin.def
16658 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16659 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16660 Require also OPTION_MASK_ISA_AVX512F in addition to
16661 OPTION_MASK_ISA_GFNI.
16662 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16663 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16664 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16665 to OPTION_MASK_ISA_GFNI.
16666 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16667 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16668 OPTION_MASK_ISA_AVX512BW.
16669 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16670 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16671 addition to OPTION_MASK_ISA_GFNI.
16672 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16673 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16674 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16675 to OPTION_MASK_ISA_GFNI.
16676 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16677 a requirement for all ISAs rather than any of them with a few
16679 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16681 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16682 bitmasks to be enabled with 3 exceptions, instead of requiring any
16683 enabled ISA with lots of exceptions.
16684 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16685 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16686 Change avx512bw in isa attribute to avx512f.
16687 * config/i386/sgxintrin.h: Add license boilerplate.
16688 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16689 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16690 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16691 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16693 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16694 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16695 temporarily sse2 rather than sse if not enabled already.
16698 * config/i386/sse.md (VI248_VLBW): Rename to ...
16699 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16700 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16701 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16702 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16703 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16704 mode iterator instead of VI248_VLBW.
16706 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16708 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16709 (record_modified): Skip clobbers; add debug output.
16710 (param_change_prob): Use sreal frequencies.
16712 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16714 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16715 punt for user-aligned variables.
16717 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16719 * tree-chrec.c (chrec_contains_symbols): Return true for
16722 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16725 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16726 of (x|y) == x for BICS pattern.
16728 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16730 PR tree-optimization/83605
16731 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16732 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16735 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16737 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16738 * config/epiphany/rtems.h: New file.
16740 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16741 Uros Bizjak <ubizjak@gmail.com>
16744 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16745 QIreg_operand instead of register_operand predicate.
16746 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16747 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16748 comments instead of -fmitigate[-_]rop.
16750 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16753 * cgraphunit.c (symbol_table::compile): Switch to text_section
16754 before calling assembly_start debug hook.
16755 * run-rtl-passes.c (run_rtl_passes): Likewise.
16758 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16760 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16761 range_int_cst_p rather than !symbolic_range_p before calling
16762 extract_range_from_multiplicative_op_1.
16764 2018-01-04 Jeff Law <law@redhat.com>
16766 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16767 redundant test in assertion.
16769 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16771 * doc/rtl.texi: Document machine_mode wrapper classes.
16773 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16775 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16776 using tree_to_uhwi.
16778 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16780 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16781 the VEC_PERM_EXPR fold to fail.
16783 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16786 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16787 to switched_sections.
16789 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16792 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16793 test for d.testing.
16795 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16798 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16799 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16801 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16804 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16805 is BLKmode and bitpos not zero or mode change is needed.
16807 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16810 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16813 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16816 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16817 instead of MULT rtx. Update all corresponding splitters.
16818 (*saddl_se): Ditto.
16819 (*ssub<modesuffix>): Ditto.
16820 (*ssubl_se): Ditto.
16821 (*cmp_sadd_di): Update split patterns.
16822 (*cmp_sadd_si): Ditto.
16823 (*cmp_sadd_sidi): Ditto.
16824 (*cmp_ssub_di): Ditto.
16825 (*cmp_ssub_si): Ditto.
16826 (*cmp_ssub_sidi): Ditto.
16827 * config/alpha/predicates.md (const23_operand): New predicate.
16828 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16829 Look for ASHIFT, not MULT inner operand.
16830 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16832 2018-01-04 Martin Liska <mliska@suse.cz>
16834 PR gcov-profile/83669
16835 * gcov.c (output_intermediate_file): Add version to intermediate
16837 * doc/gcov.texi: Document new field 'version' in intermediate
16838 file format. Fix location of '-k' option of gcov command.
16840 2018-01-04 Martin Liska <mliska@suse.cz>
16843 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16845 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16847 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16849 2018-01-03 Martin Sebor <msebor@redhat.com>
16851 PR tree-optimization/83655
16852 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16853 checking calls with invalid arguments.
16855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16857 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16858 (vectorizable_mask_load_store): Delete.
16859 (vectorizable_call): Return false for masked loads and stores.
16860 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16861 instead of gimple_assign_rhs1.
16862 (vectorizable_load): Handle IFN_MASK_LOAD.
16863 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16867 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16869 (vectorizable_mask_load_store): ...here.
16870 (vectorizable_load): ...and here.
16872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16874 * tree-vect-stmts.c (vect_build_all_ones_mask)
16875 (vect_build_zero_merge_argument): New functions, split out from...
16876 (vectorizable_load): ...here.
16878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16880 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16882 (vectorizable_mask_load_store): ...here.
16883 (vectorizable_store): ...and here.
16885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16887 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16889 (vectorizable_mask_load_store): ...here.
16891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16893 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16894 (vect_model_store_cost): Take a vec_load_store_type instead of a
16896 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16897 (vect_model_store_cost): Take a vec_load_store_type instead of a
16899 (vectorizable_mask_load_store): Update accordingly.
16900 (vectorizable_store): Likewise.
16901 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16905 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16906 IFN_MASK_LOAD calls here rather than...
16907 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16910 Alan Hayward <alan.hayward@arm.com>
16911 David Sherwood <david.sherwood@arm.com>
16913 * expmed.c (extract_bit_field_1): For vector extracts,
16914 fall back to extract_bit_field_as_subreg if vec_extract
16917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16918 Alan Hayward <alan.hayward@arm.com>
16919 David Sherwood <david.sherwood@arm.com>
16921 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16922 they are variable or constant sized.
16923 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16924 slots for constant-sized data.
16926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16927 Alan Hayward <alan.hayward@arm.com>
16928 David Sherwood <david.sherwood@arm.com>
16930 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16931 handling COND_EXPRs with boolean comparisons, try to find a better
16932 basis for the mask type than the boolean itself.
16934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16936 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16937 is calculated and how it can be overridden.
16938 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16939 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16941 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16945 Alan Hayward <alan.hayward@arm.com>
16946 David Sherwood <david.sherwood@arm.com>
16948 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16949 Remove the mode argument.
16950 (aarch64_simd_valid_immediate): Remove the mode and inverse
16952 * config/aarch64/iterators.md (bitsize): New iterator.
16953 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16954 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16955 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16956 aarch64_simd_valid_immediate.
16957 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16958 (aarch64_reg_or_bic_imm): Likewise.
16959 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16960 with an insn_type enum and msl with a modifier_type enum.
16961 Replace element_width with a scalar_mode. Change the shift
16962 to unsigned int. Add constructors for scalar_float_mode and
16963 scalar_int_mode elements.
16964 (aarch64_vect_float_const_representable_p): Delete.
16965 (aarch64_can_const_movi_rtx_p)
16966 (aarch64_simd_scalar_immediate_valid_for_move)
16967 (aarch64_simd_make_constant): Update call to
16968 aarch64_simd_valid_immediate.
16969 (aarch64_advsimd_valid_immediate_hs): New function.
16970 (aarch64_advsimd_valid_immediate): Likewise.
16971 (aarch64_simd_valid_immediate): Remove mode and inverse
16972 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16973 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16974 and aarch64_float_const_representable_p on the result.
16975 (aarch64_output_simd_mov_immediate): Remove mode argument.
16976 Update call to aarch64_simd_valid_immediate and use of
16977 simd_immediate_info.
16978 (aarch64_output_scalar_simd_mov_immediate): Update call
16981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16982 Alan Hayward <alan.hayward@arm.com>
16983 David Sherwood <david.sherwood@arm.com>
16985 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16986 (mode_nunits): Likewise CONST_MODE_NUNITS.
16987 * machmode.def (ADJUST_NUNITS): Document.
16988 * genmodes.c (mode_data::need_nunits_adj): New field.
16989 (blank_mode): Update accordingly.
16990 (adj_nunits): New variable.
16991 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16993 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16994 listed in adj_nunits.
16995 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16996 listed in adj_nunits. Don't emit case statements for such modes.
16997 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16998 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16999 nothing if adj_nunits is nonnull.
17000 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17001 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17002 (emit_mode_fbit): Update use of print_maybe_const_decl.
17003 (emit_move_size): Likewise. Treat the array as non-const
17005 (emit_mode_adjustments): Handle adj_nunits.
17007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17009 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17010 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17011 (VECTOR_MODES): Use it.
17012 (make_vector_modes): Take the prefix as an argument.
17014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17015 Alan Hayward <alan.hayward@arm.com>
17016 David Sherwood <david.sherwood@arm.com>
17018 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17019 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17020 for MODE_VECTOR_BOOL.
17021 * machmode.def (VECTOR_BOOL_MODE): Document.
17022 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17023 (make_vector_bool_mode): New function.
17024 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17026 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17027 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17029 * stor-layout.c (int_mode_for_mode): Likewise.
17030 * tree.c (build_vector_type_for_mode): Likewise.
17031 * varasm.c (output_constant_pool_2): Likewise.
17032 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17033 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17034 for MODE_VECTOR_BOOL.
17035 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17036 of mode class checks.
17037 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17038 instead of a list of mode class checks.
17039 (expand_vector_scalar_condition): Likewise.
17040 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17043 Alan Hayward <alan.hayward@arm.com>
17044 David Sherwood <david.sherwood@arm.com>
17046 * machmode.h (mode_size): Change from unsigned short to
17048 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17049 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17050 or if measurement_type is not polynomial.
17051 (fixed_size_mode::includes_p): Check for constant-sized modes.
17052 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17053 return a poly_uint16 rather than an unsigned short.
17054 (emit_mode_size): Change the type of mode_size from unsigned short
17055 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17056 (emit_mode_adjustments): Cope with polynomial vector sizes.
17057 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17059 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17061 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17062 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17063 * caller-save.c (setup_save_areas): Likewise.
17064 (replace_reg_with_saved_mem): Likewise.
17065 * calls.c (emit_library_call_value_1): Likewise.
17066 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17067 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17068 (gen_lowpart_for_combine): Likewise.
17069 * convert.c (convert_to_integer_1): Likewise.
17070 * cse.c (equiv_constant, cse_insn): Likewise.
17071 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17072 (cselib_subst_to_values): Likewise.
17073 * dce.c (word_dce_process_block): Likewise.
17074 * df-problems.c (df_word_lr_mark_ref): Likewise.
17075 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17076 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17077 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17078 (rtl_for_decl_location): Likewise.
17079 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17080 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17081 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17082 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17083 (expand_expr_real_1): Likewise.
17084 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17085 (pad_below): Likewise.
17086 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17087 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17088 * ira.c (get_subreg_tracking_sizes): Likewise.
17089 * ira-build.c (ira_create_allocno_objects): Likewise.
17090 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17091 (ira_sort_regnos_for_alter_reg): Likewise.
17092 * ira-costs.c (record_operand_costs): Likewise.
17093 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17094 (resolve_simple_move): Likewise.
17095 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17096 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17097 (lra_constraints): Likewise.
17098 (CONST_POOL_OK_P): Reject variable-sized modes.
17099 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17100 (add_pseudo_to_slot, lra_spill): Likewise.
17101 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17102 * optabs-query.c (get_best_extraction_insn): Likewise.
17103 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17104 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17105 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17106 * recog.c (offsettable_address_addr_space_p): Likewise.
17107 * regcprop.c (maybe_mode_change): Likewise.
17108 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17109 * regrename.c (build_def_use): Likewise.
17110 * regstat.c (dump_reg_info): Likewise.
17111 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17112 (find_reloads, find_reloads_subreg_address): Likewise.
17113 * reload1.c (eliminate_regs_1): Likewise.
17114 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17115 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17116 (simplify_binary_operation_1, simplify_subreg): Likewise.
17117 * targhooks.c (default_function_arg_padding): Likewise.
17118 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17119 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17120 (verify_gimple_assign_ternary): Likewise.
17121 * tree-inline.c (estimate_move_cost): Likewise.
17122 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17123 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17124 (get_address_cost_ainc): Likewise.
17125 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17126 (vect_supportable_dr_alignment): Likewise.
17127 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17128 (vectorizable_reduction): Likewise.
17129 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17130 (vectorizable_operation, vectorizable_load): Likewise.
17131 * tree.c (build_same_sized_truth_vector_type): Likewise.
17132 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17133 * var-tracking.c (emit_note_insn_var_location): Likewise.
17134 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17135 (ADDR_VEC_ALIGN): Likewise.
17137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17138 Alan Hayward <alan.hayward@arm.com>
17139 David Sherwood <david.sherwood@arm.com>
17141 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17143 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17144 or if measurement_type is polynomial.
17145 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17146 * combine.c (make_extraction): Likewise.
17147 * dse.c (find_shift_sequence): Likewise.
17148 * dwarf2out.c (mem_loc_descriptor): Likewise.
17149 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17150 (extract_bit_field, extract_low_bits): Likewise.
17151 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17152 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17153 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17154 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17155 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17156 * reload.c (find_reloads): Likewise.
17157 * reload1.c (alter_reg): Likewise.
17158 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17159 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17160 * tree-if-conv.c (predicate_mem_writes): Likewise.
17161 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17162 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17163 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17164 * valtrack.c (dead_debug_insert_temp): Likewise.
17165 * varasm.c (mergeable_constant_section): Likewise.
17166 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17169 Alan Hayward <alan.hayward@arm.com>
17170 David Sherwood <david.sherwood@arm.com>
17172 * expr.c (expand_assignment): Cope with polynomial mode sizes
17173 when assigning to a CONCAT.
17175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17176 Alan Hayward <alan.hayward@arm.com>
17177 David Sherwood <david.sherwood@arm.com>
17179 * machmode.h (mode_precision): Change from unsigned short to
17181 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17183 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17184 or if measurement_type is not polynomial.
17185 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17186 in which the mode is already known to be a scalar_int_mode.
17187 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17188 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17190 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17191 for GET_MODE_PRECISION.
17192 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17193 for GET_MODE_PRECISION.
17194 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17196 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17197 (expand_field_assignment, make_extraction): Likewise.
17198 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17199 (get_last_value): Likewise.
17200 * convert.c (convert_to_integer_1): Likewise.
17201 * cse.c (cse_insn): Likewise.
17202 * expr.c (expand_expr_real_1): Likewise.
17203 * lra-constraints.c (simplify_operand_subreg): Likewise.
17204 * optabs-query.c (can_atomic_load_p): Likewise.
17205 * optabs.c (expand_atomic_load): Likewise.
17206 (expand_atomic_store): Likewise.
17207 * ree.c (combine_reaching_defs): Likewise.
17208 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17209 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17210 * tree.h (type_has_mode_precision_p): Likewise.
17211 * ubsan.c (instrument_si_overflow): Likewise.
17213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17214 Alan Hayward <alan.hayward@arm.com>
17215 David Sherwood <david.sherwood@arm.com>
17217 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17218 polynomial numbers of units.
17219 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17220 (valid_vector_subparts_p): New function.
17221 (build_vector_type): Remove temporary shim and take the number
17222 of units as a poly_uint64 rather than an int.
17223 (build_opaque_vector_type): Take the number of units as a
17224 poly_uint64 rather than an int.
17225 * tree.c (build_vector_from_ctor): Handle polynomial
17226 TYPE_VECTOR_SUBPARTS.
17227 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17228 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17229 (build_vector_from_val): If the number of units is variable,
17230 use build_vec_duplicate_cst for constant operands and
17231 VEC_DUPLICATE_EXPR otherwise.
17232 (make_vector_type): Remove temporary is_constant ().
17233 (build_vector_type, build_opaque_vector_type): Take the number of
17234 units as a poly_uint64 rather than an int.
17235 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17237 * cfgexpand.c (expand_debug_expr): Likewise.
17238 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17239 (store_constructor, expand_expr_real_1): Likewise.
17240 (const_scalar_mask_from_tree): Likewise.
17241 * fold-const-call.c (fold_const_reduction): Likewise.
17242 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17243 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17244 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17245 (fold_relational_const): Likewise.
17246 (native_interpret_vector): Likewise. Change the size from an
17247 int to an unsigned int.
17248 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17249 TYPE_VECTOR_SUBPARTS.
17250 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17251 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17252 duplicating a non-constant operand into a variable-length vector.
17253 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17254 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17255 * ipa-icf.c (sem_variable::equals): Likewise.
17256 * match.pd: Likewise.
17257 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17258 * print-tree.c (print_node): Likewise.
17259 * stor-layout.c (layout_type): Likewise.
17260 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17261 * tree-cfg.c (verify_gimple_comparison): Likewise.
17262 (verify_gimple_assign_binary): Likewise.
17263 (verify_gimple_assign_ternary): Likewise.
17264 (verify_gimple_assign_single): Likewise.
17265 * tree-pretty-print.c (dump_generic_node): Likewise.
17266 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17267 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17268 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17269 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17270 (vect_shift_permute_load_chain): Likewise.
17271 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17272 (expand_vector_condition, optimize_vector_constructor): Likewise.
17273 (lower_vec_perm, get_compute_type): Likewise.
17274 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17275 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17276 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17277 (vect_recog_mask_conversion_pattern): Likewise.
17278 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17279 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17280 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17281 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17282 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17283 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17284 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17285 (supportable_widening_operation): Likewise.
17286 (supportable_narrowing_operation): Likewise.
17287 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17289 * varasm.c (output_constant): Likewise.
17291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17292 Alan Hayward <alan.hayward@arm.com>
17293 David Sherwood <david.sherwood@arm.com>
17295 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17296 so that both the length == 3 and length != 3 cases set up their
17297 own permute vectors. Add comments explaining why we know the
17298 number of elements is constant.
17299 (vect_permute_load_chain): Likewise.
17301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17302 Alan Hayward <alan.hayward@arm.com>
17303 David Sherwood <david.sherwood@arm.com>
17305 * machmode.h (mode_nunits): Change from unsigned char to
17307 (ONLY_FIXED_SIZE_MODES): New macro.
17308 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17309 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17310 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17312 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17313 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17314 or if measurement_type is not polynomial.
17315 * genmodes.c (ZERO_COEFFS): New macro.
17316 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17318 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17319 Use ZERO_COEFFS when emitting initializers.
17320 * data-streamer.h (bp_pack_poly_value): New function.
17321 (bp_unpack_poly_value): Likewise.
17322 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17323 for GET_MODE_NUNITS.
17324 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17325 for GET_MODE_NUNITS.
17326 * tree.c (make_vector_type): Remove temporary shim and make
17327 the real function take the number of units as a poly_uint64
17328 rather than an int.
17329 (build_vector_type_for_mode): Handle polynomial nunits.
17330 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17331 * emit-rtl.c (const_vec_series_p_1): Likewise.
17332 (gen_rtx_CONST_VECTOR): Likewise.
17333 * fold-const.c (test_vec_duplicate_folding): Likewise.
17334 * genrecog.c (validate_pattern): Likewise.
17335 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17336 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17337 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17338 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17339 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17340 * rtlanal.c (subreg_get_info): Likewise.
17341 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17342 (vect_grouped_load_supported): Likewise.
17343 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17344 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17345 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17346 (simplify_const_unary_operation, simplify_binary_operation_1)
17347 (simplify_const_binary_operation, simplify_ternary_operation)
17348 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17349 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17350 instead of CONST_VECTOR_NUNITS.
17351 * varasm.c (output_constant_pool_2): Likewise.
17352 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17353 explicit-encoded elements in the XVEC for variable-length vectors.
17355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17357 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17360 Alan Hayward <alan.hayward@arm.com>
17361 David Sherwood <david.sherwood@arm.com>
17363 * coretypes.h (fixed_size_mode): Declare.
17364 (fixed_size_mode_pod): New typedef.
17365 * builtins.h (target_builtins::x_apply_args_mode)
17366 (target_builtins::x_apply_result_mode): Change type to
17367 fixed_size_mode_pod.
17368 * builtins.c (apply_args_size, apply_result_size, result_vector)
17369 (expand_builtin_apply_args_1, expand_builtin_apply)
17370 (expand_builtin_return): Update accordingly.
17372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17374 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17375 * cselib.c (cselib_hash_rtx): Likewise.
17376 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17377 CONST_VECTOR encoding.
17379 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17380 Jeff Law <law@redhat.com>
17383 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17384 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17385 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17386 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17389 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17390 explicitly probe *sp in a noreturn function if there were any callee
17391 register saves or frame pointer is needed.
17393 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17396 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17397 BLKmode for ternary, binary or unary expressions.
17400 * var-tracking.c (delete_vta_debug_insn): New inline function.
17401 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17402 insns from get_insns () to NULL instead of each bb separately.
17403 Use delete_vta_debug_insn. No longer static.
17404 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17405 delete_vta_debug_insns callers.
17406 * rtl.h (delete_vta_debug_insns): Declare.
17407 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17408 instead of variable_tracking_main.
17410 2018-01-03 Martin Sebor <msebor@redhat.com>
17412 PR tree-optimization/83603
17413 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17414 arguments past the endof the argument list in functions declared
17415 without a prototype.
17416 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17417 Avoid checking when arguments are null.
17419 2018-01-03 Martin Sebor <msebor@redhat.com>
17422 * doc/extend.texi (attribute const): Fix a typo.
17423 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17424 issuing -Wsuggest-attribute for void functions.
17426 2018-01-03 Martin Sebor <msebor@redhat.com>
17428 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17429 offset_int::from instead of wide_int::to_shwi.
17430 (maybe_diag_overlap): Remove assertion.
17431 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17432 * gimple-ssa-sprintf.c (format_directive): Same.
17433 (parse_directive): Same.
17434 (sprintf_dom_walker::compute_format_length): Same.
17435 (try_substitute_return_value): Same.
17437 2018-01-03 Jeff Law <law@redhat.com>
17439 PR middle-end/83654
17440 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17441 non-constant residual for zero at runtime and avoid probing in
17442 that case. Reorganize code for trailing problem to mirror handling
17445 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17447 PR tree-optimization/83501
17448 * tree-ssa-strlen.c (get_string_cst): New.
17449 (handle_char_store): Call get_string_cst.
17451 2018-01-03 Martin Liska <mliska@suse.cz>
17453 PR tree-optimization/83593
17454 * tree-ssa-strlen.c: Include tree-cfg.h.
17455 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17456 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17457 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17459 (strlen_dom_walker::before_dom_children): Call
17460 gimple_purge_dead_eh_edges. Dump tranformation with details
17462 (strlen_dom_walker::before_dom_children): Update call by adding
17463 new argument cleanup_eh.
17464 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17466 2018-01-03 Martin Liska <mliska@suse.cz>
17469 * cif-code.def (VARIADIC_THUNK): New enum value.
17470 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17473 2018-01-03 Jan Beulich <jbeulich@suse.com>
17475 * sse.md (mov<mode>_internal): Tighten condition for when to use
17476 vmovdqu<ssescalarsize> for TI and OI modes.
17478 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17480 Update copyright years.
17482 2018-01-03 Martin Liska <mliska@suse.cz>
17485 * ipa-visibility.c (function_and_variable_visibility): Skip
17486 functions with noipa attribure.
17488 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17490 * gcc.c (process_command): Update copyright notice dates.
17491 * gcov-dump.c (print_version): Ditto.
17492 * gcov.c (print_version): Ditto.
17493 * gcov-tool.c (print_version): Ditto.
17494 * gengtype.c (create_file): Ditto.
17495 * doc/cpp.texi: Bump @copying's copyright year.
17496 * doc/cppinternals.texi: Ditto.
17497 * doc/gcc.texi: Ditto.
17498 * doc/gccint.texi: Ditto.
17499 * doc/gcov.texi: Ditto.
17500 * doc/install.texi: Ditto.
17501 * doc/invoke.texi: Ditto.
17503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17505 * vector-builder.h (vector_builder::m_full_nelts): Change from
17506 unsigned int to poly_uint64.
17507 (vector_builder::full_nelts): Update prototype accordingly.
17508 (vector_builder::new_vector): Likewise.
17509 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17510 (vector_builder::operator ==): Likewise.
17511 (vector_builder::finalize): Likewise.
17512 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17513 Take the number of elements as a poly_uint64 rather than an
17515 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17516 from unsigned int to poly_uint64.
17517 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17518 (vec_perm_indices::new_vector): Likewise.
17519 (vec_perm_indices::length): Likewise.
17520 (vec_perm_indices::nelts_per_input): Likewise.
17521 (vec_perm_indices::input_nelts): Likewise.
17522 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17523 number of elements per input as a poly_uint64 rather than an
17524 unsigned int. Use the original encoding for variable-length
17525 vectors, rather than clamping each individual element.
17526 For the second and subsequent elements in each pattern,
17527 clamp the step and base before clamping their sum.
17528 (vec_perm_indices::series_p): Handle polynomial element counts.
17529 (vec_perm_indices::all_in_range_p): Likewise.
17530 (vec_perm_indices_to_tree): Likewise.
17531 (vec_perm_indices_to_rtx): Likewise.
17532 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17533 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17534 (tree_vector_builder::new_binary_operation): Handle polynomial
17535 element counts. Return false if we need to know the number
17536 of elements at compile time.
17537 * fold-const.c (fold_vec_perm): Punt if the number of elements
17538 isn't known at compile time.
17540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17542 * vec-perm-indices.h (vec_perm_builder): Change element type
17543 from HOST_WIDE_INT to poly_int64.
17544 (vec_perm_indices::element_type): Update accordingly.
17545 (vec_perm_indices::clamp): Handle polynomial element_types.
17546 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17547 (vec_perm_indices::all_in_range_p): Likewise.
17548 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17550 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17551 polynomial vec_perm_indices element types.
17552 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17553 * fold-const.c (fold_vec_perm): Likewise.
17554 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17555 * tree-vect-generic.c (lower_vec_perm): Likewise.
17556 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17557 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17558 element type to HOST_WIDE_INT.
17560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17561 Alan Hayward <alan.hayward@arm.com>
17562 David Sherwood <david.sherwood@arm.com>
17564 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17565 rather than an int. Use plus_constant.
17566 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17567 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17570 Alan Hayward <alan.hayward@arm.com>
17571 David Sherwood <david.sherwood@arm.com>
17573 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17574 a HOST_WIDE_INT to a poly_int64.
17576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17577 Alan Hayward <alan.hayward@arm.com>
17578 David Sherwood <david.sherwood@arm.com>
17580 * calls.c (load_register_parameters): Cope with polynomial
17581 mode sizes. Require a constant size for BLKmode parameters
17582 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17583 forces a parameter to be padded at the lsb end in order to
17584 fill a complete number of words, require the parameter size
17585 to be ordered wrt UNITS_PER_WORD.
17587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17588 Alan Hayward <alan.hayward@arm.com>
17589 David Sherwood <david.sherwood@arm.com>
17591 * reload1.c (spill_stack_slot_width): Change element type
17592 from unsigned int to poly_uint64_pod.
17593 (alter_reg): Treat mode sizes as polynomial.
17595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17596 Alan Hayward <alan.hayward@arm.com>
17597 David Sherwood <david.sherwood@arm.com>
17599 * reload.c (complex_word_subreg_p): New function.
17600 (reload_inner_reg_of_subreg, push_reload): Use it.
17602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17603 Alan Hayward <alan.hayward@arm.com>
17604 David Sherwood <david.sherwood@arm.com>
17606 * lra-constraints.c (process_alt_operands): Reject matched
17607 operands whose sizes aren't ordered.
17608 (match_reload): Refer to this check here.
17610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17611 Alan Hayward <alan.hayward@arm.com>
17612 David Sherwood <david.sherwood@arm.com>
17614 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17615 that the mode size is in the set {1, 2, 4, 8, 16}.
17617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17618 Alan Hayward <alan.hayward@arm.com>
17619 David Sherwood <david.sherwood@arm.com>
17621 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17622 Use plus_constant instead of gen_rtx_PLUS.
17624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17625 Alan Hayward <alan.hayward@arm.com>
17626 David Sherwood <david.sherwood@arm.com>
17628 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17629 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17630 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17631 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17632 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17633 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17634 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17635 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17636 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17637 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17639 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17640 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17641 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17642 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17643 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17644 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17645 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17646 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17647 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17648 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17650 * expr.c (emit_move_resolve_push): Treat the input and result
17651 of PUSH_ROUNDING as a poly_int64.
17652 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17653 (emit_push_insn): Likewise.
17654 * lra-eliminations.c (mark_not_eliminable): Likewise.
17655 * recog.c (push_operand): Likewise.
17656 * reload1.c (elimination_effects): Likewise.
17657 * rtlanal.c (nonzero_bits1): Likewise.
17658 * calls.c (store_one_arg): Likewise. Require the padding to be
17659 known at compile time.
17661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17662 Alan Hayward <alan.hayward@arm.com>
17663 David Sherwood <david.sherwood@arm.com>
17665 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17666 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17673 rather than an int.
17675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17676 Alan Hayward <alan.hayward@arm.com>
17677 David Sherwood <david.sherwood@arm.com>
17679 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17680 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17681 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
17688 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17689 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17690 as a poly_uint64 too.
17692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17693 Alan Hayward <alan.hayward@arm.com>
17694 David Sherwood <david.sherwood@arm.com>
17696 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
17703 REGMODE_NATURAL_SIZE.
17704 * expmed.c (store_bit_field_1): Likewise.
17705 * expr.c (store_constructor): Likewise.
17706 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17707 and polynomial REGMODE_NATURAL_SIZE.
17708 (gen_lowpart_common): Likewise.
17709 * reginfo.c (record_subregs_of_mode): Likewise.
17710 * rtlanal.c (read_modify_subreg_p): Likewise.
17712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17713 Alan Hayward <alan.hayward@arm.com>
17714 David Sherwood <david.sherwood@arm.com>
17716 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17717 numbers of elements.
17719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17720 Alan Hayward <alan.hayward@arm.com>
17721 David Sherwood <david.sherwood@arm.com>
17723 * match.pd: Cope with polynomial numbers of vector elements.
17725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17726 Alan Hayward <alan.hayward@arm.com>
17727 David Sherwood <david.sherwood@arm.com>
17729 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17730 in a POINTER_PLUS_EXPR.
17732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17733 Alan Hayward <alan.hayward@arm.com>
17734 David Sherwood <david.sherwood@arm.com>
17736 * omp-simd-clone.c (simd_clone_subparts): New function.
17737 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17738 (ipa_simd_modify_function_body): Likewise.
17740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17741 Alan Hayward <alan.hayward@arm.com>
17742 David Sherwood <david.sherwood@arm.com>
17744 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17745 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17746 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17747 (expand_vector_condition, vector_element): Likewise.
17748 (subparts_gt): New function.
17749 (get_compute_type): Use subparts_gt.
17750 (count_type_subparts): Delete.
17751 (expand_vector_operations_1): Use subparts_gt instead of
17752 count_type_subparts.
17754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17755 Alan Hayward <alan.hayward@arm.com>
17756 David Sherwood <david.sherwood@arm.com>
17758 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17759 (vect_compile_time_alias): ...this new function. Do the calculation
17760 on poly_ints rather than trees.
17761 (vect_prune_runtime_alias_test_list): Update call accordingly.
17763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17764 Alan Hayward <alan.hayward@arm.com>
17765 David Sherwood <david.sherwood@arm.com>
17767 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17769 (vect_schedule_slp_instance): Likewise.
17771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17772 Alan Hayward <alan.hayward@arm.com>
17773 David Sherwood <david.sherwood@arm.com>
17775 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17776 constant and extern definitions for variable-length vectors.
17777 (vect_get_constant_vectors): Note that the number of units
17778 is known to be constant.
17780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17781 Alan Hayward <alan.hayward@arm.com>
17782 David Sherwood <david.sherwood@arm.com>
17784 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17785 of units as polynomial. Choose between WIDE and NARROW based
17788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17789 Alan Hayward <alan.hayward@arm.com>
17790 David Sherwood <david.sherwood@arm.com>
17792 * tree-vect-stmts.c (simd_clone_subparts): New function.
17793 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17796 Alan Hayward <alan.hayward@arm.com>
17797 David Sherwood <david.sherwood@arm.com>
17799 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17800 vectors as polynomial. Use build_index_vector for
17801 IFN_GOMP_SIMD_LANE.
17803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17804 Alan Hayward <alan.hayward@arm.com>
17805 David Sherwood <david.sherwood@arm.com>
17807 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17808 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17809 for variable-length vectors.
17810 (vectorizable_mask_load_store): Treat the number of units as
17811 polynomial, asserting that it is constant if the condition has
17812 already been enforced.
17813 (vectorizable_store, vectorizable_load): Likewise.
17815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17816 Alan Hayward <alan.hayward@arm.com>
17817 David Sherwood <david.sherwood@arm.com>
17819 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17820 of units as polynomial. Punt if we can't tell at compile time
17821 which vector contains the final result.
17823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17824 Alan Hayward <alan.hayward@arm.com>
17825 David Sherwood <david.sherwood@arm.com>
17827 * tree-vect-loop.c (vectorizable_induction): Treat the number
17828 of units as polynomial. Punt on SLP inductions. Use an integer
17829 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17830 cast of such a series for variable-length floating-point
17833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17834 Alan Hayward <alan.hayward@arm.com>
17835 David Sherwood <david.sherwood@arm.com>
17837 * tree.h (build_index_vector): Declare.
17838 * tree.c (build_index_vector): New function.
17839 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17840 of units as polynomial, forcibly converting it to a constant if
17841 vectorizable_reduction has already enforced the condition.
17842 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17843 to create a {1,2,3,...} vector.
17844 (vectorizable_reduction): Treat the number of units as polynomial.
17845 Choose vectype_in based on the largest scalar element size rather
17846 than the smallest number of units. Enforce the restrictions
17849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17850 Alan Hayward <alan.hayward@arm.com>
17851 David Sherwood <david.sherwood@arm.com>
17853 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17854 number of units as polynomial.
17856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17857 Alan Hayward <alan.hayward@arm.com>
17858 David Sherwood <david.sherwood@arm.com>
17860 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17861 * target.def (autovectorize_vector_sizes): Return the vector sizes
17862 by pointer, using vector_sizes rather than a bitmask.
17863 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17864 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17865 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17867 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17868 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17869 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17870 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17871 * omp-general.c (omp_max_vf): Likewise.
17872 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17873 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17874 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17875 * tree-vect-slp.c (vect_slp_bb): Likewise.
17876 * doc/tm.texi: Regenerate.
17877 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17879 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17880 the vector size as a poly_uint64 rather than an unsigned int.
17881 (current_vector_size): Change from an unsigned int to a poly_uint64.
17882 (get_vectype_for_scalar_type): Update accordingly.
17883 * tree.h (build_truth_vector_type): Take the size and number of
17884 units as a poly_uint64 rather than an unsigned int.
17885 (build_vector_type): Add a temporary overload that takes
17886 the number of units as a poly_uint64 rather than an unsigned int.
17887 * tree.c (make_vector_type): Likewise.
17888 (build_truth_vector_type): Take the number of units as a poly_uint64
17889 rather than an unsigned int.
17891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17892 Alan Hayward <alan.hayward@arm.com>
17893 David Sherwood <david.sherwood@arm.com>
17895 * target.def (get_mask_mode): Take the number of units and length
17896 as poly_uint64s rather than unsigned ints.
17897 * targhooks.h (default_get_mask_mode): Update accordingly.
17898 * targhooks.c (default_get_mask_mode): Likewise.
17899 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17900 * doc/tm.texi: Regenerate.
17902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17903 Alan Hayward <alan.hayward@arm.com>
17904 David Sherwood <david.sherwood@arm.com>
17906 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17907 * omp-general.c (omp_max_vf): Likewise.
17908 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17909 (expand_omp_simd): Handle polynomial safelen.
17910 * omp-low.c (omplow_simd_context): Add a default constructor.
17911 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17912 (lower_rec_simd_input_clauses): Update accordingly.
17913 (lower_rec_input_clauses): Likewise.
17915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17916 Alan Hayward <alan.hayward@arm.com>
17917 David Sherwood <david.sherwood@arm.com>
17919 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17920 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17921 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17922 (vect_analyze_slp_cost): Likewise.
17923 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17924 (vect_model_load_cost): Likewise.
17926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17927 Alan Hayward <alan.hayward@arm.com>
17928 David Sherwood <david.sherwood@arm.com>
17930 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17931 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17932 from an unsigned int * to a poly_uint64_pod *.
17933 (calculate_unrolling_factor): New function.
17934 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17937 Alan Hayward <alan.hayward@arm.com>
17938 David Sherwood <david.sherwood@arm.com>
17940 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17941 from an unsigned int to a poly_uint64.
17942 (_loop_vec_info::slp_unrolling_factor): Likewise.
17943 (_loop_vec_info::vectorization_factor): Change from an int
17945 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17946 (vect_get_num_vectors): New function.
17947 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17948 (vect_get_num_copies): Use vect_get_num_vectors.
17949 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17950 to an unsigned int *.
17951 (vect_analyze_data_refs): Change min_vf from an int * to a
17953 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17954 than an unsigned HOST_WIDE_INT.
17955 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17956 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17957 to an unsigned int *.
17958 (vect_analyze_data_ref_dependences): Likewise.
17959 (vect_compute_data_ref_alignment): Handle polynomial vf.
17960 (vect_enhance_data_refs_alignment): Likewise.
17961 (vect_prune_runtime_alias_test_list): Likewise.
17962 (vect_shift_permute_load_chain): Likewise.
17963 (vect_supportable_dr_alignment): Likewise.
17964 (dependence_distance_ge_vf): Take the vectorization factor as a
17965 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17966 (vect_analyze_data_refs): Change min_vf from an int * to a
17968 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17969 vfm1 as a poly_uint64 rather than an int. Make the same change
17970 for the returned bound_scalar.
17971 (vect_gen_vector_loop_niters): Handle polynomial vf.
17972 (vect_do_peeling): Likewise. Update call to
17973 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17974 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17976 * tree-vect-loop.c (vect_determine_vectorization_factor)
17977 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17978 (vect_get_known_peeling_cost): Likewise.
17979 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17980 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17981 (vect_transform_loop): Likewise. Use the lowest possible VF when
17982 updating the upper bounds of the loop.
17983 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17984 rather than an int.
17985 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17986 polynomial unroll factors.
17987 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17988 (vect_make_slp_decision): Likewise.
17989 (vect_supported_load_permutation_p): Likewise, and polynomial
17991 (vect_analyze_slp_cost): Handle polynomial vf.
17992 (vect_slp_analyze_node_operations): Likewise.
17993 (vect_slp_analyze_bb_1): Likewise.
17994 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17995 than an unsigned HOST_WIDE_INT.
17996 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17997 (vectorizable_load): Handle polynomial vf.
17998 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18000 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18003 Alan Hayward <alan.hayward@arm.com>
18004 David Sherwood <david.sherwood@arm.com>
18006 * match.pd: Handle bit operations involving three constants
18007 and try to fold one pair.
18009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18011 * tree-vect-loop-manip.c: Include gimple-fold.h.
18012 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18013 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18014 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18015 Add a path that uses a step of VF instead of 1, but disable it
18017 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18018 and niters_no_overflow parameters. Update calls to
18019 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18020 Create a new SSA name if the latter choses to use a ste other
18021 than zero, and return it via niters_vector_mult_vf_var.
18022 * tree-vect-loop.c (vect_transform_loop): Update calls to
18023 vect_do_peeling, vect_gen_vector_loop_niters and
18024 slpeel_make_loop_iterate_ntimes.
18025 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18026 (vect_gen_vector_loop_niters): Update declarations after above changes.
18028 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18030 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18031 128-bit round to integer instructions.
18032 (ceil<mode>2): Likewise.
18033 (btrunc<mode>2): Likewise.
18034 (round<mode>2): Likewise.
18036 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18038 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18039 unaligned VSX load/store on P8/P9.
18040 (expand_block_clear): Allow the use of unaligned VSX
18041 load/store on P8/P9.
18043 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18045 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18047 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18048 swap associated with both a load and a store.
18050 2018-01-02 Andrew Waterman <andrew@sifive.com>
18052 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18053 * config/riscv/riscv.md (clear_cache): Use it.
18055 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18057 * web.c: Remove out-of-date comment.
18059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18061 * expr.c (fixup_args_size_notes): Check that any existing
18062 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18063 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18064 (emit_single_push_insn): ...here.
18066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18068 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18069 (const_vector_encoded_nelts): New function.
18070 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18071 (const_vector_int_elt, const_vector_elt): Declare.
18072 * emit-rtl.c (const_vector_int_elt_1): New function.
18073 (const_vector_elt): Likewise.
18074 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18075 of CONST_VECTOR_ELT.
18077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18079 * expr.c: Include rtx-vector-builder.h.
18080 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18081 directly on the tree encoding.
18082 (const_vector_from_tree): Likewise.
18083 * optabs.c: Include rtx-vector-builder.h.
18084 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18085 sequence of "u" values.
18086 * vec-perm-indices.c: Include rtx-vector-builder.h.
18087 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18088 directly on the vec_perm_indices encoding.
18090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18092 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18093 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18094 * rtx-vector-builder.h: New file.
18095 * rtx-vector-builder.c: Likewise.
18096 * rtl.h (rtx_def::u2): Add a const_vector field.
18097 (CONST_VECTOR_NPATTERNS): New macro.
18098 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18099 (CONST_VECTOR_DUPLICATE_P): Likewise.
18100 (CONST_VECTOR_STEPPED_P): Likewise.
18101 (CONST_VECTOR_ENCODED_ELT): Likewise.
18102 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18103 (unwrap_const_vec_duplicate): Likewise.
18104 (const_vec_series_p): Check for a non-duplicated vector encoding.
18105 Say that the function only returns true for integer vectors.
18106 * emit-rtl.c: Include rtx-vector-builder.h.
18107 (gen_const_vec_duplicate_1): Delete.
18108 (gen_const_vector): Call gen_const_vec_duplicate instead of
18109 gen_const_vec_duplicate_1.
18110 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18111 (gen_const_vec_duplicate): Use rtx_vector_builder.
18112 (gen_const_vec_series): Likewise.
18113 (gen_rtx_CONST_VECTOR): Likewise.
18114 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18115 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18116 Build a new vector rather than modifying a CONST_VECTOR in-place.
18117 (handle_special_swappables): Update call accordingly.
18118 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18119 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18120 Build a new vector rather than modifying a CONST_VECTOR in-place.
18121 (handle_special_swappables): Update call accordingly.
18123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18125 * simplify-rtx.c (simplify_const_binary_operation): Use
18126 CONST_VECTOR_ELT instead of XVECEXP.
18128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18130 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18131 the selector elements to be different from the data elements
18132 if the selector is a VECTOR_CST.
18133 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18134 ssizetype for the selector.
18136 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18138 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18139 before testing each element individually.
18140 * tree-vect-generic.c (lower_vec_perm): Likewise.
18142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18144 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18145 * selftest-run-tests.c (selftest::run_tests): Call it.
18146 * vector-builder.h (vector_builder::operator ==): New function.
18147 (vector_builder::operator !=): Likewise.
18148 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18149 (vec_perm_indices::all_from_input_p): New function.
18150 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18151 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18152 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18153 instead of reading the VECTOR_CST directly. Detect whether both
18154 vector inputs are the same before constructing the vec_perm_indices,
18155 and update the number of inputs argument accordingly. Use the
18156 utility functions added above. Only construct sel2 if we need to.
18158 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18160 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18161 the broadcast of the low byte.
18162 (expand_mult_highpart): Use an explicit encoding for the permutes.
18163 * optabs-query.c (can_mult_highpart_p): Likewise.
18164 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18165 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18166 (vectorizable_bswap): Likewise.
18167 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18168 explicit encoding for the power-of-2 permutes.
18169 (vect_permute_store_chain): Likewise.
18170 (vect_grouped_load_supported): Likewise.
18171 (vect_permute_load_chain): Likewise.
18173 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18175 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18176 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18177 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18178 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18179 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18180 (vect_gen_perm_mask_any): Likewise.
18182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18184 * int-vector-builder.h: New file.
18185 * vec-perm-indices.h: Include int-vector-builder.h.
18186 (vec_perm_indices): Redefine as an int_vector_builder.
18187 (auto_vec_perm_indices): Delete.
18188 (vec_perm_builder): Redefine as a stand-alone class.
18189 (vec_perm_indices::vec_perm_indices): New function.
18190 (vec_perm_indices::clamp): Likewise.
18191 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18192 (vec_perm_indices::new_vector): New function.
18193 (vec_perm_indices::new_expanded_vector): Update for new
18194 vec_perm_indices class.
18195 (vec_perm_indices::rotate_inputs): New function.
18196 (vec_perm_indices::all_in_range_p): Operate directly on the
18197 encoded form, without computing elided elements.
18198 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18199 encoding. Update for new vec_perm_indices class.
18200 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18201 the given vec_perm_builder.
18202 (expand_vec_perm_var): Update vec_perm_builder constructor.
18203 (expand_mult_highpart): Use vec_perm_builder instead of
18204 auto_vec_perm_indices.
18205 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18206 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18207 or double series encoding as appropriate.
18208 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18209 vec_perm_indices instead of auto_vec_perm_indices.
18210 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18211 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18212 (vect_permute_store_chain): Likewise.
18213 (vect_grouped_load_supported): Likewise.
18214 (vect_permute_load_chain): Likewise.
18215 (vect_shift_permute_load_chain): Likewise.
18216 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18217 (vect_transform_slp_perm_load): Likewise.
18218 (vect_schedule_slp_instance): Likewise.
18219 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18220 (vectorizable_mask_load_store): Likewise.
18221 (vectorizable_bswap): Likewise.
18222 (vectorizable_store): Likewise.
18223 (vectorizable_load): Likewise.
18224 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18225 vec_perm_indices instead of auto_vec_perm_indices. Use
18226 tree_to_vec_perm_builder to read the vector from a tree.
18227 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18228 vec_perm_builder instead of a vec_perm_indices.
18229 (have_whole_vector_shift): Use vec_perm_builder and
18230 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18231 truncation to calc_vec_perm_mask_for_shift.
18232 (vect_create_epilog_for_reduction): Likewise.
18233 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18234 from auto_vec_perm_indices to vec_perm_indices.
18235 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18236 instead of changing individual elements.
18237 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18238 the vector in d.perm.
18239 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18240 from auto_vec_perm_indices to vec_perm_indices.
18241 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18242 instead of changing individual elements.
18243 (arm_vectorize_vec_perm_const): Use new_vector to install
18244 the vector in d.perm.
18245 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18246 Update vec_perm_builder constructor.
18247 (rs6000_expand_interleave): Likewise.
18248 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18249 (rs6000_expand_interleave): Likewise.
18251 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18253 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18254 to qimode could truncate the indices.
18255 * optabs.c (expand_vec_perm_var): Likewise.
18257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18259 * Makefile.in (OBJS): Add vec-perm-indices.o.
18260 * vec-perm-indices.h: New file.
18261 * vec-perm-indices.c: Likewise.
18262 * target.h (vec_perm_indices): Replace with a forward class
18264 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18265 * optabs.h: Include vec-perm-indices.h.
18266 (expand_vec_perm): Delete.
18267 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18268 (expand_vec_perm_const): Declare.
18269 * target.def (vec_perm_const_ok): Replace with...
18270 (vec_perm_const): ...this new hook.
18271 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18272 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18273 * doc/tm.texi: Regenerate.
18274 * optabs.def (vec_perm_const): Delete.
18275 * doc/md.texi (vec_perm_const): Likewise.
18276 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18277 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18278 expand_vec_perm for constant permutation vectors. Assert that
18279 the mode of variable permutation vectors is the integer equivalent
18280 of the mode that is being permuted.
18281 * optabs-query.h (selector_fits_mode_p): Declare.
18282 * optabs-query.c: Include vec-perm-indices.h.
18283 (selector_fits_mode_p): New function.
18284 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18285 is defined, instead of checking whether the vec_perm_const_optab
18286 exists. Use targetm.vectorize.vec_perm_const instead of
18287 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18288 fit in the vector mode before using a variable permute.
18289 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18290 vec_perm_indices instead of an rtx.
18291 (expand_vec_perm): Replace with...
18292 (expand_vec_perm_const): ...this new function. Take the selector
18293 as a vec_perm_indices rather than an rtx. Also take the mode of
18294 the selector. Update call to shift_amt_for_vec_perm_mask.
18295 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18296 Use vec_perm_indices::new_expanded_vector to expand the original
18297 selector into bytes. Check whether the indices fit in the vector
18298 mode before using a variable permute.
18299 (expand_vec_perm_var): Make global.
18300 (expand_mult_highpart): Use expand_vec_perm_const.
18301 * fold-const.c: Includes vec-perm-indices.h.
18302 * tree-ssa-forwprop.c: Likewise.
18303 * tree-vect-data-refs.c: Likewise.
18304 * tree-vect-generic.c: Likewise.
18305 * tree-vect-loop.c: Likewise.
18306 * tree-vect-slp.c: Likewise.
18307 * tree-vect-stmts.c: Likewise.
18308 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18310 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18311 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18312 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18313 (aarch64_vectorize_vec_perm_const): ...this new function.
18314 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18315 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18316 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18317 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18318 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18319 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18320 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18322 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18323 check for NEON modes.
18324 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18325 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18326 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18327 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18329 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18330 the old VEC_PERM_CONST conditions.
18331 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18332 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18333 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18334 (ia64_vectorize_vec_perm_const_ok): Merge into...
18335 (ia64_vectorize_vec_perm_const): ...this new function.
18336 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18337 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18338 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18339 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18340 * config/mips/mips.c (mips_expand_vec_perm_const)
18341 (mips_vectorize_vec_perm_const_ok): Merge into...
18342 (mips_vectorize_vec_perm_const): ...this new function.
18343 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18344 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18345 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18346 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18347 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18348 (rs6000_expand_vec_perm_const): Delete.
18349 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18351 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18352 (altivec_expand_vec_perm_const_le): Take each operand individually.
18353 Operate on constant selectors rather than rtxes.
18354 (altivec_expand_vec_perm_const): Likewise. Update call to
18355 altivec_expand_vec_perm_const_le.
18356 (rs6000_expand_vec_perm_const): Delete.
18357 (rs6000_vectorize_vec_perm_const_ok): Delete.
18358 (rs6000_vectorize_vec_perm_const): New function.
18359 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18360 an element count and rtx array.
18361 (rs6000_expand_extract_even): Update call accordingly.
18362 (rs6000_expand_interleave): Likewise.
18363 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18364 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18365 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18366 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18367 (rs6000_expand_vec_perm_const): Delete.
18368 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18369 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18370 (altivec_expand_vec_perm_const_le): Take each operand individually.
18371 Operate on constant selectors rather than rtxes.
18372 (altivec_expand_vec_perm_const): Likewise. Update call to
18373 altivec_expand_vec_perm_const_le.
18374 (rs6000_expand_vec_perm_const): Delete.
18375 (rs6000_vectorize_vec_perm_const_ok): Delete.
18376 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18377 reference to the SPE evmerge intructions.
18378 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18379 an element count and rtx array.
18380 (rs6000_expand_extract_even): Update call accordingly.
18381 (rs6000_expand_interleave): Likewise.
18382 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18383 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18385 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18387 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18389 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18390 vector mode and that that mode matches the mode of the data
18392 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18393 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18394 directly using expand_vec_perm_1 when forcing selectors into
18396 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18400 * optabs-query.h (can_vec_perm_p): Delete.
18401 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18402 * optabs-query.c (can_vec_perm_p): Split into...
18403 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18404 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18405 particular selector is valid.
18406 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18407 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18408 (vect_grouped_load_supported): Likewise.
18409 (vect_shift_permute_load_chain): Likewise.
18410 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18411 (vect_transform_slp_perm_load): Likewise.
18412 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18413 (vectorizable_bswap): Likewise.
18414 (vect_gen_perm_mask_checked): Likewise.
18415 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18416 implementations of variable permutation vectors into account
18417 when deciding which selector to use.
18418 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18419 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18420 with a false third argument.
18421 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18422 to test whether the constant selector is valid and can_vec_perm_var_p
18423 to test whether a variable selector is valid.
18425 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18427 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18428 * optabs-query.c (can_vec_perm_p): Likewise.
18429 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18430 instead of vec_perm_indices.
18431 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18432 (vect_gen_perm_mask_checked): Likewise,
18433 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18434 (vect_gen_perm_mask_checked): Likewise,
18436 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18438 * optabs-query.h (qimode_for_vec_perm): Declare.
18439 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18440 (qimode_for_vec_perm): ...this new function.
18441 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18443 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18445 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18446 does not have a conditional at the top.
18448 2018-01-02 Richard Biener <rguenther@suse.de>
18450 * ipa-inline.c (big_speedup_p): Fix expression.
18452 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18455 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18458 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18462 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18463 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18464 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18465 cond_taken_branch_cost 3->4.
18467 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18469 PR tree-optimization/83581
18470 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18471 TODO_cleanup_cfg if any changes have been made.
18473 PR middle-end/83608
18474 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18475 convert_modes if target mode has the right side, but different mode
18478 PR middle-end/83609
18479 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18480 last argument when extracting from CONCAT. If either from_real or
18481 from_imag is NULL, use expansion through memory. If result is not
18482 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18483 the parts directly to inner mode, if even that fails, use expansion
18486 PR middle-end/83623
18487 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18488 check for bswap in mode rather than HImode and use that in expand_unop
18491 Copyright (C) 2018 Free Software Foundation, Inc.
18493 Copying and distribution of this file, with or without modification,
18494 are permitted in any medium without royalty provided the copyright
18495 notice and this notice are preserved.