1 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4 statements that have been replaced by further pattern statements.
5 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7 2018-07-03 Richard Biener <rguenther@suse.de>
9 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10 always set *dt. Dump vectype in vectype overload.
11 * dumpfile.h (dump_gimple_expr): New function.
12 (dump_gimple_expr_loc): Likewise.
13 * dumpfile.c (dump_gimple_expr): New function.
14 (dump_gimple_expr_loc): Likewise.
16 2018-07-02 Jeff Law <law@redhat.com>
18 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
19 the H8/300, H8/300H and H8/S variants into a single pattern.
20 (movhi_h8300, movqi_h8300hs): Similarly.
21 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
22 (QHI mode iterator): New.
24 * config/h8300/h8300.md: Remove trailing whitespace.
26 2018-07-02 Jim Wilson <jimw@sifive.com>
28 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
29 instead of emit_insn for interrupt returns.
30 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
31 (riscv_sret, riscv_uret): Likewise.
33 2018-07-02 David Malcolm <dmalcolm@redhat.com>
35 * pretty-print.c (selftest::test_pp_format): Move save and restore
36 of quotes to class auto_fix_quotes, and add an instance.
37 * selftest.c: Include "intl.h".
38 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
39 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
40 * selftest.h (selftest::auto_fix_quotes): New class.
42 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
44 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
45 (aarch64_sve_prepare_conditional_op): Remove.
46 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
47 Allow aarch64_simd_reg_or_zero as select operand; remove
48 the aarch64_sve_prepare_conditional_op call.
49 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
50 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
51 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
52 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
53 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
54 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
55 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
56 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
57 and a splitters to match all of the *_any patterns.
58 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
60 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
61 (SVE_COND_FP_BINARY_REV): Remove.
62 (sve_int_op_rev, sve_fp_op_rev): New.
63 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
64 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
65 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
66 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
67 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
68 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
69 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
70 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
71 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
73 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
74 Remove match_dup 1 from the inner unspec.
75 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
77 * config/aarch64/aarch64.md (movprfx): New attr.
78 (length): Default movprfx to 8.
79 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
80 (*madd<SVE_I>, *msub<SVE_I): Likewise.
81 (*<su>mul<SVE_I>3_highpart): Likewise.
82 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
83 (*v<ASHIFT><SVE_I>3): Likewise.
84 (*<su><MAXMIN><SVE_I>3): Likewise.
85 (*<su><MAXMIN><SVE_F>3): Likewise.
86 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
87 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
88 (*div<SVE_F>4): Likewise.
90 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
92 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
95 2018-07-02 Richard Biener <rguenther@suse.de>
97 PR tree-optimization/86363
98 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
99 memset argument refers to a non-variable address.
101 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
103 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
104 {PLUS,MINUS}_EXPR code to...
105 (adjust_symbolic_bound): ...here,
106 (combine_bound): ...here,
107 (set_value_range_with_overflow): ...and here.
109 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
111 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
113 (extract_range_from_abs_expr): ...here.
115 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
117 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
118 -fno-omit-frame-pointer when not optimizing.
120 2018-07-02 Martin Liska <mliska@suse.cz>
123 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
124 (propagate_nothrow): Likewise.
126 2018-07-02 Martin Liska <mliska@suse.cz>
129 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
131 2018-07-02 David Malcolm <dmalcolm@redhat.com>
133 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
134 function in r262149, changing "loc" param from source_location to
135 const dump_location_t &.
136 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
137 declaration, as above.
139 2018-07-01 Paul Koning <ni1d@arrl.net>
141 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
142 -munit-asm, -mgnu-asm, -mdec-asm.
143 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
144 (pdp11_output_labelref): New.
145 (pdp11_output_def): New.
146 (pdp11_output_addr_vec_elt): New.
147 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
148 %# and %@ format codes.
149 (pdp11_option_override): New.
150 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
151 (pdp11_output_ident): New.
152 (pdp11_asm_named_section): New.
153 (pdp11_asm_init_sections): New.
154 (pdp11_file_start): New.
155 (pdp11_file_end): New.
156 (output_ascii): Use .ascii/.asciz for -mdec-asm.
157 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
158 %o, like %c but octal.
159 (pdp11_option_override): New.
160 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
162 (DATA_SECTION_ASM_OP): Ditto.
163 (READONLY_DATA_SECTION_ASM_OP): New.
164 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
165 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
166 (ASM_OUTPUT_LABELREF): Ditto.
167 (ASM_OUTPUT_DEF): Ditto.
168 (ASM_OUTPUT_EXTERNAL): New.
169 (ASM_OUTPUT_SOURCE_FILENAME): New.
170 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
171 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
172 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
173 %# and %@ format codes.
174 * config/pdp11/pdp11.opt (mgnu-asm): New.
175 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
176 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
177 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
179 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
181 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
182 dereferencing path[] beyond its length.
184 (debug_all_paths): New.
185 (rewire_first_differing_edge): New.
186 (adjust_paths_after_duplication): New.
187 (duplicate_thread_path): Call adjust_paths_after_duplication.
189 (thread_through_all_blocks): Add new argument to
190 duplicate_thread_path.
192 2018-06-30 Jim Wilson <jimw@sifive.com>
194 * config/riscv/predicates.md (p2m1_shift_operand): New.
195 (high_mask_shift_operand): New.
196 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
197 pattern using p2m1_shift_operand.
198 (lshsi3_zero_extend_3+2): New combiner pattern using
199 high_mask_shift_operand.
201 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
203 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
205 (vect_recog_rotate_pattern): ...here.
206 (vect_convert_input): Try to insert casts of invariants in the
208 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
209 preheader to be empty.
211 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
213 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
214 vector type. If given, install it in the new statement's
216 (vect_element_precision): New function.
217 (vect_unpromoted_value): New struct.
218 (vect_unpromoted_value::vect_unpromoted_value): New function.
219 (vect_unpromoted_value::set_op): Likewise.
220 (vect_look_through_possible_promotion): Likewise.
221 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
222 (vect_widened_op_tree, vect_convert_input): Likewise.
223 (vect_convert_inputs, vect_convert_output): Likewise.
224 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
225 to handle the optional cast of the multiplication result and
226 vect_widened_op_tree to detect the widened multiplication itself.
227 Do not require the input and output of promotion casts to have
228 the same sign, but base the signedness of the operation on the
229 input rather than the result. If the pattern includes two
230 promotions, check that those promotions have the same sign.
231 Do not restrict the MULT_EXPR handling to a double-width result;
232 handle quadruple-width results and wider. Use vect_convert_inputs
233 to convert the inputs to the common type.
234 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
235 to handle the optional cast of the ABS result. Also allow a sign
236 change or a sign extension between the ABS and MINUS.
237 Use vect_widened_op_tree to detect the widened subtraction and use
238 vect_convert_inputs to convert the inputs to the common type.
239 (vect_handle_widen_op_by_const): Delete.
240 (vect_recog_widen_op_pattern): New function.
241 (vect_recog_widen_mult_pattern): Use it.
242 (vect_recog_widen_shift_pattern): Likewise.
243 (vect_recog_widen_sum_pattern): Use
244 vect_look_through_possible_promotion to handle the promoted
247 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
249 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
250 the containing gimple_seq *.
251 * gimple-iterator.h (gsi_for_stmt): Declare it.
252 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
253 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
254 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
255 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
256 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
258 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
260 (vect_mark_pattern_stmts): ...here. Handle cases in which the
261 statement being replaced is part of an existing pattern
262 definition sequence, inserting the new pattern statements before
264 (vect_pattern_recog_1): Don't return a bool. If the statement
265 is already part of a pattern, instead apply pattern matching
266 to the pattern definition statements. Don't clear the
267 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
268 (vect_pattern_recog): Don't break after the first match;
269 continue processing the pattern definition statements instead.
270 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
272 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
274 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
275 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
276 (vect_recog_widen_sum_pattern): Use it.
278 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
280 * tree-vect-loop.c (vectorizable_reduction): Assert that the
281 phi is not a pattern statement and has not been replaced by
283 * tree-vect-patterns.c (type_conversion_p): Don't check
284 STMT_VINFO_IN_PATTERN_P.
285 (vect_recog_vector_vector_shift_pattern): Likewise.
286 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
287 the pattern statement rather than the original statement; check
288 directly for a WIDEN_MULT_EXPR here.
289 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
290 vect_is_simple_use to return the pattern statement rather
291 than the original statement; use is_pattern_stmt_p to check
292 for such a pattern statement.
293 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
294 to return the pattern statement rather than the original statement;
295 don't do the same transformation here.
296 (vect_is_simple_use): If the defining statement has been replaced
297 by a pattern statement, return the pattern statement instead.
298 Remove the corresponding (local) transformation from the vectype
301 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
303 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
304 end and default to null.
305 * tree-vect-loop.c (vect_create_epilog_for_reduction)
306 (vectorizable_reduction): Update calls accordingly, dropping the
307 gimple ** argument if the passed-back statement isn't needed.
308 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
309 (vect_recog_rotate_pattern): Likewise.
310 (vect_recog_mask_conversion_pattern): Likewise.
311 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
312 (vect_mask_constant_operand_p): Likewise.
313 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
314 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
315 (get_group_load_store_type, get_load_store_type): Likewise.
316 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
317 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
318 (vectorizable_conversion, vectorizable_assignment): Likewise.
319 (vectorizable_shift, vectorizable_operation): Likewise.
320 (vectorizable_store, vect_is_simple_cond): Likewise.
321 (vectorizable_condition, vectorizable_comparison): Likewise.
322 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
323 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
324 and move it to the end. Cope with null def_stmt_outs.
326 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
328 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
330 2018-06-29 Jeff Law <law@redhat.com>
332 * config/v850/v850.c (v850_legitimate_address_p): Handle large
333 displacements for TARGET_V850E2V3 and newer.
334 (TARGET_LRA_P): Remove. Defaults to LRA now.
335 * config/v850/v850.md (sign23byte_load): Remove.
336 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
337 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
339 2018-06-29 Martin Liska <mliska@suse.cz>
342 * coverage.c (coverage_init): Mangle full path name.
343 * doc/invoke.texi: Document the change.
344 * gcov-io.c (mangle_path): New.
345 * gcov-io.h (mangle_path): Likewise.
346 * gcov.c (mangle_name): Use mangle_path for path mangling.
348 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
350 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
351 if starting source register is not even.
353 2018-06-29 Martin Liska <mliska@suse.cz>
355 PR tree-optimization/86263
356 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
357 Make edge redirection.
359 2018-06-29 David Malcolm <dmalcolm@redhat.com>
361 * dumpfile.c (dump_loc): Add indentation based on scope depth.
362 (dump_scope_depth): New variable.
363 (get_dump_scope_depth): New function.
364 (dump_begin_scope): New function.
365 (dump_end_scope): New function.
366 * dumpfile.h (get_dump_scope_depth): New declaration.
367 (dump_begin_scope): New declaration.
368 (dump_end_scope): New declaration.
369 (class auto_dump_scope): New class.
370 (AUTO_DUMP_SCOPE): New macro.
371 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
374 2018-06-29 Richard Biener <rguenther@suse.de>
376 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
377 compute_all_dependences succeeds.
378 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
379 exceed --param loop-max-datarefs-for-datadeps.
381 2018-06-29 Jakub Jelinek <jakub@redhat.com>
383 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
385 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
388 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
389 alternative 0 in preferred_for_speed attribute.
391 2018-06-28 Paul Koning <ni1d@arrl.net>
393 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
394 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
395 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
396 * config/pdp11/pdp11.md: Correct "length" attribute calculation
397 for shift insn patterns.
399 2018-06-28 David Malcolm <dmalcolm@redhat.com>
401 * cgraph.c (cgraph_node::get_body): Replace assignments to
402 "dump_file" with calls to set_dump_file.
403 * dumpfile.c (alt_dump_file): Make static, and group with...
404 (alt_flags): ...this definition.
405 (dumps_are_enabled): New variable.
406 (refresh_dumps_are_enabled): New function.
407 (set_dump_file): New function.
408 (set_alt_dump_file): New function.
409 (gcc::dump_manager::dump_start): Replace assignments to
410 "dump_file" and "alt_dump_file" with calls to set_dump_file and
412 (gcc::dump_manager::dump_finish): Likewise.
413 * dumpfile.h (alt_dump_file): Delete decl.
414 (dumps_are_enabled): New variable decl.
415 (set_dump_file): New function decl.
416 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
418 * tree-nested.c (lower_nested_functions): Replace assignments to
419 "dump_file" with calls to set_dump_file.
421 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
423 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
424 goto_locus of each outgoing edge of each basic block.
426 2018-06-28 Richard Biener <rguenther@suse.de>
428 * dwarf2out.c (decl_scope_table): Remove.
429 (push_decl_scope): Likewise.
430 (pop_decl_scope): Likewise.
431 (gen_type_die_for_member): Do not call push/pop_decl_scope.
432 (gen_struct_or_union_type_die): Likewise.
433 (gen_tagged_type_die): Likewise.
434 (dwarf2out_init): Do not initialize decl_scope_table.
435 (dwarf2out_c_finalize): Do not free it.
437 2018-06-28 Richard Biener <rguenther@suse.de>
439 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
440 deciding whether to not re-use a DIE.
442 2018-06-28 Richard Biener <rguenther@suse.de>
444 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
445 DW_AT_abstract_origin attribute.
447 2018-06-28 Martin Liska <mliska@suse.cz>
449 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
450 Use newly introduced constants.
451 * tree-switch-conversion.h (struct jump_table_cluster):
452 Define max_ratio_for_size and max_ratio_for_speed.
454 2018-06-28 Martin Liska <mliska@suse.cz>
456 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
457 Add new checking assert to catch invalid state.
458 (jump_table_cluster::can_be_handled): Handle single case
460 (jump_table_cluster::is_beneficial): Bail out for such case.
461 (bit_test_cluster::find_bit_tests):
462 Add new checking assert to catch invalid state.
463 (bit_test_cluster::can_be_handled): Handle single case
465 (bit_test_cluster::is_beneficial): Bail out for such case.
466 (switch_decision_tree::analyze_switch_statement):
469 2018-06-28 Martin Liska <mliska@suse.cz>
471 * common.opt: Introduce -completion option.
472 * gcc.c (driver_handle_option): Handle it.
473 (driver::main): Print completions if completion
475 * opt-suggestions.c (option_proposer::get_completions):
477 (option_proposer::suggest_completion): Likewise.
478 (option_proposer::find_param_completions): Likewise.
479 (verify_autocompletions): Likewise.
480 (test_completion_valid_options): Likewise.
481 (test_completion_valid_params): Likewise.
482 (in_completion_p): Likewise.
483 (empty_completion_p): Likewise.
484 (test_completion_partial_match): Likewise.
485 (test_completion_garbage): Likewise.
486 (opt_proposer_c_tests): Likewise.
487 * opt-suggestions.h: Declare new functions.
488 * opts.c (common_handle_option): Handle OPT__completion_.
489 * selftest-run-tests.c (selftest::run_tests): Add
490 opt_proposer_c_tests.
491 * selftest.c (assert_str_startswith): New.
492 * selftest.h (assert_str_startswith): Likewise.
493 (opt_proposer_c_tests): New.
494 (ASSERT_STR_STARTSWITH): Likewise.
496 2018-06-28 Martin Liska <mliska@suse.cz>
498 * Makefile.in: Add opt-suggestions.o.
499 * gcc-main.c: Include opt-suggestions.h.
500 * gcc.c (driver::driver): Likewise.
501 (driver::~driver): Remove m_option_suggestions.
502 (driver::build_option_suggestions): Moved to option_proposer.
503 (driver::suggest_option): Likewise.
504 (driver::handle_unrecognized_options): Use option_proposer.
505 * gcc.h (class driver): Add new memver m_option_proposer.
506 * opt-suggestions.c: New file.
507 * opt-suggestions.h: New file.
509 2018-06-28 Martin Liska <mliska@suse.cz>
511 * vec.h (class auto_string_vec): New (moved from auto_argvec).
512 (auto_string_vec::~auto_string_vec): Likewise.
514 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
516 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
517 prevent_decl_creation_for_types fields up and add reset_location field.
518 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
519 statement if id->reset_location is true.
520 (copy_edges_for_bb): Do not set goto_locus on the new edges if
521 id->reset_location is true.
522 (copy_phis_for_bb): Force input_location on the arguments if
523 id->reset_location is true.
524 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
525 is set on the function to be inlined.
527 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
529 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
531 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
533 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
535 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
536 hard registers for the clobbered pseudo.
538 2018-06-27 Paul Koning <ni1d@arrl.net>
540 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
541 mutually exclusive options.
542 * config/pdp11/constraints.md (h): New constraint.
543 (O): Update definition to match shift code generation.
545 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
547 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
549 (output_jump): Change arguments.
550 (pdp11_fixed_cc_regs): New function.
551 (pdp11_cc_mode): Ditto.
552 (pdp11_expand_shift): Ditto.
553 (pdp11_assemble_shift): Ditto.
554 (pdp11_small_shift): Ditto.
555 (pdp11_branch_cost): Remove.
556 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
558 (pdp11_register_move_cost): Update for CC registers.
559 (pdp11_rtx_costs): Add case for LSHIFTRT.
560 (pdp11_output_jump): Add CCNZ mode conditional branches.
561 (notice_update_cc_on_set): Remove.
562 (pdp11_cc_mode): New function.
563 (simple_memory_operand): Correct pre/post decrement case.
564 (no_side_effect_operand): New function.
565 (pdp11_regno_reg_class): Add CC_REGS class.
566 (pdp11_fixed_cc_regs): New function.
567 (pdp11_small_shift): New function.
568 (pdp11_expand_shift): New function to expand shift insns.
569 (pdp11_assemble_shift): New function to output shifts.
570 (pdp11_branch_cost): Remove.
571 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
572 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
574 (PTRDIFF_TYPE): Ditto.
575 (ADJUST_INSN_LENGTH): New macro.
576 (FIXED_REGISTERS): Add CC registers.
577 (CALL_USED_REGISTERS): Ditto.
579 (REG_CLASS_NAMES): Ditto.
580 (REG_CLASS_CONTENTS): Ditto.
581 (SELECT_CC_MODE): Use new function.
582 (TARGET_FLAGS_REGNUM): New macro.
583 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
584 (cc0_reg_rtx): Remove.
585 (CC_STATUS_MDEP): Remove.
586 (CC_STATUS_MDEFP_INIT): Remove.
588 (NOTICE_UPDATE_CC): Remove.
589 (REGISTER_NAMES): Add CC registers.
590 (BRANCH_COST): Change to constant 1.
591 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
593 * config/pdp11/pdp11.opt (mbcopy): Remove.
594 (mbcopy-builtin): Remove.
595 (mbranch-cheap): Remove.
596 (mbranch-expensive): Remove.
597 * config/pdp11/predicates.md (expand_shift_operand): Update to
598 match shift code generation.
599 (ccnz_operator): New predicate.
600 * doc/invoke.texi (PDP-11 Options): Remove deleted options
601 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
602 Remove non-existent option -mabshi, -mno-abshi. Document mutually
604 * doc/md.texi (PDP-11): Document new D and h constraints. Update
605 description of O constraint.
607 2018-06-27 Jeff Law <law@redhat.com>
608 Austin Law <austinklaw@gmail.com>
610 * config/v850/v850.md (addsi3_set_flags): New pattern.
611 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
612 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
613 (zero_extendhisi2_v850_set_flags): Likewise.
614 (zero_extendqisi2_v850_set_flags): Likewise.
615 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
616 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
617 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
619 * config/v850/v850-protos.h (notice_update_cc): Remove.
620 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
621 (v850_print_operand): Handle 'D' and "d".
622 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
623 Add handling of arithmetic/logical operations compared against zero.
624 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
625 Do not look at v850_compare_op, instead get mode from last argument.
626 (v850_gen_compare): Remove
627 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
628 after reload for prologue insns.
629 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
631 (construct_save_jarl): Likewise.
632 (TARGET_FLAGS_REGNUM): Define.
633 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
634 (NOTICE_UPDATE_CC): Remove.
635 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
636 than cc0. Conditionalize on reload_completed.
637 (cmpsi_insn, setfcc_insn): Likewise.
638 (tst1 splitter): Turn into define_and_split which sets the flags
640 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
641 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
642 (cstoresf4, cstoredf4): Clobber the flags.
643 (cmpsi, cmpsf, cmpdf): Remove expanders.
644 (setf_insn): Remove pattern.
645 (addsi3): Turn into define_and_split which clobbers the flags after
646 reload and a suitable pattern (addsi3_clobber_flags) for use after
648 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
649 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
650 (ashrsi3, ashrsi3_v850e2): Likewise.
651 (bins): Clobber the flags.
652 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
653 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
654 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
655 (call_value_internal_short, call_value_internal_long): Likewise.
656 (callt_save_interrupt, callt_return_interrupt): Likewise.
657 (save_interrupt, return_interrupt): Likewise.
658 (callt_save_all_interrupt, save_all_interrupt): Likewise.
659 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
660 (restore_all_interrupt, _restore_all_interrupt): Likewise.
661 (All FP comparisons): Only allow after reload has completed.
663 (divh, divhu): Tweak output template.
664 (branch_z_normal, branch_z_invert): Remove
665 (branch_nz_normal, branch_nz_invert): Likewise.
666 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
668 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
669 * config/v850/v850.c (notice_update_cc): Remove.
670 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
671 (CC_NO_CARRY): Likewise.
672 (NOTICE_UPDATE_CC): Define to nothing.
673 * config/v850/v850.md: Remove block comment on cc0 handling
674 Remove "cc" attribute from all patterns. Remove cc_status handling
675 from all patterns. Minor formatting fixes.
677 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
679 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
680 (cortex-a76.cortex-a55): Likewise.
681 * config/aarch64/aarch64-tune.md: Regenerate.
682 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
683 cortex-a76.cortex-a55.
685 2018-06-27 Jeff Law <law@redhat.com>
687 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
688 (MULTILIB_DIRNAMES): Similarly.
690 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
692 * gimple.h (gimple_return_retbnd): Delete.
693 (gimple_return_set_retbnd): Likewise.
694 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
695 gimple_return_set_retbnd.
696 * gimple-pretty-print.c (dump_gimple_return): Remove call to
697 gimple_return_retbnd and adjust.
698 * tree-inline.h (struct copy_body_data): Remove retbnd field.
699 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
700 Explicitly return NULL in a couple more cases. Move assertion
701 on debug statements and remove unreachable code.
702 (reset_debug_binding): Do not test id->retbnd.
703 (expand_call_inline): Do not set it.
705 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
707 * configure.ac: Add --disable-gcov option.
708 * configure: Regenerate.
709 * Makefile.in: Honour @enable_gcov@.
710 * doc/install.texi: Document --disable-gcov.
712 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
714 * config/arm/arm-cpus.in (cortex-a76): New entry.
715 (cortex-a76.cortex-a55): Likewise.
716 * config/arm/arm-tables.opt: Regenerate.
717 * config/arm/arm-tune.md: Likewise.
718 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
719 * doc/invoke.texi (ARM Options): Document cortex-a76 and
720 cortex-a76.cortex-a55.
722 2018-06-27 Tamar Christina <tamar.christina@arm.com>
725 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
727 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
729 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
731 (EPILOGUE_USES): Likewise.
733 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
735 * tree-inline.c (remap_location): New function extracted from...
736 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
737 (copy_phis_for_bb): ...here. Call remap_location.
738 (copy_cfg_body): Adjust call to copy_edges_for_bb.
740 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
742 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
743 unaligned vsx for 16B memset.
745 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
748 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
749 ieee128_float_type_node to long_double_type_node unless
750 TARGET_LONG_DOUBLE_128 is set.
752 2018-06-26 David Malcolm <dmalcolm@redhat.com>
754 * cfgloop.c (get_loop_location): Convert return type from
755 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
756 by implicit construction from rtx_insn *, and using
757 dump_user_location_t::from_function_decl for the fallback case.
758 * cfgloop.h (get_loop_location): Convert return type from
759 location_t to dump_user_location_t.
760 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
761 dump_printf_loc to pass in a dump_location_t rather than a
762 location_t, via the gimple stmt.
763 * coverage.c (get_coverage_counts): Update calls to
764 dump_printf_loc to pass in dump_location_t rather than a
766 * doc/optinfo.texi (Dump types): Convert example of
767 dump_printf_loc from taking "locus" to taking "insn". Update
768 description of the "_loc" calls to cover dump_location_t.
769 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
771 (dump_user_location_t::dump_user_location_t): New constructors,
772 from gimple *stmt and rtx_insn *.
773 (dump_user_location_t::from_function_decl): New function.
774 (dump_loc): Make static.
775 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
776 const dump_location_t &.
777 (dump_generic_expr_loc): Delete.
778 (dump_printf_loc): Convert param "loc" from location_t to
779 const dump_location_t &.
780 (selftest::test_impl_location): New function.
781 (selftest::dumpfile_c_tests): New function.
782 * dumpfile.h: Include "profile-count.h".
783 (class dump_user_location_t): New class.
784 (struct dump_impl_location_t): New struct.
785 (class dump_location_t): New class.
786 (dump_printf_loc): Convert 2nd param from source_location to
787 const dump_location_t &.
788 (dump_generic_expr_loc): Delete.
789 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
790 const dump_location_t &.
791 * gimple-fold.c (fold_gimple_assign): Update call to
792 dump_printf_loc to pass in a dump_location_t rather than a
793 location_t, via the gimple stmt.
794 (gimple_fold_call): Likewise.
795 * gimple-loop-interchange.cc
796 (loop_cand::analyze_iloop_reduction_var): Update for change to
797 check_reduction_path.
798 (tree_loop_interchange::interchange): Update for change to
800 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
801 change in return-type of find_loop_location.
802 (graphite_regenerate_ast_isl): Likewise.
803 * graphite-optimize-isl.c (optimize_isl): Likewise.
804 * graphite.c (graphite_transform_loops): Likewise.
805 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
806 pass in a dump_location_t rather than a location_t, via the
808 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
809 * ipa.c (walk_polymorphic_call_targets): Likewise.
810 * loop-unroll.c (report_unroll): Convert "locus" param from
811 location_t to dump_location_t.
812 (decide_unrolling): Update for change to get_loop_location's
814 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
815 location_t to dump_user_location_t.
816 (grid_find_single_omp_among_assignments_1): Updates calls to
817 dump_printf_loc to pass in a dump_location_t rather than a
818 location_t, via the gimple stmt.
819 (grid_parallel_clauses_gridifiable): Convert "tloc" from
820 location_t to dump_location_t. Updates calls to dump_printf_loc
821 to pass in a dump_location_t rather than a location_t, via the
823 (grid_inner_loop_gridifiable_p): Likewise.
824 (grid_dist_follows_simple_pattern): Likewise.
825 (grid_gfor_follows_tiling_pattern): Likewise.
826 (grid_target_follows_gridifiable_pattern): Likewise.
827 (grid_attempt_target_gridification): Convert initialization
828 of local "grid" from memset to zero-initialization; FIXME: does
829 this require C++11? Update call to dump_printf_loc to pass in a
830 optinfo_location rather than a location_t, via the gimple stmt.
831 * profile.c (read_profile_edge_counts): Updates call to
832 dump_printf_loc to pass in a dump_location_t rather than a
834 (compute_branch_probabilities): Likewise.
835 * selftest-run-tests.c (selftest::run_tests): Call
837 * selftest.h (dumpfile_c_tests): New decl.
838 * tree-loop-distribution.c (pass_loop_distribution::execute):
839 Update for change in return type of find_loop_location.
840 * tree-parloops.c (parallelize_loops): Likewise.
841 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
842 "locus" from location_t to dump_user_location_t.
843 (canonicalize_loop_induction_variables): Likewise.
844 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
845 for change in return type of find_loop_location.
846 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
847 to dump_printf_loc to pass in a dump_location_t rather than a
848 location_t, via the stmt.
849 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
851 * tree-vect-loop-manip.c (find_loop_location): Convert return
852 type from source_location to dump_user_location_t.
853 (vect_do_peeling): Update for above change.
854 (vect_loop_versioning): Update for change in type of
856 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
857 from location_t to dump_user_location_t.
858 (vect_estimate_min_profitable_iters): Update for change in type
860 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
861 location_t to dump_location_t.
862 (vect_slp_bb): Update for change in type of vect_location.
863 * tree-vectorizer.c (vect_location): Convert from source_location
864 to dump_user_location_t.
865 (try_vectorize_loop_1): Update for change in vect_location's type.
866 (vectorize_loops): Likewise.
867 (increase_alignment): Likewise.
868 * tree-vectorizer.h (vect_location): Convert from source_location
869 to dump_user_location_t.
870 (find_loop_location): Convert return type from source_location to
871 dump_user_location_t.
872 (check_reduction_path): Convert 1st param from location_t to
873 dump_user_location_t.
874 * value-prof.c (check_counter): Update call to dump_printf_loc to
875 pass in a dump_user_location_t rather than a location_t; update
876 call to error_at for change in type of "locus".
877 (check_ic_target): Update call to dump_printf_loc to
878 pass in a dump_user_location_t rather than a location_t, via the
881 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
883 * config/s390/s390.h (enum processor_flags): Do not use
884 default tune parameter when -march was specified.
886 2018-06-26 Jakub Jelinek <jakub@redhat.com>
889 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
890 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
893 2018-06-26 Richard Biener <rguenther@suse.de>
895 PR tree-optimization/86287
897 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
898 (vect_analyze_loop): Initialize n_stmts.
900 2018-06-26 Richard Biener <rguenther@suse.de>
903 * fold-const.c (fold_convertible_p): Pointer extension
906 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
909 * dwarf2out.c (loc_list_has_views): Adjust comments.
910 (dw_loc_list): Split single cross-partition range with
913 2018-06-25 Jeff Law <law@redhat.com>
915 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
916 on -mbig-switch by default.
918 * config/v850/predicates.md (const_float_1_operand): Fix match_code
920 (const_float_0_operand): Remove unused predicate.
921 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
922 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
923 (recipsf2): New expander. Original pattern now called
925 (recipdf2, recipdf2_insn): Similarly.
926 (rsqrtsf2, rsqrtsf2_insn): Similarly
927 (rsqrtdf2, rsqrtdf2_insn): Similarly
929 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
931 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
932 Simplify logic for FreeBSD (twice).
934 2018-06-25 Martin Sebor <msebor@redhat.com>
936 PR tree-optimization/86204
937 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
938 a strnlen result if it's less than the length of the string.
940 2018-06-25 Martin Sebor <msebor@redhat.com>
942 PR tree-optimization/85700
943 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
944 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
945 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
947 2018-06-25 Martin Sebor <msebor@redhat.com>
949 * doc/extend.texi (Zero-length arrays): Update and clarify.
951 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
953 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
954 added IEEE/IBM long double multilib support on PowerPC little
955 endian Linux systems.
956 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
957 (MULTILIB_DEFAULTS): Likewise.
958 * config/rs6000/rs6000.c (rs6000_option_override_internal):
960 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
961 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
962 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
964 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
967 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
968 (REORDER_45): Likewise.
970 2018-06-25 Jeff Law <law@redhat.com>
972 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
973 dividend to 32 bits. Adjust length.
974 (udivmodhi4): Cleanup output template. Fix length.
976 2018-06-25 Carl Love <cel@us.ibm.com>
978 * config/rs6000/vsx.md: Change word selector to prefered location.
980 2018-06-25 Richard Biener <rguenther@suse.de>
982 PR tree-optimization/86304
983 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
984 epilogue-if-converted loops as well.
986 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
988 * lto-section-out.c (lto_begin_section): Do not print section
989 name for noaddr and unnumbered dumps.
991 2018-06-25 Richard Biener <rguenther@suse.de>
993 * tree-vectorizer.h (struct vec_info_shared): New structure
994 with parts split out from struct vec_info and loop_nest from
995 struct _loop_vec_info.
996 (struct vec_info): Adjust accordingly.
997 (struct _loop_vec_info): Likewise.
998 (LOOP_VINFO_LOOP_NEST): Adjust.
999 (LOOP_VINFO_DATAREFS): Likewise.
1000 (LOOP_VINFO_DDRS): Likewise.
1001 (struct _bb_vec_info): Likewise.
1002 (BB_VINFO_DATAREFS): Likewise.
1003 (BB_VINFO_DDRS): Likewise.
1004 (struct _stmt_vec_info): Add dr_aux member.
1005 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1006 (DR_MISALIGNMENT_UNINITIALIZED): New.
1007 (set_dr_misalignment): Adjust.
1008 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1009 (vect_analyze_loop): Adjust prototype.
1010 (vect_analyze_loop_form): Likewise.
1011 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1012 Compute dependences lazily.
1013 (vect_record_base_alignments): Use shared datarefs/ddrs.
1014 (vect_verify_datarefs_alignment): Likewise.
1015 (vect_analyze_data_refs_alignment): Likewise.
1016 (vect_analyze_data_ref_accesses): Likewise.
1017 (vect_analyze_data_refs): Likewise.
1018 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1019 constructor parameter for shared part.
1020 (vect_analyze_loop_form): Pass in shared part and adjust.
1021 (vect_analyze_loop_2): Pass in storage for the number of
1022 stmts. Move loop nest finding to the caller. Compute
1024 (vect_analyze_loop): Pass in shared part.
1025 (vect_transform_loop): Verify shared datarefs are unchanged.
1026 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1027 constructor parameter for shared part.
1028 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1029 (vect_slp_bb): Verify shared datarefs are unchanged before
1031 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1033 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1034 DR_MISALIGNMENT_UNINITIALIZED.
1035 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1036 parameter for shared part.
1037 (vec_info::~vec_info): Adjust.
1038 (vec_info_shared::vec_info_shared): New.
1039 (vec_info_shared::~vec_info_shared): Likewise.
1040 (vec_info_shared::save_datarefs): Likewise.
1041 (vec_info_shared::check_datarefs): Likewise.
1042 (try_vectorize_loop_1): Construct shared part live for analyses
1043 of a single loop for multiple vector sizes.
1044 * tree-parloops.c (gather_scalar_reductions): Adjust.
1046 2018-06-25 Richard Biener <rguenther@suse.de>
1048 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1049 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1050 (vect_analyze_data_refs): Remove similar code from here and
1051 simplify accordingly.
1053 2018-06-25 Richard Biener <rguenther@suse.de>
1055 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1056 for reverse storage order accesses rather than asserting
1057 they cannot happen here.
1059 2018-06-25 Tom de Vries <tdevries@suse.de>
1062 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1063 Use data16 instead of .byte for insn prefix.
1065 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1068 * parser.c (make_char_string_pack): Pass this literal chars
1069 through cpp_interpret_string.
1070 (cp_parser_userdef_numeric_literal): Check the result of
1071 make_char_string_pack.
1073 2018-06-24 Maya Rashish <coypu@sdf.org>
1075 * ginclude/stddef.h: Simplify conditions around avoiding
1076 re-definition of __size_t.
1078 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1080 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1081 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1083 2018-06-22 Maya Rashish <coypu@sdf.org>
1085 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1086 made to OpenBSD and NetBSD through the years.
1088 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1090 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1091 behavior of vec_pack (vector double, vector double) to match
1092 behavior of vec_float2 (vector double, vector double).
1094 2018-06-22 Olivier Hainque <hainque@adacore.com>
1096 * gimplify.c (gimplify_function_tree): Prevent creation
1097 of a trampoline for the address of the current function
1098 passed to entry/exit instrumentation hooks.
1100 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1103 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1106 2018-06-22 Martin Liska <mliska@suse.cz>
1108 PR tree-optimization/86263
1109 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1110 Bail out if is_enabled is false.
1111 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1113 (jump_table_cluster::is_enabled): New function.
1115 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1117 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1118 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1119 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1120 (lto_input_ts_binfo_tree_pointers): Likewise.
1121 * tree-streamer-out.c (streamer_write_tree_bitfields,
1122 write_ts_binfo_tree_pointers): Likewise.
1123 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1125 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1127 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1129 2018-06-22 Martin Liska <mliska@suse.cz>
1131 * symbol-summary.h (get): Make it pure and inline move
1132 functionality from ::get function.
1133 (get): Remove and inline into ::get and ::get_create.
1134 (get_create): Move code from ::get function.
1136 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1139 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1140 -x assembler-with-cpp.
1142 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1144 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1145 _FILE_OFFSET_BITS=64 for C++.
1147 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1149 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1150 conversion insn that shows up when pr85657-3.c is compiled using
1151 IEEE 128-bit long double.
1152 (neg<mode>2_internal): Use the correct mode to check whether the
1153 mode is IBM extended.
1154 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1155 multiply and divide external functions from being created more
1158 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1160 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1162 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1163 the edge can be forwarded.
1164 (cfg_layout_merge_blocks): Likewise.
1166 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1168 * except.c (finish_eh_generation): Commit edge insertions only after
1169 the EH edges have been redirected from post-landing to landing pads.
1171 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1173 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1174 create_tmp_var_for to create the FRAME decl.
1175 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1177 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1179 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1180 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1181 debug statement when resetting its value.
1182 (expand_call_inline): Copy the locus of the call onto the assignment
1183 of the return value, if any. Use local variable in more cases.
1185 2018-06-21 Martin Liska <mliska@suse.cz>
1187 * ipa-pure-const.c (propagate_nothrow): Use
1188 funct_state_summaries->get.
1189 (dump_malloc_lattice): Likewise.
1190 (propagate_malloc): Likewise.
1192 2018-06-21 Richard Biener <rguenther@suse.de>
1194 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1195 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1196 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1198 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1199 BLOCK_ABSTRACT_ORIGIN unconditionally.
1201 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1203 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1205 * ipa-reference.c (ipa_reference_c_finalize): Delete
1206 ipa_ref_opt_sum_summaries and set it to NULL.
1208 2018-06-21 Tom de Vries <tdevries@suse.de>
1210 PR tree-optimization/85859
1211 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1212 test with comment from bb_no_side_effects_p.
1214 2018-06-21 Richard Biener <rguenther@suse.de>
1216 PR tree-optimization/86232
1217 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1218 max for constant niter.
1220 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1222 * config/aarch64/aarch64-simd.md
1223 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1225 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1227 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1228 Make opernads of the unspec commutative.
1230 2018-06-21 Richard Biener <rguenther@suse.de>
1232 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1233 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1234 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1235 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1237 (vect_record_base_alignments): Do not record base alignment
1238 for gathers or scatters.
1239 (vect_compute_data_ref_alignment): Drop return value that is always
1240 true. Bail out early for gathers or scatters.
1241 (vect_enhance_data_refs_alignment): Bail out early for gathers
1243 (vect_find_same_alignment_drs): Likewise.
1244 (vect_analyze_data_refs_alignment): Remove dead code.
1245 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1246 (vect_analyze_data_refs): For possible gathers or scatters do
1247 not create an alternate DR, just check their possible validity
1248 and mark them. Adjust DECL_NONALIASED handling to not rely
1250 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1251 update inits of gathers or scatters.
1252 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1253 Also copy gather/scatter flag to pattern vinfo.
1255 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1257 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1258 behavior of vec_packsu (vector unsigned long long, vector unsigned
1259 long long) to match behavior of vec_packs with same signature.
1261 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1262 Thomas Schwinge <thomas@codesourcery.com>
1263 Cesar Philippidis <cesar@codesourcery.com>
1265 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1266 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1267 (gimplify_adjust_omp_clauses): Likewise.
1268 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1269 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1270 (gimplify_omp_target_update): Update handling of acc update and
1272 * omp-low.c (install_var_field): Remove unused parameter
1273 base_pointers_restrict.
1274 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1275 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1277 (omp_target_base_pointers_restrict_p): Delete.
1278 (scan_omp_target): Update call to scan_sharing_clauses.
1279 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1281 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1282 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1283 (convert_local_omp_clauses): Likewise.
1284 * tree-pretty-print.c (dump_omp_clause): Likewise.
1285 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1287 (omp_clause_code_name): Likewise.
1289 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1292 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1295 PR tree-optimization/86231
1296 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1297 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1299 2018-06-20 Tom de Vries <tdevries@suse.de>
1301 PR tree-optimization/86097
1302 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1303 iv type if signedness of iv type is not the same as that of *nit.
1305 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1307 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1308 EDGE_EH edges, verify they are all EDGE_EH.
1310 2018-06-20 Maya Rashish <coypu@sdf.org>
1312 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1314 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1316 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1317 * config/aarch64/aarch64.c (xgene1_tunings): Add
1318 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1319 (aarch64_mode_valid_for_sched_fusion_p):
1320 Allow 16-byte modes.
1321 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1322 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1324 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1326 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1327 * config/aarch64/iterators.md (VQ2): New mode iterator.
1329 2018-06-20 Martin Liska <mliska@suse.cz>
1331 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1332 Change default ratio from 10 to 8.
1334 2018-06-20 Martin Liska <mliska@suse.cz>
1336 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1338 (bit_test_cluster::find_bit_tests): Likewise.
1339 (switch_decision_tree::analyze_switch_statement): Find clusters.
1340 * tree-switch-conversion.h (struct jump_table_cluster): Document
1343 2018-06-20 Martin Liska <mliska@suse.cz>
1345 * tree-switch-conversion.c (switch_conversion::collect):
1346 Record m_uniq property.
1347 (switch_conversion::expand): Bail out for special conditions.
1348 (group_cluster::~group_cluster): New.
1349 (group_cluster::group_cluster): Likewise.
1350 (group_cluster::dump): Likewise.
1351 (jump_table_cluster::emit): New.
1352 (switch_decision_tree::fix_phi_operands_for_edges): New.
1353 (struct case_node): Remove struct.
1354 (jump_table_cluster::can_be_handled): New.
1355 (case_values_threshold): Moved to header.
1356 (reset_out_edges_aux): Likewise.
1357 (jump_table_cluster::is_beneficial): New.
1358 (bit_test_cluster::can_be_handled): Likewise.
1359 (add_case_node): Remove.
1360 (bit_test_cluster::is_beneficial): New.
1361 (case_bit_test::cmp): New.
1362 (bit_test_cluster::emit): New.
1363 (expand_switch_as_decision_tree_p): Remove.
1364 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1365 (fix_phi_operands_for_edge): Likewise.
1366 (switch_decision_tree::analyze_switch_statement): New.
1367 (compute_cases_per_edge): Move ...
1368 (switch_decision_tree::compute_cases_per_edge): ... here.
1369 (try_switch_expansion): Likewise.
1370 (switch_decision_tree::try_switch_expansion): Likewise.
1371 (record_phi_operand_mapping): Likewise.
1372 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1373 (emit_case_decision_tree): Likewise.
1374 (switch_decision_tree::emit): Likewise.
1375 (balance_case_nodes): Likewise.
1376 (switch_decision_tree::balance_case_nodes): Likewise.
1377 (dump_case_nodes): Likewise.
1378 (switch_decision_tree::dump_case_nodes): Likewise.
1379 (emit_jump): Likewise.
1380 (switch_decision_tree::emit_jump): Likewise.
1381 (emit_cmp_and_jump_insns): Likewise.
1382 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1383 (emit_case_nodes): Likewise.
1384 (switch_decision_tree::emit_case_nodes): Likewise.
1385 (conditional_probability): Remove.
1386 * tree-switch-conversion.h (enum cluster_type): New.
1388 (struct cluster): Likewise.
1389 (cluster::cluster): Likewise.
1390 (struct simple_cluster): Likewise.
1391 (simple_cluster::simple_cluster): Likewise.
1392 (struct group_cluster): Likewise.
1393 (struct jump_table_cluster): Likewise.
1394 (struct bit_test_cluster): Likewise.
1395 (struct min_cluster_item): Likewise.
1396 (struct case_tree_node): Likewise.
1397 (case_tree_node::case_tree_node): Likewise.
1398 (jump_table_cluster::case_values_threshold): Likewise.
1399 (struct case_bit_test): Likewise.
1400 (struct switch_decision_tree): Likewise.
1401 (struct switch_conversion): Likewise.
1402 (switch_decision_tree::reset_out_edges_aux): Likewise.
1404 2018-06-20 Martin Liska <mliska@suse.cz>
1406 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1407 (hoist_edge_and_branch_if_true): Likewise.
1408 (expand_switch_using_bit_tests_p): Likewise.
1409 (struct case_bit_test): Likewise.
1410 (case_bit_test_cmp): Likewise.
1411 (emit_case_bit_tests): Likewise.
1412 (switch_conversion::switch_conversion): New class.
1413 (struct switch_conv_info): Remove old struct.
1414 (collect_switch_conv_info): More to ...
1415 (switch_conversion::collect): ... this.
1416 (check_range): Likewise.
1417 (switch_conversion::check_range): Likewise.
1418 (check_all_empty_except_final): Likewise.
1419 (switch_conversion::check_all_empty_except_final): Likewise.
1420 (check_final_bb): Likewise.
1421 (switch_conversion::check_final_bb): Likewise.
1422 (create_temp_arrays): Likewise.
1423 (switch_conversion::create_temp_arrays): Likewise.
1424 (free_temp_arrays): Likewise.
1425 (gather_default_values): Likewise.
1426 (switch_conversion::gather_default_values): Likewise.
1427 (build_constructors): Likewise.
1428 (switch_conversion::build_constructors): Likewise.
1429 (constructor_contains_same_values_p): Likewise.
1430 (switch_conversion::contains_same_values_p): Likewise.
1431 (array_value_type): Likewise.
1432 (switch_conversion::array_value_type): Likewise.
1433 (build_one_array): Likewise.
1434 (switch_conversion::build_one_array): Likewise.
1435 (build_arrays): Likewise.
1436 (switch_conversion::build_arrays): Likewise.
1437 (gen_def_assigns): Likewise.
1438 (switch_conversion::gen_def_assigns): Likewise.
1439 (prune_bbs): Likewise.
1440 (switch_conversion::prune_bbs): Likewise.
1441 (fix_phi_nodes): Likewise.
1442 (switch_conversion::fix_phi_nodes): Likewise.
1443 (gen_inbound_check): Likewise.
1444 (switch_conversion::gen_inbound_check): Likewise.
1445 (process_switch): Use the newly created class.
1446 (switch_conversion::expand): New.
1447 (switch_conversion::~switch_conversion): New.
1448 * tree-switch-conversion.h: New file.
1450 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1452 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1453 tree-vect-patterns.c.
1454 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1455 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1456 (vect_recog_sad_pattern): Likewise.
1457 (vect_recog_widen_sum_pattern): Likewise.
1458 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1459 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1460 (vect_recog_rotate_pattern): Likewise.
1461 (vect_recog_mult_pattern): Likewise.
1462 (vect_recog_vector_vector_shift_pattern): Likewise.
1463 (vect_recog_divmod_pattern): Likewise.
1464 (vect_recog_mixed_size_cond_pattern): Likewise.
1465 (vect_recog_bool_pattern): Likewise.
1466 (vect_recog_mask_conversion_pattern): Likewise.
1467 (vect_try_gather_scatter_pattern): Likewise.
1468 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1469 (vect_recog_over_widening_pattern): Likewise.
1470 (vect_recog_gather_scatter_pattern): Likewise.
1471 (vect_recog_func_ptr): Move from tree-vectorizer.h
1472 (vect_vect_recog_func_ptrs): Move further down the file.
1473 (vect_recog_func): Likewise. Remove the third argument.
1474 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1475 (vect_pattern_recog_1): Expect the pattern function to do any
1476 necessary target tests. Also expect it to provide a vector type.
1477 Remove the type_in handling.
1479 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1481 * tree-vect-patterns.c (vect_pattern_detected): New function.
1482 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1483 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1484 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1485 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1486 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1487 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1488 (vect_recog_mask_conversion_pattern)
1489 (vect_try_gather_scatter_pattern): Likewise.
1491 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1493 * tree-vect-patterns.c (vect_get_internal_def): New function.
1494 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1495 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1496 (search_type_for_mask_1): Use it.
1498 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1500 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1501 redundant WIDEN_SUM_EXPR handling.
1502 (vect_recog_sad_pattern): Likewise.
1504 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1506 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1507 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1508 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1509 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1510 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1512 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1514 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1515 use the stmt_vec_info of the original bb statement for the
1516 new zero assignment, even if the call is part of a pattern.
1518 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1520 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1521 that the sequence is attached to the original statement rather
1522 than the pattern statement.
1523 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1524 PATTERN_DEF_SEQ from the original statement rather than
1525 the main pattern statement.
1526 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1527 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1528 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1530 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1532 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1533 definition statements before the early exit for statements that aren't
1535 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1537 (vect_transform_loop): ...here. Process pattern definition
1538 statements without first checking whether the main pattern
1539 statement is live or relevant.
1541 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1543 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1544 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1546 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1548 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1549 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1550 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1551 (expand_block_compare): Change select_block_compare_mode call.
1552 (expand_strncmp_align_check): Use new functions, fix comment.
1553 (emit_final_str_compare_gpr): New function.
1554 (expand_strn_compare): Refactor and clean up code.
1555 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1557 2018-06-19 Tony Reix <tony.reix@atos.com>
1558 Damien Bergamini <damien.bergamini@atos.com>
1559 David Edelsohn <dje.gcc@gmail.com>
1561 * collect2.c (static_obj): New variable.
1562 (static_libs): New variable.
1563 (is_in_list): Uncomment declaration.
1564 (main): Track AIX libraries linked statically.
1565 (is_in_list): Uncomment definition.
1566 (scan_prog_file): Don't add AIX shared libraries initializer
1567 to constructor list if linking statically.
1569 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1571 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1573 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1575 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1577 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1580 2018-06-19 Martin Liska <mliska@suse.cz>
1582 * config/i386/i386.c (ix86_can_inline_p): Do not use
1583 ipa_fn_summaries::get_create.
1584 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1586 (devirtualization_time_bonus): Likewise.
1587 (ipcp_propagate_stage): Likewise.
1588 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1589 (edge_set_predicate): Likewise.
1590 (evaluate_conditions_for_known_args): Likewise.
1591 (evaluate_properties_for_edge): Likewise.
1592 (ipa_call_summary::reset): Tranform to ...
1593 (ipa_call_summary::~ipa_call_summary): ... this.
1594 (ipa_fn_summary::reset): Transform to ...
1595 (ipa_fn_summary::~ipa_fn_summary): ... this.
1596 (ipa_fn_summary_t::remove): Rename to ...
1597 (ipa_fn_summary_t::remove_callees): ... this.
1598 (ipa_fn_summary_t::duplicate): Use placement new
1599 instead of memory copy.
1600 (ipa_call_summary_t::duplicate): Likewise.
1601 (ipa_call_summary_t::remove): Remove.
1602 (dump_ipa_call_summary): Change get_create to get.
1603 (ipa_dump_fn_summary): Dump only when summary exists.
1604 (analyze_function_body): Use symbol_summary::get instead
1606 (compute_fn_summary): Likewise.
1607 (estimate_edge_devirt_benefit): Likewise.
1608 (estimate_edge_size_and_time): Likewise.
1609 (inline_update_callee_summaries): Likewise.
1610 (remap_edge_change_prob): Likewise.
1611 (remap_edge_summaries): Likewise.
1612 (ipa_merge_fn_summary_after_inlining): Likewise.
1613 (write_ipa_call_summary): Likewise.
1614 (ipa_fn_summary_write): Likewise.
1615 (ipa_free_fn_summary): Likewise.
1616 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1617 (struct ipa_call_summary): Likewise.
1618 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1620 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1621 (estimate_size_after_inlining): Likewise.
1622 (estimate_growth): Likewise.
1623 (growth_likely_positive): Likewise.
1624 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1625 (inline_call): Likewise.
1626 * ipa-inline.c (caller_growth_limits): Likewise.
1627 (can_inline_edge_p): Likewise.
1628 (can_inline_edge_by_limits_p): Likewise.
1629 (compute_uninlined_call_time): Likewise.
1630 (compute_inlined_call_time): Likewise.
1631 (want_inline_small_function_p): Likewise.
1632 (edge_badness): Likewise.
1633 (update_caller_keys): Likewise.
1634 (update_callee_keys): Likewise.
1635 (inline_small_functions): Likewise.
1636 (inline_to_all_callers_1): Likewise.
1637 (dump_overall_stats): Likewise.
1638 (early_inline_small_functions): Likewise.
1639 (early_inliner): Likewise.
1640 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1641 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1642 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1643 * ipa-split.c (execute_split_functions): Likewise.
1644 * symbol-summary.h: Likewise.
1645 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1647 2018-06-19 Richard Biener <rguenther@suse.de>
1649 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1650 (vectorize_loops): ... here. Fix dbgcnt handling.
1651 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1653 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1656 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1657 ieee128 argument takes up only one (vector) register, not two (floating
1660 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1662 * gimplify.c (gimplify_init_constructor): Really never clear for an
1663 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1665 2018-06-19 Richard Biener <rguenther@suse.de>
1667 PR tree-optimization/86179
1668 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1669 after failed recognition.
1671 2018-06-18 Martin Sebor <msebor@redhat.com>
1674 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1675 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1676 Handle integer subtraction.
1677 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1678 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1680 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1682 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1683 param from rtx to rtx_insn *.
1684 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1686 (frv_ifcvt_modify_insn): Likwise.
1687 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1688 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1689 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1690 as_a <rtx_insn *> cast to local "unprotected_region" once
1691 it's been established that it's not NULL or pc_rtx.
1692 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1693 param "sethi" from rtx to rtx_insn *.
1694 (nds32_group_float_insns): Likewise for param "insn".
1695 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1697 (vax_output_int_subtract): Likewise.
1698 * config/vax/vax.c (vax_output_int_add): Likewise for param
1700 (vax_output_int_subtract): Likewise.
1701 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1702 (emit_pattern_after): Likewise for param "after".
1703 (emit_insn_after): Likewise.
1704 (emit_jump_insn_after): Likewise.
1705 (emit_call_insn_after): Likewise.
1706 (emit_debug_insn_after): Likewise.
1707 (emit_pattern_before): Likewise for param "before".
1708 (emit_insn_before): Likewise.
1709 (emit_jump_insn_before): Likewise.
1710 * final.c (get_insn_template): Likewise for param "insn", removing
1712 * output.h (get_insn_template): Likewise for 2nd param.
1713 * rtl.h (emit_insn_before): Likewise.
1714 (emit_jump_insn_before): Likewise.
1715 (emit_debug_insn_before_noloc): Likewise.
1716 (emit_insn_after): Likewise.
1717 (emit_jump_insn_after): Likewise.
1718 (emit_call_insn_after): Likewise.
1719 (emit_debug_insn_after): Likewise.
1720 (set_insn_deleted): Likewise for param.
1722 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1725 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1726 floating point modes, so that IFmode is numerically greater than
1727 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1728 to declare the ordering. This prevents IFmode from being
1729 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1730 machine. Include rs6000-modes.h to share the fractional values
1731 between genmodes* and the rest of the compiler.
1735 * config/rs6000/rs6000-modes.h: New file.
1736 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1737 meaning of rs6000_long_double_size so that 126..128 selects an
1738 appropriate 128-bit floating point type.
1739 (rs6000_option_override_internal): Likewise.
1740 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1741 (TARGET_LONG_DOUBLE_128): Change the meaning of
1742 rs6000_long_double_size so that 126..128 selects an appropriate
1743 128-bit floating point type.
1744 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1745 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1746 source and destination to match the standard usage.
1747 (truncifkf2): Likewise.
1748 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1749 ISA 2.07 to use an explicit clobber, instead of passing in a
1751 (copysign<mode>3_soft): Likewise.
1753 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1755 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1756 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1757 (vect_slp_analyze_instance_dependence): Likewise.
1758 (vect_enhance_data_refs_alignment): Likewise.
1759 (vect_analyze_data_refs_alignment): Likewise.
1760 (vect_slp_analyze_and_verify_instance_alignment
1761 (vect_analyze_data_ref_accesses): Likewise.
1762 (vect_prune_runtime_alias_test_list): Likewise.
1763 (vect_analyze_data_refs): Likewise.
1764 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1765 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1766 (vect_analyze_scalar_cycles_1): Likewise.
1767 (vect_get_loop_niters): Likewise.
1768 (vect_analyze_loop_form_1): Likewise.
1769 (vect_update_vf_for_slp): Likewise.
1770 (vect_analyze_loop_operations): Likewise.
1771 (vect_analyze_loop): Likewise.
1772 (vectorizable_induction): Likewise.
1773 (vect_transform_loop): Likewise.
1774 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1775 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1776 (vect_make_slp_decision): Likewise.
1777 (vect_detect_hybrid_slp): Likewise.
1778 (vect_slp_analyze_operations): Likewise.
1779 (vect_slp_bb): Likewise.
1780 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1781 (vectorizable_bswap): Likewise.
1782 (vectorizable_call): Likewise.
1783 (vectorizable_simd_clone_call): Likewise.
1784 (vectorizable_conversion): Likewise.
1785 (vectorizable_assignment): Likewise.
1786 (vectorizable_shift): Likewise.
1787 (vectorizable_operation): Likewise.
1788 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1790 2018-06-18 Martin Sebor <msebor@redhat.com>
1792 PR tree-optimization/81384
1793 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1794 * builtins.c (expand_builtin_strnlen): New function.
1795 (expand_builtin): Call it.
1796 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1797 * builtins.def (BUILT_IN_STRNLEN): New.
1798 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1799 Warn for bounds in excess of maximum object size.
1800 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1801 single-value ranges. Handle strnlen.
1802 (handle_builtin_strlen): Handle strnlen.
1803 (strlen_check_and_optimize_stmt): Same.
1804 * doc/extend.texi (Other Builtins): Document strnlen.
1806 2018-06-18 Maya Rashish <coypu@sdf.org>
1808 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1809 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1810 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1812 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1814 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1816 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1818 * tree.c (escaped_string::escape): Replace cast to char * by
1819 const_cast<char *> (unescaped).
1821 2018-06-18 Nick Clifton <nickc@redhat.com>
1824 * tree.c (escaped_string): New class. Converts an unescaped
1825 string into its escaped equivalent.
1826 (warn_deprecated_use): Use the new class to convert the
1827 deprecation message, if present.
1828 (test_escaped_strings): New self test.
1829 (test_c_tests): Add test_escaped_strings.
1830 * doc/extend.texi (deprecated): Add a note that the
1831 deprecation message is affected by the -fmessage-length
1832 option, and that control characters will be escaped.
1833 (#pragma GCC error): Document this pragma.
1834 (#pragma GCC warning): Likewise.
1835 * doc/invoke.texi (-fmessage-length): Document this option's
1836 effect on the #warning and #error preprocessor directives and
1837 the deprecated attribute.
1839 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1841 * tree.c (decl_value_expr_lookup): Revert latest change.
1842 (decl_value_expr_insert): Likewise.
1844 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1846 * gimplify.c (nonlocal_vlas): Delete.
1847 (nonlocal_vla_vars): Likewise.
1848 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1850 (gimplify_body): Do not create and destroy nonlocal_vlas.
1851 * tree-nested.c: Include diagnostic.h.
1852 (use_pointer_in_frame): Tweak.
1853 (lookup_field_for_decl): Add assertion and declare the transformation.
1854 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1855 internal error when the reference is in a wrong context. Do not
1856 create a debug decl by default.
1857 (note_nonlocal_block_vlas): Delete.
1858 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1859 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1860 create a debug decl by default.
1861 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1862 call is in a wrong context.
1863 (fixup_vla_decls): New function.
1864 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1865 debug variables were created.
1866 * tree.c (decl_value_expr_lookup): Add checking assertion.
1867 (decl_value_expr_insert): Likewise.
1869 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1872 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1873 * tree-scalar-evolution.c (interpret_expr): Likewise.
1874 (expression_expensive_p): Likewise.
1875 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1876 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1877 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1878 (ssa_defined_by_minus_one_stmt_p): New.
1880 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1883 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1884 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1885 * dojump.c (do_jump): Likewise.
1886 * expr.c (expand_expr_real_2): Check operand type's sign.
1887 * fold-const.c (const_unop): Handle ABSU_EXPR.
1888 (fold_abs_const): Likewise.
1889 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1890 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1891 (strip_sign_op_1): Likesise.
1892 * match.pd: Add new pattern to generate ABSU_EXPR.
1893 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1894 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1895 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1896 * tree-inline.c (estimate_operator_cost): Likewise.
1897 * tree-pretty-print.c (dump_generic_node): Likewise.
1898 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1899 * tree.def (ABSU_EXPR): New.
1901 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1904 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1905 documented as preserved for backward compatibility only.
1906 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1908 PR rtl-optimization/86108
1909 * bb-reorder.c (create_forwarder_block): Renamed to ...
1910 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1911 jump from new landing pad to the second part.
1912 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1915 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1918 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1919 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1920 Only call store_expr for halves if the mode is the same.
1923 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1924 Fix up comment formatting.
1926 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1928 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1929 (typed_splay_tree::closure,
1930 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1931 (typed_splay_tree::typed_splay_tree,
1932 typed_splay_tree::operator =): Declared private.
1933 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1934 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1935 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1936 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1937 typed_splay_tree::splay_tree_splay,
1938 typed_splay_tree::splay_tree_foreach_helper,
1939 typed_splay_tree::splay_tree_insert,
1940 typed_splay_tree::splay_tree_remove,
1941 typed_splay_tree::splay_tree_lookup,
1942 typed_splay_tree::splay_tree_predecessor,
1943 typed_splay_tree::splay_tree_successor,
1944 typed_splay_tree::splay_tree_min,
1945 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1946 (typed_splay_tree::root, typed_splay_tree::comp,
1947 typed_splay_tree::delete_key,
1948 typed_splay_tree::delete_value): New data members.
1949 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1950 typed_splay_tree::remove.
1952 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1954 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1955 -mginv and -mno-ginv to the assembler.
1956 * config/mips/mips.opt (-mcrc): New option.
1958 * doc/invoke.text (-mcrc): Document.
1961 2018-06-15 Nick Clifton <nickc@redhat.com>
1964 * tree.c (escaped_string): New class. Converts an unescaped
1965 string into its escaped equivalent.
1966 (warn_deprecated_use): Use the new class to convert the
1967 deprecation message, if present.
1968 (test_escaped_strings): New self test.
1969 (test_c_tests): Add test_escaped_strings.
1970 * doc/extend.texi (deprecated): Add a note that the
1971 deprecation message is affected by the -fmessage-length
1972 option, and that control characters will be escaped.
1973 (#pragma GCC error): Document this pragma.
1974 (#pragma GCC warning): Likewise.
1975 * doc/invoke.texi (-fmessage-length): Document this option's
1976 effect on the #warning and #error preprocessor directives and
1977 the deprecated attribute.
1979 2018-06-15 Richard Biener <rguenther@suse.de>
1981 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1982 here, also noting vector size used.
1983 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1984 size used in MSG_OPTIMIZED_LOCATIONS dump.
1985 (pass_slp_vectorize::execute): Adjust.
1987 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1990 * config/arc/arc.c (arc_return_address_register): Fix
1993 2018-06-15 Richard Biener <rguenther@suse.de>
1996 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1997 leave useless conversion stripping to force_gimple_operand_gsi.
1998 (gimplify_build2): Likewise.
1999 (gimplify_build1): Likewise.
2001 2018-06-15 Richard Biener <rguenther@suse.de>
2004 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2005 before adjusting their block.
2007 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2009 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2010 multilibs for *-*-rtems*.
2011 * config/riscv/t-rtems: New file.
2013 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2016 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2017 unsigned_type_for returns NULL.
2020 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2021 subregs of multi-word pseudos unless the float mode has word size.
2023 2018-06-14 Richard Biener <rguenther@suse.de>
2026 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2027 duplicate and harmful type_hash_canon.
2028 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2030 2018-06-14 Richard Biener <rguenther@suse.de>
2033 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2036 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2038 * config/rtems.h (STDINT_LONG32): Define.
2040 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2041 Prachi Godbole <prachi.godbole@imgtec.com>
2043 * config/mips/mips-cpus.def: Define P6600.
2044 * config/mips/mips-tables.opt: Regenerate.
2045 * config/mips/mips.c (mips_ucbranch_type): New enum.
2046 (mips_rtx_cost_data): Add support for P6600.
2047 (mips_issue_rate): Likewise.
2048 (mips_multipass_dfa_lookahead): Likewise.
2049 (mips_avoid_hazard): Likewise.
2050 (mips_reorg_process_insns): Likewise.
2051 (mips_classify_branch_p6600): New function.
2052 * config/mips/mips.h (TUNE_P6600): New define.
2053 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2054 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2055 * config/mips/mips.md: Include p6600.md.
2056 (processor): Add p6600.
2057 * config/mips/p6600.md: New file.
2058 * doc/invoke.texi: Add p6600 to supported architectures.
2060 2018-06-13 Martin Sebor <msebor@redhat.com>
2062 PR tree-optimization/86114
2063 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2065 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2067 2018-06-13 Richard Biener <rguenther@suse.de>
2069 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2070 Properly set vector type of the intermediate stmt.
2071 * tree-vect-stmts.c (vectorizable_operation): The destination
2072 var always has vectype_out type.
2074 2018-06-13 Jeff Law <law@redhat.com>
2076 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2077 integer 0 for argument to print_rtl_with_bb.
2078 (rl78_reorg): Likewise.
2080 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2082 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2083 from rtx to rtx_insn *.
2084 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2086 (add_sched_insns_for_speculation): Likewise for local "target",
2087 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2088 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2089 from rtx_insn ** to rtx_code_label **.
2090 (reorg_emit_nops): Likewise.
2091 (c6x_reorg): Likewise for local "call_labels".
2092 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2094 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2095 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2096 the loops over LABEL_REFS.
2097 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2099 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2100 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2101 (split_branches): Strengthen local "olabel" from rtx to
2102 rtx_insn *, adding a safe_as_a cast.
2103 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2105 (add_insn_after): Likewise for first two params.
2106 (add_insn_before): Likewise.
2107 (remove_insn): Likewise for param.
2108 (emit_pattern_before_noloc): Likewise for second and third params.
2109 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2110 (emit_call_insn_before_noloc): Likewise.
2111 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2113 (emit_barrier_before): Likewise.
2114 (emit_label_before): Strengthen "label" param from "rtx" to
2115 "rtx_code_label *". Strengthen "before" param from "rtx" to
2117 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2119 (emit_pattern_after_noloc): Likewise.
2120 (emit_insn_after_noloc): Likewise.
2121 (emit_jump_insn_after_noloc): Likewise.
2122 (emit_call_insn_after_noloc): Likewise.
2123 (emit_debug_insn_after_noloc): Likewise.
2124 (emit_barrier_after): Likewise.
2125 (emit_label_after): Likewise for both params.
2126 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2127 "loc" param from "int" to "location_t".
2128 (emit_insn_after_setloc): Likewise.
2129 (emit_jump_insn_after_setloc): Likewise.
2130 (emit_call_insn_after_setloc): Likewise.
2131 (emit_debug_insn_after_setloc): Likewise.
2132 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2133 "loc" param from "int" to "location_t".
2134 (emit_pattern_before): Convert NULL_RTX to NULL.
2135 (emit_insn_before_setloc): Convert "loc" param from "int" to
2137 (emit_jump_insn_before_setloc): Likewise.
2138 (emit_call_insn_before_setloc): Likewise.
2139 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2140 rtx_insn *. Convert "loc" param from "int" to "location_t".
2141 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2142 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2143 Convert 3rd param from "int" to "location_t".
2144 (emit_barrier_before, emit_barrier_after, next_real_insn):
2145 Strengthen param from rtx to rtx_insn *.
2146 (emit_label_before): Strengthen 1st param from "rtx" to
2147 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2149 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2150 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2151 Strengthen 2nd param from "rtx" to "rtx_insn *".
2152 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2153 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2154 Likewise. Convert 3rd param from "int" to "location_t".
2155 (emit_label_after): Strengthen 1st param from "rtx" to
2157 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2159 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2160 from "rtx" to "rtx_insn *".
2162 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2164 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2165 bodies streamed in with -Q.
2166 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2167 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2168 * gimple-streamer-out.c: Include gimple-pretty-print.h
2169 (output_bb): Dump stmts streamed.
2170 * lto-section-out.c: Include print-tree.h
2171 (lto_begin_section): Dump sections created.
2172 (lto_output_decl_index): Dump decl encoded.
2173 * lto-streamer-out.c: Include print-tree.h
2174 (create_output_block): Dump output block created.
2175 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2176 (output_function): Dump function output.
2177 (output_constructor): Dump constructor streamed.
2178 (write_global_stream): Output indexes encoded.
2179 (produce_asm_for_decls): Dump streams encoded.
2180 * lto-streamer.c (streamer_dump_file): New global var.
2181 * lto-streamer.h (streamer_dump_file): Declare.
2182 * passes.c (ipa_write_summaries): Initialize streamer dump.
2183 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2186 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2189 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2190 offsets for register save directives. Emit a second batch of save
2191 directives, if need be, when the function accesses prior frames.
2193 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2195 * config/arc/fpu.md (fmasf4): Force operand to register.
2196 (fnmasf4): Likewise.
2198 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2200 * config/arc/arc-protos.h (arc_pad_return): Remove.
2201 * config/arc/arc.c (machine_function): Remove force_short_suffix
2203 (arc_print_operand): Adjust printing of '&'.
2204 (arc_verify_short): Remove conditional printing of short suffix.
2205 (arc_final_prescan_insn): Remove reference to size_reason.
2206 (pad_return): New function.
2207 (arc_reorg): Call pad_return.
2208 (arc_pad_return): Remove.
2209 (arc_init_machine_status): Remove reference to force_short_suffix.
2210 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2211 (attr length): When attribute iscompact is true force to 2
2212 regardless; in the case of maybe check if we want to force the
2213 instruction to have 4 bytes length.
2214 (nopv): Change it to generate 4 byte long nop as well.
2215 (blockage): New pattern.
2216 (simple_return): Remove call to arc_pad_return.
2217 (p_return_i): Likewise.
2219 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2221 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2223 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2225 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2228 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2230 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2231 for ARC700 and ARCv2.
2233 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2236 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2237 operands[2] instead of operands[1].
2240 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2242 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2243 case, check whether the outer register overlaps an unallocatable
2244 register, not just whether it fits the required class.
2246 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2248 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2250 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2251 (memrefs_conflict_p): Likewise.
2252 (init_alias_analysis): Likewise.
2253 * cfgexpand.c (expand_debug_expr): Likewise.
2254 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2255 * cse.c (fold_rtx): Likewise.
2256 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2257 * expr.c (emit_block_move_hints): Likewise.
2258 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2259 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2260 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2261 (emit_group_store): Likewise.
2262 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2263 to read the PRE/POST_MODIFY increment.
2264 * calls.c (store_one_arg): Use strip_offset.
2265 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2267 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2269 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2270 (simplify_binary_operation_1): Extend CONST_INT handling to
2272 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2273 than a HOST_WIDE_INT.
2274 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2276 (adjust_mems, add_stores): Update accodingly.
2277 (vt_canonicalize_addr): Track polynomial offsets.
2278 (emit_note_insn_var_location): Likewise.
2279 (vt_add_function_parameter): Likewise.
2280 (vt_initialize): Likewise.
2282 2018-06-12 Jeff Law <law@redhat.com>
2284 * config.gcc (alpha*-*-freebsd*): Remove.
2285 * config/alpha/freebsd.h: Remove.
2287 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2290 * spellcheck-tree.c (levenshtein_distance): Rename to...
2291 (get_edit_distance): ...this, and update for underlying renaming.
2292 * spellcheck-tree.h (levenshtein_distance): Rename to...
2293 (get_edit_distance): ...this.
2294 * spellcheck.c (levenshtein_distance): Rename to...
2295 (get_edit_distance): ...this. Convert from Levenshtein distance
2296 to Damerau-Levenshtein distance by supporting transpositions of
2297 adjacent characters. Rename "v1" to "v_next" and "v0" to
2299 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2300 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2301 update for underlying renaming.
2302 (selftest::levenshtein_distance_unit_test): Rename to...
2303 (selftest::test_get_edit_distance_unit): ...this, and update for
2304 underlying renaming.
2305 (selftest::test_find_closest_string): Add example from PR 69968
2306 where transposition helps
2307 (selftest::test_metric_conditions): Update for renaming.
2308 (selftest::test_metric_conditions): Likewise.
2309 (selftest::spellcheck_c_tests): Likewise.
2310 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2311 (get_edit_distance): ...this.
2312 (best_match::consider): Update for renaming.
2314 2018-06-12 Martin Sebor <msebor@redhat.com>
2316 PR tree-optimization/85259
2317 * builtins.c (compute_objsize): Handle constant offsets.
2318 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2319 true iff a warning has been issued.
2320 * gimple.h (gimple_nonartificial_location): New function.
2321 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2322 gimple_nonartificial_location and handle -Wno-system-headers.
2323 (handle_builtin_stxncpy): Same.
2325 2018-06-12 Martin Sebor <msebor@redhat.com>
2328 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2330 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2332 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2333 BUILTIN_VEC_XST entries for pointer to double and long long.
2335 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2338 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2340 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2343 2018-06-12 Martin Liska <mliska@suse.cz>
2345 * doc/options.texi: Document IntegerRange.
2347 2018-06-12 Martin Liska <mliska@suse.cz>
2349 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2350 * opt-functions.awk: Handle Deprecated flag.
2351 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2353 (read_cmdline_option): Report warning for a deprecated option.
2354 * opts.h (struct cl_option): Add new field cl_deprecated.
2355 (CL_ERR_DEPRECATED): New.
2357 2018-06-12 Martin Liska <mliska@suse.cz>
2359 * doc/options.texi: Document Deprecated option flag.
2361 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2363 * config/arc/arc-arch.h (arc_extras): New enum.
2364 (arc_cpu_t):Add field extra.
2365 (arc_cpu_types): Consider the extras.
2366 * config/arc/arc-cpus.def: Add extras info.
2367 * config/arc/arc-opts.h (processor_type): Consider extra field.
2368 * config/arc/arc.c (arc_override_options): Handle extra field.
2370 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2372 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2373 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2375 * config/arc/arc.c (arc_init): Update pic warning.
2376 (irq_range): Update irq range parsing warnings.
2377 (arc_override_options): Update various warning messages.
2378 (arc_handle_aux_attribute): Likewise.
2380 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2382 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2384 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2386 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2387 compiler messages directives.
2389 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2391 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2392 * config/mips/mips-tables.opt: Regenerate.
2393 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2395 * doc/invoke.texi: Document -march=i6500.
2397 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2399 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2400 (i6400_gpmul): Add cpu_unit.
2401 (i6400_gpdiv): Likewise.
2402 (i6400_msa_add_d): Update reservations.
2403 (i6400_msa_int_add) Likewise.
2404 (i6400_msa_short_logic3) Likewise.
2405 (i6400_msa_short_logic2) Likewise.
2406 (i6400_msa_short_logic) Likewise.
2407 (i6400_msa_move) Likewise.
2408 (i6400_msa_cmp) Likewise.
2409 (i6400_msa_short_float2) Likewise.
2410 (i6400_msa_div_d) Likewise.
2411 (i6400_msa_long_logic1) Likewise.
2412 (i6400_msa_long_logic2) Likewise.
2413 (i6400_msa_mult) Likewise.
2414 (i6400_msa_long_float2) Likewise.
2415 (i6400_msa_long_float4) Likewise.
2416 (i6400_msa_long_float5) Likewise.
2417 (i6400_msa_long_float8) Likewise.
2418 (i6400_fpu_fadd): Include frint type.
2419 (i6400_fpu_store): New define_insn_reservation.
2420 (i6400_fpu_load): Likewise.
2421 (i6400_fpu_move): Likewise.
2422 (i6400_fpu_fcmp): Likewise.
2423 (i6400_fpu_fmadd): Likewise.
2424 (i6400_int_mult): Include imul3nc type and update reservation.
2425 (i6400_int_div): Include idiv3 type and update reservation.
2426 (i6400_int_load): Update to check type not move_type.
2427 (i6400_int_store): Likewise.
2428 (i6400_int_prefetch): Set zero latency.
2430 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2432 * gcc.c: Document new %@{...} sequence.
2433 (LINK_COMMAND_SPEC): Use it for the -L switches.
2434 (cpp_unique_options): Use it for the -I switches.
2435 (at_file_argbuf): New global variable.
2436 (in_at_file): Likewise.
2437 (alloc_args): Create at_file_argbuf.
2438 (clear_args): Truncate at_file_argbuf.
2439 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2440 (open_at_file): New function.
2441 (close_at_file): Likewise.
2442 (create_at_file): Delete.
2443 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2446 (validate_switches_from_spec): Deal with %@{...} sequence.
2447 (validate_switches): Likewise.
2448 (driver::finalize): Call clear_args.
2450 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2452 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2454 2018-06-11 Martin Sebor <msebor@redhat.com>
2456 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2457 (Wno-class-memaccess): Add @opindex.
2458 (Wno-templates, Wno-multiple-inheritance): Same.
2459 (Wno-virtual-inheritance, Wno-namespaces): Same.
2460 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2461 (Wno-format-overflow, Wno-format-truncation): Same.
2462 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2463 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2464 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2465 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2466 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2467 (Wno-misspelled-isr): Same.
2469 2018-06-11 Martin Sebor <msebor@redhat.com>
2471 * PR tree-optimization/86083
2472 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2474 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2476 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2478 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2481 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2482 on the correct operand.
2483 (*movdi_internal64): Ditto.
2485 2018-06-11 Martin Liska <mliska@suse.cz>
2487 PR tree-optimization/86089
2488 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2490 2018-06-11 Julia Koval <julia.koval@intel.com>
2492 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2493 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2494 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2496 2018-06-11 Olivier Hainque <hainque@adacore.com>
2498 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2499 for Ada with strict dwarf2.
2501 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2504 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2507 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2509 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2511 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2513 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2516 2018-06-08 Martin Liska <mliska@suse.cz>
2518 * tree-cfg.h (debug_function): Fix argument type to match
2521 2018-06-08 Martin Liska <mliska@suse.cz>
2523 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2524 Remove usage of MPX-related (and removed) fields.
2525 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2527 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2529 * cfg.c (debug): Use TDF_NONE rather than 0.
2530 * cfghooks.c (debug): Likewise.
2531 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2532 (struct dump_option_value_info): Convert to...
2533 (struct kv_pair): ...this template type.
2534 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2536 (optinfo_verbosity_options): Likewise.
2537 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2539 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2540 than int for "optgroup_flags" param.
2541 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2543 (dump_dec): Likewise.
2544 (dump_finish): Use TDF_NONE rather than 0.
2545 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2546 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2547 than 0. Update for change to option_ptr.
2548 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2549 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2550 0. Update for changes to optinfo_verbosity_options and
2552 (opt_info_switch_p): Convert optgroup_flags from int to
2554 (dump_basic_block): Use dump_flags_t rather than int
2555 for "dump_kind" param.
2556 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2557 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2558 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2559 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2560 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2561 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2562 TDF_NONE): Convert from macros to...
2563 (enum dump_flag): ...this new enum.
2564 (dump_flags_t): Update to use enum.
2565 (operator|, operator&, operator~, operator|=, operator&=):
2566 Implement for dump_flags_t.
2567 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2568 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2569 Convert from macros to...
2570 (enum optgroup_flag): ...this new enum.
2571 (optgroup_flags_t): New typedef.
2572 (operator|, operator|=): Implement for optgroup_flags_t.
2573 (struct dump_file_info): Convert field "alt_flags" to
2574 dump_flags_t. Convert field "optgroup_flags" to
2576 (dump_basic_block): Use dump_flags_t rather than int for param.
2577 (dump_generic_expr_loc): Likewise.
2578 (dump_dec): Likewise.
2579 (dump_register): Convert param "optgroup_flags" to
2581 (opt_info_enable_passes): Likewise.
2582 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2584 * gimple-pretty-print.c (debug): Likewise.
2585 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2586 (merged_store_group::apply_stores): Likewise.
2587 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2588 * gimple.c (verify_gimple_pp): Likewise.
2589 * graphite-poly.c (print_pbb_body): Likewise.
2590 * passes.c (pass_manager::register_one_dump_file): Convert
2591 local "optgroup_flags" to optgroup_flags_t.
2592 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2594 (debug_body): Likewise.
2595 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2596 to optgroup_flags_t.
2597 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2599 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2600 (convert_mult_to_fma): Likewise.
2601 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2602 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2603 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2604 "dump_kind" to dump_flags_t.
2606 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2608 * config/rs6000/rs6000.c (min, max): Delete.
2610 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2612 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2613 -mabi=spe and -mabi=no-spe.
2615 2018-06-08 Martin Liska <mliska@suse.cz>
2617 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2618 where we expect an existing summary.
2620 2018-06-08 Martin Liska <mliska@suse.cz>
2622 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2623 * ipa-inline.h (estimate_edge_growth): Likewise.
2625 2018-06-08 Martin Liska <mliska@suse.cz>
2627 * cgraph.c (function_version_hasher::hash): Use
2628 cgraph_node::get_uid ().
2629 (function_version_hasher::equal):
2630 * cgraph.h (cgraph_node::get_uid): New method.
2631 * ipa-inline.c (update_caller_keys): Use
2632 cgraph_node::get_uid ().
2633 (update_callee_keys): Likewise.
2634 * ipa-utils.c (searchc): Likewise.
2635 (ipa_reduced_postorder): Likewise.
2636 * lto-cgraph.c (input_node): Likewise.
2637 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2638 * symbol-summary.h (symtab_insertion): Likewise.
2639 (symtab_removal): Likewise.
2640 (symtab_duplication): Likewise.
2641 * tree-pretty-print.c (dump_function_header): Likewise.
2642 * tree-sra.c (convert_callers_for_node): Likewise.
2644 2018-06-08 Martin Liska <mliska@suse.cz>
2646 * cgraph.c (symbol_table::create_edge): Always assign a new
2648 (symbol_table::free_edge): Do not recycle numbers.
2649 * cgraph.h (cgraph_edge::get): New method.
2650 * symbol-summary.h (symtab_removal): Use it.
2651 (symtab_duplication): Likewise.
2652 (call_summary::hashable_uid): Remove.
2654 2018-06-08 Martin Liska <mliska@suse.cz>
2656 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2657 (initialize_growth_caches): Remove.
2658 (free_growth_caches): Likewise.
2659 (do_estimate_edge_time): Use edge_growth_cache.
2660 (do_estimate_edge_size): Likewise.
2661 (do_estimate_edge_hints): Likewise.
2662 * ipa-inline.c (reset_edge_caches): Likewise.
2663 (recursive_inlining): Likewise.
2664 (inline_small_functions): Likewise.
2665 * ipa-inline.h (initialize_growth_caches): Remove.
2666 (estimate_edge_size): Likewise.
2667 (estimate_edge_time): Likewise.
2668 (estimate_edge_hints): Likewise.
2669 (reset_edge_growth_cache): Likewise.
2670 * symbol-summary.h (call_summary::remove): New method.
2672 2018-06-08 Martin Liska <mliska@suse.cz>
2674 * ipa-cp.c (class edge_clone_summary): New summary.
2675 (grow_edge_clone_vectors): Remove.
2676 (ipcp_edge_duplication_hook): Remove.
2677 (class edge_clone_summary_t): New call_summary class.
2678 (ipcp_edge_removal_hook): Remove.
2679 (edge_clone_summary_t::duplicate): New function.
2680 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2681 (create_specialized_node): Likewise.
2682 (ipcp_driver): Initialize edge_clone_summaries and do not
2685 2018-06-08 Martin Liska <mliska@suse.cz>
2687 * symbol-summary.h (get): New function.
2688 (call_summary::m_initialize_when_cloning): New class member.
2690 2018-06-08 Martin Liska <mliska@suse.cz>
2692 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2693 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2694 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2695 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2696 (remove_cgraph_node_from_order): Use the removed_nodes set.
2697 (do_per_function_toporder): Likwise.
2698 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2699 instead of summary_uid.
2700 (symtab_removal): Likewise.
2701 (symtab_duplication): Likewise.
2703 2018-06-08 Martin Liska <mliska@suse.cz>
2705 * ipa-cp.c (ipcp_store_bits_results): Use
2706 ipcp_transformation_sum.
2707 (ipcp_store_vr_results): Likewise.
2708 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2710 (ipcp_transformation_initialize): ... this.
2711 (ipa_set_node_agg_value_chain):
2712 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2713 (write_ipcp_transformation_info): Likewise.
2714 (read_ipcp_transformation_info): Likewise.
2715 (ipcp_update_bits): Likewise.
2716 (ipcp_update_vr): Likewise.
2717 (ipcp_transform_function): Likewise.
2718 * ipa-prop.h: Rename ipcp_transformation_summary to
2719 ipcp_transformation.
2720 (class ipcp_transformation_t): New function summary.
2721 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2722 (ipa_get_agg_replacements_for_node): Likewise.
2724 2018-06-08 Martin Liska <mliska@suse.cz>
2726 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2728 (class funct_state_summary_t): New function_summary class.
2729 (has_function_state): Remove.
2730 (get_function_state): Likewise.
2731 (set_function_state): Likewise.
2732 (add_new_function): Likewise.
2733 (funct_state_summary_t::insert): New function.
2734 (duplicate_node_data): Remove.
2735 (remove_node_data): Remove.
2736 (funct_state_summary_t::duplicate): New function.
2737 (register_hooks): Create new funct_state_summaries.
2738 (pure_const_generate_summary): Use it.
2739 (pure_const_write_summary): Likewise.
2740 (pure_const_read_summary): Likewise.
2741 (propagate_pure_const): Likewise.
2742 (propagate_nothrow): Likewise.
2743 (dump_malloc_lattice): Likewise.
2744 (propagate_malloc): Likewise.
2745 (execute): Do not register hooks, just remove summary
2747 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2750 2018-06-08 Martin Liska <mliska@suse.cz>
2752 * ipa-reference.c (remove_node_data): Remove.
2753 (duplicate_node_data): Likewise.
2754 (class ipa_ref_var_info_summary_t): New class.
2755 (class ipa_ref_opt_summary_t): Likewise.
2756 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2757 (get_reference_optimization_summary): Use
2758 ipa_ref_opt_sum_summaries.
2759 (set_reference_vars_info): Remove.
2760 (set_reference_optimization_summary): Likewise.
2761 (ipa_init): Create summaries.
2762 (init_function_info): Use function summary.
2763 (ipa_ref_opt_summary_t::duplicate): New function.
2764 (ipa_ref_opt_summary_t::remove): New function.
2765 (get_read_write_all_from_node): Fix GNU coding style.
2766 (propagate): Use function summary.
2767 (write_node_summary_p): Fix GNU coding style.
2768 (stream_out_bitmap): Likewise.
2769 (ipa_reference_read_optimization_summary): Use function summary.
2770 (ipa_reference_c_finalize): Do not release hooks.
2772 2018-06-08 Martin Liska <mliska@suse.cz>
2774 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2775 (analyze_function_body): Extract multiple calls of get_create.
2776 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2777 * ipa-inline.c (recursive_inlining): Use ::get method.
2778 * ipa-inline.h (estimate_edge_growth): Likewise.
2780 2018-06-08 Martin Liska <mliska@suse.cz>
2782 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2784 (hsa_function_summary::hsa_function_summary): Use the new enum
2786 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2787 * hsa-gen.c (hsa_get_host_function): Likewise.
2788 (get_brig_function_name): Likewise.
2789 * ipa-hsa.c (process_hsa_functions): Likewise.
2790 (ipa_hsa_write_summary): Likewise.
2791 * symbol-summary.h (symtab_duplication): Use ::get function/
2792 (get): New function.
2794 2018-06-08 Martin Liska <mliska@suse.cz>
2796 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2798 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2799 (hsa_register_kernel): Likewise.
2800 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2801 * hsa-gen.c (hsa_get_host_function): Likewise.
2802 (get_brig_function_name): Likewise.
2803 (generate_hsa): Likewise.
2804 (pass_gen_hsail::execute): Likewise.
2805 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2806 (devirtualization_time_bonus): Likewise.
2807 (ipcp_propagate_stage): Likewise.
2808 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2809 (edge_set_predicate): Likewise.
2810 (evaluate_conditions_for_known_args): Likewise.
2811 (evaluate_properties_for_edge): Likewise.
2812 (ipa_fn_summary::reset): Likewise.
2813 (ipa_fn_summary_t::duplicate): Likewise.
2814 (dump_ipa_call_summary): Likewise.
2815 (ipa_dump_fn_summary): Likewise.
2816 (analyze_function_body): Likewise.
2817 (compute_fn_summary): Likewise.
2818 (estimate_edge_devirt_benefit): Likewise.
2819 (estimate_edge_size_and_time): Likewise.
2820 (estimate_calls_size_and_time): Likewise.
2821 (estimate_node_size_and_time): Likewise.
2822 (inline_update_callee_summaries): Likewise.
2823 (remap_edge_change_prob): Likewise.
2824 (remap_edge_summaries): Likewise.
2825 (ipa_merge_fn_summary_after_inlining): Likewise.
2826 (ipa_update_overall_fn_summary): Likewise.
2827 (read_ipa_call_summary): Likewise.
2828 (inline_read_section): Likewise.
2829 (write_ipa_call_summary): Likewise.
2830 (ipa_fn_summary_write): Likewise.
2831 (ipa_free_fn_summary): Likewise.
2832 * ipa-hsa.c (process_hsa_functions): Likewise.
2833 (ipa_hsa_write_summary): Likewise.
2834 (ipa_hsa_read_section): Likewise.
2835 * ipa-icf.c (sem_function::merge): Likewise.
2836 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2837 (do_estimate_edge_time): Likewise.
2838 (estimate_size_after_inlining): Likewise.
2839 (estimate_growth): Likewise.
2840 (growth_likely_positive): Likewise.
2841 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2842 (inline_call): Likewise.
2843 * ipa-inline.c (caller_growth_limits): Likewise.
2844 (can_inline_edge_p): Likewise.
2845 (can_inline_edge_by_limits_p): Likewise.
2846 (compute_uninlined_call_time): Likewise.
2847 (compute_inlined_call_time): Likewise.
2848 (want_inline_small_function_p): Likewise.
2849 (edge_badness): Likewise.
2850 (update_caller_keys): Likewise.
2851 (update_callee_keys): Likewise.
2852 (recursive_inlining): Likewise.
2853 (inline_small_functions): Likewise.
2854 (inline_to_all_callers_1): Likewise.
2855 (dump_overall_stats): Likewise.
2856 (early_inline_small_functions): Likewise.
2857 (early_inliner): Likewise.
2858 * ipa-inline.h (estimate_edge_growth): Likewise.
2859 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2860 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2861 * ipa-prop.h (IPA_NODE_REF): Likewise.
2862 (IPA_EDGE_REF): Likewise.
2863 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2864 (propagate_malloc): Likewise.
2865 * ipa-split.c (execute_split_functions): Likewise.
2866 * symbol-summary.h: Rename get to get_create.
2868 (get_create): Likewise.
2869 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2871 2018-06-08 Martin Liska <mliska@suse.cz>
2873 * symbol-summary.h (release): Move definition out of class
2875 (symtab_removal): Likewise.
2876 (symtab_duplication): Likewise.
2878 2018-06-08 Martin Liska <mliska@suse.cz>
2880 * symbol-summary.h (function_summary): Move constructor
2881 implementation out of class declaration.
2882 (release): Likewise.
2883 (symtab_insertion): Likewise.
2884 (symtab_removal): Likewise.
2885 (symtab_duplication): Likewise.
2888 2018-06-08 Martin Liska <mliska@suse.cz>
2890 * Makefile.in: Remove support for MPX (macros, related functions,
2891 fields in cgraph_node, ...).
2892 * builtin-types.def (BT_BND): Likewise.
2893 (BT_FN_BND_CONST_PTR): Likewise.
2894 (BT_FN_CONST_PTR_BND): Likewise.
2895 (BT_FN_VOID_PTR_BND): Likewise.
2896 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2897 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2898 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2899 (expand_builtin_mempcpy_with_bounds): Likewise.
2900 (expand_builtin_memset_with_bounds): Likewise.
2901 (expand_builtin_memset_args): Likewise.
2902 (std_expand_builtin_va_start): Likewise.
2903 (expand_builtin): Likewise.
2904 (expand_builtin_with_bounds): Likewise.
2905 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2906 (DEF_LIB_BUILTIN_CHKP): Likewise.
2907 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2908 (DEF_CHKP_BUILTIN): Likewise.
2909 (BUILT_IN_MEMCPY): Likewise.
2910 (BUILT_IN_MEMMOVE): Likewise.
2911 (BUILT_IN_MEMPCPY): Likewise.
2912 (BUILT_IN_MEMSET): Likewise.
2913 (BUILT_IN_STPCPY): Likewise.
2914 (BUILT_IN_STRCAT): Likewise.
2915 (BUILT_IN_STRCHR): Likewise.
2916 (BUILT_IN_STRCPY): Likewise.
2917 (BUILT_IN_STRLEN): Likewise.
2918 (BUILT_IN_MEMCPY_CHK): Likewise.
2919 (BUILT_IN_MEMMOVE_CHK): Likewise.
2920 (BUILT_IN_MEMPCPY_CHK): Likewise.
2921 (BUILT_IN_MEMSET_CHK): Likewise.
2922 (BUILT_IN_STPCPY_CHK): Likewise.
2923 (BUILT_IN_STRCAT_CHK): Likewise.
2924 (BUILT_IN_STRCPY_CHK): Likewise.
2925 * calls.c (store_bounds): Likewise.
2926 (emit_call_1): Likewise.
2927 (special_function_p): Likewise.
2928 (maybe_warn_nonstring_arg): Likewise.
2929 (initialize_argument_information): Likewise.
2930 (finalize_must_preallocate): Likewise.
2931 (compute_argument_addresses): Likewise.
2932 (expand_call): Likewise.
2933 * cfgexpand.c (expand_call_stmt): Likewise.
2934 (expand_return): Likewise.
2935 (expand_gimple_stmt_1): Likewise.
2936 (pass_expand::execute): Likewise.
2937 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2938 (cgraph_node::remove): Likewise.
2939 (cgraph_node::dump): Likewise.
2940 (cgraph_node::verify_node): Likewise.
2941 * cgraph.h (chkp_function_instrumented_p): Likewise.
2942 (symtab_node::get_alias_target): Likewise.
2943 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2944 (cgraph_local_p): Likewise.
2945 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2946 (cgraph_edge::rebuild_references): Likewise.
2947 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2948 (walk_polymorphic_call_targets): Likewise.
2949 (cgraph_node::expand_thunk): Likewise.
2950 (symbol_table::output_weakrefs): Likewise.
2951 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2952 (ix86_handle_option): Likewise.
2953 * config/i386/constraints.md: Likewise.
2954 * config/i386/i386-builtin-types.def (BND): Likewise.
2958 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2959 (BDESC_FIRST): Likewise.
2961 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2962 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2963 * config/i386/i386.c (enum reg_class): Likewise.
2964 (ix86_target_string): Likewise.
2965 (ix86_option_override_internal): Likewise.
2966 (ix86_conditional_register_usage): Likewise.
2967 (ix86_valid_target_attribute_inner_p): Likewise.
2968 (ix86_set_indirect_branch_type): Likewise.
2969 (ix86_set_current_function): Likewise.
2970 (ix86_function_arg_regno_p): Likewise.
2971 (init_cumulative_args): Likewise.
2972 (ix86_function_arg_advance): Likewise.
2973 (ix86_function_arg): Likewise.
2974 (ix86_pass_by_reference): Likewise.
2975 (ix86_function_value_regno_p): Likewise.
2976 (ix86_function_value_1): Likewise.
2977 (ix86_function_value_bounds): Likewise.
2978 (ix86_return_in_memory): Likewise.
2979 (ix86_setup_incoming_vararg_bounds): Likewise.
2980 (ix86_va_start): Likewise.
2981 (indirect_thunk_need_prefix): Likewise.
2982 (print_reg): Likewise.
2983 (ix86_print_operand): Likewise.
2984 (ix86_expand_call): Likewise.
2985 (ix86_output_function_return): Likewise.
2986 (reg_encoded_number): Likewise.
2987 (BDESC_VERIFYS): Likewise.
2988 (ix86_init_mpx_builtins): Likewise.
2989 (ix86_init_builtins): Likewise.
2990 (ix86_emit_cmove): Likewise.
2991 (ix86_emit_move_max): Likewise.
2992 (ix86_expand_builtin): Likewise.
2993 (ix86_builtin_mpx_function): Likewise.
2994 (ix86_get_arg_address_for_bt): Likewise.
2995 (ix86_load_bounds): Likewise.
2996 (ix86_store_bounds): Likewise.
2997 (ix86_load_returned_bounds): Likewise.
2998 (ix86_store_returned_bounds): Likewise.
2999 (ix86_class_likely_spilled_p): Likewise.
3000 (ix86_hard_regno_mode_ok): Likewise.
3001 (x86_order_regs_for_local_alloc): Likewise.
3002 (ix86_mitigate_rop): Likewise.
3003 (ix86_bnd_prefixed_insn_p): Likewise.
3004 (ix86_mpx_bound_mode): Likewise.
3005 (ix86_make_bounds_constant): Likewise.
3006 (ix86_initialize_bounds): Likewise.
3007 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3008 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3009 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3010 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3011 (TARGET_CHKP_BOUND_MODE): Likewise.
3012 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3013 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3014 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3015 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3016 * config/i386/i386.h (TARGET_MPX): Likewise.
3017 (TARGET_MPX_P): Likewise.
3018 (VALID_BND_REG_MODE): Likewise.
3019 (FIRST_BND_REG): Likewise.
3020 (LAST_BND_REG): Likewise.
3021 (enum reg_class): Likewise.
3022 (BND_REG_P): Likewise.
3023 (BND_REGNO_P): Likewise.
3024 (BNDmode): Likewise.
3025 (ADJUST_INSN_LENGTH): Likewise.
3026 * config/i386/i386.md: Likewise.
3027 * config/i386/i386.opt: Likewise.
3028 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3029 (defined): Likewise.
3030 (LINK_MPX): Likewise.
3031 (MPX_SPEC): Likewise.
3032 (LIBMPX_SPEC): Likewise.
3033 (LIBMPXWRAPPERS_SPEC): Likewise.
3034 (CHKP_SPEC): Likewise.
3035 * config/i386/predicates.md: Likewise.
3036 * dbxout.c (dbxout_type): Likewise.
3037 * doc/extend.texi: Likewise.
3038 * doc/invoke.texi: Likewise.
3039 * doc/md.texi: Likewise.
3040 * doc/tm.texi: Likewise.
3041 * doc/tm.texi.in: Likewise.
3042 * dwarf2out.c (is_base_type): Likewise.
3043 (gen_formal_types_die): Likewise.
3044 (gen_subprogram_die): Likewise.
3045 (gen_type_die_with_usage): Likewise.
3046 (gen_decl_die): Likewise.
3047 (dwarf2out_late_global_decl): Likewise.
3048 * expr.c (expand_assignment): Likewise.
3049 (emit_storent_insn): Likewise.
3050 (store_expr_with_bounds): Likewise.
3051 (store_expr): Likewise.
3052 (expand_expr_real_1): Likewise.
3053 * expr.h (store_expr_with_bounds): Likewise.
3054 * function.c (use_register_for_decl): Likewise.
3055 (struct bounds_parm_data): Likewise.
3056 (assign_parms_augmented_arg_list): Likewise.
3057 (assign_parm_find_entry_rtl): Likewise.
3058 (assign_parm_is_stack_parm): Likewise.
3059 (assign_parm_load_bounds): Likewise.
3060 (assign_bounds): Likewise.
3061 (assign_parms): Likewise.
3062 (expand_function_start): Likewise.
3063 * gcc.c (CHKP_SPEC): Likewise.
3064 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3065 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3066 (wrestrict_dom_walker::check_call): Likewise.
3067 * gimple.c (gimple_build_call_from_tree): Likewise.
3068 * gimple.h (enum gf_mask): Likewise.
3069 (gimple_call_with_bounds_p): Likewise.
3070 (gimple_call_set_with_bounds): Likewise.
3071 * gimplify.c (gimplify_init_constructor): Likewise.
3072 * ipa-cp.c (initialize_node_lattices): Likewise.
3073 (propagate_constants_across_call): Likewise.
3074 (find_more_scalar_values_for_callers_subset): Likewise.
3075 * ipa-hsa.c (process_hsa_functions): Likewise.
3076 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3077 * ipa-icf.c (sem_function::merge): Likewise.
3078 * ipa-inline.c (early_inliner): Likewise.
3079 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3080 (warn_function_cold): Likewise.
3081 (propagate_pure_const): Likewise.
3082 * ipa-ref.h (enum GTY): Likewise.
3083 * ipa-split.c (find_retbnd): Likewise.
3084 (consider_split): Likewise.
3085 (split_function): Likewise.
3086 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3087 * ipa.c (walk_polymorphic_call_targets): Likewise.
3088 (symbol_table::remove_unreachable_nodes): Likewise.
3089 (process_references): Likewise.
3090 (cgraph_build_static_cdtor_1): Likewise.
3091 * lto-cgraph.c (lto_output_node): Likewise.
3092 (output_refs): Likewise.
3093 (compute_ltrans_boundary): Likewise.
3094 (input_overwrite_node): Likewise.
3095 (input_node): Likewise.
3096 (input_cgraph_1): Likewise.
3097 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3098 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3099 (class pass_chkp_instrumentation_passes): Likewise.
3100 (make_pass_chkp_instrumentation_passes): Likewise.
3101 * passes.def: Likewise.
3102 * rtl.h (struct GTY): Likewise.
3103 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3104 * stor-layout.c (layout_type): Likewise.
3105 * symtab.c: Likewise.
3106 * target.def: Likewise.
3107 * targhooks.c (default_chkp_bound_type): Likewise.
3108 (default_chkp_bound_mode): Likewise.
3109 (default_builtin_chkp_function): Likewise.
3110 (default_chkp_function_value_bounds): Likewise.
3111 (default_chkp_make_bounds_constant): Likewise.
3112 (default_chkp_initialize_bounds): Likewise.
3113 * targhooks.h (default_chkp_bound_type): Likewise.
3114 (default_chkp_bound_mode): Likewise.
3115 (default_builtin_chkp_function): Likewise.
3116 (default_chkp_function_value_bounds): Likewise.
3117 (default_chkp_make_bounds_constant): Likewise.
3118 (default_chkp_initialize_bounds): Likewise.
3119 * toplev.c (compile_file): Likewise.
3120 (process_options): Likewise.
3121 * tree-core.h (DEF_BUILTIN): Likewise.
3122 (DEF_BUILTIN_CHKP): Likewise.
3123 * tree-inline.c (declare_return_variable): Likewise.
3124 (remap_gimple_stmt): Likewise.
3125 (copy_bb): Likewise.
3126 (initialize_inlined_parameters): Likewise.
3127 (expand_call_inline): Likewise.
3128 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3129 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3130 (make_pass_ipa_chkp_produce_thunks): Likewise.
3131 (make_pass_chkp): Likewise.
3132 (make_pass_chkp_opt): Likewise.
3133 (make_pass_chkp_instrumentation_passes): Likewise.
3134 * tree-pretty-print.c (dump_generic_node): Likewise.
3135 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3136 * tree-ssa-dce.c (propagate_necessity): Likewise.
3137 (eliminate_unnecessary_stmts): Likewise.
3138 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3139 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3140 * tree-ssa-sccvn.h: Likewise.
3141 * tree-ssa-strlen.c (get_string_length): Likewise.
3142 (valid_builtin_call): Likewise.
3143 (adjust_last_stmt): Likewise.
3144 (handle_builtin_strchr): Likewise.
3145 (handle_builtin_strcpy): Likewise.
3146 (handle_builtin_stxncpy): Likewise.
3147 (handle_builtin_memcpy): Likewise.
3148 (handle_builtin_strcat): Likewise.
3149 (strlen_check_and_optimize_stmt): Likewise.
3150 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3151 * tree-streamer-in.c: Likewise.
3152 * tree-streamer.c (record_common_node): Likewise.
3153 * tree.c (tree_code_size): Likewise.
3154 (wide_int_to_tree_1): Likewise.
3155 (type_contains_placeholder_1): Likewise.
3156 (build_common_tree_nodes): Likewise.
3157 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3158 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3159 (POINTER_BOUNDS_P): Likewise.
3160 (BOUNDED_TYPE_P): Likewise.
3161 (BOUNDED_P): Likewise.
3162 (CALL_WITH_BOUNDS_P): Likewise.
3163 (pointer_bounds_type_node): Likewise.
3164 * value-prof.c (gimple_ic): Likewise.
3165 * var-tracking.c (vt_add_function_parameters): Likewise.
3166 * varasm.c (make_decl_rtl): Likewise.
3167 (assemble_start_function): Likewise.
3168 (output_constant): Likewise.
3169 (maybe_assemble_visibility): Likewise.
3170 * varpool.c (ctor_for_folding): Likewise.
3171 * chkp-builtins.def: Remove.
3172 * ipa-chkp.c: Remove.
3173 * ipa-chkp.h: Remove.
3174 * rtl-chkp.c: Remove.
3175 * rtl-chkp.h: Remove.
3176 * tree-chkp-opt.c: Remove.
3177 * tree-chkp.c: Remove.
3178 * tree-chkp.h: Remove.
3180 2018-06-07 Carl Love <cel@us.ibm.com>
3182 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3183 vextract_fp_from_shortl): Add BE support.
3185 2018-06-07 Paul Koning <ni1d@arrl.net>
3187 * compare-elim.c (try_merge_compare): Don't merge compare if
3188 address contains a side effect.
3189 (try_eliminate_compare): Likewise.
3191 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3193 * config.gcc: Support "tremont".
3194 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3195 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3197 * config/i386/i386.c (m_TREMONT): Define.
3198 (processor_target_table): Add "tremont".
3199 (PTA_TREMONT): Define.
3200 (ix86_lea_outperforms): Add TARGET_TREMONT.
3201 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3202 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3203 and M_INTEL_GOLDMONT_PLUS.
3204 (fold_builtin_cpu): Add "tremont".
3205 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3206 (ix86_option_override_internal): Add "tremont".
3207 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3208 (processor_type): Add PROCESSOR_TREMONT.
3209 * config/i386/x86-tune.def: Add m_TREMONT.
3210 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3212 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3214 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3215 symbol defined for msp430i* devices to be lower case.
3217 gcc/testsuite/gcc.target/msp430/
3218 * msp430i-device-symbol.c: New test.
3219 * msp430f-device-symbol.c: New test.
3220 * msp430.h: New test header file.
3222 2018-06-07 Richard Biener <rguenther@suse.de>
3224 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3225 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3226 Properly wrap signed arithmetic if overflow wraps.
3228 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3230 PR tree-optimization/69615
3231 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3232 of a cast from a same precision integral SSA_NAME in a bb dominated
3233 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3234 cast to utype if rhs2 has already a compatible type.
3236 2018-06-07 Richard Biener <rguenther@suse.de>
3238 PR tree-optimization/85935
3239 * graphite-scop-detection.c (find_params_in_bb): Analyze
3240 condition operands with respect to the correct loop. Assert
3241 the analysis doesn't fail.
3243 2018-06-04 Carl Love <cel@us.ibm.com>
3245 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3246 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3247 as it is slightly cheaper.
3248 (first_match_or_eos_index_<mode>):
3249 Calculate index using natural element order.
3250 (first_match_index_<mode>):
3251 Calculate index using natural element order.
3252 (first_match_or_eos_index_<mode>):
3253 Calculate index using natural order.
3254 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3255 for BE and LE modes.
3256 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3257 P9V_BUILTIN_VCLZLSBB_V16QI.
3258 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3261 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3263 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3264 indentation and line wrap for many prototypes. Add missing
3265 @smallexample directives around block of prototypes for vec_xl and
3268 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3270 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3271 track if we pass or return IEEE 128-bit floating point.
3272 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3273 C++ mangling that is compatible with GCC 8.1.
3274 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3275 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3276 floating point types.
3277 (rs6000_function_arg_advance_1): Likewise.
3278 (rs6000_mangle_type): Optionally generate mangled names that match
3279 what GCC 8.1 generated for IEEE 128-bit floating point types.
3280 (rs6000_globalize_decl_name): If we have an external function that
3281 passes or returns IEEE 128-bit types, generate a weak reference
3282 from the mangled name used in GCC 8.1 to the current mangled
3284 (rs6000_init_builtins): Make __ibm128 use the long double type if
3285 long double is IBM extended double. Make __float128 use the long
3286 double type if long double is IEEE 128-bit.
3289 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3290 macro for __ibm128 built-in functions.
3291 (PACK_IF): Add __ibm128 pack/unpack functions.
3292 (UNPACK_IF): Likewise.
3293 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3294 enable long double built-in functions if long double is IEEE
3295 128-bit floating point.
3296 (rs6000_invalid_builtin): Update long double built-in function
3298 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3299 functions, adjust the built-in function to use the long double
3300 built-in function if __ibm128 and long double are the same type.
3301 * doc/extend.texi (PowerPC builtins): Update documention for
3302 __builtin_{,un}pack_longdouble. Add documentation for
3303 __builtin_{,un}pack_ibm128.
3305 2018-06-06 Jim Wilson <jimw@sifive.com>
3307 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3308 (struct machine_function): New field interrupt_mode.
3309 (riscv_handle_type_attribute): New function. Add forward declaration.
3310 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3311 (riscv_expand_epilogue): Check interrupt_mode field.
3312 (riscv_set_current_function): Check interrupt attribute args and
3313 set interrupt_mode field.
3314 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3315 (riscv_sret, riscv_uret): New.
3316 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3317 new arguments to interrupt attribute.
3319 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3322 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3323 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3325 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3327 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3328 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3329 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3330 several redundant entries.
3332 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3334 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3335 type from "rtx" to "rtx_insn *".
3336 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3337 for local "call_insn", removing cast.
3338 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3340 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3342 PR tree-optimization/86066
3343 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3344 for BIT_INSERT_EXPR stores.
3346 2018-06-06 Richard Biener <rguenther@suse.de>
3348 PR tree-optimization/86062
3349 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3350 component refs ontop
3351 of to be offsetted base.
3353 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3355 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3356 to be static and remove check on interrupt attribute name.
3358 gcc/testsuite/gcc.target/msp430/
3359 * function-attributes-4.c: New test.
3360 * static-interrupts.c: New test.
3362 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3364 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3365 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3367 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3370 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3372 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3373 Remove second argument, change how error is called.
3374 (aarch64_layout_arg): Remove second argument from
3375 aarch64_err_no_fpadvsimd call.
3376 (aarch64_init_cumulative_args): Ditto.
3377 (aarch64_gimplify_va_arg_expr): Ditto.
3378 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3380 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3382 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3383 (*simple_return_indirect_internal<mode>): Rename from
3384 simple_return_indirect_internal. Use W mode iterator.
3385 (rstorssp): New expander.
3386 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3387 (clrssbsy): New expander.
3388 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3390 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3392 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3394 (cmse_check_pointed_object): Likewise.
3396 2018-06-05 Martin Liska <mliska@suse.cz>
3398 PR gcov-profile/47618
3399 * doc/invoke.texi: Document how -fprofile-dir format
3402 2018-06-05 Richard Biener <rguenther@suse.de>
3404 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3405 removal pretend DOM info isn't available so we do not update
3406 it and only remove edges, not dominated blocks. Actually free
3407 DOM info in case we removed something. Remove unreachable blocks.
3408 (mfb_keep_latches): Work with either DOM info or marked backedges.
3409 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3410 first. Mark backedges if DOM info isn't available.
3411 (Re-)compute DOM info after cleanup_control_flow_pre.
3413 2018-06-05 Richard Biener <rguenther@suse.de>
3415 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3416 (locus_discrim_hasher::hash): Adjust.
3417 (locus_discrim_hasher::equal): Likewise.
3418 (next_discriminator_for_locus): Work on line directly.
3419 (same_line_p): Pass in expanded locus1 as well.
3420 (assign_discriminators): Avoid redundant location expansions.
3422 2018-06-05 Richard Biener <rguenther@suse.de>
3424 PR tree-optimization/86046
3425 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3426 if required after clearing TREE_ADDRESSABLE.
3428 2018-06-05 Richard Biener <rguenther@suse.de>
3430 PR tree-optimization/86047
3431 * tree-ssa-loop.c (for_each_index): Glob handling of all
3432 decls and constants and really handle all of them.
3434 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3437 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3438 qualifier_void_pointer and qualifier_const_void_pointer.
3439 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3440 (arm_init_builtins): Handle the above.
3441 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3442 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3443 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3446 2018-06-05 Martin Liska <mliska@suse.cz>
3448 * auto-profile.c (read_autofdo_file): Do not use
3449 gcov_ctr_summary struct.
3450 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3451 * coverage.c (struct counts_entry): Likewise.
3452 (read_counts_file): Read just single summary entry.
3453 (get_coverage_counts): Use gcov_summary struct.
3454 * coverage.h (get_coverage_counts): Likewise.
3455 * gcov-dump.c (dump_working_sets): Likewise.
3456 (tag_summary): Dump just single summary.
3457 * gcov-io.c (gcov_write_summary): Write just histogram
3459 (gcov_read_summary): Read just single summary.
3460 (compute_working_sets): Use gcov_summary struct.
3461 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3462 of GCOV_COUNTERS_SUMMABLE.
3463 (GCOV_COUNTERS_SUMMABLE): Remove.
3464 (GCOV_FIRST_VALUE_COUNTER): Replace with
3465 GCOV_COUNTER_V_INTERVAL.
3466 (struct gcov_ctr_summary): Remove.
3467 (struct gcov_summary): Directly use fields of former
3469 (compute_working_sets): Use gcov_summary struct.
3470 * gcov.c (read_count_file): Do not use ctrs fields.
3471 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3473 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3475 * profile.c: Likewise.
3476 * profile.h: Likewise.
3478 2018-06-05 Martin Liska <mliska@suse.cz>
3480 PR gcov-profile/84846
3481 * gcov.c (output_lines): Print working directory only
3482 in intermediate format.
3484 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3486 * config/s390/s390-builtin-types.def: Add void function type.
3487 * config/s390/s390-builtins.def: Use the function type for the
3490 2018-06-04 Jim Wilson <jimw@sifive.com>
3492 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3494 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3495 and maybe_eh_return. Change regno to unsigned int. Use new args to
3496 handle EH_RETURN_DATA_REGNO registers properly.
3497 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3498 (riscv_expand_epilogue): Update comment. Change argument name and
3499 type. Update code to use new name and type. Pass new args to
3500 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3502 * config/riscv/riscv.md (NORMAL_RETURN): New.
3503 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3504 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3505 (eh_return): Call gen_eh_return_internal and emit barrier.
3506 (eh_return_internal): Call riscv_expand_epilogue.
3508 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3510 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3511 bit_insertion field and declare can_be_merged_into method.
3512 (merged_store_group::can_be_merged_into): New method.
3513 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3514 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3515 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3517 2018-06-04 Richard Biener <rguenther@suse.de>
3519 PR tree-optimization/85955
3520 * builtins.c (fold_builtin_sincos): Convert pointers to
3521 destination to appropriate type before dereferencing.
3523 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3525 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3527 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3529 * expr.c (expand_expr_real_1): Force the operand into memory if
3530 its TYPE_MODE is BLKmode and if there is no integer mode for
3531 the number of bits being extracted.
3533 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3537 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3538 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3540 2018-06-04 Richard Biener <rguenther@suse.de>
3542 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3543 (cleanup_tree_cfg_noloop): ... single caller. Do
3544 start_recording_case_labels later.
3546 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3548 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3549 to _IMMINTRIN_H_INCLUDED.
3550 * config/i386/pconfigintrin.h: Ditto.
3551 * config/i386/waitpkgintrin.h: Ditto.
3552 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3553 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3554 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3555 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3556 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3557 waitpkgintrin.h and cldemoteintrin.h.
3559 2018-06-04 Richard Biener <rguenther@suse.de>
3561 PR tree-optimization/86038
3562 * tracer.c (find_best_successor): Check probability for
3563 being initialized, bail out if not.
3565 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3568 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3569 of bits to ignore when comparing architectures.
3571 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3573 PR tree-optimization/69615
3574 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3575 maximum or minimum of the type, try to merge it also as if
3576 range1 is + [-, x - 1] or + [x + 1, -].
3579 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3581 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3583 PR tree-optimization/86034
3584 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3585 the unsigned bitfield type in a bit insertion sequence if it does not
3586 have a larger precision than the bitfield size.
3587 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3589 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3591 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3593 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3595 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3596 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3597 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3598 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3600 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3602 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3603 Disable -fdelete-null-pointer-checks for ELF toolchain.
3605 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3606 Kito Cheng <kito.cheng@gmail.com>
3608 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3609 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3610 (nds32*-*-*): Add glibc and uclibc conditions.
3611 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3612 (TARGET_EXCEPT_UNWIND_INFO): Define.
3613 * config/nds32/elf.h: New file.
3614 * config/nds32/linux.h: New file.
3615 * config/nds32/nds32-elf.opt: New file.
3616 * config/nds32/nds32-linux.opt: New file.
3617 * config/nds32/nds32-fp-as-gp.c
3618 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3619 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3621 (nds32_asm_file_end): Ditto.
3622 (nds32_print_operand): Ditto.
3623 (nds32_insert_attributes): Ditto.
3624 (nds32_init_libfuncs): New function.
3625 (TARGET_HAVE_TLS): Define.
3626 (TARGET_INIT_LIBFUNCS): Define.
3627 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3629 (TARGET_ELF): Apply different mcmodel setting.
3630 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3631 been migrated into elf.h and linux.h files.
3632 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3633 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3634 (mcmodel): The content has been migrated into nds32-elf.opt and
3635 nds32-linux.opt files.
3636 * config/nds32/t-elf: New file.
3637 * config/nds32/t-linux: New file.
3639 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3640 Shiva Chen <shiva0217@gmail.com>
3642 * config/nds32/constants.md (unspec_volatile_element): Add
3643 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3644 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3646 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3647 (make_pass_nds32_fp_as_gp): Declare.
3648 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3650 (nds32_asm_function_end_prologue): Remove unused asm output.
3651 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3652 (nds32_asm_file_start): Output necessary fp_as_gp information.
3653 (nds32_option_override): Adjust register usage.
3654 (nds32_expand_prologue): Consider fp_as_gp situation.
3655 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3656 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3659 (simple_return): Ditto.
3660 (omit_fp_begin): Output special directive for fp_as_gp.
3661 (omit_fp_end): Output special directive for fp_as_gp.
3662 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3663 mforbid-fp-as-gp): New options.
3665 2018-06-01 Mark Wielaard <mark@klomp.org>
3667 * dwarf2out.c (dwarf2out_finish): Remove generation of
3668 DW_AT_loclists_base.
3670 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3672 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3673 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3674 (struct merged_store_group): Add bit_insertion field.
3675 (dump_char_array): Use standard hexadecimal format.
3676 (merged_store_group::merged_store_group): Set bit_insertion to false.
3677 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3678 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3679 also print the mask in the dump file.
3680 (pass_store_merging::gate): Minor tweak.
3681 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3682 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3683 stores with INTEGER_CST stores.
3684 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3685 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3686 and use it throughout. Generate bit insertion sequences if need be.
3687 (pass_store_merging::process_store): Remove redundant condition.
3688 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3690 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3692 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3693 the 128-bit floating point types. Fix function comment.
3695 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3697 * config/aarch64/aarch64-simd.md
3698 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3700 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3703 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3705 PR tree-optimization/85989
3706 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3708 (backprop::intersect_uses): Check it when deciding whether this
3709 is a backedge reference.
3710 (backprop::process_block): Add each phi to m_visited_phis
3711 after visiting it, then clear it at the end.
3713 2018-06-01 Richard Biener <rguenther@suse.de>
3715 * tree-vectorizer.h (vect_dr_stmt): New function.
3716 (vect_get_load_cost): Adjust.
3717 (vect_get_store_cost): Likewise.
3718 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3719 Use vect_dr_stmt instead of DR_SMTT.
3720 (vect_record_base_alignments): Likewise.
3721 (vect_calculate_target_alignment): Likewise.
3722 (vect_compute_data_ref_alignment): Likewise and make static.
3723 (vect_update_misalignment_for_peel): Likewise.
3724 (vect_verify_datarefs_alignment): Likewise.
3725 (vector_alignment_reachable_p): Likewise.
3726 (vect_get_data_access_cost): Likewise. Pass down
3727 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3728 (vect_get_peeling_costs_all_drs): Likewise.
3729 (vect_peeling_hash_get_lowest_cost): Likewise.
3730 (vect_enhance_data_refs_alignment): Likewise.
3731 (vect_find_same_alignment_drs): Likewise.
3732 (vect_analyze_data_refs_alignment): Likewise.
3733 (vect_analyze_group_access_1): Likewise.
3734 (vect_analyze_group_access): Likewise.
3735 (vect_analyze_data_ref_access): Likewise.
3736 (vect_analyze_data_ref_accesses): Likewise.
3737 (vect_vfa_segment_size): Likewise.
3738 (vect_small_gap_p): Likewise.
3739 (vectorizable_with_step_bound_p): Likewise.
3740 (vect_prune_runtime_alias_test_list): Likewise.
3741 (vect_analyze_data_refs): Likewise.
3742 (vect_supportable_dr_alignment): Likewise.
3743 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3744 (vect_gen_prolog_loop_niters): Likewise.
3745 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3746 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3748 (vect_recog_mask_conversion_pattern): Likewise.
3749 (vect_try_gather_scatter_pattern): Likewise.
3750 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3751 to vect_get_store_cost.
3752 (vect_get_store_cost): Get stmt_info instead of DR.
3753 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3754 (vect_get_load_cost): Get stmt_info instead of DR.
3756 2018-06-01 Richard Biener <rguenther@suse.de>
3759 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3760 vars wrapped in MEM_REFs.
3762 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3764 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3765 Fix subreg tests so that we only return a choice between
3766 GENERAL_REGS and FP_REGS if the original classes included both.
3768 2018-06-01 Richard Biener <rguenther@suse.de>
3771 * tree-ssa-structalias.c (get_function_part_constraint):
3772 Handle NULL fi->decl.
3773 (find_func_aliases_for_call): Properly handle indirect
3774 fi from direct call.
3775 (find_func_clobbers): Likewise.
3776 (ipa_pta_execute): Likewise.
3777 (create_variable_info_for): For functions that are ifunc_resolver
3778 resolve to a varinfo that contains the result of the resolver call.
3779 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3782 2018-05-31 Michael Collison <michael.collison@arm.com>
3784 * config/aarch64/aarch64.md:
3785 (*fix_to_zero_extenddfdi2): New pattern.
3786 * gcc.target/aarch64/fix_extend1.c: New testcase.
3788 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3792 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3793 and BUILT_IN_STRNCMP_EQ.
3794 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3795 BUILT_IN_STRNCMP_EQ.
3796 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3797 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3798 (gimple_fold_builtin): Likewise.
3799 * tree-ssa-strlen.c (compute_string_length): New function.
3800 (determine_min_obsize): New function.
3801 (handle_builtin_string_cmp): New function to handle calls to
3802 string compare functions.
3803 (strlen_optimize_stmt): Add handling to builtin string compare
3805 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3806 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3807 * tree.c (build_common_builtin_nodes): Add new defines of
3808 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3810 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3813 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3814 functions with naked attribute.
3816 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3818 * config/i386/sse.md (avx_vec_concat<mode>):
3819 Substitute concat_tg_mode mode attribute with xtg_mode.
3820 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3821 (concat_tg_mode): Remove mode attribute.
3823 2018-05-31 Martin Sebor <msebor@redhat.com>
3826 * calls.c (alloc_max_size): Correct a logic error/typo.
3827 Treat excessive arguments as infinite. Warn for invalid arguments.
3828 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3830 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3833 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3834 and movx for Haswell.
3836 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3837 Cesar Philippidis <cesar@codesourcery.com>
3840 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3841 when emitting error on private/firstprivate reductions.
3842 * omp-low.c (lower_omp_target): Avoid reference-type processing
3843 on pointers for firstprivate clause.
3845 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3847 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3850 * config/aarch64/aarch64-simd.md
3851 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3852 (aarch64_ld1_x3_<mode>): Likewise
3853 (aarch64_st1x2<VALLDIF:mode>): Likewise
3854 (aarch64_st1_x2_<mode>): Likewise
3855 (aarch64_st1x3<VALLDIF:mode>): Likewise
3856 (aarch64_st1_x3_<mode>): Likewise
3857 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3858 (vld1_s8_x3): Likewise.
3859 (vld1_u16_x3): Likewise.
3860 (vld1_s16_x3): Likewise.
3861 (vld1_u32_x3): Likewise.
3862 (vld1_s32_x3): Likewise.
3863 (vld1_u64_x3): Likewise.
3864 (vld1_s64_x3): Likewise.
3865 (vld1_f16_x3): Likewise.
3866 (vld1_f32_x3): Likewise.
3867 (vld1_f64_x3): Likewise.
3868 (vld1_p8_x3): Likewise.
3869 (vld1_p16_x3): Likewise.
3870 (vld1_p64_x3): Likewise.
3871 (vld1q_u8_x3): Likewise.
3872 (vld1q_s8_x3): Likewise.
3873 (vld1q_u16_x3): Likewise.
3874 (vld1q_s16_x3): Likewise.
3875 (vld1q_u32_x3): Likewise.
3876 (vld1q_s32_x3): Likewise.
3877 (vld1q_u64_x3): Likewise.
3878 (vld1q_s64_x3): Likewise.
3879 (vld1q_f16_x3): Likewise.
3880 (vld1q_f32_x3): Likewise.
3881 (vld1q_f64_x3): Likewise.
3882 (vld1q_p8_x3): Likewise.
3883 (vld1q_p16_x3): Likewise.
3884 (vld1q_p64_x3): Likewise.
3885 (vst1_s64_x2): Likewise.
3886 (vst1_u64_x2): Likewise.
3887 (vst1_f64_x2): Likewise.
3888 (vst1_s8_x2): Likewise.
3889 (vst1_p8_x2): Likewise.
3890 (vst1_s16_x2): Likewise.
3891 (vst1_p16_x2): Likewise.
3892 (vst1_s32_x2): Likewise.
3893 (vst1_u8_x2): Likewise.
3894 (vst1_u16_x2): Likewise.
3895 (vst1_u32_x2): Likewise.
3896 (vst1_f16_x2): Likewise.
3897 (vst1_f32_x2): Likewise.
3898 (vst1_p64_x2): Likewise.
3899 (vst1q_s8_x2): Likewise.
3900 (vst1q_p8_x2): Likewise.
3901 (vst1q_s16_x2): Likewise.
3902 (vst1q_p16_x2): Likewise.
3903 (vst1q_s32_x2): Likewise.
3904 (vst1q_s64_x2): Likewise.
3905 (vst1q_u8_x2): Likewise.
3906 (vst1q_u16_x2): Likewise.
3907 (vst1q_u32_x2): Likewise.
3908 (vst1q_u64_x2): Likewise.
3909 (vst1q_f16_x2): Likewise.
3910 (vst1q_f32_x2): Likewise.
3911 (vst1q_f64_x2): Likewise.
3912 (vst1q_p64_x2): Likewise.
3913 (vst1_s64_x3): Likewise.
3914 (vst1_u64_x3): Likewise.
3915 (vst1_f64_x3): Likewise.
3916 (vst1_s8_x3): Likewise.
3917 (vst1_p8_x3): Likewise.
3918 (vst1_s16_x3): Likewise.
3919 (vst1_p16_x3): Likewise.
3920 (vst1_s32_x3): Likewise.
3921 (vst1_u8_x3): Likewise.
3922 (vst1_u16_x3): Likewise.
3923 (vst1_u32_x3): Likewise.
3924 (vst1_f16_x3): Likewise.
3925 (vst1_f32_x3): Likewise.
3926 (vst1_p64_x3): Likewise.
3927 (vst1q_s8_x3): Likewise.
3928 (vst1q_p8_x3): Likewise.
3929 (vst1q_s16_x3): Likewise.
3930 (vst1q_p16_x3): Likewise.
3931 (vst1q_s32_x3): Likewise.
3932 (vst1q_s64_x3): Likewise.
3933 (vst1q_u8_x3): Likewise.
3934 (vst1q_u16_x3): Likewise.
3935 (vst1q_u32_x3): Likewise.
3936 (vst1q_u64_x3): Likewise.
3937 (vst1q_f16_x3): Likewise.
3938 (vst1q_f32_x3): Likewise.
3939 (vst1q_f64_x3): Likewise.
3940 (vst1q_p64_x3): Likewise.
3942 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3944 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3945 user_label_prefix to name.
3947 * tree-core.h: Update comment about the format of NAME string
3948 passed to handler in attribute_spec.
3950 * config/msp430/msp430.md: Remove erroneous subreg expression from
3951 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3952 zero_extend{q,h}isi2.
3954 2018-05-30 Borislav Petkov <bp@suse.de>
3956 * doc/extend.texi: Document some architecture specific
3957 constraints and sort entries.
3959 2018-05-30 Martin Sebor <msebor@redhat.com>
3962 * builtins.c (expand_builtin_stpcpy_1): New function.
3963 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3964 only if the former succeeds.
3966 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3968 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3971 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3973 * doc/invoke.texi (-flinker-output): Document
3975 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3977 * passes.c (ipa_write_summaries): Only modify statements if body
3979 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3980 incrementally linking.
3981 (ipa_passes): Likewise.
3982 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3983 pass down resolution info.
3984 * common.opt (flag_incremental_link): Update info.
3985 * gcc.c (plugin specs): Turn flinker-output=* to
3986 -plugin-opt=-linker-output-known
3987 * toplev.c (compile_file): Also cut compilation when doing incremental
3989 * flag-types. (enum lto_partition_model): Add
3990 LTO_LINKER_OUTPUT_NOLTOREL.
3991 (invoke.texi): Add -flinker-output docs.
3992 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3993 link same way as WPA; do not stream in dead initializers.
3995 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3996 darf2out_register_external_decl): Support incremental link.
3998 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4000 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4002 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4004 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4005 it down to simple_object_copy_lto_debug_sections.
4006 (run_gcc): Determine incremental LTO link time and configure
4007 lto1 into non-wpa mode, disable renaming of debug sections.
4009 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4011 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4012 descriptions of various incorrectly documented functions.
4014 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4017 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4018 address check not strict.
4020 2018-05-30 Richard Biener <rguenther@suse.de>
4022 PR tree-optimization/85964
4023 * tracer.c (better_p): Drop initialized count check, we only
4024 call the function with initialized counts now.
4025 (find_best_successor): Do find a best edge if one
4026 has uninitialized count.
4027 (find_best_predecessor): Likewise. Do BB frequency check only
4028 if count is initialized.
4030 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4032 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4033 (aarch64_ldrstr_offset_compare): New.
4034 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4035 load/store orderings.
4036 (aarch64_gen_adjusted_ldpstp): Likewise.
4038 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4040 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4041 Check for subset of GENERAL_REGS and FP_REGS.
4042 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4045 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4047 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4048 and wi::to_poly_offset. Add the current offset and then check
4049 whether the sum fits, rather than using an unchecked addition of
4050 a checked term. Check for a shwi rather than a uhwi.
4051 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4052 (store_constructor): Use poly_int_tree_p.
4053 (expand_expr_real_1): Likewise.
4054 * function.c (assign_temp): Likewise.
4055 * fold-const.c (const_binop): Use poly_int_tree_p and
4057 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4059 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4060 to_poly_offset for MEM offsets.
4061 * ipa-icf.c (sem_variable::equals): Likewise.
4062 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4063 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4064 wi::to_poly_offset for BIT_FIELD_REF offsets.
4065 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4067 * var-tracking.c (emit_note_insn_var_location): Use
4068 tree_to_poly_uint64.
4070 2018-05-29 Jim Wilson <jimw@sifive.com>
4072 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4074 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4077 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4078 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4080 (sse4_1_round<mode>2): Use nonimmediate_operand
4081 for operand 1 predicate.
4083 2018-05-29 Martin Sebor <msebor@redhat.com>
4084 Richard Biener <rguenther@suse.de>
4087 * calls.c (get_size_range): Call determine_value_range instead
4088 of get_value_range..
4089 * tree-vrp.h (determine_value_range): Declared new function.
4090 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4092 2018-05-29 Richard Biener <rguenther@suse.de>
4094 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4095 sure to use non-pattern stmts for get_earlier_stmt arguments.
4096 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4097 called on pattern stmts.
4098 (get_later_stmt): Likewise.
4100 2018-05-29 Martin Liska <mliska@suse.cz>
4102 PR gcov-profile/85759
4103 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4106 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4108 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4110 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4114 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4115 VEC_PACK_FLOAT_EXPR): New tree codes.
4116 * tree-pretty-print.c (op_code_prio): Handle
4117 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4118 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4119 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4120 * tree-inline.c (estimate_operator_cost): Likewise.
4121 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4122 * fold-const.c (const_binop): Likewise.
4123 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4124 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4125 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4126 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4127 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4128 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4129 * expr.c (expand_expr_real_2): Likewise.
4130 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4131 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4132 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4134 * optabs.c (expand_widen_pattern_expr): For
4135 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4136 sign from result type rather than operand's type.
4137 (expand_binop_directly): For vec_packu_float_optab and
4138 vec_packs_float_optab allow result type to be different from operand's
4140 * optabs-tree.c (optab_for_tree_code): Handle
4141 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4142 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4143 * tree-vect-generic.c (expand_vector_operations_1): Handle
4144 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4145 VEC_PACK_FLOAT_EXPR.
4146 * tree-vect-stmts.c (supportable_widening_operation): Handle
4148 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4149 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4150 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4151 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4152 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4154 (vec_pack<floatprefix>_float_<mode>): New expander.
4155 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4157 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4158 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4159 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4160 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4161 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4163 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4164 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4165 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4166 VEC_PACK_FLOAT_EXPR): Document.
4168 2018-05-29 Richard Biener <rguenther@suse.de>
4170 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4172 (stmt_vec_info_vec): Make pointer.
4173 (init_stmt_vec_info_vec): Remove.
4174 (free_stmt_vec_info_vec): Likewise.
4175 (set_stmt_vec_info_vec): New function.
4176 (free_stmt_vec_infos): Likewise.
4177 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4178 (set_vinfo_for_stmt): Likewise.
4179 (get_earlier_stmt): Likewise.
4180 (get_later_stmt): Likewise.
4181 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4182 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4183 (vec_info::~vec_info): Free stmt_vec_infos.
4184 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4185 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4186 (pass_slp_vectorize::execute): Likewise.
4187 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4188 (free_stmt_vec_info_vec): Likewise.
4189 (set_stmt_vec_info_vec): New function.
4190 (free_stmt_vec_infos): Likewise.
4191 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4192 the global stmt_vec_info_vec.
4193 * tree-parloops.c (gather_scalar_reductions): Use
4194 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4197 2018-05-29 Richard Biener <rguenther@suse.de>
4199 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4201 2018-05-29 Martin Liska <mliska@suse.cz>
4202 David Malcolm <dmalcolm@redhat.com>
4204 * vec.c (test_reverse): New.
4205 (vec_c_tests): Add new test.
4206 * vec.h (vl_ptr>::reverse): New function.
4208 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4210 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4212 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4215 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4217 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4219 2018-05-28 Richard Biener <rguenther@suse.de>
4221 PR tree-optimization/85933
4222 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4223 look at stmts marked as vectorizable.
4225 2018-05-28 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/85934
4228 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4229 vector boolean check before scalar optimization.
4231 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4233 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4236 2018-05-28 Mark Wielaard <mark@klomp.org>
4238 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4239 if it is an expression containing a minus sign.
4241 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4243 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4245 2018-05-27 Paul Koning <ni1d@arrl.net>
4247 * config/pdp11/pdp11.md (truncsihi2): Remove.
4249 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4250 Chung-Ju Wu <jasonwucj@gmail.com>
4252 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4254 (unaligned_store_dw): Ditto.
4255 * config/nds32/nds32-memory-manipulation.c
4256 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4257 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4258 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4259 (emit_setmem_word_loop): Rename to ...
4260 (emit_setmem_doubleword_loop): ... this.
4261 (nds32_gen_dup_4_byte_to_word_value): New function.
4262 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4263 (nds32_expand_setmem_loop): Refine implementation.
4264 (nds32_expand_setmem_loop_v3m): Ditto.
4265 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4268 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4270 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4272 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4274 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4275 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4276 machine->attr_no_prologue_p.
4277 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4278 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4279 (nds32_expand_epilogue): Consider attr_naked_p.
4280 (nds32_expand_epilogue_v3pop): Likewise.
4281 (nds32_can_use_return_insn): Likewise.
4282 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4283 attr_no_prologue_p fields.
4284 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4286 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4289 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4291 * config/i386/sse.md
4292 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4294 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4296 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4298 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4300 (*<floatsuffix>floatv2div2sf2): Rename to ...
4301 (*float<floatunssuffix>v2div2sf2): ... this.
4302 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4303 (float<floatunssuffix>v2div2sf2_mask): ... this.
4304 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4305 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4306 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4308 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4310 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4312 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4314 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4316 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4318 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4319 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4320 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4321 gen_ufix_truncv8dfv8si2.
4322 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4323 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4324 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4325 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4326 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4327 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4328 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4329 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4331 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4335 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4337 2018-05-25 Jim Wilson <jimw@sifive.com>
4339 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4340 * config/riscv/riscv.c (struct machine_function): Add
4341 interrupt_handler_p and attribute_checked_p fields.
4342 (riscv_attribute_table): Add interrupt.
4343 (riscv_interrupt_type_p): New.
4344 (riscv_save_reg_p): Save extra regs for interrupt handler.
4345 (riscv_use_save_libcall): Return false for interrupt handler.
4346 (riscv_first_stack_step): Add forward declaration.
4347 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4348 for interrupt handler with large frame. Use it for saved reg list.
4349 (riscv_expand_prologue): Move flag_stack_usage_info support to
4350 eliminate duplication.
4351 (riscv_expand_epilogue): Generate mret for interrupt handler.
4352 (riscv_epilogue_uses): New.
4353 (riscv_can_use_return_insn): Return false for interrupt handler.
4354 (riscv_function_ok_for_sibcall): Likewise.
4355 (riscv_set_current_function): Add interrupt handler support.
4356 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4357 * config/riscv/riscv.md (UNSPECV_MRET): New.
4359 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4361 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4363 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4365 PR tree-optimization/85712
4366 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4367 this candidate has already been replaced in-situ by a copy.
4369 2018-05-25 Jason Merrill <jason@redhat.com>
4371 PR c++/80485 - inline function non-zero address.
4372 * symtab.c (nonzero_address): Check DECL_COMDAT.
4374 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4377 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4378 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4379 extension of SImode operation. Use const123_operand predicate.
4380 (*saddsi_1): Remove.
4381 (*saddl_se_1): Ditto.
4383 (*ssubl_se_1): Ditto.
4384 * config/alpha/predicates.md (const123_operand): New predicate.
4385 * config/alpha/constraints.md (P): Use IN_RANGE.
4387 2018-05-25 Richard Biener <rguenther@suse.de>
4389 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4391 (ref_maybe_used_by_stmt_p): Likewise.
4392 (stmt_may_clobber_ref_p): Likewise.
4393 (stmt_may_clobber_ref_p_1): Likewise.
4394 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4396 (ref_maybe_used_by_stmt_p): Likewise.
4397 (stmt_may_clobber_ref_p): Likewise.
4398 (stmt_may_clobber_ref_p_1): Likewise.
4399 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4400 the alias oracle to disambiguate DRs with stmts DR analysis
4402 (vect_analyze_data_refs): Do not give up on not analyzable
4403 DRs for BB vectorization. Remove code truncating the dataref
4406 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4409 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4410 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4411 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4413 2018-05-25 Richard Biener <rguenther@suse.de>
4415 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4416 function, combining stmt data ref gathering and fatal analysis
4418 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4419 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4420 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4421 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4422 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4424 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4426 PR tree-optimization/85720
4427 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4428 SCC if all partitions are builtins.
4429 (version_loop_by_alias_check): New parameter. Generate cancelable
4430 runtime alias check if all partitions are builtins.
4431 (distribute_loop): Update call to above function.
4433 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4435 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4436 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4437 (parm_default_def_partition_arg): Ditto.
4438 (set_parm_default_def_partition): Ditto.
4439 (get_parm_default_def_partitions): Ditto and make it static.
4440 (get_undefined_value_partitions): Ditto and make it static.
4441 (remove_ssa_form): Refactor call to init_var_map here.
4442 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4443 computation for loop region.
4444 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4445 (register_default_def): Delete.
4446 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4447 (parm_default_def_partition_arg): Ditto.
4448 (set_parm_default_def_partition): Ditto.
4449 (get_parm_default_def_partitions): Ditto and make it static.
4450 (get_undefined_value_partitions): Ditto and make it static.
4451 (coalesce_with_default, coalesce_with_default): Update comment.
4452 (create_coalesce_list_for_region): New func factored out from
4453 create_outofssa_var_map.
4454 (populate_coalesce_list_for_outofssa): New func factored out from
4455 create_outofssa_var_map and coalesce_ssa_name.
4456 (create_outofssa_var_map): Delete.
4457 (coalesce_ssa_name): Refactor to support live range computation.
4458 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4459 (get_parm_default_def_partitions): Delete.
4460 (get_undefined_value_partitions): Ditto.
4461 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4462 computation for loop region.
4463 (new_tree_live_info, loe_visit_block): Ditto.
4464 (live_worklist, set_var_live_on_entry): Ditto.
4465 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4466 * tree-ssa-live.h (struct _var_map): New fields.
4467 (init_var_map): Change decl.
4468 (region_contains_p): New.
4470 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4472 * tree-ssa-live.h (live_merge_and_clear): Delete.
4474 2018-05-25 Richard Biener <rguenther@suse.de>
4477 * tree-dump.c (dequeue_and_dump): Remove access to removed
4478 operand 2 of a SWITCH_EXPR.
4480 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4482 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4483 multiplication and division.
4484 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4485 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4486 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4487 (cond_udiv_optab, cond_umod_optab): New optabs.
4488 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4489 (IFN_COND_RDIV): New internal functions.
4490 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4491 TRUNC_MOD_EXPR and RDIV_EXPR.
4492 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4493 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4495 (SVE_INT_BINARY): Include mult.
4496 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4497 (optab, sve_int_op): Handle mult.
4498 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4500 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4501 for SVE_INT_BINARY_SD.
4503 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4505 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4506 (optab, sve_int_op): Handle div and udiv.
4507 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4508 for SVE_INT_BINARY_SD.
4509 (*<optab><mode>3): New insn for the same.
4511 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4513 * tree-vect-patterns.c: Include predict.h.
4514 (vect_recog_divmod_pattern): Restrict check for division support
4515 to when optimizing for size.
4517 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4519 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4520 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4521 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4522 (gimple_match_op::set_op): Likewise.
4523 (gimple_resimplify4): Declare.
4524 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4525 (expr::gen_transform): Likewise.
4526 (decision_tree::gen): Generate a simplification routine for 4 operands.
4527 * gimple-match-head.c (gimple_simplify): Add an overload for
4528 4 operands. In the top-level function, handle up to 4 call
4529 arguments and call gimple_resimplify4.
4530 (gimple_resimplify4): New function.
4531 (build_call_internal): Pass a fourth operand.
4532 (maybe_push_to_seq): Likewise.
4533 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4534 Fold VEC_COND_EXPRs of an operation and a default value into
4535 an IFN_COND_* function if possible.
4536 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4538 (SVE_COND_FP_BINARY): Include them.
4539 (optab, sve_fp_op): Handle them.
4540 (SVE_INT_BINARY_REV): New code iterator.
4541 (SVE_COND_FP_BINARY_REV): New int iterator.
4542 (commutative): New int attribute.
4543 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4545 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4547 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4548 (*cond_<optab><mode>): New patterns for reversed operands.
4550 2018-05-25 Richard Biener <rguenther@suse.de>
4552 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4553 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4554 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4555 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4556 * tree-vect-data-refs.c (everywhere): Adjust users.
4557 * tree-vect-loop.c (everywhere): Likewise.
4558 * tree-vect-slp.c (everywhere): Likewise.
4559 * tree-vect-stmts.c (everywhere): Likewise.
4560 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4562 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4564 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4566 (gcc_cv_as_section_exclude): ... this.
4567 Try Solaris as #exclude syntax.
4568 * configure: Regenerate.
4569 * config.in: Regenerate.
4570 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4572 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4573 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4575 * varasm.c (default_elf_asm_named_section): Don't check if
4576 HAVE_GAS_SECTION_EXCLUDE is defined.
4578 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4580 * doc/md.texi: Update the documentation of the cond_* optabs
4581 to mention the new final operand. Fix GET_MODE_NUNITS call.
4582 Describe the scalar case too.
4583 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4584 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4586 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4587 (get_conditional_internal_fn): Update comment.
4588 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4589 accumulator value as a final argument to conditional functions.
4590 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4591 a define_expand and add an "else" operand. Assert for now that
4592 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4593 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4594 (*cond_<optab><mode>): New patterns.
4595 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4596 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4597 (UNSPEC_COND_EOR): Delete.
4598 (optab): Remove associated mappings.
4599 (SVE_INT_BINARY): New code iterator.
4600 (sve_int_op): Remove int attribute and add "minus" to the code
4602 (SVE_COND_INT_OP): Delete.
4603 (SVE_COND_FP_OP): Rename to...
4604 (SVE_COND_FP_BINARY): ...this.
4606 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4608 * optabs.c (can_reuse_operands_p): New function.
4609 (maybe_legitimize_operands): Try to reuse the results for
4612 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4614 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4615 Add {q} suffix to insn mnemonic.
4617 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4619 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4620 (msp430_warn_func_return): New.
4622 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4624 * fold-const.c (tree_nonzero_bits): New function.
4625 * fold-const.h (tree_nonzero_bits): Likewise.
4626 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4627 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4629 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4633 * varasm.c (assemble_alias): Check ifunc_resolver only on
4636 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4639 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4640 when memory input operand is handled.
4642 2018-05-24 Luis Machado <luis.machado@linaro.org>
4644 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4646 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4648 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4650 * match.pd: Delay FMA folds until after vectorization.
4652 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4655 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4656 address check not strict.
4658 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4660 * gimple-match.h (gimple_match_op): New class.
4661 (mprts_hook): Replace parameters with a gimple_match_op *.
4662 (maybe_build_generic_op): Likewise.
4663 (gimple_simplified_result_is_gimple_val): Replace parameters with
4664 a const gimple_match_op *.
4665 (gimple_simplify): Replace code_helper * and tree * parameters with
4666 a gimple_match_op * parameter.
4667 (gimple_resimplify1): Replace code_helper *, tree and tree *
4668 parameters with a gimple_match_op * parameter.
4669 (gimple_resimplify2): Likewise.
4670 (gimple_resimplify3): Likewise.
4671 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4672 parameters with a gimple_match_op * parameter.
4673 * gimple-match-head.c (gimple_simplify): Change prototypes of
4674 auto-generated functions to take a gimple_match_op * instead of
4675 separate code_helper * and tree * parameters. Make the same
4676 change in the top-level overload and update calls to the
4677 gimple_resimplify routines. Update calls to the auto-generated
4678 functions and to maybe_push_res_to_seq in the publicly-facing
4679 operation-specific gimple_simplify overloads.
4680 (gimple_match_op::MAX_NUM_OPS): Define.
4681 (gimple_resimplify1): Replace rcode and ops with a single res_op
4682 parameter. Update call to gimple_simplify.
4683 (gimple_resimplify2): Likewise.
4684 (gimple_resimplify3): Likewise.
4685 (mprts_hook): Replace parameters with a gimple_match_op *.
4686 (maybe_build_generic_op): Likewise.
4687 (build_call_internal): Replace type, nargs and ops with
4688 a gimple_match_op *.
4689 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4690 with a single gimple_match_op *. Update calls to mprts_hook,
4691 build_call_internal and gimple_simplified_result_is_gimple_val.
4692 Factor out code that is common to the tree_code and combined_fn cases.
4693 * genmatch.c (expr::gen_transform): Replace tem_code and
4694 tem_ops with a gimple_match_op called tem_op. Update calls
4695 to the gimple_resimplify functions and maybe_push_res_to_seq.
4696 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4697 res_ops. Update call to the gimple_resimplify functions.
4698 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4699 (decision_tree::gen): Make the functions take a gimple_match_op *
4700 called res_op instead of separate res_code and res_ops parameters.
4701 Update call accordingly.
4702 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4703 and ops with a single res_op parameter. Update calls to
4704 maybe_build_generic_op and maybe_push_res_to_seq.
4705 (fold_stmt_1): Update calls to gimple_simplify and
4706 replace_stmt_with_simplification.
4707 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4708 and gimple_simplified_result_is_gimple_val.
4709 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4711 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4712 with a gimple_match_op *.
4713 (vn_nary_build_or_lookup): Likewise. Update call to
4714 vn_nary_build_or_lookup_1.
4715 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4716 gimple_match_op *. Update calls to the gimple_resimplify routines
4717 and to gimple_simplified_result_is_gimple_val.
4718 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4719 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4720 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4721 (visit_nary_op): Likewise.
4722 (visit_reference_op_load): Likewise.
4724 2018-05-23 Luis Machado <luis.machado@linaro.org>
4726 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4727 modifier for printing the step amount.
4729 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4732 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4735 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4737 * doc/sourcebuild.texi (Endianness): New subsubsection.
4739 2018-05-23 Luis Machado <luis.machado@linaro.org>
4741 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4742 <prefetch_dynamic_strides>: New const bool field.
4743 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4744 prefetch_dynamic_strides.
4745 (exynosm1_prefetch_tune): Likewise.
4746 (thunderxt88_prefetch_tune): Likewise.
4747 (thunderx_prefetch_tune): Likewise.
4748 (thunderx2t99_prefetch_tune): Likewise.
4749 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4751 (aarch64_override_options_internal): Update to set
4752 PARAM_PREFETCH_DYNAMIC_STRIDES.
4753 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4754 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4755 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4756 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4757 prefetch-dynamic-strides setting.
4759 2018-05-23 Luis Machado <luis.machado@linaro.org>
4761 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4762 <minimum_stride>: New const int field.
4763 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4764 minimum_stride field defaulting to -1.
4765 (exynosm1_prefetch_tune): Likewise.
4766 (thunderxt88_prefetch_tune): Likewise.
4767 (thunderx_prefetch_tune): Likewise.
4768 (thunderx2t99_prefetch_tune): Likewise.
4769 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4770 <default_opt_level>: Set to 3.
4771 (aarch64_override_options_internal): Update to set
4772 PARAM_PREFETCH_MINIMUM_STRIDE.
4773 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4774 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4775 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4776 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4777 stride is constant and is below the minimum stride threshold.
4779 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4781 * config/arm/arm-cpus.in (mode26): Delete.
4782 (armv4): Delete mode26 reference.
4783 * config/arm/arm.c (arm_configure_build_target): Delete use of
4786 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4788 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4790 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4791 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4793 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4794 (floatunsdidf2): Ditto.
4796 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4798 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4799 (fixuns_trunc<mode>si2_avx512f): Ditto.
4800 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4801 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4802 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4804 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4806 PR rtl-optimization/79985
4807 * df-scan.c (df_insn_refs_collect): Remove special case for
4808 global registers and asm statements.
4810 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4812 * extend.texi (Global Register Variables): Rewrite the bullet list.
4813 Note that the register is available for allocation. Note that access
4814 via inline asm must use constraints. Add note about async-signal
4815 handlers. Remove paragraph about automagic register selection.
4817 2018-05-23 Richard Biener <rguenther@suse.de>
4819 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4820 of fixed offset from memset VN.
4822 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4824 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4826 (alloc_cand_and_find_basis): Initialize first_interp field.
4827 (slsr_process_mul): Modify first_interp field.
4828 (slsr_process_add): Likewise.
4829 (slsr_process_cast): Modify first_interp field for each new
4831 (slsr_process_copy): Likewise.
4832 (dump_candidate): Dump first_interp field.
4833 (replace_mult_candidate): Process all interpretations, not just
4835 (replace_rhs_if_not_dup): Likewise.
4836 (replace_one_candidate): Likewise.
4838 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4840 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4842 (aarch64_needs_frame_chain): New function.
4843 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4845 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4848 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4849 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4850 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4851 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4852 as true for strict-align.
4853 (aarch64_can_inline_p): Perform checks even when callee has no
4854 attributes to check for strict alignment.
4855 * doc/extend.texi (AArch64 Function Attributes): Document
4857 * doc/invoke.texi: (AArch64 Options): Likewise.
4859 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4861 PR tree-optimization/85853
4862 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4863 the handling of the root of the node to...
4864 (vect_slp_analyze_node_operations_1): ...this new function,
4865 and run the whole thing with the child nodes' def types
4866 set according to their SLP node's def type.
4868 2018-05-23 Richard Biener <rguenther@suse.de>
4871 * tree-data-ref.c (create_runtime_alias_checks): Defer
4872 and ignore overflow warnings.
4874 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4876 PR tree-optimization/85822
4877 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4880 2018-05-23 Richard Biener <rguenther@suse.de>
4882 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4883 memset constants via native_interpret_expr.
4885 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4888 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4890 (cgraph_node::create_alias): Likewise.
4891 (cgraph_node::get_availability): Check ifunc_resolver instead
4892 of looking up ifunc attribute.
4893 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4894 * varasm.c (do_assemble_alias): Likewise.
4895 (assemble_alias): Likewise.
4896 (default_binds_local_p_3): Likewise.
4897 * cgraph.h (cgraph_node): Add ifunc_resolver.
4898 (cgraph_node::only_called_directly_or_aliased_p): Return false
4900 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4902 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4903 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4904 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4905 instead of looking up ifunc attribute.
4907 2018-05-22 Luis Machado <luis.machado@linaro.org>
4909 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4911 2018-05-22 Martin Sebor <msebor@redhat.com>
4914 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4915 only when expasion succeeds.
4916 (expand_builtin_strcmp): Same.
4917 (expand_builtin_strncmp): Same.
4919 2018-05-22 Martin Sebor <msebor@redhat.com>
4921 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4923 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4924 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4926 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4927 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4929 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4930 Add check for legitimate_address.
4931 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4932 (aarch64_swap_ldrstr_operands): New.
4933 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4936 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4937 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4939 * config/aarch64/aarch64.md: New patterns to generate stp
4941 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4942 single words and double words.
4943 (load_pair_sw, load_pair_dw): Likewise.
4944 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4946 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4948 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4949 for different mode ldpstp and add peephole for merged zero stores.
4951 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4953 (aarch64_gen_store_pair): Rename calls to match new patterns.
4954 (aarch64_gen_load_pair): Rename calls to match new patterns.
4955 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4956 (load_pair<DREG:mode><DREG2:mode>): ... This.
4957 (store_pair<mode>): Rename to...
4958 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4959 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4961 (V_INT_EQUIV): Handle SImode.
4962 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4965 2018-05-22 Martin Sebor <msebor@redhat.com>
4968 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4969 or ajust the presumed bound on an operation to avoid unnecessary
4972 2018-05-22 Martin Sebor <msebor@redhat.com>
4974 PR tree-optimization/85826
4975 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4976 assuming that a DECL necesarily has a constant size.
4978 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4981 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4983 2018-05-22 Richard Biener <rguenther@suse.de>
4985 PR tree-optimization/85834
4986 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4987 non-constant and non-zero memset arguments.
4989 2018-05-22 Martin Liska <mliska@suse.cz>
4992 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4994 2018-05-22 Richard Biener <rguenther@suse.de>
4996 PR tree-optimization/85863
4997 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4998 comparisons when vectype is specified.
4999 (vectorizable_condition): Do not specify vectype for
5000 vect_is_simple_cond when SLP vectorizing.
5002 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5005 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5006 define __ibm128 as long double.
5007 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5008 as a distinct type when IEEE 128-bit support is enabled.
5009 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5010 128-bit types to use the correct functions.
5011 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5012 convert between 128-bit floating point types that have different
5013 modes but the same representation, instead of using gen_lowpart to
5015 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5017 (IFKF_reg): New attributes to give the register constraints for
5019 (extend<mode>tf2_internal): New insns to mark an explicit
5020 conversion between 128-bit floating point types that have a
5021 different mode but share the same representation.
5023 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5025 PR tree-optimization/85814
5026 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5027 a null return from get_strinfo when unsharing the next
5028 strinfo in the chain.
5030 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5033 * varasm.c (weak_finish): Clean up weak_decls.
5035 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5037 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5038 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5039 UNSPEC_UADALP values.
5040 * config/aarch64/iterators.md (ABAL): New int iterator.
5043 (sur): Add mappings for the above.
5044 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5046 (aarch64_<sur>abal<mode>_4): Likewise.
5047 (aarch64_<sur>adalp<mode>_3): Likewise.
5048 (<sur>sadv16qi): New define_expand.
5050 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5052 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5053 (*movdf_internal): Ditto.
5054 (*rcpsf2_sse): Ditto.
5055 (*rsqrtsf2_sse): Ditto.
5056 (*sqrt<mode>2_sse): Ditto.
5058 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5060 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5062 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5063 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5064 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5065 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5067 * config/aarch64/arm_neon.h: Likewise.
5068 * config/aarch64/iterators.md (VQ_I): New.
5070 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5072 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5074 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5077 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5079 * config/nds32/constraints.md (S): New constraint.
5080 * config/nds32/nds32.md (call_internal): Use constraint S.
5081 (call_value_internal): Likewise.
5082 (sibcall_internal): Likewise.
5083 (sibcall_value_internal): Likewise.
5085 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5086 Chung-Ju Wu <jasonwucj@gmail.com>
5088 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5091 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5092 Chung-Ju Wu <jasonwucj@gmail.com>
5094 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5095 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5096 (nds32_rtx_costs_impl): Simplify.
5097 (nds32_address_cost_impl): Simplify.
5098 (nds32_init_rtx_costs): New function.
5099 (nds32_rtx_costs_speed_prefer): Likewise.
5100 (nds32_rtx_costs_size_prefer): Likewise.
5101 (nds32_address_cost_speed_prefer): Likewise.
5102 (nds32_address_cost_speed_fwprop): Likewise.
5103 (nds32_address_cost_size_prefer): Likewise.
5104 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5105 * config/nds32/nds32.c (nds32_option_override): Use
5106 nds32_init_rtx_costs function.
5108 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5110 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5111 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5112 (TARGET_PIPELINE_N8): Likewise.
5113 (TARGET_PIPELINE_N10): Likewise.
5114 (TARGET_PIPELINE_N13): Likewise.
5115 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5117 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5119 * config/nds32/nds32-fpu.md: Update copyright year.
5121 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5123 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5125 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5127 * config/nds32/nds32.c
5128 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5129 * config/nds32/nds32.opt (minline-asm-r15): New option.
5131 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5133 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5135 * config/nds32/nds32.md (abssi2): New pattern.
5137 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5139 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5140 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5141 Merge insn pattern from sse_cvtsi2ss<round_name> and
5142 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5143 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5144 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5145 using SWI48 mode iterator.
5146 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5147 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5148 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5149 pattern from sse_cvttss2si<round_saeonly_name>
5150 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5151 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5152 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5153 using SWI48 mode iterator.
5154 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5155 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5156 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5157 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5158 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5159 using SWI48 mode iterator.
5160 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5161 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5162 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5163 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5164 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5165 SWI48 mode iterator.
5166 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5167 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5168 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5169 pattern from sse_cvttsd2si<round_saeonly_name>
5170 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5172 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5174 * config/nds32/nds32-md-auxiliary.c
5175 (nds32_valid_smw_lwm_base_p): Refine.
5176 (nds32_output_smw_single_word): Refine.
5177 (nds32_output_smw_double_word): New.
5178 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5180 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5182 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5183 (nds32_output_stack_pop): Refine.
5184 (nds32_expand_unaligned_load): Refine.
5185 (nds32_expand_unaligned_store): Refine.
5187 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5188 Chung-Ju Wu <jasonwucj@gmail.com>
5190 * config/nds32/constants.md: Add TP_REGNUM constant.
5191 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5192 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5194 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5195 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5196 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5197 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5198 and PIC code generation.
5199 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5201 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5203 * config/nds32/nds32.md: Support TLS and PIC.
5204 * config/nds32/nds32.c: Support TLS and PIC.
5205 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5206 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5209 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5211 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5212 mode with E_ prefix.
5214 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5215 Chung-Ju Wu <jasonwucj@gmail.com>
5217 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5218 * config/nds32/nds32-md-auxiliary.c
5219 (symbolic_reference_mentioned_p): New.
5220 (nds32_legitimize_ict_address): New.
5221 (nds32_expand_ict_move): New.
5222 (nds32_indirect_call_referenced_p): New.
5223 (nds32_symbol_binds_local_p): Delete.
5224 (nds32_long_call_p): Modify.
5225 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5226 * config/nds32/nds32-protos.h
5227 (symbolic_reference_mentioned_p): Declare.
5228 (nds32_legitimize_ict_address): Declare.
5229 (nds32_expand_ict_move): Declare.
5230 (nds32_indirect_call_referenced_p): Declare.
5231 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5232 (nds32_relax_group): Use nds32_ict_const_p as condition.
5233 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5234 (nds32_asm_file_start): Output ict_model directive in asm code.
5235 (nds32_legitimate_address_p): Consider indirect call.
5236 (nds32_print_operand): Consider indirect call.
5237 (nds32_print_operand_address): Consider indirect call.
5238 (nds32_insert_attributes): Handle "indirect_call" attribute.
5239 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5240 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5241 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5242 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5243 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5244 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5245 (TARGET_ICT_MODEL_SMALL): Define.
5246 (TARGET_ICT_MODEL_LARGE): Define.
5247 * config/nds32/nds32.md (movsi): Consider ict model.
5248 (call, call_value): Consider ict model.
5249 (sibcall, sibcall_value): Consider ict model.
5250 * config/nds32/nds32.opt (mict-model): New option.
5251 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5254 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5255 Monk Chiang <sh.chiang04@gmail.com>
5256 Jim Wilson <jimw@sifive.com>
5258 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5259 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5260 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5261 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5262 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5263 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5264 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5265 compute save_libcall_adjustment properly.
5266 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5267 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5268 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5269 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5270 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5271 (ABI_SPEC): Handle mabi=ilp32e.
5272 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5273 (RVE): Add RVE mask.
5274 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5275 <-march>: Add rv32e as an example.
5277 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5280 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5281 (intra_create_variable_infos): Handle C++ constructors.
5283 2018-05-18 Martin Liska <mliska@suse.cz>
5285 * passes.def: Remove a redundant pass.
5287 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5290 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5292 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5294 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5296 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5298 (armv2, armv2a, armv3, armv3m): Delete architectures.
5299 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5300 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5301 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5303 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5304 (*mulsidi3adddi): Likewise.
5305 (mulsidi3): Likewise.
5306 (*mulsidi3_nov6): Likewise.
5307 (umulsidi3): Likewise.
5308 (umulsidi3_nov6): Likewise.
5309 (umaddsidi4): Likewise.
5310 (*umulsidi3adddi): Likewise.
5311 (smulsi3_highpart): Likewise.
5312 (*smulsi3_highpart_nov6): Likewise.
5313 (umulsi3_highpart): Likewise.
5314 (*umulsi3_highpart_nov6): Likewise.
5315 * config/arm/arm.h (arm_arch3m): Delete.
5316 * config/arm/arm.c (arm_arch3m): Delete.
5317 (arm_option_override_internal): Update armv3-related comment.
5318 (arm_configure_build_target): Delete use of isa_bit_mode32.
5319 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5320 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5321 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5323 (mulusa3): Likewise.
5324 * config/arm/arm-protos.h (arm_arch3m): Delete.
5325 * config/arm/arm-tables.opt: Regenerate.
5326 * config/arm/arm-tune.md: Likewise.
5327 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5328 deleted architectures.
5330 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5332 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5333 (armv5t, armv5te): New features.
5334 (ARMv5, ARMv5e): Delete fgroups.
5335 (ARMv5t, ARMv5te): Adjust for above changes.
5337 (armv5, armv5e): Delete arches.
5338 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5340 (*call_reg_arm): Likewise.
5341 (*call_value_reg_armv5): Likewise.
5342 (*call_value_reg_arm): Likewise.
5343 (*call_symbol): Likewise.
5344 (*call_value_symbol): Likewise.
5345 (*sibcall_insn): Likewise.
5346 (*sibcall_value_insn): Likewise.
5348 (prefetch): Likewise.
5349 (define_split and define_peephole2 dependent on arm_arch5):
5351 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5353 (TARGET_ARM_QBIT): Likewise.
5354 (TARGET_DSP_MULTIPLY): Likewise.
5355 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5356 (arm_arch5, arm_arch5e): Delete.
5357 (arm_arch5t, arm_arch5te): Declare.
5358 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5359 (arm_arch5t): Declare.
5360 (arm_option_reconfigure_globals): Update for the above.
5361 (arm_options_perform_arch_sanity_checks): Update comment, replace
5362 use of arm_arch5 with arm_arch5t.
5363 (use_return_insn): Likewise.
5364 (arm_emit_call_insn): Likewise.
5365 (output_return_instruction): Likewise.
5366 (arm_final_prescan_insn): Likewise.
5367 (arm_coproc_builtin_available): Likewise.
5368 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5369 arm_arch5e with arm_arch5t and arm_arch5te.
5370 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5371 (arm_arch5t, arm_arch5te): Declare.
5372 * config/arm/arm-tables.opt: Regenerate.
5373 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5374 * config/arm/t-multilib: Likewise.
5375 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5376 instead of arm_arch5.
5377 (*call_reg_thumb1): Likewise.
5378 (*call_value_reg_thumb1_v5): Likewise.
5379 (*call_value_reg_thumb1): Likewise.
5380 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5382 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5384 2018-05-18 Martin Liska <mliska@suse.cz>
5386 PR gcov-profile/84846
5387 * doc/gcov.texi: Document -t option of gcov tool.
5389 2018-05-18 Martin Liska <mliska@suse.cz>
5391 PR gcov-profile/84846
5392 * gcov.c (print_usage): Add new -t option.
5393 (process_args): Handle the option.
5394 (generate_results): Use stdout as output when requested by
5397 2018-05-18 Martin Liska <mliska@suse.cz>
5399 PR gcov-profile/84846
5400 * coverage.c (coverage_init): Write PWD to .gcno file.
5401 * doc/gcov.texi: Document how working directory is printed.
5402 * gcov-dump.c (dump_gcov_file): Print PWD.
5403 * gcov.c (output_intermediate_file): Likewise.
5404 (read_graph_file): Read PWD string.
5405 (output_lines): Print PWD.
5407 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5410 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5411 for retval and return false if all args to phi are zero.
5413 2018-05-18 Richard Biener <rguenther@suse.de>
5415 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5417 (evrp_dom_walker::before_dom_children): Call it.
5419 2018-05-18 Richard Biener <rguenther@suse.de>
5421 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5422 results when processing array refs with variable index.
5424 2018-05-18 Toon Moene <toon@moene.org>
5426 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5427 directly after that of -floop-interchange. Indicate that both
5428 options are enabled by default when specifying -O3.
5430 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5432 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5433 iterator. Delete separate integer-mode vec_set<mode> expander.
5434 (aarch64_simd_vec_setv2di): Delete.
5435 (vec_setv2di): Delete.
5436 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5437 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5438 the "w, r" alternative.
5440 2018-05-18 Martin Liska <mliska@suse.cz>
5442 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5443 * tree-pass.h (make_pass_lower_switch_O0): New function.
5444 * tree-switch-conversion.c (node_has_low_bound): Remove.
5445 (node_has_high_bound): Likewise.
5446 (node_is_bounded): Likewise.
5447 (class pass_lower_switch): Make it a template type and create
5449 (pass_lower_switch::execute): Add template argument.
5450 (make_pass_lower_switch): New function.
5451 (make_pass_lower_switch_O0): New function.
5452 (do_jump_if_equal): Remove.
5453 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5454 all the hard work to tree optimization passes.
5456 2018-05-18 Martin Liska <mliska@suse.cz>
5458 * dbgcnt.c (limit_low): Renamed from limit.
5459 (limit_high): New variable.
5460 (dbg_cnt_is_enabled): Check for upper limit.
5461 (dbg_cnt): Adjust dumping.
5462 (dbg_cnt_set_limit_by_index): Add new argument for high
5464 (dbg_cnt_set_limit_by_name): Likewise.
5465 (dbg_cnt_process_single_pair): Parse new format.
5466 (dbg_cnt_process_opt): Use strtok.
5467 (dbg_cnt_list_all_counters): Remove 'value' and add
5469 * doc/invoke.texi: Document changes.
5471 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5473 * doc/sourcebuild.texi (scalar_all_fma): Document.
5474 * tree.def (FMA_EXPR): Delete.
5475 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5476 * internal-fn.c (ternary_direct): New macro.
5477 (expand_ternary_optab_fn): Likewise.
5478 (direct_ternary_optab_supported_p): Likewise.
5479 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5480 * builtins.c (fold_builtin_fma): Delete.
5481 (fold_builtin_3): Don't call it.
5482 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5483 * expr.c (expand_expr_real_2): Likewise.
5484 * fold-const.c (operand_equal_p): Likewise.
5485 (fold_ternary_loc): Likewise.
5486 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5487 * gimple.c (DEFTREECODE): Likewise.
5488 * gimplify.c (gimplify_expr): Likewise.
5489 * optabs-tree.c (optab_for_tree_code): Likewise.
5490 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5491 * tree-eh.c (operation_could_trap_p): Likewise.
5492 (stmt_could_throw_1_p): Likewise.
5493 * tree-inline.c (estimate_operator_cost): Likewise.
5494 * tree-pretty-print.c (dump_generic_node): Likewise.
5495 (op_code_prio): Likewise.
5496 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5497 * tree-ssa-operands.c (get_expr_operands): Likewise.
5498 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5499 * fold-const-call.h (fold_fma): Delete.
5500 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5501 CFN_FNMA and CFN_FNMS.
5503 * genmatch.c (combined_fn): New enum.
5504 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5505 (commutative_op): New function.
5506 (commutate): Use it. Handle more than 2 operands.
5507 (dt_operand::gen_gimple_expr): Use commutative_op.
5508 (parser::parse_expr): Allow :c to be used with non-binary
5509 operators if the commutative operand is known.
5510 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5511 CFN_FMS, CFN_FNMA and CFN_FNMS.
5512 (backprop::process_assign_use): Remove FMA_EXPR handling.
5513 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5514 (gen_hsa_fma): New function.
5515 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5516 IFN_FNMA and IFN_FNMS.
5517 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5518 * gimple-fold.h (follow_all_ssa_edges): Declare.
5519 * gimple-fold.c (follow_all_ssa_edges): New function.
5520 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5521 gimple_build interface and use follow_all_ssa_edges to fold the result.
5522 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5523 instead of checking for optabs directly.
5524 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5525 rather than FMA_EXPRs.
5526 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5527 call to IFN_FMA instead of an FMA_EXPR.
5529 2018-05-17 Jim Wilson <jimw@sifive.com>
5531 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5532 sign extended promoted subreg, and the range does not have the sign bit
5533 set, then do a sign extend.
5535 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5536 test, check for sign extended subreg and/or constant operands, and
5537 do a sign extend in that case.
5539 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5541 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5542 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5544 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5545 Change logics_shift_reg to logics_shift_imm.
5546 (thunderx2t99_fp_loadpair_basic): Delete.
5547 (thunderx2t99_fp_storepair_basic): Delete.
5548 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5549 (thunderx2t99_asimd_polynomial): Delete.
5550 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5551 and neon_fp_mul_d_scalar_q.
5552 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5553 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5554 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5555 (thunderx2t99_asimd_lut): Add missing tbl types.
5556 (thunderx2t99_asimd_ext): Delete.
5557 (thunderx2t99_asimd_load1_1_mult): Delete.
5558 (thunderx2t99_asimd_load1_2_mult): Delete.
5559 (thunderx2t99_asimd_load1_ldp): New.
5560 (thunderx2t99_asimd_load1): New.
5561 (thunderx2t99_asimd_load2): Add missing *load2* types.
5562 (thunderx2t99_asimd_load3): New.
5563 (thunderx2t99_asimd_load4): New.
5564 (thunderx2t99_asimd_store1_1_mult): Delete.
5565 (thunderx2t99_asimd_store1_2_mult): Delete.
5566 (thunderx2t99_asimd_store2_mult): Delete.
5567 (thunderx2t99_asimd_store2_onelane): Delete.
5568 (thunderx2t99_asimd_store_stp): New.
5569 (thunderx2t99_asimd_store1): New.
5570 (thunderx2t99_asimd_store2): New.
5571 (thunderx2t99_asimd_store3): New.
5572 (thunderx2t99_asimd_store4): New.
5574 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5576 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5577 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5579 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5580 Segher Boessenkool <segher@kernel.crashing.org>
5583 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5586 2018-05-17 Richard Biener <rguenther@suse.de>
5588 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5589 for pruning loop and prune defs feeding only already visited PHIs.
5591 2018-05-17 Richard Biener <rguenther@suse.de>
5593 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5595 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5596 Richard Biener <rguenther@suse.de>
5598 PR tree-optimization/85793
5599 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5600 for VMAT_ELEMENTWISE.
5602 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5604 * internal-fn.h (lookup_internal_fn): Declare
5605 * internal-fn.c (lookup_internal_fn): New function.
5606 * gimple.c (gimple_build_call_from_tree): Handle calls to
5608 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5609 internal function names.
5610 * tree-pretty-print.c (dump_generic_node): Likewise.
5611 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5613 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5615 * gimple-fold.h (gimple_build): Make the function forms take
5616 combined_fn rather than built_in_function.
5617 (gimple_simplify): Likewise.
5618 * gimple-match-head.c (gimple_simplify): Likewise.
5619 * gimple-fold.c (gimple_build): Likewise.
5620 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5621 rather than gimple_build_call_internal.
5622 (get_initial_defs_for_reduction): Likewise.
5623 (vect_create_epilog_for_reduction): Likewise.
5624 (vectorizable_live_operation): Likewise.
5626 2018-05-17 Martin Liska <mliska@suse.cz>
5628 * gimple-ssa-sprintf.c (format_directive): Do not use
5629 space in between 'G_' and '('.
5631 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5634 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5635 even if the mask is not all ones.
5638 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5640 (ix86_gimple_fold_builtin): Likewise.
5643 * config/i386/i386.c: Include tree-vector-builder.h.
5644 (ix86_vector_shift_count): New function.
5645 (ix86_fold_builtin): Fold shift builtins by scalar count.
5646 (ix86_gimple_fold_builtin): Likewise.
5648 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5649 _mm512_setzero): New intrinsics.
5651 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5652 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5654 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5655 code generation for cases where splatting a value is not useful.
5656 * simplify-rtx.c (simplify_ternary_operation): Simplify
5657 vec_merge across a vec_duplicate and a paradoxical subreg forming
5658 a vector mode to a vec_concat.
5660 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5662 * config.gcc: Support "goldmont-plus".
5663 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5665 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5666 PROCESSOR_GOLDMONT_PLUS.
5667 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5668 (processor_target_table): Add "goldmont-plus".
5669 (PTA_GOLDMONT_PLUS): Define.
5670 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5671 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5672 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5673 (fold_builtin_cpu): Add "goldmont-plus".
5674 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5675 (ix86_option_override_internal): Add "goldmont-plus".
5676 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5677 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5678 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5679 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5681 2018-05-17 Richard Biener <rguenther@suse.de>
5683 PR tree-optimization/85757
5684 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5685 remove defs that only feed that PHI from further processing.
5687 2018-05-16 Jim Wilson <jimw@sifive.com>
5689 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5691 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5693 2018-05-16 Mark Wielaard <mark@klomp.org>
5695 * dwarf2out.c (count_index_strings): New function.
5696 (output_indirect_strings): Call count_index_strings and generate
5697 header for dwarf_version >= 5.
5699 2018-05-16 Mark Wielaard <mark@klomp.org>
5701 * dwarf2out.c (dwarf_FORM): New function.
5702 (set_indirect_string): Use dwarf_FORM.
5703 (reset_indirect_string): Likewise.
5704 (size_of_die): Likewise.
5705 (value_format): Likewise.
5706 (output_die): Likewise.
5707 (add_skeleton_AT_string): Likewise.
5708 (output_macinfo_op): Likewise.
5709 (index_string): Likewise.
5710 (output_index_string_offset): Likewise.
5711 (output_index_string): Likewise.
5712 (count_index_strings): Likewise.
5714 2018-05-16 Carl Love <cel@us.ibm.com>
5716 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5717 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5719 2018-05-16 Martin Jambor <mjambor@suse.cz>
5721 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5722 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5724 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5726 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5727 (fnma<mode>4): Likewise.
5728 (fms<mode>4): Likewise.
5729 (fnms<mode>4): Likewise.
5730 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5731 (aarch64_fnma<mode>4): Likewise.
5732 (aarch64_fms<mode>4): Likewise.
5733 (aarch64_fnms<mode>4): Likewise.
5734 (aarch64_fnmadd<mode>4): Likewise.
5736 2018-05-16 Jason Merrill <jason@redhat.com>
5738 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5740 2018-05-16 Richard Biener <rguenther@suse.de>
5742 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5743 (dump_stmt_cost): Declare.
5744 (add_stmt_cost): Dump cost we add.
5745 (add_stmt_costs): New function.
5746 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5748 (vect_analyze_stmt): Adjust prototype.
5749 (vectorizable_condition): Likewise.
5750 (vectorizable_live_operation): Likewise.
5751 (vectorizable_reduction): Likewise.
5752 (vectorizable_induction): Likewise.
5753 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5754 cost vector to pass to vectorizable_ and record afterwards.
5755 (vect_model_reduction_cost): Take cost vector argument and adjust.
5756 (vect_model_induction_cost): Likewise.
5757 (vectorizable_reduction): Likewise.
5758 (vectorizable_induction): Likewise.
5759 (vectorizable_live_operation): Likewise.
5760 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5761 SLP_TREE_NUMBER_OF_VEC_STMTS.
5762 (vect_analyze_slp_cost_1): Remove.
5763 (vect_analyze_slp_cost): Likewise.
5764 (vect_slp_analyze_node_operations): Take visited args and
5765 a target cost vector. Avoid processing already visited stmt sets.
5766 (vect_slp_analyze_operations): Use a local cost vector to gather
5767 costs and register those of non-discarded instances.
5768 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5769 (vect_schedule_slp_instance): Remove copying of
5770 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5772 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5773 adding cost. Record cost entry location.
5774 (vect_prologue_cost_for_slp_op): Function to compute cost of
5775 a constant or invariant generated for SLP vect in the prologue,
5776 split out from vect_analyze_slp_cost_1.
5777 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5778 (vect_model_promotion_demotion_cost): Likewise.
5779 (vect_model_store_cost): Likewise, make static.
5780 (vect_model_load_cost): Likewise.
5781 (vectorizable_bswap): Add cost vector arg and adjust.
5782 (vectorizable_call): Likewise.
5783 (vectorizable_simd_clone_call): Likewise.
5784 (vectorizable_conversion): Likewise.
5785 (vectorizable_assignment): Likewise.
5786 (vectorizable_shift): Likewise.
5787 (vectorizable_operation): Likewise.
5788 (vectorizable_store): Likewise.
5789 (vectorizable_load): Likewise.
5790 (vectorizable_condition): Likewise.
5791 (vectorizable_comparison): Likewise.
5792 (can_vectorize_live_stmts): Likewise.
5793 (vect_analyze_stmt): Likewise.
5794 (vect_transform_stmt): Adjust calls to vectorizable_*.
5795 * tree-vectorizer.c: Include gimple-pretty-print.h.
5796 (dump_stmt_cost): New function.
5798 2018-05-16 Richard Biener <rguenther@suse.de>
5800 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5801 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5802 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5803 (check_name): New callback.
5804 (dse_classify_store): Track cycles via a visited bitmap of PHI
5805 defs and simplify handling of in-loop and across loop dead stores
5806 and properly fail for loop-variant refs. Handle byte-tracking with
5807 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5810 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5812 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5813 (vect_get_mask_type_for_stmt): Likewise.
5814 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5816 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5817 to determine the statement's vector type and the vector type that
5818 should be used for calculating nunits. Deal with cases in which
5819 the type has to be deferred.
5820 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5821 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5822 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5823 (vect_determine_vf_for_stmt): New functions, split out from...
5824 (vect_determine_vectorization_factor): ...here.
5825 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5826 (vect_get_mask_type_for_stmt): New functions, split out from
5827 vect_determine_vectorization_factor.
5829 2018-05-16 Richard Biener <rguenther@suse.de>
5831 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5832 verify the [VEC_]COND_EXPR embedded comparison.
5834 2018-05-15 Martin Sebor <msebor@redhat.com>
5836 PR tree-optimization/85753
5837 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5838 RECORD_TYPE in addition to ARRAY_TYPE.
5840 2018-05-15 Martin Sebor <msebor@redhat.com>
5843 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5845 2018-05-15 Richard Biener <rguenther@suse.de>
5847 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5848 add by_clobber_p one. Change algorithm to collect all defs
5849 representing uses we need to walk and try reducing them to
5850 a single one before failing.
5851 (dse_dom_walker::dse_optimize_stmt): Adjust.
5853 2018-05-13 Mark Wielaard <mark@klomp.org>
5855 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5856 (size_of_loc_descr): Likewise.
5857 (output_loc_operands): Likewise.
5858 (output_loc_operands_raw): Likewise.
5859 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5860 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5861 (hash_loc_operands): Likewise.
5862 (compare_loc_operands): Likewise.
5864 2018-05-14 Mark Wielaard <mark@klomp.org>
5866 * dwarf2out.c (count_index_addrs): New function.
5867 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5869 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5871 PR tree-optimization/83648
5872 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5873 return value as malloc candidate.
5875 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5878 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5879 param as true in call to suggest_attribute.
5881 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5883 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5884 -mreadonly-in-sdata.
5886 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5888 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5890 (aarch64_crypto_aesd_fused): Likewise.
5892 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5894 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5895 (movsi_aarch64): Likewise.
5896 (load_pairsi): Likewise.
5897 (load_pairdi): Likewise.
5898 (store_pairsi): Likewise.
5899 (store_pairdi): Likewise.
5900 (load_pairsf): Likewise.
5901 (load_pairdf): Likewise.
5902 (store_pairsf): Likewise.
5903 (store_pairdf): Likewise.
5904 (zero_extend): Likewise.
5905 (trunc): Swap alternatives.
5906 (fcvt_target): Add '?' to prefer w over r.
5908 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5911 * config/i386/i386.md: Disallow non-commutative arithmetics in
5912 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5913 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5914 in the peephole2 before it.
5916 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5918 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5919 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5920 (ix86_handle_option): Handle -mcldemote.
5921 * config.gcc: New header.
5922 * config/i386/cldemoteintrin.h: New file.
5923 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5924 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5927 OPTION_MASK_ISA_CLDEMOTE.
5928 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5929 (ix86_valid_target_attribute_inner_p): Ditto.
5930 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5931 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5932 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5933 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5934 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5936 * config/i386/i386.opt: Add -mcldemote.
5937 * config/i386/x86intrin.h: New header.
5938 * doc/invoke.texi: Add -mcldemote.
5940 2018-05-14 Richard Biener <rguenther@suse.de>
5942 * doc/match-and-simplify.texi: Adjust :s documentation.
5944 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5946 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5947 intended memcpy size.
5948 (REORDER_45): Likewise.
5950 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5952 * sort.cc: New file.
5953 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5954 * vec.c (qsort_chk): Use gcc_qsort.
5955 * Makefile.in (OBJS-libcommon): Add sort.o.
5956 (build/sort.o): New target. Use it...
5957 (BUILD_RTL): ... here, and...
5958 (build/gencfn-macros): ... here, and...
5959 (build/genmatch): ... here.
5961 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5962 Chung-Ju Wu <jasonwucj@gmail.com>
5964 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5965 * config/nds32/nds32-graywolf.md: New file.
5966 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5967 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5969 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5970 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5971 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5972 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5973 * config/nds32/pipelines.md: Include n15 settings.
5975 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5976 Chung-Ju Wu <jasonwucj@gmail.com>
5978 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5979 * config/nds32/nds32-n13.md: New file.
5980 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5981 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5983 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5984 * config/nds32/nds32.md (pipeline_model): Add n13.
5985 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5986 * config/nds32/pipelines.md: Include n13 settings.
5988 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5989 Chung-Ju Wu <jasonwucj@gmail.com>
5991 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5992 * config/nds32/nds32-n10.md: New file.
5993 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5994 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5996 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5997 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5998 * config/nds32/nds32.md (pipeline_model): Add n10.
5999 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6000 * config/nds32/pipelines.md: Include n10 settings.
6002 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6003 Kito Cheng <kito.cheng@gmail.com>
6004 Chung-Ju Wu <jasonwucj@gmail.com>
6006 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6007 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6008 Add enum values for DSP extension instructions.
6009 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6011 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6012 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6014 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6015 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6016 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6017 * config/nds32/nds32-intrinsic.md: Likewise.
6018 * config/nds32/nds32_intrinsic.h: Likewise.
6019 * config/nds32/nds32-md-auxiliary.c: Likewise.
6020 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6021 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6022 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6023 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6024 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6025 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6026 TYPE_DMAC in switch statement.
6027 * config/nds32/nds32.c: New checking and implementation for DSP
6028 extension instructions.
6029 * config/nds32/nds32.h: Likewise.
6030 * config/nds32/nds32.md: Likewise.
6031 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6032 * config/nds32/predicates.md: Implement new predicates for DSP
6035 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6037 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6038 Reformat alternatives and attributes so it is easier to identify
6039 which constraints/attributes go with which instruction.
6040 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6041 (mov<mode>_softfloat32, FMOVE64): Likewise.
6042 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6043 (mov<mode>_softfloat64, FMOVE64): Likewise.
6045 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6047 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6049 (Basic PowerPC Built-in Functions): The new name of the
6050 subsection previously known as "PowerPC Built-in Functions".
6051 (Basic PowerPC Built-in Functions Available on all Configurations):
6053 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6054 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6055 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6056 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6058 2018-05-11 Martin Jambor <mjambor@suse.cz>
6061 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6064 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6067 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6069 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6071 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6072 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6073 (ix86_handle_option): Handle -mwaitpkg.
6074 * config.gcc: New header.
6075 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6076 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6077 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6079 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6080 OPTION_MASK_ISA_WAITPKG.
6081 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6082 (ix86_option_override_internal): Add PTA_WAITPKG.
6083 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6084 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6085 IX86_BUILTIN_TPAUSE.
6086 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6087 __builtin_ia32_umwait and __builtin_ia32_tpause.
6088 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6089 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6090 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6091 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6092 UNSPECV_TPAUSE): New.
6093 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6094 * config/i386/i386.opt: Add -mwaitpkg.
6095 * config/i386/waitpkgintrin.h: New file.
6096 * config/i386/x86intrin.h: New header.
6097 * doc/invoke.texi: Add -mwaitpkg.
6099 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6102 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6104 (cortex-m0): Use armv6s-m isa.
6105 (cortex-m0plus): Likewise.
6106 (cortex-m1): Likewise.
6107 (cortex-m0.small-multiply): Likewise.
6108 (cortex-m0plus.small-multiply): Likewise.
6109 (cortex-m1.small-multiply): Likewise.
6111 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6112 Jakub Jelinek <jakub@redhat.com>
6114 PR tree-optimization/85692
6115 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6116 source permute as well.
6118 2018-05-11 Martin Liska <mliska@suse.cz>
6121 * doc/extend.texi: Document LLVM style format for no_sanitize
6124 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6126 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6127 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6128 (mode_supports_vsx_dform_quad): Likewise.
6129 (mode_supports_vmx_dform): Move these functions to be next to the
6130 other mode_supports functions.
6131 (mode_supports_dq_form): Likewise.
6132 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6133 mode_supports_dq_form.
6134 (reg_offset_addressing_ok_p): Likewise.
6135 (offsettable_ok_by_alignment): Likewise.
6136 (rs6000_legitimate_offset_address_p): Likewise.
6137 (legitimate_lo_sum_address_p): Likewise.
6138 (rs6000_legitimize_address): Likewise.
6139 (rs6000_legitimize_reload_address): Likewise.
6140 (rs6000_secondary_reload_inner): Likewise.
6141 (rs6000_preferred_reload_class): Likewise.
6142 (rs6000_output_move_128bit): Likewise.
6144 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6146 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6147 Generate SImode target register for null target.
6148 <case IX86_BUILTIN_XGETBV>: Ditto.
6149 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6150 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6152 2018-05-10 Carl Love <cel@us.ibm.com>
6154 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6155 dcbtt and dcbtstt if operands[2] is 0.
6157 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6160 * config/i386/sse.md (usadv64qi): New expander.
6162 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6164 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6165 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6166 -maltivec=be support.
6167 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6168 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6169 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6170 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6171 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6172 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6173 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6174 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6175 altivec_vsumsws): Adjust.
6176 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6177 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6178 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6180 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6181 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6182 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6183 (altivec_lve<VI_char>x): Delete expand.
6184 (*altivec_lve<VI_char>x_internal): Rename to...
6185 (altivec_lve<VI_char>x): ... this.
6186 (altivec_lvxl_<mode>): Delete expand.
6187 (*altivec_lvxl_<mode>_internal): Rename to ...
6188 (altivec_lvxl_<mode>): ... this.
6189 (altivec_stvxl_<mode>): Delete expand.
6190 (*altivec_stvxl_<mode>_internal): Rename to ...
6191 (altivec_stvxl_<mode>): ... this.
6192 (altivec_stve<VI_char>x): Delete expand.
6193 (*altivec_stve<VI_char>x_internal): Rename to ...
6194 (altivec_stve<VI_char>x): ... this.
6195 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6196 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6197 reduc_plus_scal_<mode>): Adjust.
6198 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6200 (rs6000_cpu_cpp_builtins): Adjust.
6201 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6202 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6203 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6204 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6205 -maltivec=be support.
6206 (rs6000_split_vec_extract_var): Adjust.
6207 (rs6000_split_v4si_init): Adjust.
6208 (swap_selector_for_mode): Delete.
6209 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6210 altivec_expand_stvex_be): Delete.
6211 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6212 -maltivec=be support.
6213 (rs6000_gimple_fold_builtin): Ditto.
6214 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6216 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6217 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6218 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6219 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6220 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6221 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6222 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6223 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6224 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6225 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6226 anonymous split): Adjust.
6227 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6228 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6230 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6232 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6233 when --with-gxx-include-dir is also specified.
6234 * configure: Regenerate.
6236 2018-05-09 Jim Wilson <jimw@sifive.com>
6239 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6240 * config/riscv/t-withmultilib: New.
6241 * config/riscv/withmultilib.h: New.
6242 * doc/install.texi: Document RISC-V --with-multilib-list support.
6244 2018-05-09 Richard Biener <rguenther@suse.de>
6246 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6248 (vect_bb_vectorization_profitable_p): Adjust. Compute
6249 actual scalar cost using the cost vector and the add_stmt_cost
6252 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6254 PR rtl-optimization/85645
6255 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6256 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6258 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6260 PR rtl-optimization/85645
6261 * shrink-wrap.c (spread_components): Return a boolean saying if
6262 anything was changed.
6263 (try_shrink_wrapping_separate): Iterate spread_components until
6264 nothing changes anymore.
6266 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6268 PR rtl-optimization/85645
6269 * regrename.c (build_def_use): Also kill the chains that include the
6270 destination of a REG_CFA_REGISTER note.
6272 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6274 PR rtl-optimization/85645
6275 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6276 insn that has a REG_CFA_REGISTER note.
6278 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6280 * cfgexpand.c (expand_clobber): New function.
6281 (expand_gimple_stmt_1): Use it.
6282 * tree-vect-stmts.c (vect_clobber_variable): New function,
6284 (vectorizable_simd_clone_call): ...here.
6285 (vectorizable_store): Emit a clobber either side of an
6286 IFN_STORE_LANES sequence.
6287 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6289 2018-05-09 Tom de Vries <tom@codesourcery.com>
6292 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6293 (define_insn "trap_if_false"): Add exit after trap.
6295 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6297 PR rtl-optimization/85638
6298 * bb-reorder.c: Include common/common-target.h.
6299 (create_forwarder_block): New function extracted from...
6300 (fix_up_crossing_landing_pad): ...here. Rename into...
6301 (dw2_fix_up_crossing_landing_pad): ...this.
6302 (sjlj_fix_up_crossing_landing_pad): New function.
6303 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6304 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6305 from both partitions and exit the loop after one iteration.
6307 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6310 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6312 (Basic PowerPC Built-in Functions): The new name of the
6313 subsection previously known as "PowerPC Built-in Functions".
6314 (Basic PowerPC Built-in Functions Available on all Configurations):
6316 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6318 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6320 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6322 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6325 2018-05-08 Jim Wilson <jimw@sifive.com>
6327 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6328 (LD_EMUL_SUFFIX): New.
6329 (LINK_SPEC): Use it.
6331 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6333 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6335 (Basic PowerPC Built-in Functions): The new name of the
6336 subsection previously known as "PowerPC Built-in Functions".
6337 (Basic PowerPC Built-in Functions Available on all Configurations):
6339 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6341 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6343 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6345 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6348 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6351 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6352 after cmpelim optimization.
6354 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6356 * config.gcc: Support "goldmont".
6357 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6358 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6360 * config/i386/i386.c (m_GOLDMONT): Define.
6361 (processor_target_table): Add "goldmont".
6362 (PTA_GOLDMONT): Define.
6363 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6364 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6365 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6366 (fold_builtin_cpu): Add "goldmont".
6367 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6368 (ix86_option_override_internal): Add "goldmont".
6369 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6370 (processor_type): Add PROCESSOR_GOLDMONT.
6371 * config/i386/i386.md: Add CPU "glm".
6372 * config/i386/glm.md: New file.
6373 * config/i386/x86-tune.def: Add m_GOLDMONT.
6374 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6376 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6379 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6381 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6382 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6383 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6386 * config/i386/i386.c (ix86_fold_builtin): Handle
6387 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6390 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6391 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6393 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6396 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6397 (check_arch): Likewise.
6398 (check_fpu): Return the result rather than printing it.
6399 (end arch): Fix operator precedence.
6400 (end cpu): Likewise.
6401 (END): Print the result from check_fpu.
6403 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6404 Alan Hayward <alan.hayward@arm.com>
6405 David Sherwood <david.sherwood@arm.com>
6407 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6408 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6409 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6410 (*fcmuo<mode>_and): New patterns.
6412 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6414 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6415 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6416 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6417 (cmp_op, sve_imm_con): New code attributes.
6418 (SVE_COND_INT_CMP, imm_con): Delete.
6419 (cmp_op): Remove above unspecs from int attribute.
6420 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6422 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6423 comparison-specific unspecs.
6424 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6425 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6426 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6427 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6428 (*vec_fcm<cmp_op><mode>): Rename to...
6429 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6430 (*vec_fcmuo<mode>): Rename to...
6431 (*fcmuo<mode>): ...this and adjust likewise.
6432 (*pred_fcm<cmp_op><mode>): New pattern.
6433 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6434 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6436 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6438 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6439 (aarch64_emit_sve_predicated_cond): New function.
6440 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6441 (aarch64_emit_unspec_cond_or): Replace with...
6442 (aarch64_emit_sve_or_conds): ...this new function. Use
6443 aarch64_emit_sve_ptrue_op for the individual comparisons and
6444 aarch64_emit_binop to OR them together.
6445 (aarch64_emit_inverted_unspec_cond): Replace with...
6446 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6447 aarch64_emit_sve_ptrue_op for the comparison and
6448 aarch64_emit_unop to invert the result.
6449 (aarch64_expand_sve_vec_cmp_float): Update after the above
6450 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6452 2018-05-07 Nathan Sidwell <nathan@acm.org>
6454 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6455 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6456 (Backwards Compatibility): Likewise.
6458 2018-05-07 Luis Machado <luis.machado@linaro.org>
6462 2018-05-07 Luis Machado <luis.machado@linaro.org>
6464 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6465 <prefetch_dynamic_strides>: New const bool field.
6466 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6467 prefetch_dynamic_strides.
6468 (exynosm1_prefetch_tune): Likewise.
6469 (thunderxt88_prefetch_tune): Likewise.
6470 (thunderx_prefetch_tune): Likewise.
6471 (thunderx2t99_prefetch_tune): Likewise.
6472 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6474 (aarch64_override_options_internal): Update to set
6475 PARAM_PREFETCH_DYNAMIC_STRIDES.
6476 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6477 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6478 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6479 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6480 prefetch-dynamic-strides setting.
6482 2018-05-07 Luis Machado <luis.machado@linaro.org>
6484 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6485 <minimum_stride>: New const int field.
6486 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6487 minimum_stride field.
6488 (exynosm1_prefetch_tune): Likewise.
6489 (thunderxt88_prefetch_tune): Likewise.
6490 (thunderx_prefetch_tune): Likewise.
6491 (thunderx2t99_prefetch_tune): Likewise.
6492 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6493 (aarch64_override_options_internal): Update to set
6494 PARAM_PREFETCH_MINIMUM_STRIDE.
6495 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6496 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6497 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6498 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6499 stride is constant and is below the minimum stride threshold.
6501 2018-05-07 Luis Machado <luis.machado@linaro.org>
6503 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6506 2018-05-07 Luis Machado <luis.machado@linaro.org>
6508 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6509 <prefetch_dynamic_strides>: New const bool field.
6510 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6511 prefetch_dynamic_strides.
6512 (exynosm1_prefetch_tune): Likewise.
6513 (thunderxt88_prefetch_tune): Likewise.
6514 (thunderx_prefetch_tune): Likewise.
6515 (thunderx2t99_prefetch_tune): Likewise.
6516 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6518 (aarch64_override_options_internal): Update to set
6519 PARAM_PREFETCH_DYNAMIC_STRIDES.
6520 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6521 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6522 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6523 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6524 prefetch-dynamic-strides setting.
6526 2018-05-07 Luis Machado <luis.machado@linaro.org>
6528 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6529 <minimum_stride>: New const int field.
6530 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6531 minimum_stride field.
6532 (exynosm1_prefetch_tune): Likewise.
6533 (thunderxt88_prefetch_tune): Likewise.
6534 (thunderx_prefetch_tune): Likewise.
6535 (thunderx2t99_prefetch_tune): Likewise.
6536 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6537 (aarch64_override_options_internal): Update to set
6538 PARAM_PREFETCH_MINIMUM_STRIDE.
6539 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6540 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6541 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6542 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6543 stride is constant and is below the minimum stride threshold.
6545 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6548 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6549 the type is addressable. Don't force op into register if it has
6552 2018-05-05 Roland McGrath <mcgrathr@google.com>
6555 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6556 any section for which we don't know a specific type it should have,
6557 regardless of name. Previously this was done only for the exact
6558 names ".init_array", ".fini_array", and ".preinit_array".
6559 (default_elf_asm_named_section): Add comment about
6560 relationship with default_section_type_flags and SECTION_NOTYPE.
6561 (get_section): Don't consider it a type conflict if one side has
6562 SECTION_NOTYPE and the other doesn't, as long as neither has the
6563 SECTION_BSS et al used in the default_section_type_flags logic.
6565 2018-05-05 Tom de Vries <tom@codesourcery.com>
6568 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6569 (workaround_barsyncs): New function.
6570 (nvptx_reorg): Use workaround_barsyncs.
6571 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6572 (define_expand "nvptx_membar_cta"): New define_expand.
6573 (define_insn "*nvptx_membar_cta"): New insn.
6575 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6577 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6578 To improve optimization opportunities.
6579 * builtin-types.def: The new needed builtin types for the above.
6581 2018-05-04 Richard Biener <rguenther@suse.de>
6583 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6584 * gimple-ssa-store-merging.c
6585 (imm_store_chain_info::output_merged_store): Remove redundant create,
6586 release split_store vector contents on failure.
6587 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6588 scalar stmt vector on cache hit.
6590 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6592 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6594 * config.gcc (powerpc-xilinx-eabi*): Remove.
6595 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6597 (fusion_addis_mem_combo_load): Ditto.
6598 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6600 (rs6000_cpu_cpp_builtins): Ditto.
6601 * config/rs6000/rs6000-linux.c
6602 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6603 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6604 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6606 (rs6000_setup_reg_addr_masks): Ditto.
6607 (rs6000_init_hard_regno_mode_ok): Ditto.
6608 (rs6000_option_override_internal): Ditto.
6609 (legitimate_lo_sum_address_p): Ditto.
6610 (rs6000_legitimize_address): Ditto.
6611 (rs6000_legitimize_reload_address): Ditto.
6612 (rs6000_legitimate_address_p): Ditto.
6613 (abi_v4_pass_in_fpr): Ditto.
6614 (setup_incoming_varargs): Ditto.
6615 (rs6000_gimplify_va_arg): Ditto.
6616 (rs6000_split_multireg_move): Ditto.
6617 (rs6000_savres_strategy): Ditto.
6618 (rs6000_emit_prologue_components): Ditto.
6619 (rs6000_emit_epilogue_components): Ditto.
6620 (rs6000_emit_prologue): Ditto.
6621 (rs6000_emit_epilogue): Ditto.
6622 (rs6000_elf_file_end): Ditto.
6623 (rs6000_function_value): Ditto.
6624 (rs6000_libcall_value): Ditto.
6625 * config/rs6000/rs6000.h: Ditto.
6626 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6627 (TARGET_MINMAX): ... this. New.
6628 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6629 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6630 (*movsi_internal1_single): Delete.
6631 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6632 mfpu=, mxilinx-fpu): Delete.
6633 * config/rs6000/singlefp.h: Delete.
6634 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6635 * config/rs6000/t-rs6000: Ditto.
6636 * config/rs6000/t-xilinx: Delete.
6637 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6638 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6639 (VStype_simple): Delete.
6640 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6641 * config/rs6000/xfpu.h: Delete.
6642 * config/rs6000/xfpu.md: Delete.
6643 * config/rs6000/xilinx.h: Delete.
6644 * config/rs6000/xilinx.opt: Delete.
6645 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6646 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6648 2018-05-04 Tom de Vries <tom@codesourcery.com>
6651 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6654 2018-05-04 Richard Biener <rguenther@suse.de>
6657 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6658 (expand_complex_div_wide): Likewise.
6659 (expand_complex_operations_1): Likewise.
6660 (expand_complex_libcall): Preserve EH info of the original stmt.
6661 (tree_lower_complex): Handle removed blocks.
6662 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6663 on complex multiplication and division libcall builtins.
6665 2018-05-04 Richard Biener <rguenther@suse.de>
6668 * fold-const.c (negate_expr_p): Restrict negation of operand
6669 zero of a division to when we know that can happen without
6671 (fold_negate_expr_1): Likewise.
6673 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6676 * real.h (real_nextafter): Declare.
6677 * real.c (real_nextafter): New function.
6678 * fold-const-call.c (fold_const_nextafter): New function.
6679 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6680 CASE_CFN_NEXTTOWARD.
6681 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6682 even when arg1_mode is different from arg0_mode.
6684 2018-05-03 Nathan Sidwell <nathan@acm.org>
6686 * doc/extend.texi (Deprecated Features): Remove
6688 (Backwards Compatibility): Likewise.
6689 * doc/invoke.texi (C++ Language Options): Likewise.
6690 (C++ Dialect Options): Likewise.
6692 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6695 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6696 _mm512_mask_mullox_epi64): New intrinsics.
6698 2018-05-03 Tom de Vries <tom@codesourcery.com>
6701 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6702 dump files): Add offload-tree.
6704 2018-05-03 Richard Biener <rguenther@suse.de>
6706 PR tree-optimization/85615
6707 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6708 to loops not nested in BBs loop father to avoid creating multi-entry
6711 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6713 PR tree-optimization/70291
6714 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6715 arguments. Change return type to tree. Emit libcall as a new
6716 statement rather than replacing existing one when inplace_p is true.
6717 (expand_complex_multiplication_components): New function.
6718 (expand_complex_multiplication): Expand floating-point complex
6719 multiplication using the above.
6720 (expand_complex_division): Rename inner_type parameter to type.
6721 Update expand_complex_libcall call-site.
6722 (expand_complex_operations_1): Update expand_complex_multiplication
6723 and expand_complex_division call-sites.
6725 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6728 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6729 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6730 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6731 the highest significant bit of the shift count mask is clear. In
6732 check whether and[sq]i3 is needed verify that all significant bits
6733 of the shift count other than the highest are set.
6735 2018-05-02 Tom de Vries <tom@codesourcery.com>
6738 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6739 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6740 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6741 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6742 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6743 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6744 __builtin_goacc_parlevel_size.
6746 2018-05-02 Richard Biener <rguenther@suse.de>
6748 PR tree-optimization/85597
6749 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6750 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6753 2018-05-02 Tom de Vries <tom@codesourcery.com>
6756 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6757 dump files): Add ltrans-tree.
6759 2018-05-02 Tom de Vries <tom@codesourcery.com>
6762 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6763 dump files): Add wpa-ipa.
6765 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6767 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6768 powerpc*-*-linux*paired* target.
6769 * config/rs6000/750cl.h: Delete.
6770 * config/rs6000/paired.h: Delete.
6771 * config/rs6000/paired.md: Delete.
6772 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6774 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6775 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6776 comment. Remove paired float support.
6777 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6778 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6780 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6781 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6783 * config/rs6000/rs6000.c: Remove paired float support.
6784 (paired_expand_vector_init, paired_expand_vector_move,
6785 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6786 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6787 paired_expand_builtin, paired_expand_predicate_builtin,
6788 paired_init_builtins): Delete.
6789 * config/rs6000/rs6000.h: Remove paired float support.
6790 * config/rs6000/rs6000.md: Remove paired float support.
6791 (move_from_CR_ov_bit): Delete.
6792 * config/rs6000/rs6000.opt (mpaired): Delete.
6793 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6794 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6796 2018-05-02 Richard Biener <rguenther@suse.de>
6799 * gimplify.c (gimplify_save_expr): When in SSA form allow
6800 SAVE_EXPRs to compute to SSA vars.
6802 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6805 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6806 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6807 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6808 clobber operands[2], instead use a new pseudo. Formatting fixes.
6810 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6812 PR tree-optimization/85586
6813 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6814 exit early for statements in the same group if the accesses are
6817 2018-05-02 Tom de Vries <tom@codesourcery.com>
6820 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6823 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6825 PR tree-optimization/85143
6826 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6828 2018-05-01 Tom de Vries <tom@codesourcery.com>
6831 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6832 not found" error message.
6834 2018-05-01 Tom de Vries <tom@codesourcery.com>
6837 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6838 * vec.c (test_ordered_remove_if): New function.
6839 (vec_c_tests): Call test_ordered_remove_if.
6840 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6841 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6842 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6843 VEC_ORDERED_REMOVE_IF.
6845 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6847 PR tree-optimization/82665
6848 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6849 pointer subtraction where arguments come from a memchr call.
6851 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6853 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6854 --push-state --as-needed and --pop-state instead of --as-needed and
6855 --no-as-needed if ld supports it.
6856 * configure: Regenerated.
6859 * doc/install.texi2html: Replace _002d with - and _002a with * in
6860 generated html files using sed.
6862 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6865 * gcc-rich-location.c (blank_line_before_p): New function.
6866 (use_new_line): New function.
6867 (gcc_rich_location::add_fixit_insert_formatted): New function.
6868 * gcc-rich-location.h
6869 (gcc_rich_location::add_fixit_insert_formatted): New function.
6871 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6873 * selftest.c (assert_streq): Rename "expected" and "actual" to
6874 "val1" and "val2". Extend NULL-handling to cover both inputs
6875 symmetrically, while still requiring both to be non-NULL for a pass.
6876 * selftest.h (assert_streq): Rename "expected" and "actual" to
6878 (ASSERT_EQ): Likewise.
6879 (ASSERT_EQ_AT): Likewise.
6880 (ASSERT_KNOWN_EQ): Likewise.
6881 (ASSERT_KNOWN_EQ_AT): Likewise.
6882 (ASSERT_NE): Likewise.
6883 (ASSERT_MAYBE_NE): Likewise.
6884 (ASSERT_MAYBE_NE_AT): Likewise.
6885 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6886 the assertion to pass.
6887 (ASSERT_STREQ_AT): Likewise.
6889 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6891 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6892 interaction with -pie.
6894 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6896 * selftest.h: Fix alphabetization of per-source-file selftest
6899 2018-04-30 Jason Merrill <jason@redhat.com>
6901 PR c++/61982 - dead stores to destroyed objects.
6902 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6905 2018-04-30 Jason Merrill <jason@redhat.com>
6907 * tree.c (build_clobber): New.
6908 * tree.h: Declare it.
6909 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6911 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6913 * diagnostic-show-locus.c (layout::layout): Update for
6914 location_get_source_line returning a char_span.
6915 (struct char_span): Move to input.h.
6916 (struct correction): Update for fields in char_span becoming
6918 (struct source_line): Update for location_get_source_line
6919 returning a char_span.
6920 (layout::print_line): Likewise.
6921 * edit-context.c (edited_file::print_content): Likewise.
6922 (edited_file::print_diff_hunk): Likewise.
6923 (edited_file::print_run_of_changed_lines): Likewise.
6924 (edited_file::get_num_lines): Likewise.
6925 (edited_line::edited_line): Likewise.
6926 * final.c (asm_show_source): Likewise.
6927 * input.c (location_get_source_line): Convert return type
6928 from const char * to char_span, losing the final "line_len"
6930 (dump_location_info): Update for the above.
6931 (get_substring_ranges_for_loc): Likewise. Use a char_span
6932 when handling the literal within the line.
6933 (test_reading_source_line): Update for location_get_source_line
6934 returning a char_span.
6935 * input.h (class char_span): Move here from
6936 diagnostic-show-locus.c, converting from a struct to a class.
6937 Make data members private.
6938 (char_span::operator bool): New.
6939 (char_span::length): New.
6940 (char_span::get_buffer): New.
6941 (char_span::operator[]): New.
6942 (char_span::subspan): Make const.
6943 (char_span::xstrdup): New.
6944 (location_get_source_line): Convert return type from const char *
6945 to char_span, losing the final "line_size" param.
6947 2018-04-30 Jan Hubicka <jh@suse.cz>
6949 * lto-wrapper.c (ltrans_priorities): New static var.
6950 (cmp_priority): New.
6951 (run_gcc): Read priorities and if doing parallel build order
6952 the Makefile by them.
6954 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6956 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6958 2018-04-30 Richard Biener <rguenther@suse.de>
6960 * tree-cfg.c (verify_address): Remove base argument, add
6961 flag whether to check TREE_ADDRESSABLE and do that.
6962 (verify_expr): Remove.
6963 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6964 (verify_gimple_assign_single): Likewise.
6965 (verify_gimple_switch): Likewise.
6966 (verify_expr_location_1): Dereference tp once. Add (disabled)
6967 piece from verify_expr.
6968 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6970 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6972 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6974 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6976 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6977 (small_data_pattern): Likewise.
6978 (arc_rewrite_small_data): Likewise.
6979 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6980 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6981 (get_symbol_alignment): New function.
6982 (legitimate_small_data_address_p): Likewise.
6983 (legitimate_scaled_address): Update, call
6984 legitimate_small_data_address_p.
6985 (output_sdata): New static variable.
6986 (arc_print_operand): Update how we handle small data operands.
6987 (arc_print_operand_address): Likewise.
6988 (arc_legitimate_address_p): Update, use
6989 legitimate_small_data_address_p.
6990 (arc_rewrite_small_data_p): Remove.
6991 (arc_rewrite_small_data_1): Likewise.
6992 (arc_rewrite_small_data): Likewise.
6993 (small_data_pattern): Likewise.
6994 (compact_sda_memory_operand): Update to use
6995 legitimate_small_data_address_p and get_symbol_alignment.
6996 (prepare_move_operands): Don't rewite sdata pattern.
6997 (prepare_extend_operands): Remove.
6998 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7000 (zero_extendqisi2): Likewise.
7001 (zero_extendhisi2): Likewise.
7002 (extendqihi2): Likewise.
7003 (extendqisi2): Likewise.
7004 (extendhisi2): Likewise.
7008 * config/arc/constraints.md (Usd): Change it to memory constraint.
7010 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7012 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7013 as source of std instructions.
7014 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7015 allow 6-bit constants as source for store instructions.
7016 (movdi_insn): Update instruction pattern to allow 6-bit constants
7017 as source for store instructions.
7019 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7021 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7023 2018-04-30 Nathan Sidwell <nathan@acm.org>
7024 Sandra Loosemore <sandra@codesourcery.com>
7026 * dumpfile.c (dump_open): Allow '-' for stdout.
7027 * doc/invoke.texi (Developer Options): Document dump filename
7028 determination early. Document stdin/stdout selection.
7030 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7032 Microblaze Target: PIC data text relative
7034 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7035 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7037 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7038 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7039 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7040 New addressing mode for data-text relative position indepenedent code.
7041 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7042 'ADDRESS_SYMBOLIC_TXT_REL'.
7043 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7044 (microblaze_legitimate_pic_operand): Exclude function calls from
7045 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7046 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7048 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7049 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7050 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7051 for 'address + offset'.
7052 (microblaze_expand_prologue): Add new function prologue call for
7054 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7055 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7056 table in case of TARGET_PIC_DATA_TEXT_REL.
7057 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7058 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7059 Add new macros 'UNSPEC_TEXT',
7060 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7061 + exclude function calls from 'UNSPEC_PLT' in case of data text
7063 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7064 new target hook for generating address diff vector tables in case of
7066 * doc/tm.texi : Regenerate.
7067 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7068 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7069 of addr diff vector generation.
7070 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7071 target hook definition.
7072 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7073 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7074 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7075 option for data text relative.
7077 2018-04-30 Richard Biener <rguenther@suse.de>
7079 * tree-chrec.h (evolution_function_is_constant_p): Remove
7081 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7083 2018-04-30 Richard Biener <rguenther@suse.de>
7086 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7088 2018-04-30 Richard Biener <rguenther@suse.de>
7090 PR tree-optimization/28364
7091 PR tree-optimization/85275
7092 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7093 copying first exit test.
7095 2018-04-28 Mark Wielaard <mark@klomp.org>
7097 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7099 (dwarf_AT): Handle DW_AT_addr_base.
7100 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7102 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7105 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7106 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7107 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7108 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7110 2018-04-28 Richard Biener <rguenther@suse.de>
7112 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7113 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7114 to reflect use. Only add interesting stmts.
7116 2018-04-27 Martin Jambor <mjambor@suse.cz>
7119 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7120 the jump function allows for passing through aggregate values.
7122 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7124 * input.h (in_system_header_at): Convert from macro to inline
7126 (from_macro_expansion_at): Likewise.
7127 (from_macro_definition_at): Likewise.
7129 2018-04-27 Jeff Law <law@redhat.com>
7131 * config.gcc: Mark tile* targets as deprecated/obsolete.
7133 2018-04-27 Richard Biener <rguenther@suse.de>
7135 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7138 2018-04-27 Richard Biener <rguenther@suse.de>
7140 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7142 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7144 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7145 with Yd constraint. Set "preferred_for_speed" attribute from
7146 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7148 (*movdi_internal): Ditto.
7149 (movti_interunit splitters): Remove
7150 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7151 (movdi_interunit splitters): Ditto.
7152 * config/i386/constraints.md (Ye): Remove.
7153 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7155 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7158 * config/aarch64/constraints.md (Usg): Limit to 31.
7161 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7163 PR tree-optimization/85529
7164 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7165 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7166 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7167 zero extension or masking of the MSB bit.
7168 (optimize_range_tests): Add FIRST_BB argument, pass it through
7169 to optimize_range_tests_var_bound.
7170 (maybe_optimize_range_tests, reassociate_bb): Adjust
7171 optimize_range_tests callers.
7173 2018-04-26 Richard Biener <rguenther@suse.de>
7174 Jakub Jelinek <jakub@redhat.com>
7176 * cgraph.h (symbol_table): Just declare debug method here.
7177 * symtab.c (symbol_table::debug): Define.
7179 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7181 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7183 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7185 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7186 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7187 (*movdi_internal): Substitute Yi and Yj constraint with x
7188 and Ym and Yn constraint with y constraint. Update "isa"
7189 attribute and set "preferred_for_speed" attribute from
7190 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7191 (*movsi_internal): Ditto.
7192 (*movdf_internal): Ditto.
7193 (*movsf_internal): Ditto.
7194 (*zero_extendsidi2): Ditto.
7195 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7196 (sse2_loadld): Ditto.
7197 (*vec_extract<ssevecmodelower>_0): Ditto.
7198 (*vec_extractv4si_0_zext_sse4): Ditto.
7199 (vec_concatv2di): Ditto.
7200 (*vec_dup<mode>): Ditto.
7201 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7202 * config/i386/constraints.md (Yi): Remove.
7207 2018-04-26 Nathan Sidwell <nathan@acm.org>
7209 * dumpfile.c (dump_open): New.
7210 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7211 (dump_finish): Detect stdio/stderr by value not name.
7213 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7215 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7217 2018-04-26 Tom de Vries <tom@codesourcery.com>
7220 * config/nvptx/nvptx.c (verify_neutering_jumps)
7221 (verify_neutering_labels): New function
7222 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7224 2018-04-26 Tom de Vries <tom@codesourcery.com>
7227 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7228 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7231 2018-04-26 Richard Biener <rguenther@suse.de>
7232 Tom de Vries <tom@codesourcery.com>
7235 * lto-streamer-out.c (output_function): Fixup loops if required to match
7236 discovery done in the reader.
7238 2018-04-26 Richard Biener <rguenther@suse.de>
7240 PR tree-optimization/85116
7241 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7242 have a loop exit from the single latch predecessor. Remove
7243 case of header with just condition.
7244 (ch_base::copy_headers): Exclude infinite loops from any
7246 (pass_ch::execute): Record exits.
7248 2018-04-26 Richard Biener <rguenther@suse.de>
7250 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7251 prologue cost vector and pass it to vect_get_load_cost.
7252 (vect_get_peeling_costs_all_drs): Likewise.
7253 (vect_peeling_hash_get_lowest_cost): Likewise.
7254 (vect_enhance_data_refs_alignment): Likewise.
7256 2018-04-26 Richard Biener <rguenther@suse.de>
7259 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7260 checking of integer<->pointer conversions.
7261 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7262 sign-/zero-extending pointer types.
7263 (expand_omp_for_static_chunk): Likewise.
7265 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7266 Jean Lee <xiaoyur347@gmail.com>
7268 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7269 (TARGET_ASAN_SHADOW_OFFSET): Define.
7270 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7271 true for -fsanitize=address.
7273 2018-04-25 Mark Wielaard <mark@klomp.org>
7275 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7278 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7280 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7281 than "alu", remove explicit "memory" and "imm_disp" attributes.
7282 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7285 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7286 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7287 gen_lowpart_no_emit.
7289 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7292 * config/i386/i386.c (ix86_expand_builtin): Change memory
7293 operand to XI, extend p0 to Pmode.
7294 * config/i386/i386.md: Change unspec volatile and operand
7295 1 mode to XI, change operand 0 mode to P.
7297 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7299 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7300 GET_MODE_MASK before any checking.
7301 (nds32_can_use_bset_p): Likewise.
7302 (nds32_can_use_btgl_p): Likewise.
7304 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7306 * config/nds32/nds32-doubleword.md: New define_split pattern for
7307 illegal register number.
7309 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7311 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7313 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7315 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7317 2018-04-25 Richard Biener <rguenther@suse.de>
7319 * lto-streamer.h (LTO_major_version): Bump to 8.
7321 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7323 * BASE-VER: Set to 9.0.0.
7325 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7327 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7328 in __abskf2 and __powikf2.
7330 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7333 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7334 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7335 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7336 Use the above on operand 2. Reindent.
7337 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7339 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7342 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7344 * config/i386/i386.opt (mcet): Removed.
7345 * doc/install.texi: Remove -mcet documentation.
7346 * doc/invoke.texi: Likewise.
7348 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7351 * doc/install.texi: Remove -mcet from bootstrap-cet.
7353 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7356 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7357 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7361 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7362 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7363 containing a CONST_VECTOR.
7365 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7367 * doc/install.texi: Update newlib dependency for nvptx.
7369 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7372 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7373 instead of INTVAL when shifting x left.
7375 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7377 PR tree-optimization/85478
7378 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7379 vect_grouped_store_supported for single element vectors.
7381 2018-04-24 Richard Biener <rguenther@suse.de>
7384 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7385 load cost increase to the case of non-constant step.
7387 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7390 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7391 destination if any_malformed_asm.
7393 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7396 * expr.c (store_field): In the bitfield case, if the value comes from
7397 a function call and is returned in registers by means of a PARALLEL,
7398 do not change the mode of the temporary unless BLKmode and VOIDmode.
7400 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7402 PR rtl-optimization/85423
7403 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7404 dependencies to debug insns when the previous insn is non-debug.
7406 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7408 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7409 enums into a single definition.
7410 (fls): Fix predicates and printing.
7413 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7415 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7416 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7417 and short u6 immediate.
7418 (check_if_valid_sleep_operand): Remove.
7419 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7421 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7423 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7424 flag_always_save_lp condition.
7425 * config/nds32/nds32.opt (malways-save-lp): New option.
7427 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7429 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7430 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7431 * config/nds32/nds32.h
7432 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7433 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7435 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7437 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7438 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7440 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7441 Chung-Ju Wu <jasonwucj@gmail.com>
7443 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7445 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7446 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7448 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7450 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7452 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7454 * config/nds32/nds32-protos.h (nds32_data_alignment,
7455 nds32_local_alignment): Declare.
7456 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7457 nds32_local_alignment): New functions.
7458 (TARGET_CONSTANT_ALIGNMENT): Define.
7459 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7461 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7463 * config/nds32/nds32.c
7464 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7465 (TARGET_MODES_TIEABLE_P): Likewise.
7467 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7469 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7472 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7473 Chung-Ju Wu <jasonwucj@gmail.com>
7475 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7476 for unaligned access.
7477 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7479 * config/nds32/nds32-intrinsic.md: Likewise.
7480 * config/nds32/nds32_intrinsic.h: Likewise.
7481 * config/nds32/nds32.h (nds32_builtins): Likewise.
7482 * config/nds32/nds32.opt (munaligned-access): New option.
7483 * config/nds32/nds32.c (nds32_asm_file_start): Display
7484 flag_unaligned_access status.
7486 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7488 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7489 -mno-relax is present.
7490 * config/riscv/linux.h (LINK_SPEC): Ditto.
7492 2018-04-20 Martin Sebor <msebor@redhat.com>
7495 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7497 (gimple_fold_builtin_stxcpy_chk): Same.
7498 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7500 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7503 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7504 __powikf2 when long double is IEEE 128-bit.
7506 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7508 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7509 step to make sure stack always aligned.
7511 2018-04-20 Carl Love <cel@us.ibm.com>
7514 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7515 size check for arg0.
7517 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7518 Tom de Vries <tom@codesourcery.com>
7521 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7522 Emit insns for calls too.
7523 (nvptx_find_par): Always look for worker-level predecessor insn.
7524 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7526 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7527 (nvptx_process_pars): Propagate frames for calls.
7529 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7532 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7534 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7535 (ix86_handle_option): Don't handle OPT_mibt.
7536 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7538 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7540 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7541 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7542 (ix86_target_macros): Define __CET__ with flag_cf_protection
7543 for -fcf-protection.
7544 * config/i386/i386.c (isa2_opts): Remove -mibt.
7545 * config/i386/i386.h (TARGET_IBT): Removed.
7546 (TARGET_IBT_P): Likewise.
7547 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7548 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7549 * config/i386/i386.opt (mcet): Update help message.
7552 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7553 -mcet as an alias for -mshstk.
7555 2018-04-20 Richard Biener <rguenther@suse.de>
7558 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7559 complexity by forcing a single use of the multiply operand.
7561 2018-04-20 Martin Jambor <mjambor@suse.cz>
7564 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7565 recursion dependency to only apply to non-clones.
7567 2018-04-20 Martin Jambor <mjambor@suse.cz>
7570 * ipa-cp.c (create_specialized_node): Check that clones of
7571 self-recursive edges exist during IPA-CP.
7573 2018-04-19 Toon Moene <toon@moene.org>
7575 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7578 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7580 PR tree-optimization/85467
7581 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7582 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7583 VECTOR_CST element to type.
7585 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7588 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7589 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7590 (builtin_longjmp): Likewise.
7591 (save_stack_nonlocal): New pattern.
7592 (restore_stack_nonlocal): Likewise.
7594 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7597 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7598 Replace ASM_OUTPUT_LABEL with fprintf.
7600 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7603 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7604 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7605 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7606 define __IBT__ and __SHSTK__ for -fcf-protection.
7607 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7609 (ix86_trampoline_init): Likewise.
7610 (x86_output_mi_thunk): Likewise.
7611 (ix86_notrack_prefixed_insn_p): Likewise.
7612 (ix86_option_override_internal): Don't disallow -fcf-protection.
7613 * config/i386/i386.md (rdssp<mode>): Also enable for
7615 (incssp<mode>): Likewise.
7616 (nop_endbr): Likewise.
7617 * config/i386/i386.opt (mcet): Change help message to built-in
7621 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7622 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7623 enable CET built-in functions.
7625 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7627 * common/config/i386/i386-common.c
7628 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7629 OPTION_MASK_ISA_MOVDIRI_UNSET,
7630 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7631 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7632 * config.gcc (movdirintrin.h): New header.
7633 * config/i386/cpuid.h (bit_MOVDIRI,
7634 bit_MOVDIR64B): New bits.
7635 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7637 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7638 (VOID, PVOID, PCVOID)): New function types.
7639 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7640 __builtin_ia32_directstoreu_u64,
7641 __builtin_ia32_movdir64b): New builtins.
7642 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7643 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7645 (ix86_valid_target_attribute_inner_p): Ditto.
7646 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7647 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7648 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7649 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7650 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7651 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7652 (movdiri<mode>, movdir64b_<mode>): New.
7653 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7654 * config/i386/immintrin.h: Include movdirintrin.h.
7655 * config/i386/movdirintrin.h: New file.
7656 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7658 2018-04-19 Richard Biener <rguenther@suse.de>
7661 * cfg.c (clear_bb_flags): When loop state says we have
7662 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7664 2018-04-19 Richard Biener <rguenther@suse.de>
7666 PR tree-optimization/84737
7667 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7668 copying restrict info.
7669 (vect_setup_realignment): Use it.
7670 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7671 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7672 the first DR to all generated stores.
7673 (vectorizable_load): Likewise for loads.
7675 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7677 PR tree-optimization/85446
7678 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7679 the integral and pointer types to have the same precision.
7681 * doc/install.texi: Document --disable-cet being the default and
7684 2018-04-18 Martin Liska <mliska@suse.cz>
7686 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7689 2018-04-18 Martin Liska <mliska@suse.cz>
7692 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7695 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7696 arguments if they are comparable.
7698 2018-04-18 Martin Liska <mliska@suse.cz>
7701 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7704 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7707 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7710 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7711 ENDBR after calling __morestack.
7713 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7716 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7717 by using gcc_base_ver to generate a gcc_driver_version, and use
7718 it when generating GCC_DRIVER_NAME.
7719 * configure: Regenerate.
7721 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7724 * config.gcc: Obsolete powerpc*-*-*spe*.
7726 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7729 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7730 (stabstr_D): Change type of unum from unsigned int to
7731 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7734 2018-04-17 Jim Wilson <jimw@sifive.com>
7737 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7738 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7739 Set arg_pointer_offset after using pretend_args_size.
7741 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7743 PR rtl-optimization/85431
7744 * dse.c (record_store): Ignore zero width stores.
7747 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7748 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7749 __builtin_stack_restore rather than after it.
7750 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7751 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7752 argument instead of virtual_dynamic_stack_rtx.
7754 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7756 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7758 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7759 Add note to error message to explain internal mapping of overloaded
7760 built-in function name to non-overloaded built-in function name.
7761 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7764 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7767 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7768 where the inputs overlap with the output.
7770 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7772 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7773 (=v, v) alternative and explicit "memory" attribute.
7774 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7775 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7777 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7778 "sselog1" type instead of "sselog".
7779 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7780 "sselog". Remove explicit "memory" attribute.
7781 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7782 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7784 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7785 "sselog1" type instead of "sselog". Remove explicit "memory"
7787 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7788 use "sselog1" type instead of "sselog". Remove explicit "memory"
7790 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7791 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7793 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7794 "sselog1" type instead of "sselog". Remove explicit "memory"
7796 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7797 use "sselog1" type instead of "sselog". Remove explicit "memory"
7801 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7804 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7807 2018-04-17 Martin Jambor <mjambor@suse.cz>
7810 * ipa-cp.c (create_specialized_node): Call
7811 expand_all_artificial_thunks if necessary.
7813 2018-04-17 Martin Liska <mliska@suse.cz>
7816 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7817 in message, remote space in between '_G' and '('.
7819 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7822 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7823 avx512f_vmcmp<mode>3<round_saeonly_name>,
7824 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7825 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7826 avx512f_rndscale<mode><round_saeonly_name>,
7827 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7828 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7829 Use %<iptr>2 instead of %2 for -masm=intel.
7830 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7831 avx512f_vcvttss2usi<round_saeonly_name>,
7832 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7834 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7835 avx512f_vcvttsd2usi<round_saeonly_name>,
7836 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7837 Use %q1 instead of %1 for -masm=intel.
7838 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7839 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7840 of %3 for -masm=intel.
7841 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7843 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7845 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7847 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7848 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7850 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7852 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7853 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7854 %g1 and one with %0 and %1.
7855 (avx512er_vmrcp28<mode><round_saeonly_name>,
7856 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7858 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7859 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7860 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7861 of %0 and %{%4%} for -masm=intel.
7862 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7863 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7864 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7865 order of %0 and %{%5%}%{z%} for -masm=intel.
7867 2018-04-17 Jan Hubicka <jh@suse.cz>
7870 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7872 2018-04-17 Martin Liska <mliska@suse.cz>
7875 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7876 for target_clone error message. Make default implementation
7877 clone to be a local declaration.
7878 (separate_attrs): Add new argument and check for an empty
7880 (expand_target_clones): Handle it.
7881 (ipa_target_clone): Make redirection just for target_clones
7884 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7885 Tom de Vries <tom@codesourcery.com>
7888 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7889 tiled basic blocks without omp continue statements.
7891 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7894 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7895 vec_extract expression as having side effects to make sure it gets
7898 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7901 * config/i386/i386.c (get_builtin_code_for_version): Check
7904 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7907 * gcc/config.gcc: Support "skylake".
7908 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7910 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7911 (processor_target_table): Add "skylake".
7912 (ix86_option_override_internal): Add "skylake".
7913 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7914 PROCESSOR_CANNONLAKE.
7915 (get_builtin_code_for_version): Fix priority for
7916 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7917 PROCESSOR_SKYLAKE-AVX512.
7918 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7919 (processor_type): Add PROCESSOR_SKYLAKE.
7921 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7922 Jason Merrill <jason@redhat.com>
7925 * convert.c (convert_to_integer_1): Use direct recursion for
7926 enumeral types and types with a precision less than the number
7927 of bits in their mode.
7929 2018-04-16 Julia Koval <julia.koval@intel.com>
7932 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7933 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7935 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7938 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7939 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7940 and -mno-direct-move.
7942 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7945 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7946 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7947 Ensure negative shifts result in {0}.
7949 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7951 PR rtl-optimization/79916
7952 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7953 regs (if any) to define how to gnerate SD moves when LRA is in
7956 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7958 PR rtl-optimization/85393
7959 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7960 * except.c (expand_dw2_landing_pad_for_region): Make static.
7961 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7962 a label and unconditional jump to old_bb, rather than
7963 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7966 PR rtl-optimization/85376
7967 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7968 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7969 instead of a specific value.
7971 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7972 Bin Cheng <bin.cheng@arm.com>
7974 PR tree-optimization/82965
7975 PR tree-optimization/83991
7976 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7977 by_profile_only parameter.
7978 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7979 information if the loop was predicted to iterate too many times.
7980 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7982 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7985 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7988 2018-04-13 Martin Liska <mliska@suse.cz>
7989 Jakub Jelinek <jakub@redhat.com>
7992 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7993 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7994 * builtins.c (expand_builtin_memory_copy_args): Use
7995 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7996 handle dest_addr == pc_rtx.
7998 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8001 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8002 asked to not generate direct moves.
8003 (fix_trunc<mode>si2_stfiwx): Similar.
8004 (fix_trunc<mode>si2_internal): Similar.
8006 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8009 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8010 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8011 lookup if dest in some wider mode is known to be const0_rtx and
8012 if so, record permanent equivalence for it to be ZERO_EXTEND of
8013 the narrower mode destination.
8015 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8017 * lto-streamer-out.c (output_function): Revert 259346.
8018 * omp-expand.c (expand_oacc_for): Likewise.
8020 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8022 PR rtl-optimization/85354
8023 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8024 * sel-sched.c (sel_global_init): ... here.
8026 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8029 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8030 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8031 mode for PE-COFF targets.
8032 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8033 (i386_pe_asm_lto_end): Likewise.
8034 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8035 (TARGET_ASM_LTO_END): Likewise.
8036 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8037 (i386_pe_asm_lto_start): New function.
8038 (i386_pe_asm_lto_end): Likewise.
8040 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8041 Richard Biener <rguenther@suse.de>
8044 * lto-streamer-out.c (output_function): Fix CFG loop state before
8046 * omp-expand.c (expand_oacc_for): Handle calls to internal
8047 functions like regular functions.
8049 2018-04-12 Richard Biener <rguenther@suse.de>
8052 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8053 for the early LTO debug to properly generate references to it
8054 during DIE emission. Do not re-use that for the skeleton for
8056 (dwarf2out_early_finish): Likewise.
8058 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8061 * config/i386/sse.md
8062 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8063 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8064 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8065 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8066 and output is a reg, avoid creating invalid lowpart subreg, but
8067 instead split into a 512-bit move. Don't split if not AVX512VL,
8068 input is xmm16+ reg and output is a mem.
8069 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8070 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8071 xmm16+ reg and output is a mem.
8073 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8075 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8076 also for flag_dwarf2_cfi_asm.
8078 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8080 PR rtl-optimization/85342
8081 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8082 a bool scalar var inside of the loop instead. Don't try to update
8083 recog_data.operand after failed apply_change_group.
8085 2018-04-12 Tom de Vries <tom@codesourcery.com>
8088 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8089 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8090 array with flexible array member as array without given dimension.
8091 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8092 argument for undefined param to true.
8094 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8097 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8098 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8099 from PowerPC section.
8100 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8101 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8102 help text that is too long.
8103 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8104 help text that is too long.
8105 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8106 help text that is too long.
8108 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8110 * config/alpha/alpha.md (stack_probe_internal): Rename
8111 from "probe_stack". Update all callers.
8113 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8115 PR rtl-optimization/84566
8116 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8117 sched_macro_fuse_insns.
8119 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8122 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8123 (compute_block_dependences): ... from here.
8125 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8127 PR tree-optimization/85331
8128 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8129 from int to HOST_WIDE_INT.
8131 2018-04-11 Martin Jambor <mjambor@suse.cz>
8134 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8135 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8136 not the same as the source val.
8137 (cgraph_edge_brings_value_p): New parameter.
8138 (gather_edges_for_value): Pass destination value to
8139 cgraph_edge_brings_value_p.
8140 (perhaps_add_new_callers): Likewise.
8141 (get_info_about_necessary_edges): Likewise and exclude values brought
8142 only by self-recursive edges.
8143 (create_specialized_node): Redirect only clones of self-calling edges.
8144 (+self_recursive_pass_through_p): New function.
8145 (find_more_scalar_values_for_callers_subset): Use it.
8146 (find_aggregate_values_for_callers_subset): Likewise.
8147 (known_aggs_to_agg_replacement_list): Removed.
8148 (decide_whether_version_node): Re-calculate known constants for all
8149 remaining context clones.
8151 2018-04-11 Richard Biener <rguenther@suse.de>
8154 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8155 from early DWARF output.
8156 (dwarf2out_early_finish): Output line info unconditionally into
8157 early DWARF and add reference to it.
8159 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8162 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8163 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8164 other than V2DFmode using iptr mode attribute.
8165 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8167 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8169 PR rtl-optimization/84659
8170 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8172 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8175 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8177 (output_loc_list): Pass address of a dummy size variable even in the
8178 locview handling loop.
8179 (index_location_lists): Add comment on why skip_loc_list_entry can't
8182 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8185 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8188 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8191 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8192 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8193 and -mstring-compare-inline-limit.
8195 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8198 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8199 for stack clash protection in a register whenever we need it to be in
8202 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8204 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8205 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8207 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8210 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8212 (mlong-double-): Ditto.
8213 * config/rs6000/sysv4.opt (msdata=): Ditto.
8214 (mtls-size=): Ditto.
8216 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8218 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8219 erroneous entries for
8220 "vector int vec_ldl (int, long int *)", and
8221 "vector unsigned int vec_ldl (int, unsigned long int *)".
8222 Add comments and entries for
8223 "vector bool char vec_ldl (int, bool char *)",
8224 "vector bool short vec_ldl (int, bool short *)",
8225 "vector bool int vec_ldl (int, bool int *)",
8226 "vector bool long long vec_ldl (int, bool long long *)",
8227 "vector pixel vec_ldl (int, pixel *)",
8228 "vector long long vec_ldl (int, long long *)",
8229 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8230 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8231 type tree bool_long_long_type_node and correct definition of
8232 bool_V2DI_type_node to make reference to this new type tree.
8233 (rs6000_mangle_type): Replace erroneous reference to
8234 bool_long_type_node with bool_long_long_type_node.
8235 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8236 comments to emphasize sign distinctions for char and int types and
8237 replace RS6000_BTI_bool_long constant with
8238 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8239 use of RS6000_BTI_pixel.
8240 (bool_long_type_node): Remove this macro definition.
8241 (bool_long_long_type_node): New macro definition
8243 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8245 PR rtl-optimization/85300
8246 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8247 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8248 simplify_unary_operation fails.
8250 2018-04-10 Martin Liska <mliska@suse.cz>
8252 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8253 cgraph_edge and ipa_ref.
8255 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8259 * config/i386/sse.md
8260 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8261 computation of the VEC_MERGE selector from mask.
8262 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8263 Fix decoding of the VEC_MERGE selector into mask.
8265 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8267 PR tree-optimization/85286
8268 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8270 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8272 * final.c (final_1): Set insn_last_address as well as
8273 insn_current_address.
8275 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8278 * explow.c (emit_stack_probe): Call validize_mem on memory location
8279 before passing it to gen_probe_stack. Create address operand and
8280 legitimize it for the probe_stack_address case.
8282 2018-04-09 Jan Hubicka <jh@suse.cz>
8285 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8286 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8287 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8289 (free_lang_data): Rebuild type inheritance graph.
8291 2018-04-09 Martin Sebor <msebor@redhat.com>
8293 * invoke.texi (-finline-small-functions): Mention other optimization
8295 (-findirect-inlining, -fpartial-inlining): Same.
8296 (-finline-functions-called-once): Same.
8297 (-freorder-blocks-and-partition): Same.
8299 2018-04-09 Jan Hubicka <jh@suse.cz>
8302 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8303 jumps; choose last target that matches the criteria (i.e.
8304 no partition changes for non-crossing jumps).
8305 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8306 support for redirecting crossing jumps to non-crossing.
8308 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8310 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8311 also for naked functions.
8313 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8315 * config/arc/arc.md (add_shift): New pattern.
8316 (add_shift2): Likewise.
8317 (sub_shift): Likewise.
8318 (sub_shift_cmp0_noout): Likewise.
8319 (compare_si_ashiftsi): Likewise.
8320 (xbfu_cmp0_noout): New combine pattern.
8321 (xbfu_cmp0"): Likewise.
8322 (movsi_set_cc_insn): Place the predicable variant first.
8323 (commutative_binary_cmp0_noout): Remove clobber.
8324 (commutative_binary_cmp0): New pattern.
8325 (noncommutative_binary_cmp0): Likewise.
8326 (noncommutative_binary_cmp0_noout): Likewise.
8327 (noncommutative_binary_comparison_result_used): Removed.
8328 (rsub_cmp0): New pattern.
8329 (rsub_cmp0_noout): Likewise.
8330 (extzvsi): Changed, keep only meaningful variants.
8331 (SQH, SEZ): New iterators.
8332 (SQH_postfix): New mode attribute.
8333 (SEZ_prefix): New code attribute.
8334 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8335 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8336 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8338 (noncommutative_operator): Check the availability of barrel
8341 2018-04-09 Richard Biener <rguenther@suse.de>
8343 PR tree-optimization/85284
8344 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8345 Only use the niter constraining form of simple_iv when the exit
8348 2018-04-09 Tom de Vries <tom@codesourcery.com>
8351 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8352 (define_expand "*memory_barrier"): New define_expand.
8353 (define_insn "memory_barrier"): New insn.
8355 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8357 PR rtl-optimization/80463
8358 PR rtl-optimization/83972
8359 PR rtl-optimization/83480
8361 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8362 correct producer for the insn.
8363 (tidy_control_flow): Fixup seqnos in case of debug insns.
8365 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8367 PR rtl-optimization/83913
8369 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8370 different sched-times when merging exprs.
8372 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8374 PR rtl-optimization/83962
8376 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8377 tidy_fallthru_edge and tidy_control_flow.
8379 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8381 PR rtl-optimization/83530
8383 * sel-sched.c (force_next_insn): New global variable.
8384 (remove_insn_for_debug): When force_next_insn is true, also leave only
8385 next insn in the ready list.
8386 (sel_sched_region): When the region wasn't scheduled, make another pass
8387 over it with force_next_insn set to 1.
8389 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8391 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8393 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8394 for interrupt control.
8395 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8396 functions for interrupt control.
8397 * config/nds32/nds32-intrinsic.md: Likewise.
8398 * config/nds32/nds32_intrinsic.h: Likewise.
8399 * config/nds32/nds32.h (nds32_builtins): Likewise.
8401 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8403 * config/nds32/nds32.c (nds32_init_machine_status,
8404 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8405 strict_aligned_p field.
8406 (nds32_expand_to_rtl_hook): New function.
8407 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8408 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8410 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8411 Chung-Ju Wu <jasonwucj@gmail.com>
8413 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8414 * config/nds32/nds32-n7.md: New file.
8415 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8416 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8418 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8419 * config/nds32/nds32.md (pipeline_model): Add n7.
8420 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8421 * config/nds32/pipelines.md: Include n7 settings.
8423 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8424 Chung-Ju Wu <jasonwucj@gmail.com>
8426 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8427 * config/nds32/nds32-e8.md: New file.
8428 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8429 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8431 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8432 * config/nds32/nds32.md (pipeline_model): Add e8.
8433 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8434 * config/nds32/pipelines.md: Include e8 settings.
8436 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8437 Chung-Ju Wu <jasonwucj@gmail.com>
8439 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8440 * config/nds32/nds32-n8.md: New file.
8441 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8442 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8444 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8445 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8446 * config/nds32/nds32.md (pipeline_model): Add n8.
8447 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8448 * config/nds32/pipelines.md: Include n8 settings.
8450 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8451 Chung-Ju Wu <jasonwucj@gmail.com>
8453 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8454 * config/nds32/nds32-n9-2r1w.md: New file.
8455 * config/nds32/nds32-n9-3r2w.md: New file.
8456 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8457 nds32_register_ports): New or modify for cpu n9.
8458 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8460 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8461 * config/nds32/nds32-utils.c: New file.
8462 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8463 TARGET_MUL_SLOW): Define.
8464 * config/nds32/nds32.md (pipeline_model): New attribute.
8465 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8466 New options that support cpu n9.
8467 * config/nds32/pipelines.md: Include n9 settings.
8468 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8470 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8472 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8473 information if necessary.
8474 (output_cond_branch_compare_zero): Likewise.
8475 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8476 (nds32_target_alignment): Refine for alignment.
8477 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8478 (FUNCTION_BOUNDARY): Modify.
8479 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8481 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8483 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8485 * config/nds32/constants.md (unspec_volatile_element): Add values for
8486 TLB operation and data prefetch.
8487 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8488 functions for TLB operation and data prefetch.
8489 * config/nds32/nds32-intrinsic.md: Likewise.
8490 * config/nds32/nds32_intrinsic.h: Likewise.
8491 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8492 (nds32_print_operand): Likewise.
8493 * config/nds32/nds32.h (nds32_builtins): Likewise.
8495 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8496 Andrew Pinski <pinsika@gcc.gnu.org>
8499 * match.pd: Use constant_boolean_node of correct type instead of
8500 boolean_true_node or boolean_false_node for simplifying
8501 pointer comparisons to zero.
8503 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8505 PR tree-optimization/80021
8506 * tree.c (verify_type_variant): Make error call in verify_variant_match
8507 translatable and remove final full stop.
8509 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8511 * config/nds32/constants.md (unspec_volatile_element): Add
8512 UNSPEC_VOLATILE_EH_RETURN.
8513 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8514 nds32_output_stack_pop): Support dwarf exception handling process.
8515 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8516 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8517 exception handling process.
8518 (nds32_compute_stack_frame): Likewise.
8519 (nds32_return_addr_rtx): Likewise.
8520 (nds32_initial_elimination_offset): Likewise.
8521 (nds32_expand_prologue): Likewise.
8522 (nds32_expand_epilogue): Likewise.
8523 (nds32_dynamic_chain_address): New function.
8524 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8526 (DYNAMIC_CHAIN_ADDRESS): Define.
8527 (EH_RETURN_DATA_REGNO): Define.
8528 (EH_RETURN_STACKADJ_RTX): Define.
8529 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8530 patterns for dwarf exception handling.
8532 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8534 * config/nds32/nds32.h: Clean up obsolete macros.
8536 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8538 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8539 Add enum values for particular instructions.
8540 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8541 particular intrinsic functions.
8542 * config/nds32/nds32-intrinsic.md: Likewise.
8543 * config/nds32/nds32_intrinsic.h: Likewise.
8544 * config/nds32/nds32.h (nds32_builtins): Likewise.
8545 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8546 (btst, ave): New patterns for particular instructions.
8548 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8550 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8551 Add enum values for atomic load/store and memory sync.
8552 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8554 * config/nds32/nds32-intrinsic.md: Likewise.
8555 * config/nds32/nds32_intrinsic.h: Likewise.
8556 * config/nds32/nds32.h (nds32_builtins): Likewise.
8558 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8560 PR tree-optimization/85257
8561 * fold-const.c (native_encode_vector): If not all elts could fit
8562 and off is -1, return 0 rather than offset.
8563 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8564 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8565 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8566 adjust buffer in native_interpret_expr call.
8568 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8570 * config/nds32/constants.md (unspec_volatile_element): Add cache
8571 control enum values.
8572 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8573 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8574 * config/nds32/nds32.c (nds32_cctl_names): New.
8575 (nds32_print_operand): Handle cache control register names.
8576 * config/nds32/nds32.h (nds32_builtins): New enum values.
8577 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8579 * config/nds32/nds32.md (type): Add mmu.
8580 * config/nds32/pipelines.md (simple_insn): Add mmu.
8582 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8584 * config/nds32/nds32.md (type): Remove call.
8585 * config/nds32/pipelines.md (simple_insn): Likewise.
8587 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8589 * config/nds32/constants.md (unspec_volatile_element): Add
8590 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8591 UNSPEC_VOLATILE_FMFCFG.
8592 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8593 description for fmfcfg and fmfcsr.
8594 (bdesc_1arg): Add fmtcsr.
8595 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8596 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8597 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8598 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8599 unspec_fmfcfg): New patterns.
8600 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8601 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8602 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8603 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8604 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8605 __nds32__fmfcfg): Define.
8607 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8609 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8610 intrinsic register names.
8611 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8612 intrinsic register enum values and macros.
8614 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8616 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8617 for load/store addressing form.
8618 (nds32_print_operand_address): Likewise.
8620 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8623 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8624 based on LABEL_REF. Remove useless assertion.
8625 (pic_address_needs_scratch): Fix formatting.
8626 (sparc_legitimize_pic_address): Minor tweaks.
8627 (sparc_delegitimize_address): Adjust assertion accordingly.
8628 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8629 into symbolic_operand.
8630 (movsi_high_pic_label_ref): Likewise.
8631 (movsi_lo_sum_pic_label_ref): Likewise.
8632 (movdi_pic_label_ref): Likewise.
8633 (movdi_high_pic_label_ref): Likewise.
8634 (movdi_lo_sum_pic_label_ref): Likewise.
8636 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8638 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8639 custom LIB_SPEC setup.
8641 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8642 Kito Cheng <kito.cheng@gmail.com>
8644 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8645 * config/riscv/freebsd.h: New.
8647 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8649 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8650 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8653 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8654 Kito Cheng <kito.cheng@gmail.com>
8656 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8657 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8658 * config/nds32/nds32-protos.h (nds32_output_call,
8659 nds32_output_return): Declare.
8660 * config/nds32/nds32.md: Refine all the call and return patterns.
8662 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8665 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8666 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8668 PR rtl-optimization/84872
8669 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8670 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8673 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8675 * expr.c (copy_blkmode_to_reg): Revert 254862.
8676 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8678 2018-04-06 Richard Biener <rguenther@suse.de>
8681 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8682 after seeing a component reference with an adjacent field. Treat
8683 refs to arrays at struct end of external decls similar to
8684 refs to unconstrained commons.
8686 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8689 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8690 look through SAVE_EXPRs with non-side-effects argument. Adjust
8692 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8695 2018-04-06 Richard Biener <rguenther@suse.de>
8698 * alias.c (find_base_term): New wrapper around find_base_term
8699 unwinding CSELIB_VAL_PTR changes.
8700 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8703 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8705 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8707 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8708 constant definitions.
8709 ("nop"): lr 0,0 -> nopr r0
8710 ("nop_lr0", "nop_lr1"): New insn definitions.
8712 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8714 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8715 NDS32_V3PUSH_AVAILABLE_P macro.
8717 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8718 Chung-Ju Wu <jasonwucj@gmail.com>
8720 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8721 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8722 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8723 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8724 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8725 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8726 * config/nds32/constraints.md: New constraints and checking for hard
8727 float configuration.
8728 * config/nds32/iterators.md: New mode iterator and attribute for hard
8729 float configuration.
8730 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8732 * config/nds32/nds32-fpu.md: New file.
8733 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8734 deal with hard float code generation.
8735 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8737 (abi_type, float_reg_number): New enum type.
8738 * config/nds32/nds32-predicates.c: New predicates for hard float.
8739 * config/nds32/nds32-protos.h: Declare functions for hard float.
8740 * config/nds32/nds32.c: Implementation for hard float configuration.
8741 * config/nds32/nds32.h: Definitions for hard float configuration.
8742 * config/nds32/nds32.md: Include hard float machine description and
8743 modify patterns for hard float configuration.
8744 * config/nds32/nds32.opt: New options for hard float configuration.
8745 * config/nds32/predicates.md: New predicates for hard float
8748 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8750 * common/config/nds32/nds32-common.c
8751 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8753 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8756 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8757 CONSTRUCTOR_ELT (ctor, ...)->value.
8759 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8762 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8764 2018-04-05 Tom de Vries <tom@codesourcery.com>
8767 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8770 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8771 Kito Cheng <kito.cheng@gmail.com>
8773 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8774 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8775 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8776 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8778 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8779 Kito Cheng <kito.cheng@gmail.com>
8781 * config/nds32/constraints.md (Ufe): New memory constraint.
8782 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8783 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8784 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8786 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8787 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8789 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8791 * config/nds32/nds32.md: Use optimize_size in the condition for
8792 alu-shift instructions.
8794 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8796 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8798 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8800 * config/nds32/nds32.md (negsi2): Refine pattern.
8802 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8803 Chung-Ju Wu <jasonwucj@gmail.com>
8805 * config/nds32/iterators.md (shift_rotate): New code iterator.
8806 (shift): New code attribute.
8807 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8808 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8809 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8810 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8811 bit-wise operations.
8812 (andsi3, *andsi3): Ditto.
8813 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8814 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8815 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8816 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8817 nds32_ior_operand, nds32_xor_operand): New predicates.
8819 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8821 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8822 (addsi3, subsi3): ... this.
8824 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8826 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8828 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8830 * config/nds32/nds32.md: Adjust indention.
8832 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8834 * config/nds32/nds32.md (feature): New attribute.
8836 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8838 * config/nds32/nds32.md (subtype): New attribute.
8840 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8843 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8844 expansion to perform a bitwise AND of the argument followed by a
8845 boolean negation of the result.
8847 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8849 PR rtl-optimization/84878
8850 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8851 the basic block. Assert the use reference is not artificial and that
8852 it has an associated insn.
8854 2018-04-04 Michael Matz <matz@suse.de>
8856 * builtins.c (compute_objsize): Pass correct operand
8857 to array_at_struct_end_p.
8859 2018-04-04 Richard Biener <rguenther@suse.de>
8862 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8863 from contexts for DINFO_LEVEL_TERSE and below.
8865 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8867 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8868 resiter_operand condition.
8869 * config/nds32/nds32.md (*move<mode>): Ditto.
8871 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8872 Monk Chiang <sh.chiang04@gmail.com>
8874 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8876 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8878 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8880 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8881 Kito Cheng <kito.cheng@gmail.com>
8883 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8884 nds32_cond_code_str, output_cond_branch,
8885 output_cond_branch_compare_zero, nds32_expand_cbranch,
8886 nds32_expand_cstore, nds32_expand_movcc,
8887 nds32_output_cbranchsi4_equality_zero,
8888 nds32_output_cbranchsi4_equality_reg,
8889 nds32_output_cbranchsi4_equality_reg_or_const_int,
8890 nds32_output_cbranchsi4_greater_less_zero: New functions.
8891 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8892 nds32_expand_cstore, nds32_expand_movcc,
8893 nds32_output_cbranchsi4_equality_zero,
8894 nds32_output_cbranchsi4_equality_reg,
8895 nds32_output_cbranchsi4_equality_reg_or_const_int,
8896 nds32_output_cbranchsi4_greater_less_zero): Declare.
8897 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8898 nds32_rimm11s_operand): New predicates.
8899 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8900 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8903 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8905 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8906 * config/nds32/nds32.md: Ditto.
8907 * config/nds32/pipelines.md: Ditto.
8909 2018-04-04 Richard Biener <rguenther@suse.de>
8911 PR tree-optimization/85168
8912 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8913 propagating abnormals.
8915 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8917 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8919 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8920 Kito Cheng <kito.cheng@gmail.com>
8922 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8923 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8924 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8925 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8926 * config/nds32/nds32.md (sibcall_internal): New.
8927 (sibcall_register): Remove.
8928 (sibcall_immediate): Remove.
8929 (sibcall_value_internal): New.
8930 (sibcall_value_register): Remove.
8931 (sibcall_value_immediate): Remove.
8932 * config/nds32/predicates.md (nds32_general_register_operand): New.
8933 (nds32_call_address_operand): New.
8935 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8937 PR rtl-optimization/85167
8938 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8939 bb_defs if *split_p, instead preinitialize it to NULL.
8941 PR tree-optimization/85156
8942 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8943 evaluating the argument multiple times.
8945 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8947 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8949 (_mm_cvtpd_ps): Likewise.
8950 (_mm_cvttpd_epi32): Likewise.
8951 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8952 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8953 vector, pixel, and bool following altivec.h include.
8955 2018-04-03 Martin Sebor <msebor@redhat.com>
8957 * doc/extend.texi (Common Function Attributes): Clarify.
8958 (const attribute): Likewise.
8959 (pure attribute): Likewise.
8961 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8964 * config/i386/i386.c (ix86_expand_vector_set): Use
8965 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8967 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8969 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8970 instructions when changing rounding bits to preserve precision bits
8971 in the x87 control word.
8973 2018-04-03 Martin Liska <mliska@suse.cz>
8975 PR tree-optimization/82491
8976 * rtl.h (strip_offset_and_add): Replace += suboffset with
8977 poly_uint64 () + suboffset.
8979 2018-03-29 Martin Liska <mliska@suse.cz>
8980 Martin Jambor <mjambor@suse.cz>
8983 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8984 param_type is not an integral or pointer type.
8986 2018-04-03 Richard Biener <rguenther@suse.de>
8988 * sese.h (recompute_all_dominators): Remove.
8990 2018-04-02 Martin Sebor <msebor@redhat.com>
8992 * doc/invoke.texi (-Wrestrict): Fix typos.
8994 2018-04-02 Jim Wilson <jimw@sifive.com>
8996 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8997 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8998 (<optab>di3, <optab>si3_extend): Likewise.
8999 (<optab>si3_mask, <optab>si3_mask_1): New.
9000 (<optab>di3_mask, <optab>di3_mask_1): New.
9001 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9002 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9003 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9005 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9007 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9010 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9012 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9013 (nds32_canonicalize_comparison): New function.
9015 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9016 Kito Cheng <kito.cheng@gmail.com>
9017 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9019 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9020 * config/nds32/constants.md (unspec_volatile_element): Add
9021 UNSPEC_VOLATILE_RELAX_GROUP.
9022 * config/nds32/nds32-relax-opt.c: New file.
9023 * config/nds32/nds32-predicates.c
9024 (nds32_symbol_load_store_p): New function.
9025 * config/nds32/nds32-protos.h
9026 (nds32_symbol_load_store_p): Declare function.
9027 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9028 * config/nds32/nds32.c
9029 (nds32_register_pass): New function to register pass.
9030 (nds32_register_passes): New function to register passes.
9031 * config/nds32/nds32.md (relax_group): New pattern.
9032 * config/nds32/nds32.opt (mrelax-hint): New option.
9033 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9035 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9037 * config/nds32/t-nds32: Modify files dependency.
9039 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9041 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9042 (PROFILE_HOOK): Define its implementation.
9044 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9046 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9047 type and 32-bit size.
9049 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9052 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9053 (V_128_256): New mode iterator.
9054 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9055 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9056 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9058 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9059 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9061 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9064 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9065 NaN inputs correctly.
9067 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9070 * config/rs6000/vsx.md (??r): New mode attribute.
9071 (*vsx_mov<mode>_64bit): Use it.
9072 (*vsx_mov<mode>_32bit): Likewise.
9074 2018-03-30 Martin Sebor <msebor@redhat.com>
9076 PR tree-optimization/84818
9077 * builtins.c (check_access): Use warning_n.
9079 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9082 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9084 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9087 2018-03-30 Julia Koval <julia.koval@intel.com>
9090 * x86-tune.def (movx, partial_reg_dependency): Enable for
9093 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9096 * lra-constraints.c (process_alt_operands): Move setting
9097 this_alternative_matches below.
9099 2018-03-29 Martin Liska <mliska@suse.cz>
9102 * doc/invoke.texi: Document how LTO works with debug info.
9103 Describe auto-load support of binutils. Mention 'x86-64'
9104 as valid option value of -march option.
9106 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9108 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9111 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9112 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9113 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9116 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9119 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9120 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9121 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9122 for RS6000_BTM_POWERPC64.
9123 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9124 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9125 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9130 2018-03-28 Carl Love <cel@us.ibm.com>
9133 2017-09-27 Carl Love <cel@us.ibm.com>
9135 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9136 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9137 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9140 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9142 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9143 instead of __vector bool.
9144 (_mm_max_pu8): Likewise.
9145 (_mm_min_pi16): Likewise.
9147 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9150 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9151 (DIVWEUO): Likewise.
9153 (DIVDEUO): Likewise.
9154 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9155 DIVWEUO and DIVDEUO.
9156 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9157 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9158 (div_extend): Likewise.
9159 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9161 (__builtin_divweuo): Likewise.
9162 (__builtin_divdeo): Likewise.
9163 (__builtin_divdeuo): Likewise.
9165 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9168 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9169 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9171 PR tree-optimization/82004
9172 * gimple-match-head.c (optimize_pow_to_exp): New function.
9173 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9174 Don't fold to exp if optimize_pow_to_exp is false.
9176 2018-03-28 Martin Liska <mliska@suse.cz>
9179 * calls.c (initialize_argument_information): Fix trailing space.
9180 * common.opt: Fix typo and provide better explanation for
9181 -fsanitize-coverage option.
9182 * config/i386/i386.opt: Fix typo.
9184 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9185 Martin Liska <mliska@suse.cz>
9188 * gimplify.c (asan_poison_variable): Don't do the check for
9189 gimplify_omp_ctxp here.
9190 (gimplify_decl_expr): Do it here.
9191 (gimplify_target_expr): Likewise.
9193 2018-03-28 Martin Liska <mliska@suse.cz>
9196 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9197 chkp_type_bounds_count if MPX is not enabled.
9199 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9201 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9203 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9206 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9207 function to create the function decl for complex long double
9208 multiply and divide for -mabi=ieeelongdouble.
9209 (init_float128_ieee): Call it.
9211 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9214 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9215 -fcf-protection=branch -mibt.
9216 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9218 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9221 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9222 arm_word_relocations.
9224 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9227 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9228 extern array declarations.
9230 2018-03-27 Richard Biener <rguenther@suse.de>
9233 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9234 explicit single_use checks.
9236 2018-03-27 Richard Biener <rguenther@suse.de>
9238 PR tree-optimization/85082
9239 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9242 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9244 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9245 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9246 Turn on fasynchronous-unwind-tables and funwind-tables.
9248 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9251 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9252 (*bmi_blsr_<mode>_ccz): Ditto.
9254 2018-03-26 Tom de Vries <tom@codesourcery.com>
9256 PR tree-optimization/85063
9257 * omp-general.c (offloading_function_p): New function. Factor out
9259 * omp-offload.c (pass_omp_target_link::gate): ... here.
9260 * omp-general.h (offloading_function_p): Declare.
9261 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9262 with attribute omp declare target for offloading functions.
9264 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9266 PR tree-optimization/84005
9267 * tree-data-ref.h (get_base_for_alignment): Declare.
9268 * tree-data-ref.c (get_base_for_alignment_1): New function.
9269 (get_base_for_alignment): Likewise.
9270 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9271 get_base_for_alignment to find a suitable base object, instead
9272 of always using drb->base_address.
9274 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9277 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9278 known size by default.
9280 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9283 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9284 and non BLKmode operands.
9286 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9289 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9290 Clean up attributes.
9292 2018-03-23 Richard Biener <rguenther@suse.de>
9295 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9296 we are going to emit early debug for LTO.
9298 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9301 * function.c (match_asm_constraints_1): Don't optimize if input
9302 doesn't satisfy general_operand predicate for output's mode.
9305 * alias.c (write_dependence_p): Don't require for x_canonicalized
9306 non-VOIDmode if x has VOIDmode.
9309 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9310 just don't try to optimize it rather than assert it never happens.
9312 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9314 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9315 macro expansions for definition of ST_INTERNAL_<mode> and
9316 LD_INTERNAL_<mode> builtins.
9317 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9319 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9321 (altivec_expand_st_builtin): Likewise.
9322 (altivec_expand_builtin): Remove calls to deleted functions.
9323 (rs6000_address_for_altivec): Delete this function.
9324 * config/rs6000/vector.md: Remove expands for
9325 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9327 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9330 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9331 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9332 re-computing once computed.
9333 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9334 (arm_init_machine_status): Initialize
9335 machine->static_chain_stack_bytes.
9337 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9340 * doc/extend.texi: Add four new prototypes for vec_ld.
9341 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9342 definitions for more logical presentation.
9343 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9344 entries for V1TI variants of __builtin_altivec_ld builtin.
9345 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9346 handling of V1TI variant of LVX icode pattern.
9347 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9348 (rs6000_gimple_fold_builtin): Likewise.
9349 (altivec_init_builtins): Add code to define
9350 __builtin_altivec_lvx_v1ti function.
9352 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9355 * function.c (match_asm_constraints_1): Don't do the optimization
9356 if input isn't a REG, SUBREG, MEM or constant.
9358 2018-03-22 Tom de Vries <tom@codesourcery.com>
9360 PR tree-optimization/84956
9361 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9362 bb_has_abnormal_pred.
9364 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9367 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9368 DECL_INITIAL (decl) to decl at the end.
9369 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9372 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9374 * doc/extend.texi (__builtin_tgmath): Document when complex
9375 integer types are treated as _Complex _Float64.
9377 2018-03-21 Tom de Vries <tom@codesourcery.com>
9379 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9381 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9383 PR tree-optimization/84960
9384 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9385 if it is ENTRY block, move them into single succ of ENTRY in that case.
9387 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9389 PR tree-optimization/84811
9390 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9391 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9392 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9393 (known_subrange_p): Update accordingly. Cast each value involved
9394 in the size comparison, rather than casting the result of the
9397 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9399 PR tree-optimization/84982
9400 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9401 by flipping the least significant bit rather than all bits from
9402 bitpos to bitpos + bitsize - 1.
9404 2018-03-21 Nathan Sidwell <nathan@acm.org>
9406 * doc/extend.texi (Deprecated Features): Remove mention of
9407 long-deleted deprecations.
9409 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9412 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9413 * configure: Regenerate.
9415 2018-03-21 Tom de Vries <tom@codesourcery.com>
9417 PR tree-optimization/83126
9418 * tree-parloops.c (num_phis): New function.
9419 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9421 2018-03-21 Nathan Sidwell <nathan@acm.org>
9423 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9424 mention anon-struct/union members and trailing attributes.
9426 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9428 PR tree-optimization/84969
9429 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9430 builtin memset partitions if they set different rhs values.
9432 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9434 PR rtl-optimization/84989
9435 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9436 VEC_DUPLICATE with scalar result mode.
9438 2018-03-21 Martin Liska <mliska@suse.cz>
9441 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9442 not intended return statement.
9444 2018-03-21 Martin Liska <mliska@suse.cz>
9447 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9448 (chkp_find_bound_slots_1): Limit number of iterations.
9450 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9453 * Minor grammar fixes for x86 options.
9455 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9458 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9459 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9461 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9464 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9465 (altivec_lvx_<mode>_1op): Likewise.
9466 (altivec_stvx_<mode>_2op): Likewise.
9467 (altivec_stvx_<mode>_1op): Likewise.
9468 (altivec_lvx_<VM2:mode>): New define_expand.
9469 (altivec_stvx_<VM2:mode>): Likewise.
9470 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9471 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9472 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9473 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9474 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9475 (rs6000_gen_lvx): Likewise.
9476 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9477 (altivec_expand_stv_builtin): Likewise.
9478 (altivec_expand_builtin): Likewise.
9479 * config/rs6000/vector.md: Likewise.
9481 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9484 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9487 2018-03-20 Richard Biener <rguenther@suse.de>
9490 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9491 sign-conversions as zero, fall back to standard scalar_stmt
9494 2018-03-20 Martin Liska <mliska@suse.cz>
9497 * predict.c (rebuild_frequencies): Handle case when we have
9498 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9500 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9503 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9504 flag_section_anchors.
9505 * varasm.c (use_blocks_for_decl_p): Remove hack for
9506 dw2_force_const_mem.
9509 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9511 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9512 be created, use lowpart_subreg of operands[0] rather than operands[0]
9514 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9515 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9516 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9517 and n constraint instead of aarch64_shift_imm_di and Usd.
9518 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9519 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9521 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9524 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9525 to favor GPR over NEON registers.
9526 (<shift>di3_neon): Likewise.
9528 2018-03-20 Tom de Vries <tom@codesourcery.com>
9531 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9532 (nvptx_process_pars): Emit bar.sync asap and alap.
9534 2018-03-20 Tom de Vries <tom@codesourcery.com>
9537 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9538 seen_label if seen_label is already set.
9540 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9543 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9544 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9545 Use 1U instead of 1. Formatting fixes.
9548 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9549 instead of TREE_TYPE (s1) for the return value.
9551 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9553 PR tree-optimization/84946
9554 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9555 bitsize + bitsize in poly_uint64 rather than poly_int64.
9558 * dwarf2asm.c: Include fold-const.c.
9559 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9560 of decl rather than decl itself.
9562 PR rtl-optimization/84643
9563 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9565 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9568 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9569 calling assemble_variable.
9571 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9574 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9575 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9577 2018-03-19 Jim Wilson <jimw@sifive.com>
9580 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9581 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9582 (riscv_first_stack_step): Likewise.
9583 (riscv_option_override): Use STACK_BOUNDARY instead of
9585 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9587 (BIGGEST_ALIGNMENT): Set to 128.
9588 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9589 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9592 2018-03-19 Richard Biener <rguenther@suse.de>
9594 PR tree-optimization/84933
9595 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9596 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9598 2018-03-19 Richard Biener <rguenther@suse.de>
9600 PR tree-optimization/84859
9601 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9602 (cond_if_else_store_replacement): Perform sinking operation on
9603 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9604 Generalize what a BB with a single eligible store is.
9606 2018-03-19 Richard Biener <rguenther@suse.de>
9608 PR tree-optimization/84929
9609 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9610 chrec_is_positive against non-chrec arg.
9612 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9615 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9617 2018-03-18 Martin Liska <mliska@suse.cz>
9619 PR rtl-optimization/84635
9620 * regrename.c (build_def_use): Use matches_mode only when
9623 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9625 PR tree-optimization/84913
9626 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9627 vectorize chains of COND_EXPRs.
9629 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9631 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9633 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9635 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9637 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9639 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9641 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9642 Kito Cheng <kito.cheng@gmail.com>
9644 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9645 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9646 (nds32_adjust_reg_alloc_order): New function.
9647 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9649 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9651 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9652 nds32_print_operand, nds32_print_operand_address): Use
9653 HOST_WIDE_INT_PRINT_DEC instead.
9655 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9657 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9659 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9662 * config/i386/i386.c (initial_ix86_tune_features,
9663 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9665 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9666 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9667 rather than 1u << ix86_tune. Formatting fix.
9668 (ix86_option_override_internal): Change ix86_arch_mask from
9669 unsigned int to unsigned HOST_WIDE_INT, initialize to
9670 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9671 (ix86_function_specific_restore): Likewise.
9673 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9676 * postreload.c (reload_combine_recognize_pattern): Perform
9677 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9678 truncate_int_for_mode the result for the destination's mode.
9681 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9682 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9684 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9687 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9690 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9691 regno_allocno_class_array and sorted_pseudos.
9692 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9693 insns where regno is used.
9695 2018-03-16 Martin Liska <mliska@suse.cz>
9698 * multiple_target.c (create_dispatcher_calls): Redirect
9699 reference in the symbol table.
9701 2018-03-16 Martin Liska <mliska@suse.cz>
9704 * multiple_target.c (create_dispatcher_calls): Redirect also
9707 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9711 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9712 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9713 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9715 2018-03-16 Julia Koval <julia.koval@intel.com>
9717 * doc/invoke.texi (Skylake Server): Add CLWB.
9718 Cannonlake): Remove CLWB.
9720 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9722 PR tree-optimization/84841
9723 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9725 (FLOAT_ONE_CONST_TYPE): Define.
9726 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9727 (sort_by_operand_rank): Put entries with higher constant_type last
9728 rather than first to match comments.
9730 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9732 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9735 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9738 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9739 check and use address of target always.
9741 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9744 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9745 (indirect_thunk_bnd_needed): Likewise.
9746 (indirect_thunks_used): Likewise.
9747 (indirect_thunks_bnd_used): Likewise.
9748 (indirect_return_needed): New.
9749 (indirect_return_bnd_needed): Likewise.
9750 (output_indirect_thunk_function): Add a bool argument for
9752 (output_indirect_thunk_function): Don't generate alias for
9753 function return thunk.
9754 (ix86_code_end): Call output_indirect_thunk_function to generate
9755 function return thunks.
9756 (ix86_output_function_return): Set indirect_return_bnd_needed
9757 and indirect_return_needed instead of indirect_thunk_bnd_needed
9758 and indirect_thunk_needed.
9760 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9762 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9763 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9764 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9766 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9767 Paul Hua <paul.hua.gm@gmail.com>
9770 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9772 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9774 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9775 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9778 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9781 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9782 instead of GET_MODE_SIZE when comparing Units.
9784 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9787 * varasm.c (hash_section): Return an unchangeble hash value
9788 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9789 Return !aarch64_can_use_per_function_literal_pools_p ().
9791 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9794 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9795 rather than address of cmode as last argument to prepare_cmp_insn.
9797 2018-03-15 Julia Koval <julia.koval@intel.com>
9799 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9800 F_AVX512VNNI, F_AVX512BITALG): New.
9802 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9805 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9806 insn for floating-point loads and stores.
9808 2018-03-14 Carl Love <cel@us.ibm.com>
9810 * config/rs6000/rs6000-c.c: Add macro definitions for
9811 ALTIVEC_BUILTIN_VEC_PERMXOR.
9812 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9813 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9814 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9815 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9817 * config/doc/extend.texi: Add prototypes for vec_permxor.
9819 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9822 * diagnostic-show-locus.c (class layout_point): Convert m_line
9823 from int to linenum_type.
9824 (line_span::comparator): Use linenum "compare" function when
9825 comparing line numbers.
9826 (test_line_span): New function.
9827 (layout_range::contains_point): Convert param "row" from int to
9829 (layout_range::intersects_line_p): Likewise.
9830 (layout::will_show_line_p): Likewise.
9831 (layout::print_source_line): Likewise.
9832 (layout::should_print_annotation_line_p): Likewise.
9833 (layout::print_annotation_line): Likewise.
9834 (layout::print_leading_fixits): Likewise.
9835 (layout::annotation_line_showed_range_p): Likewise.
9836 (struct line_corrections): Likewise for field m_row.
9837 (line_corrections::line_corrections): Likewise for param "row".
9838 (layout::print_trailing_fixits): Likewise.
9839 (layout::get_state_at_point): Likewise.
9840 (layout::get_x_bound_for_row): Likewise.
9841 (layout::print_line): Likewise.
9842 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9843 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9844 * input.c (selftest::test_linenum_comparisons): New function.
9845 (selftest::input_c_tests): Call it.
9846 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9847 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9848 * selftest.h (ASSERT_GT): New macro.
9849 (ASSERT_GT_AT): New macro.
9850 (ASSERT_LT): New macro.
9851 (ASSERT_LT_AT): New macro.
9853 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9855 PR rtl-optimization/84780
9856 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9858 2018-03-14 Martin Liska <mliska@suse.cz>
9860 * tree.c (record_node_allocation_statistics): Use
9861 get_stats_node_kind.
9862 (get_stats_node_kind): New function extracted from
9863 record_node_allocation_statistics.
9864 (free_node): Use get_stats_node_kind.
9866 2018-03-14 Richard Biener <rguenther@suse.de>
9868 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9869 that the value-set of ANTIC_IN doesn't grow.
9872 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9874 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9875 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9877 2018-03-14 Julia Koval <julia.koval@intel.com>
9879 * config.gcc (icelake-client, icelake-server): New.
9881 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9882 (initial_ix86_arch_features): Ditto.
9883 (PTA_SKYLAKE): Add SGX.
9884 (PTA_ICELAKE): Remove.
9885 (PTA_ICELAKE_CLIENT): New.
9886 (PTA_ICELAKE_SERVER): New.
9887 (ix86_option_override_internal): Split up icelake on icelake client and
9889 (get_builtin_code_for_version): Ditto.
9890 (fold_builtin_cpu): Ditto.
9891 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9892 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9893 * config/i386/i386.h (processor_type): Ditto.
9894 * doc/invoke.texi: Ditto.
9896 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9899 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9900 INTEGER_CST offset, add it together with bitpos / 8 and
9901 sign extend based on POINTER_SIZE.
9905 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9908 * config/i386/constraints.md (Yc): New register constraint.
9909 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9910 Use Yc constraint for alternative 2 of operand 0. Remove
9911 preferred_for_speed attribute.
9913 2018-03-14 Richard Biener <rguenther@suse.de>
9915 PR tree-optimization/84830
9916 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9917 with the old one to avoid oscillations.
9919 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9922 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9924 (assign_by_spills): Return a flag of reload assignment failure.
9925 Do not process the reload assignment failures. Do not spill other
9926 reload pseudos if they has the same reg class. Update n if
9928 (lra_assign): Add a return arg. Set up from the result of
9929 assign_by_spills call.
9930 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9931 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9932 usage_insns if it is not NULL.
9933 (spill_hard_reg_in_range): New function.
9934 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9935 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9936 function prototypes.
9937 (lra_assign): Change prototype.
9938 * lra.c (lra): Add code to deal with fails by splitting hard reg
9941 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9943 * config/riscv/riscv.opt (mrelax): New option.
9944 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9945 "norelax" when riscv_mrelax is disabled.
9946 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9948 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9951 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9952 reassociation for int modes.
9954 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9956 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9957 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9959 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9960 * config/aarch64/aarch64-sve.md
9961 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9962 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9963 (*extend<mode><Vwide>2): Rename to...
9964 (aarch64_sve_extend<mode><Vwide>2): ...this.
9965 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9966 renaming the old pattern to...
9967 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9969 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9970 define_expand, renaming the old pattern to...
9971 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9972 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9973 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9974 account when deciding which SVE instruction the optab should use.
9975 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9977 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9979 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9980 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9981 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9982 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9983 (tlsdesc_small_<mode>): Turn a define_expand and use
9984 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9985 (tlsdesc_small_advsimd_<mode>): ...this.
9986 (tlsdesc_small_sve_<mode>): New pattern.
9988 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9990 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9991 (UNSPEC_UMUL_HIGHPART): New constants.
9992 (MUL_HIGHPART): New int iteraor.
9993 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9994 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9996 (*<su>mul<mode>3_highpart): New define_insn.
9998 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10001 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10004 2018-03-13 Martin Liska <mliska@suse.cz>
10007 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10009 (sem_item_optimizer::~sem_item_optimizer): Release it.
10010 (sem_item_optimizer::merge_classes): Register variable aliases.
10011 (sem_item_optimizer::fixup_pt_set): New function.
10012 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10013 * ipa-icf.h: Declare new variables and functions.
10015 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10017 PR middle-end/84834
10018 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10019 integer_pow2p@2 and test integer_pow2p in condition.
10020 (A < 0 ? C : 0): Similarly for @1.
10022 PR middle-end/84831
10023 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10024 characters starting at p contain '\0' character, don't look beyond
10028 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10029 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10032 * reg-stack.c (change_stack): Change update_end var from int to
10033 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10034 also call set_block_for_insn on the newly added insns and rescan.
10037 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10038 on the last operand.
10041 * tree.c (stabilize_reference_1): Return save_expr (e) for
10042 STATEMENT_LIST even if it doesn't have side-effects.
10044 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10046 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10048 2018-03-12 Renlin Li <renlin.li@arm.com>
10050 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10051 aarch64_output_scalar_simd_mov_immediate.
10053 2018-03-12 Martin Sebor <msebor@redhat.com>
10055 PR tree-optimization/83456
10056 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10057 for perfectly overlapping calls to memcpy.
10058 (gimple_fold_builtin_memory_chk): Same.
10059 (gimple_fold_builtin_strcpy): Handle no-warning.
10060 (gimple_fold_builtin_stxcpy_chk): Same.
10061 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10063 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10065 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10066 parameter. Use it for SFmode.
10067 (rs6000_function_arg_advance_1): Adjust.
10068 (rs6000_function_arg): Adjust.
10069 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10071 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10073 PR rtl-optimization/84169
10074 PR rtl-optimization/84780
10075 * combine.c (can_combine_p): Check for a 2-insn combination whether
10076 the destination register is used between the two insns, too.
10078 2018-03-12 Richard Biener <rguenther@suse.de>
10080 PR tree-optimization/84803
10081 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10082 for refs DR analysis didn't process.
10084 2018-03-12 Richard Biener <rguenther@suse.de>
10086 PR tree-optimization/84777
10087 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10088 force-vectorize loops ignore whether we are optimizing for size.
10090 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10092 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10093 (TARGET_MD_ASM_ADJUST): Define.
10095 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10096 Kito Cheng <kito.cheng@gmail.com>
10097 Chung-Ju Wu <jasonwucj@gmail.com>
10099 * config/nds32/nds32.c (nds32_compute_stack_frame,
10100 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10101 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10102 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10103 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10104 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10105 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10106 * config/nds32/nds32.md (prologue, epilogue): Use macro
10107 NDS32_V3PUSH_AVAILABLE_P to do checking.
10109 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10112 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10113 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10114 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10115 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10116 addition of most attributes on !orig_type_die or the attribute not
10117 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10119 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10120 Chung-Ju Wu <jasonwucj@gmail.com>
10122 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10123 __NDS32_VH__ macro.
10124 * config/nds32/nds32.opt (mvh): New option.
10126 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10127 Chung-Ju Wu <jasonwucj@gmail.com>
10129 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10131 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10132 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10135 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10136 Chung-Ju Wu <jasonwucj@gmail.com>
10138 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10140 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10141 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10143 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10144 Kito Cheng <kito.cheng@gmail.com>
10145 Chung-Ju Wu <jasonwucj@gmail.com>
10147 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10148 UNSPEC_FFMISM and UNSPEC_FLMISM.
10149 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10150 for ffb, ffmism and flmism.
10151 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10152 (unspec_ffmism): Ditto.
10153 (unspec_flmism): Ditto.
10154 (nds32_expand_builtin_impl): Check if string extension is available.
10155 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10156 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10158 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10161 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10164 * lra-assigns.c (assign_by_spills): Return a flag of reload
10165 assignment failure. Do not process the reload assignment
10166 failures. Do not spill other reload pseudos if they has the same
10168 (lra_assign): Add a return arg. Set up from the result of
10169 assign_by_spills call.
10170 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10171 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10172 usage_insns if it is not NULL.
10173 (spill_hard_reg_in_range): New function.
10174 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10175 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10176 function prototypes.
10177 (lra_assign): Change prototype.
10178 * lra.c (lra): Add code to deal with fails by splitting hard reg
10181 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10184 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10186 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10189 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10190 (dw_val_node): Add val_symbolic_view.
10191 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10192 (symview_upper_bound): New.
10193 (new_line_info_table): Initialize symviews_since_reset.
10194 (dwarf2out_source_line): Count symviews_since_reset and set
10195 symview_upper_bound.
10196 (dw_val_equal_p): Handle symview.
10197 (add_AT_symview): New.
10198 (print_dw_val): Handle symview.
10199 (attr_checksum, attr_checksum_ordered): Likewise.
10200 (same_dw_val_p, size_of_die): Likewise.
10201 (value_format, output_die): Likewise.
10202 (add_high_low_attributes): Use add_AT_symview for entry_view.
10203 (dwarf2out_finish): Reset symview_upper_bound, clear
10206 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10209 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10210 Add strict argument and use it.
10211 (rs6000_split_multireg_move): Update for new strict argument.
10212 (mem_operand_gpr): Disallow all non-offsettable addresses.
10213 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10215 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10218 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10219 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10220 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10223 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10224 decl, use remap_type if we want to use the type.
10226 2018-03-09 Martin Sebor <msebor@redhat.com>
10228 PR tree-optimization/84526
10229 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10231 (builtin_access::generic_overlap): Be prepared to handle non-array
10234 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10236 PR rtl-optimization/84682
10237 * lra-constraints.c (process_address_1): Check is_address flag
10238 for address constraints.
10239 (process_alt_operands): Likewise.
10240 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10241 preprocess_constraints.
10242 * recog.h (preprocess_constraints): Add oploc parameter.
10244 * recog.c (preprocess_constraints): Test address_operand for
10245 CT_ADDRESS constraints.
10247 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10250 * lra-assigns.c (assign_by_spills): Return a flag of reload
10251 assignment failure. Do not process the reload assignment
10252 failures. Do not spill other reload pseudos if they has the same
10254 (lra_assign): Add a return arg. Set up from the result of
10255 assign_by_spills call.
10256 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10257 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10258 usage_insns if it is not NULL.
10259 (spill_hard_reg_in_range): New function.
10260 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10261 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10262 function prototypes.
10263 (lra_assign): Change prototype.
10264 * lra.c (lra): Add code to deal with fails by splitting hard reg
10267 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10270 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10271 Accept complain bool parameter. Only emit errors if it is true.
10272 (arm_parse_cpu_option_name): Likewise.
10273 (arm_target_thumb_only): Adjust callers of the above.
10274 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10275 prototype to take a default true bool parameter.
10276 (arm_parse_arch_option_name): Likewise.
10278 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10279 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10283 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10284 * configure: Regenerate.
10285 * configure.ac ("linker --version-script option"): New.
10286 ("linker soname option"): New.
10288 2018-03-09 Richard Biener <rguenther@suse.de>
10290 PR tree-optimization/84775
10291 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10292 immediate uses of predicate stmts and mark them modified.
10295 PR tree-optimization/84178
10296 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10298 (version_loop_for_if_conversion): Delay update_ssa call.
10299 (tree_if_conversion): Delay update_ssa until after predicate
10302 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10305 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10306 when the function accesses prior frames.
10308 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10311 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10312 gen_llsym, otherwise call maybe_gen_llsym.
10314 PR inline-asm/84742
10315 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10316 has ',' character inside of it.
10318 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10321 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10322 as clobbering CC_REGNUM.
10324 2018-03-08 Richard Biener <rguenther@suse.de>
10326 PR middle-end/84552
10327 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10328 (follow_copies_to_constant): Do not follow SSA names registered
10331 2018-03-08 Richard Biener <rguenther@suse.de>
10333 PR tree-optimization/84178
10334 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10336 (version_loop_for_if_conversion): Delay update_ssa call.
10337 (tree_if_conversion): Delay update_ssa until after predicate
10340 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10342 PR tree-optimization/84178
10343 * tree-if-conv.c (release_bb_predicate): Remove the
10344 the assertion that the stmts have NULL use_ops.
10345 Discard the statements, asserting that they haven't
10346 yet been added to a BB.
10348 2018-03-08 Richard Biener <rguenther@suse.de>
10350 PR tree-optimization/84746
10351 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10352 (phi_translate): Pass in destination ANTIC_OUT set.
10353 (phi_translate_1): Likewise. For a simplified result lookup
10354 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10355 (phi_translate_set): Adjust.
10356 (do_pre_regular_insertion): Likewise.
10357 (do_pre_partial_partial_insertion): Likewise.
10359 2018-03-08 Martin Liska <mliska@suse.cz>
10361 PR gcov-profile/84735
10362 * doc/gcov.texi: Document usage of profile files.
10363 * gcov-io.h: Document changes in the format.
10365 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10369 * dwarf2out.c (struct dw_line_info_table): Update comments for
10371 (FORCE_RESET_NEXT_VIEW): New.
10372 (FORCE_RESETTING_VIEW_P): New.
10373 (RESETTING_VIEW_P): Check for -1 too.
10374 (ZERO_VIEW_P): Likewise.
10375 (new_line_info_table): Force-reset next view.
10376 (dwarf2out_begin_function): Likewise.
10377 (dwarf2out_source_line): Simplify zero_view_p initialization.
10378 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10379 view directly. Omit view when omitting .loc at line 0.
10381 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10383 PR tree-optimization/84740
10384 * tree-switch-conversion.c (process_switch): Call build_constructors
10385 only if info.phi_count is non-zero.
10387 PR tree-optimization/84739
10388 * tree-tailcall.c (find_tail_calls): Check call arguments against
10389 DECL_ARGUMENTS (current_function_decl) rather than
10390 DECL_ARGUMENTS (func) when checking for tail recursion.
10392 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10394 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10395 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10396 Volker Reichelt's entry and add entries for people that perform
10397 GCC fuzzy testing and report numerous bugs.
10399 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10402 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10403 readonly data in sdata, if that is disabled.
10404 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10405 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10406 -mreadonly-in-sdata option.
10408 2018-03-07 Martin Sebor <msebor@redhat.com>
10410 PR tree-optimization/84468
10411 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10412 basic block when looking for nul assignment.
10414 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10417 * except.h (output_function_exception_table): Adjust prototype.
10418 * except.c (output_function_exception_table): Remove FNNAME parameter
10419 and add SECTION parameter. Ouput one part of the table at a time.
10420 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10421 the first part of the exception table and emit unwind directives.
10422 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10423 (i386_pe_seh_cold_init): Likewise.
10424 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10425 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10426 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10427 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10428 * config/i386/winnt.c: Include except.h.
10429 (struct seh_frame_state): Add reg_offset, after_prologue and
10430 in_cold_section fields.
10431 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10432 (i386_pe_seh_cold_init): New function.
10433 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10434 to seh->in_cold_section.
10435 (seh_emit_push): Record the offset of the push.
10436 (seh_emit_save): Record the offet of the save.
10437 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10438 Test seh->after_prologue to disregard the epilogue.
10439 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10440 (i386_pe_end_cold_function): New function.
10442 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10445 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10446 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10449 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10451 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10452 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10454 PR middle-end/84723
10455 * multiple_target.c: Include tree-inline.h and intl.h.
10456 (expand_target_clones): Diagnose and fail if node->definition and
10457 !tree_versionable_function_p (node->decl).
10459 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10461 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10463 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10464 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10465 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10467 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10470 * combine.c (try_combine): Use reg_or_subregno instead of handling
10471 just paradoxical SUBREGs and REGs.
10473 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10475 * config/arc/arc.c (arc_finalize_pic): Remove function.
10476 (arc_must_save_register): We use single base PIC register, remove
10477 checks to save/restore the PIC register.
10478 (arc_expand_prologue): Likewise.
10479 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10481 (arc_verify_short): Likewise.
10482 (arc_attr_type): Likewise.
10483 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10484 (walk_stores): Likewise.
10485 (arc_address_cost): Make it static.
10486 (arc_verify_short): Likewise.
10487 (branch_dest): Likewise.
10488 (arc_attr_type): Likewise.
10489 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10490 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10491 (arc_final_prescan_insn): Remove inserting the nops due to
10492 hardware hazards. It is done in reorg step.
10493 (insn_length_variant_t): Remove.
10494 (insn_length_parameters_t): Likewise.
10495 (arc_insn_length_parameters): Likewise.
10496 (arc_get_insn_variants): Likewise.
10497 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10499 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10501 PR inline-asm/84683
10502 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10505 PR tree-optimization/84687
10506 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10508 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10510 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10512 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10513 Rename to ppc_speculation_barrier.
10514 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10515 __builtin_ppc_speculation_barrier.
10517 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10520 * combine.c (combine_simplify_rtx): Don't try to simplify if
10521 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10524 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10526 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10527 to 32 bytes when compiling for POWER9.
10529 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10532 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10533 regparm >= 3 with no arg reg available also for calls with
10534 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10537 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10539 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10541 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10544 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10546 2018-03-05 Richard Biener <rguenther@suse.de>
10548 PR tree-optimization/84486
10549 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10550 When inserting a __builtin_assume_aligned call set the LHS
10551 SSA name alignment info accordingly.
10553 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10555 PR tree-optimization/84114
10556 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10557 Avoid reassociation of FLOAT_MODE addition.
10559 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10561 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10562 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10563 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10564 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10565 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10566 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10567 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10569 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10570 __builtin_ia32_wbinvd): New builtins.
10571 (SPECIAL_ARGS2): New.
10572 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10573 (SPECIAL_ARGS2): New.
10574 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10575 (ix86_valid_target_attribute_inner_p): Ditto.
10576 (ix86_init_mmx_sse_builtins): Add special_args2.
10577 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10578 TARGET_WBNOINVD_P): New.
10579 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10580 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10581 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10582 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10583 * config/i386/pconfigintrin.h: New file.
10584 * config/i386/wbnoinvdintrin.h: Ditto.
10585 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10587 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10589 2018-03-05 Richard Biener <rguenther@suse.de>
10591 PR tree-optimization/84670
10592 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10594 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10595 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10596 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10597 doesn't grow if all successors (recursively) were visited at least
10600 2018-03-05 Richard Biener <rguenther@suse.de>
10602 PR tree-optimization/84650
10603 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10604 if executed in the loop pipeline.
10606 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10608 * doc/configfiles.texi (Configuration Files): Move info about
10609 conditionalizing $target-protos.h to...
10610 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10611 differs from $target-protos.h.
10613 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10614 Chung-Ju Wu <jasonwucj@gmail.com>
10616 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10617 * config/nds32/nds32-multiple.md (setmemsi): Define.
10618 * config/nds32/nds32-memory-manipulation.c
10619 (nds32_gen_dup_4_byte_to_word_value): New.
10620 (emit_setmem_word_loop): New.
10621 (emit_setmem_byte_loop): New.
10622 (nds32_expand_setmem_loop): New.
10623 (nds32_expand_setmem_loop_v3m): New.
10624 (nds32_expand_setmem_unroll): New.
10625 (nds32_expand_setmem): New.
10627 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10628 Chung-Ju Wu <jasonwucj@gmail.com>
10630 * config/nds32/nds32-memory-manipulation.c
10631 (nds32_emit_load_store): New.
10632 (nds32_emit_post_inc_load_store): New.
10633 (nds32_emit_mem_move): New.
10634 (nds32_emit_mem_move_block): New.
10635 (nds32_expand_movmemsi_loop_unknown_size): New.
10636 (nds32_expand_movmemsi_loop_known_size): New.
10637 (nds32_expand_movmemsi_loop): New.
10638 (nds32_expand_movmemsi_unroll): New.
10639 (nds32_expand_movmemqi): Rename ...
10640 (nds32_expand_movmemsi): ... to this.
10641 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10642 (movmemsi): ... to this.
10643 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10644 (nds32_expand_movmemsi): ... to this.
10646 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10647 Monk Chiang <sh.chiang04@gmail.com>
10648 Chung-Ju Wu <jasonwucj@gmail.com>
10650 * config/nds32/nds32-protos.h
10651 (nds32_expand_load_multiple): New arguments.
10652 (nds32_expand_store_multiple): Ditto.
10653 (nds32_valid_multiple_load_store): Rename ...
10654 (nds32_valid_multiple_load_store_p): ... to this.
10655 * config/nds32/nds32-memory-manipulation.c
10656 (nds32_expand_load_multiple): Refine implementation.
10657 (nds32_expand_store_multiple): Ditto.
10658 * config/nds32/nds32-multiple.md
10659 (load_multiple): Update nds32_expand_load_multiple interface.
10660 (store_multiple): Update nds32_expand_store_multiple interface.
10661 * config/nds32/nds32-predicates.c
10662 (nds32_valid_multiple_load_store): Rename ...
10663 (nds32_valid_multiple_load_store_p): ... to this and refine
10665 * config/nds32/predicates.md
10666 (nds32_load_multiple_and_update_address_operation): New predicate.
10667 (nds32_store_multiple_and_update_address_operation): New predicate.
10669 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10670 Chung-Ju Wu <jasonwucj@gmail.com>
10672 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10673 (combo): New attribute.
10674 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10676 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10678 * config/nds32/nds32.opt: Change -mcmodel= default value.
10680 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10681 Monk Chiang <sh.chiang04@gmail.com>
10682 Chung-Ju Wu <jasonwucj@gmail.com>
10684 * config/nds32/constants.md (unspec_element): New enum.
10685 * config/nds32/constraints.md (Umw): New constraint.
10686 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10687 * config/nds32/nds32-intrinsic.md: Likewise.
10688 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10689 (nds32_valid_smw_lwm_base_p): New.
10690 (nds32_output_smw_single_word): New.
10691 (nds32_output_lmw_single_word): New.
10692 (nds32_expand_unaligned_load): New.
10693 (nds32_expand_unaligned_store): New.
10694 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10695 (nds32_output_smw_single_word): Declare.
10696 (nds32_output_lmw_single_word): Declare.
10697 (nds32_expand_unaligned_load): Declare.
10698 (nds32_expand_unaligned_store): Declare.
10699 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10700 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10701 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10702 NDS32_BUILTIN_UASTORE_DW.
10703 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10706 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10707 Kito Cheng <kito.cheng@gmail.com>
10708 Chung-Ju Wu <jasonwucj@gmail.com>
10710 * config/nds32/nds32-intrinsic.c
10711 (nds32_expand_builtin_null_ftype_reg): Delete.
10712 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10713 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10714 (nds32_read_argument): New.
10715 (nds32_legitimize_target): Ditto.
10716 (nds32_legitimize_argument): Ditto.
10717 (nds32_check_constant_argument): Ditto.
10718 (nds32_expand_unop_builtin): Ditto.
10719 (nds32_expand_unopimm_builtin): Ditto.
10720 (nds32_expand_binop_builtin): Ditto.
10721 (nds32_builtin_decl_impl): Ditto.
10722 (builtin_description): Ditto.
10723 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10724 (nds32_init_builtins_impl): Ditto.
10725 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10726 (nds32_builtin_decl): New.
10727 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10728 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10730 2018-03-02 Jeff Law <law@redhat.com>
10732 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10733 (redundant_insn, fill_simple_delay_slots): Likewise.
10734 (fill_slots_from_thread): Likewise.
10735 * resource.c (mark_referenced_resources): Likewise.
10736 (mark_set_resources, find_dead_or_set_registers): Likewise.
10738 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10740 * substring-locations.h (format_warning_va): Formatting fix for
10741 ATTRIBUTE_GCC_DIAG.
10742 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10744 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10745 * substring-locations.c: Include intl.h.
10746 (format_warning_va): Turned into small wrapper around
10747 format_warning_n_va, renamed to ...
10748 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10749 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10751 (format_warning_at_substring_n): New function.
10752 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10753 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10754 format_warning_at_substring with just a shorter name instead of
10755 const function pointer.
10756 (fmtwarn_n): New function.
10757 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10758 appropriate, get rid of all the fmtstr temporaries, move conditionals
10759 with G_() wrapped string literals directly into fmtwarn arguments,
10760 cast dir.len to (int), formatting fixes.
10762 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10764 * doc/invoke.texi: Remove "Cilk Plus" references.
10766 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10767 Richard Biener <rguenther@suse.de>
10770 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10771 for error or warning attributes if CALL_FROM_THUNK_P is set.
10774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10777 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10778 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10781 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10782 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10784 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10785 instead of -1U in last predictors element's probability member.
10787 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10790 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10791 arguments if they are comparable.
10793 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10795 PR tree-optimization/84634
10796 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10797 masks and masked_loop_p with a single loop_masks, making sure it's
10798 null for bb vectorization.
10800 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10802 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10803 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10804 loop->force_vectorize to check whether there is no alias.
10806 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10809 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10811 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10813 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10814 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10815 instead of a loop around prev_real_insn.
10816 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10819 PR inline-asm/84625
10820 * config/i386/i386.c (ix86_print_operand): Use conditional
10821 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10824 2018-03-02 Richard Biener <rguenther@suse.de>
10826 PR tree-optimization/84427
10827 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10828 (bitmap_set_subtract_values): Rewrite to handle multiple
10831 (prune_clobbered_mems): Likewise.
10832 (phi_translate): Take edge instead of pred/phiblock.
10833 (phi_translate_1): Likewise.
10834 (phi_translate_set): Likewise. Insert all translated
10835 exprs for a value into the set, keeping possibly multiple
10836 expressions per value.
10837 (compute_antic_aux): Adjust for phi_translate changes.
10838 When intersecting union the expressions and prune those
10839 not in the final value set, keeping possibly multiple
10840 expressions per value. Do not use value-insertion
10841 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10842 all expressions. Add verification that the value-sets
10843 only shrink during iteration.
10844 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10845 (do_pre_regular_insertion): Likewise.
10846 (do_pre_partial_partial_insertion): Likewise.
10848 2018-03-02 Richard Biener <rguenther@suse.de>
10851 * config/darwin.c (saved_debug_info_level): New static global.
10852 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10853 (darwin_asm_lto_end): Restore debug info generation settings.
10855 2018-03-01 Martin Liska <mliska@suse.cz>
10858 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10859 volatile arguments.
10861 2018-03-01 Richard Biener <rguenther@suse.de>
10864 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10865 type annotation with LTO when debug was disabled at compile-time.
10867 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10869 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10871 (mips_final_postscan_insn): Likewise.
10873 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10875 PR rtl-optimization/84528
10876 * alias.c (init_alias_target): Add commentary.
10877 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10878 a unique base value if the frame pointer is not eliminated
10879 to the stack pointer.
10881 2018-03-01 Tom de Vries <tom@codesourcery.com>
10883 PR rtl-optimization/83327
10884 * lra-int.h (hard_regs_spilled_into): Declare.
10885 * lra.c (hard_regs_spilled_into): Define.
10886 (init_reg_info): Init hard_regs_spilled_into.
10887 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10888 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10889 (process_bb_lives): Handle hard_regs_spilled_into.
10890 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10891 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10893 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10895 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10896 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10897 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10898 * config/rs6000/aix72.h: New file.
10900 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10902 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10903 instead of warning_at with conditional singular and plural messages
10907 * stor-layout.c (update_alignment_for_field): For
10908 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10909 && !DECL_PACKED (field), do the alignment update, just use
10910 only desired_align instead of MAX (type_align, desired_align)
10912 (place_field): Don't do known_align < desired_align handling
10913 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10914 is non-NULL, instead do it after rli->prev_field handling and
10915 only if not within a bitfield word. For DECL_PACKED (field)
10916 use type_align of BITS_PER_UNIT.
10918 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10920 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10921 superfluous parentheses and trailing spaces.
10923 2018-02-28 Richard Biener <rguenther@suse.de>
10925 PR tree-optimization/84584
10926 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10927 SCoPs with fake exit edge.
10929 2018-02-28 Martin Liska <mliska@suse.cz>
10932 * timevar.c (timer::print): Fix format to properly print 100%
10935 2018-02-28 Richard Biener <rguenther@suse.de>
10937 PR middle-end/84607
10938 * genmatch.c (capture_info::walk_match): Do not mark
10939 captured expressions without operands as expr_p given
10940 they act more like predicates and should be subject to
10941 "lost tail" side-effect preserving.
10943 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10945 PR rtl-optimization/81611
10946 * auto-inc-dec.c (attempt_change): Move dead note from
10947 mem_insn if it's the next use of regno
10948 (find_address): Take address use of reg holding
10949 non-incremented value. Add parm to limit search to the named
10951 (merge_in_block): Attempt to use a mem insn that is the next
10952 use of the original regno.
10954 2018-02-27 Martin Sebor <msebor@redhat.com>
10957 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10958 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10960 2018-02-27 Martin Sebor <msebor@redhat.com>
10962 PR translation/84207
10963 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10964 n argument to unsigned HOST_WIDE_INT.
10965 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10966 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10967 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10970 2018-02-27 Richard Biener <rguenther@suse.de>
10972 PR tree-optimization/84512
10973 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10974 Do not use the estimate returned from record_stmt_cost for
10975 the scalar iteration cost but sum properly using add_stmt_cost.
10977 2018-02-27 Richard Biener <rguenther@suse.de>
10979 PR tree-optimization/84466
10980 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10981 Adjust last change to less strictly validate use operands.
10983 2018-02-27 Martin Liska <mliska@suse.cz>
10985 PR gcov-profile/84548
10986 * gcov.c (process_file): Allow partial overlap and consider it
10987 also as group functions.
10988 (output_lines): Properly calculate range of lines for a group.
10990 2018-02-27 Martin Liska <mliska@suse.cz>
10992 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10993 'ggc' suffixes. Change first column width.
10994 (timer::print): Fix formatting of the column.
10996 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10998 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10999 preserve inline entry blocks for the sake of debug inline
11000 entry point markers alone.
11001 (remove_unused_locals): Suggest in comments a better place to
11002 force the preservation of inline entry blocks that are
11003 otherwise unused, but do not preserve them.
11005 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11007 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11009 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11012 * config/i386/constraints.md (Bs): Replace
11013 ix86_indirect_branch_register with
11014 TARGET_INDIRECT_BRANCH_REGISTER.
11016 * config/i386/i386.md (indirect_jump): Likewise.
11017 (tablejump): Likewise.
11018 (*sibcall_memory): Likewise.
11019 (*sibcall_value_memory): Likewise.
11020 Peepholes of indirect call and jump via memory: Likewise.
11021 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11022 (*sibcall_value_GOT_32): Likewise.
11023 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11024 (GOT_memory_operand): Likewise.
11025 (call_insn_operand): Likewise.
11026 (sibcall_insn_operand): Likewise.
11027 (GOT32_symbol_operand): Likewise.
11028 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11030 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11032 PR rtl-optimization/83496
11033 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11034 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11035 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11036 redundant insn, if any.
11037 (relax_delay_slots): Likewise.
11038 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11040 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11042 PR tree-optimization/83965
11043 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11044 that grouped statements are part of a reduction chain. Return
11045 true if the statement is not marked as a reduction itself but
11046 is part of a group.
11047 (vect_recog_dot_prod_pattern): Don't check whether the statement
11048 is part of a group here.
11049 (vect_recog_sad_pattern): Likewise.
11050 (vect_recog_widen_sum_pattern): Likewise.
11052 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11055 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11057 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11060 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11062 (ix86_output_indirect_function_return): New prototype.
11063 (ix86_split_simple_return_pop_internal): Likewise.
11064 * config/i386/i386.c (indirect_return_via_cx): New.
11065 (indirect_return_via_cx_bnd): Likewise.
11066 (indirect_thunk_name): Handle return va CX_REG.
11067 (output_indirect_thunk_function): Create alias for
11068 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11069 (ix86_output_indirect_jmp): Remove the bool argument.
11070 (ix86_output_indirect_function_return): New function.
11071 (ix86_split_simple_return_pop_internal): Likewise.
11072 * config/i386/i386.md (*indirect_jump): Don't pass false
11073 to ix86_output_indirect_jmp.
11074 (*tablejump_1): Likewise.
11075 (simple_return_pop_internal): Change it to define_insn_and_split.
11076 Call ix86_split_simple_return_pop_internal to split it for
11077 -mfunction-return=.
11078 (simple_return_indirect_internal): Call
11079 ix86_output_indirect_function_return instead of
11080 ix86_output_indirect_jmp.
11082 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11085 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11086 memset and value initialization afterwards.
11088 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11090 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11092 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11095 * common/config/aarch64/aarch64-common.c
11096 (aarch_option_optimization_table[]): Switch
11097 off fomit-frame-pointer
11099 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11100 Chung-Ju Wu <jasonwucj@gmail.com>
11102 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11104 (store_multiple): Ditto.
11106 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11108 * config.gcc: Add --with-cpu support for nds32 target.
11109 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11110 * config/nds32/nds32.opt: Add -mcpu= option.
11112 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11114 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11115 isel=yes): Warn for these deprecated options.
11117 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11119 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11120 ISA_2_5_MASKS_EMBEDDED.
11122 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11124 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11125 p->max as pointers rather than using iterative_hash_expr.
11127 2018-02-23 Carl Love <cel@us.ibm.com>
11129 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11130 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11132 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11133 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11134 P8V_BUILTIN_VEC_VUNSIGNED2.
11136 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11139 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11140 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11142 * lra-constraints.c (curr_insn_transform): Set up
11143 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11145 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11146 * lra-eliminations.c (spill_pseudos): Ditto.
11147 (process_insn_for_elimination): Ditto.
11148 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11149 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11150 LRA_NON_CLOBBERED_ALT.
11152 2018-02-22 Martin Sebor <msebor@redhat.com>
11154 PR tree-optimization/84480
11155 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11156 to maybe_diag_stxncpy_trunc. Call it.
11157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11158 from gimple_fold_builtin_strcpy. Print inlining stack.
11159 (handle_builtin_stxncpy): Print inlining stack.
11160 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11162 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11165 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11166 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11167 and -fcheck-pointer-bounds are used together.
11168 (indirect_thunk_prefix): New enum.
11169 (indirect_thunk_need_prefix): New function.
11170 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11171 "_nt" instead of "_bnd" for NOTRACK prefix.
11172 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11173 (output_indirect_thunk_function): Likewise.
11175 (ix86_code_end): Update output_indirect_thunk_function calls.
11176 (ix86_output_indirect_branch_via_reg): Replace
11177 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11178 (ix86_output_indirect_branch_via_push): Likewise.
11179 (ix86_output_function_return): Likewise.
11180 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11181 incompatible with -fcf-protection=branch and
11182 -fcheck-pointer-bounds.
11184 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11187 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11188 Change gcc_assert call to output_operand_lossage.
11190 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11192 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11194 2018-02-22 DJ Delorie <dj@redhat.com>
11195 Sebastian Perta <sebastian.perta@renesas.com>
11196 Oleg Endo <olegendo@gcc.gnu.org>
11198 * config/rx/rx.c (rx_rtx_costs): New function.
11199 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11201 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11203 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11205 2018-02-22 Martin Liska <mliska@suse.cz>
11208 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11209 Add "native" as a possible value.
11211 2018-02-22 Martin Liska <mliska@suse.cz>
11214 * config/i386/i386.c (ix86_option_override_internal):
11215 Add "native" as a possible value for -march and -mtune.
11217 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11220 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11221 to all type variants.
11223 PR tree-optimization/84503
11224 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11225 width as info->bitpos + info->bitsize - start.
11226 (merged_store_group::merge_overlapping): Simplify width computation.
11227 (check_no_overlap): New function.
11228 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11229 start + width and last_order of the group, fail if check_no_overlap
11231 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11232 to group if check_no_overlap fails.
11234 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11236 * config/rs6000/altivec.md: Delete contraint arguments to
11237 define_expand, define_split, and define_peephole2, and in
11238 define_insn_and_split if always unused.
11239 * config/rs6000/darwin.md: Ditto.
11240 * config/rs6000/dfp.md: Ditto.
11241 * config/rs6000/rs6000.md: Ditto.
11242 * config/rs6000/sync.md: Ditto.
11243 * config/rs6000/vector.md: Ditto.
11244 * config/rs6000/vsx.md: Ditto.
11246 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11248 * config/rs6000/altivec.md: Write output control strings as braced
11249 blocks instead of double-quoted strings.
11250 * config/rs6000/darwin.md: Ditto.
11251 * config/rs6000/rs6000.md: Ditto.
11252 * config/rs6000/vector.md: Ditto.
11253 * config/rs6000/vsx.md: Ditto.
11255 2018-02-21 Jason Merrill <jason@redhat.com>
11257 PR c++/84314 - ICE with templates and fastcall attribute.
11258 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11260 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11262 * ipa-cp.c (determine_versionability): Fix comment typos.
11264 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11267 * ipa-cp.c (determine_versionability): Do not version functions caling
11270 2018-02-21 Martin Liska <mliska@suse.cz>
11273 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11274 Add "native" as a possible value.
11275 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11276 the macro when native cpu detection is available.
11278 2018-02-21 Martin Liska <mliska@suse.cz>
11281 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11282 Add "native" as a possible value.
11283 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11284 when native cpu detection is available.
11286 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11287 Martin Sebor <msebor@redhat.com>
11289 PR tree-optimization/84478
11290 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11292 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11293 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11294 support which is conservatively correct, for 2 only stay conservative
11295 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11296 argument to the 2 argument get_range_strlen, adjust 6 arg
11297 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11299 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11300 (gimple_fold_builtin_strlen): Pass true as last argument to
11303 2018-02-20 Martin Sebor <msebor@redhat.com>
11305 PR middle-end/84095
11306 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11307 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11308 (builtin_memref::builtin_memref): Factor out parts into
11309 set_base_and_offset and call it.
11311 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11313 PR middle-end/84406
11314 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11315 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11316 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11317 search at the associated MODE_INT.
11319 2018-02-20 Jeff Law <law@redhat.com>
11321 PR middle-end/82123
11322 PR tree-optimization/81592
11323 PR middle-end/79257
11324 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11325 for range data rather than using global data.
11326 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11327 range data rather than using global data.
11328 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11329 pass it to children as needed.
11330 (struct directive::fmtresult): Similarly.
11331 (struct directive::set_width): Similarly.
11332 (struct directive::set_precision): Similarly.
11333 (format_integer, format_directive, parse_directive): Similarly.
11334 (format_none): Accept unnamed vr_values parameter.
11335 (format_percent, format_floating, format_character): Similarly.
11336 (format_string, format_plain): Similarly.
11337 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11338 the EVRP range analyzer for range data rather than using global data.
11339 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11340 gimple-ssa-evrp-analyze.h
11341 (class sprintf_dom_walker): Add after_dom_children member function.
11342 Add evrp_range_analyzer member.
11343 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11344 range analyzer as needed.
11345 (sprintf_dom_walker::after_dom_children): New member function.
11346 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11348 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11349 (evrp_range_analyzer::pop_to_marker): Likewise.
11351 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11353 PR tree-optimization/84419
11354 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11355 with the required type if its current type is compatible but
11358 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11360 PR middle-end/82004
11361 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11362 after vectorization.
11364 2018-02-20 Martin Liska <mliska@suse.cz>
11367 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11368 possible values if we don't have a hint.
11370 2018-02-20 Martin Liska <mliska@suse.cz>
11374 * final.c (shorten_branches): Build align_tab array with one
11376 * opts.c (finish_options): Add alignment option limit check.
11377 (MAX_CODE_ALIGN): Likewise.
11378 (MAX_CODE_ALIGN_VALUE): Likewise.
11379 * doc/invoke.texi: Document maximum allowed option value for
11380 all -falign-* options.
11382 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11385 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11386 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11387 * var-tracking.c (emit_note_insn_var_location): Remove all references
11388 to NOTE_INSN_CALL_ARG_LOCATION.
11389 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11390 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11391 Use copy_rtx_if_shared.
11392 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11393 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11394 (dwarf2out_var_location): Remove handling of
11395 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11397 * final.c (final_scan_insn): Remove all references to
11398 NOTE_INSN_CALL_ARG_LOCATION.
11399 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11400 before dumping final insns.
11401 * except.c (emit_note_eh_region_end): Remove all references to
11402 NOTE_INSN_CALL_ARG_LOCATION.
11403 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11404 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11405 * config/arc/arc.c (hwloop_optimize): Likewise.
11406 * config/arm/arm.c (create_fix_barrier): Likewise.
11407 * config/s390/s390.c (s390_chunkify_start): Likewise.
11408 * config/sh/sh.c (find_barrier): Likewise.
11409 * config/i386/i386.c (rest_of_insert_endbranch,
11410 ix86_seh_fixup_eh_fallthru): Likewise.
11411 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11412 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11413 * config/frv/frv.c (frv_function_prologue): Likewise.
11414 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11416 (note_outside_basic_block_p): Remove all references to
11417 NOTE_INSN_CALL_ARG_LOCATION.
11418 * gengtype.c (adjust_field_rtx_def): Likewise.
11419 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11421 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11422 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11425 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11428 PR tree-optimization/84452
11429 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11430 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11433 2018-02-19 Martin Liska <mliska@suse.cz>
11436 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11438 2018-02-19 Martin Liska <mliska@suse.cz>
11439 Richard Sandiford <richard.sandiford@linaro.org>
11441 PR tree-optimization/82491
11442 * gimple-fold.c (get_base_constructor): Make earlier bail out
11445 2018-02-19 Carl Love <cel@us.ibm.com>
11447 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11448 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11450 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11451 P8V_BUILTIN_VEC_NEG.
11453 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11455 * config/rl78/rl78.md (movdf): New define expand.
11457 2018-02-19 Martin Liska <mliska@suse.cz>
11460 * doc/invoke.texi: Fix typo.
11461 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11463 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11465 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11466 handle rs6000_single_float and rs6000_double_float specially for
11469 2018-02-16 Jeff Law <law@redhat.com>
11471 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11472 (pop_regs): Use it.
11474 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11477 * ipa-inline.c (inline_small_functions): Fix a typo.
11479 2018-02-16 Nathan Sidwell <nathan@acm.org>
11481 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11483 2018-02-16 Carl Love <cel@us.ibm.com>
11485 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11486 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11487 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11488 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11489 expansion to P8V_BUILTIN_VEC_FLOAT2.
11491 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11493 PR rtl-optimization/70023
11494 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11495 src_regno into account.
11497 2018-02-16 Carl Love <cel@us.ibm.com>
11499 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11500 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11501 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11502 * config/rs6000/rs6000.c: Remove case statements for
11503 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11504 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11505 and P9V_BUILTIN_VEC_VINSERT4B.
11506 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11507 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11508 * config/rs6000/vsx.md:
11509 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11512 2018-02-16 Carl Love <cel@us.ibm.com>
11514 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11516 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11518 * config/rs6000/rs6000-c.c: Add the definitions for
11519 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11520 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11521 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11522 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11523 definition for insert4b and define insn *insert3b_internal.
11524 * doc/extend.texi: Add documentation for vec_extract4b.
11526 2018-02-16 Nathan Sidwell <nathan@acm.org>
11528 * doc/extend.texi (Backwards Compatibility): Mention friend
11529 injection. Note for-scope is deprecated.
11530 * doc/invoke.texi (-ffriend-injection): Deprecate.
11532 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11534 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11535 that moved to I2, also allow destinations that are a paradoxical
11536 subreg (instead of a normal reg).
11538 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11541 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11544 2018-02-16 Richard Biener <rguenther@suse.de>
11546 PR tree-optimization/84037
11547 PR tree-optimization/84016
11549 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11550 Adjust vec_construct for the fact we need additional higher latency
11551 128bit inserts for AVX256 and AVX512 vector builds.
11552 (ix86_add_stmt_cost): Scale vector construction cost for
11555 2018-02-16 Richard Biener <rguenther@suse.de>
11557 PR tree-optimization/84417
11558 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11559 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11560 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11562 2018-02-16 Martin Liska <mliska@suse.cz>
11565 * internal-fn.def (ASAN_CHECK): Set proper flags.
11566 (ASAN_MARK): Likewise.
11568 2018-02-16 Julia Koval <julia.koval@intel.com>
11570 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11571 from PTA_CANNONLAKE.
11573 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11576 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11577 Use ++iter rather than iter++ for std::list iterators.
11578 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11579 defer deleting them until all nodes in the forest are processed. Do
11580 free even leaf nodes. Change to_process into auto_vec.
11583 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11584 * vec.h (vec_default_construct): Use memset instead of placement new
11585 if BROKEN_VALUE_INITIALIZATION is defined.
11586 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11587 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11590 PR rtl-optimization/83723
11591 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11592 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11593 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11595 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11597 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11599 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11601 PR rtl-optimization/81443
11602 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11603 from inner REGs to paradoxical SUBREGs.
11605 2018-02-16 Richard Biener <rguenther@suse.de>
11607 PR tree-optimization/84399
11608 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11609 For operands we can analyze at their definition make sure we can
11610 analyze them at each use as well.
11612 2018-02-16 Richard Biener <rguenther@suse.de>
11614 PR tree-optimization/84190
11615 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11616 volatile accesses if the decl isn't volatile.
11618 2018-02-15 Jason Merrill <jason@redhat.com>
11620 PR c++/84314 - ICE with templates and fastcall attribute.
11621 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11622 TYPE_CANONICAL on an existing type.
11624 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11626 PR tree-optimization/84383
11627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11628 dstoff nor call operand_equal_p if dstbase is NULL.
11630 PR tree-optimization/84334
11631 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11632 also a CONSTANT_CLASS_P, punt.
11634 2018-02-14 Jim Wilson <jimw@sifive.com>
11636 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11637 first SMALL_OPERAND check. New local min_second_step. Move assert
11638 to where locals are set. Add TARGET_RVC support.
11639 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11641 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11643 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11645 2018-02-14 Martin Sebor <msebor@redhat.com>
11647 PR tree-optimization/83698
11648 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11649 arrays constrain the offset range to their bounds.
11650 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11651 (builtin_access::overlap): Avoid setting the size of overlap if it's
11653 (maybe_diag_overlap): Also consider arrays when deciding what values
11654 of offsets to include in diagnostics.
11656 2018-02-14 Martin Sebor <msebor@redhat.com>
11659 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11660 that correspond to the kind of a declaration.
11662 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11665 * config/pa/pa.md: Load address of PIC label using the linkage table
11666 if the label is nonlocal.
11668 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11670 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11671 warning message if user requests -maltivec=be.
11672 * doc/invoke.texi: Document deprecation of -maltivec=be.
11674 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11677 * config/rs6000/rs6000-c.c: Update definitions for
11678 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11679 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11681 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11684 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11685 add _get_ssp intrinsics. Remove argument from
11686 __builtin_ia32_rdssp[d|q].
11687 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11688 * config/i386/i386-builtin.def: Remove argument from
11689 __builtin_ia32_rdssp[d|q].
11690 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11691 ix86_expand_special_args_builtin for _rdssp[d|q].
11692 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11693 Clear register before usage.
11694 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11695 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11697 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11699 PR tree-optimization/84357
11700 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11701 operand 1 of an ARRAY_REF too.
11703 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11706 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11707 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11709 (set_of_reg): New struct.
11710 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11711 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11712 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11714 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11715 Split into bitclr, bitset, bitinvert patterns if appropriate.
11716 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11717 use rx_fuse_in_memory_bitop.
11718 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11719 to named insn, correct maximum insn length.
11721 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11724 * machmode.def: Define a complex mode for PARTIAL_INT.
11725 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11727 * doc/rtl.texi: Document CSPImode.
11728 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11730 (msp430_hard_regno_nregs_with_padding): Likewise.
11732 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11735 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11737 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11739 PR rtl-optimization/84169
11740 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11741 we generated a parallel as new i3 and we split that to new i2 and i3
11742 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11743 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11744 those to i2, not i1. Partially rewrite this scan code.
11746 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11749 * stor-layout.c (place_field): For variable length fields, adjust
11750 offset_align afterwards not just based on the field's alignment,
11751 but also on the size.
11753 PR middle-end/84309
11754 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11755 of exps and logs in the use_exp2 case.
11757 2018-02-13 Jeff Law <law@redhat.com>
11759 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11760 entry for "vector".
11762 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11765 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11769 * common.opt (gas-loc-support, gas-locview-support): New.
11770 (ginline-points, ginternal-reset-location-views): New.
11771 * doc/invoke.texi: Document them. Use @itemx where intended.
11772 (gvariable-location-views): Adjust.
11773 * target.def (reset_location_view): New.
11774 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11775 (TARGET_RESET_LOCATION_VIEW): New.
11776 * doc/tm.texi: Rebuilt.
11777 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11778 (dwarf2out_default_as_locview_support): New.
11779 (output_asm_line_debug_info): Use option variables.
11780 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11781 (output_loc_list): Likewise.
11782 (add_high_low_attributes): Check option variables.
11783 Don't output entry view attribute in strict mode.
11784 (gen_inlined_subroutine_die): Check option variables.
11785 (dwarf2out_inline_entry): Likewise.
11786 (init_sections_and_labels): Likewise.
11787 (dwarf2out_early_finish): Likewise.
11788 (maybe_reset_location_view): New, from...
11789 (dwarf2out_var_location): ... here. Call it.
11790 * debug.h (dwarf2out_default_as_loc_support): Declare.
11791 (dwarf2out_default_as_locview_support): Declare.
11792 * hooks.c (hook_int_rtx_insn_0): New.
11793 * hooks.h (hook_int_rtx_insn_0): Declare.
11794 * toplev.c (process_options): Take -gas-loc-support and
11795 -gas-locview-support from dwarf2out. Enable
11796 -gvariable-location-views by default only with locview
11797 assembler support. Enable -ginternal-reset-location-views by
11798 default only if the target defines the corresponding hook.
11799 Enable -ginline-points by default if location views are
11800 enabled; force it disabled if statement frontiers are
11802 * tree-inline.c (expand_call_inline): Check option variables.
11803 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11805 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11807 PR tree-optimization/84321
11808 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11809 handling. Also check whether the anti-range contains any values
11810 that satisfy the mask; switch to a VR_RANGE if not.
11812 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11815 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11817 2018-02-13 Martin Jambor <mjambor@suse.cz>
11820 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11821 of call statements, also set location of a load to a temporary.
11823 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11825 * config/rl78/rl78.c (add_vector_labels): New function.
11826 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11827 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11828 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11829 which checks that no arguments are passed.
11830 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11831 * doc/extend.texi: Documentation for the new attribute.
11833 2018-02-13 Andreas Schwab <schwab@suse.de>
11835 * config/riscv/linux.h (CPP_SPEC): Define.
11837 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11840 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11841 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11842 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11843 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11844 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11845 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11846 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11847 temporarily for AES and PCLMUL builtins.
11849 PR tree-optimization/84339
11850 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11851 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11854 PR middle-end/84309
11855 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11856 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11857 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11859 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11861 * omp-simd-clone.h: New file.
11862 * omp-simd-clone.c: Include omp-simd-clone.h.
11863 (expand_simd_clones): No longer static.
11864 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11865 cgraph.h and omp-simd-clone.h.
11866 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11867 (vect_recog_widen_shift_pattern): Formatting fix.
11868 (vect_pattern_recog_1): Don't check optab for calls.
11871 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11872 operands[2] into a REG before using gen_lowpart on it.
11874 2018-02-12 Jeff Law <law@redhat.com>
11877 * config/sh/sh.c (find_barrier): Consider a sibling call
11880 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11881 successfully back substituting a reg.
11883 2018-02-12 Richard Biener <rguenther@suse.de>
11885 PR tree-optimization/84037
11886 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11887 parameter, move visited init to caller.
11888 (vect_slp_analyze_operations): Separate cost from validity
11889 check, initialize visited once for all instances.
11890 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11892 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11894 (vect_model_promotion_demotion_cost): Likewise.
11895 (vectorizable_bswap): Guard cost modeling with !slp_node
11896 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11898 (vectorizable_call): Likewise.
11899 (vectorizable_conversion): Likewise.
11900 (vectorizable_assignment): Likewise.
11901 (vectorizable_shift): Likewise.
11902 (vectorizable_operation): Likewise.
11903 (vectorizable_store): Likewise.
11904 (vectorizable_load): Likewise.
11905 (vectorizable_condition): Likewise.
11906 (vectorizable_comparison): Likewise.
11908 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11911 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11912 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11913 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11915 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11917 PR middle-end/83665
11918 * params.def (inline-min-speedup): Increase from 8 to 15.
11919 (max-inline-insns-auto): Decrease from 40 to 30.
11920 * ipa-split.c (consider_split): Add some buffer for function to
11921 be considered inlining candidate.
11922 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11925 2018-02-12 Richard Biener <rguenther@suse.de>
11927 PR tree-optimization/84037
11928 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11929 matched stmts if we cannot swap the non-matched ones.
11931 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11933 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11934 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11935 _mm_maskz_scalef_round_ss): New intrinsics.
11936 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11937 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11938 __builtin_ia32_scalefss_round): Remove.
11939 (__builtin_ia32_scalefsd_mask_round,
11940 __builtin_ia32_scalefss_mask_round): New intrinsics.
11941 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11942 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11943 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11944 "<round_constraint>")): Changed to ...
11945 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11946 "<round_scalar_constraint>")): ... this.
11947 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11948 %0, %1, %2<round_op3>}"): Changed to ...
11949 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11950 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11951 %2<round_scalar_mask_op3>}"): ... this.
11952 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11954 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11956 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11957 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11958 (_mm_maskz_sqrt_round_ss): New intrinsics.
11959 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11960 (__builtin_ia32_sqrtsd_mask_round)
11961 (__builtin_ia32_sqrtss_mask_round): New builtins.
11962 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11963 (__builtin_ia32_sqrtss_round): Remove.
11964 (__builtin_ia32_sqrtsd_mask_round)
11965 (__builtin_ia32_sqrtss_mask_round): New builtins.
11966 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11967 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11968 ((match_operand:VF_128 1 "vector_operand"
11969 "xBm,<round_constraint>")): Changed to ...
11970 ((match_operand:VF_128 1 "vector_operand"
11971 "xBm,<round_scalar_constraint>")): ... this.
11972 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11973 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11974 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11975 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11976 %<iptr>1<round_scalar_mask_op3>}): ... this.
11977 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11978 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11980 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11983 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11984 Cast vec_cmpeq result to correct type.
11985 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11986 Cast vec_cmpgt result to correct type.
11988 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11990 * final.c (final_scan_insn_1): Renamed from...
11991 (final_scan_insn): ... this. New wrapper, to recover
11992 seen from the outermost call in recursive ones.
11993 * config/sparc/sparc.c (output_return): Drop seen from call.
11994 (output_sibcall): Likewise.
11995 * config/visium/visium.c (output_branch): Likewise.
11997 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11999 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12002 2018-02-10 Alan Modra <amodra@gmail.com>
12005 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12006 Specify LR as an input.
12008 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12011 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12012 remove_member_access_dummy_vars): New functions.
12013 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12014 lower_omp_1, execute_lower_omp): Use them.
12016 PR rtl-optimization/84308
12017 * shrink-wrap.c (spread_components): Release todo vector.
12019 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12021 PR rtl-optimization/57193
12022 * ira-color.c (struct allocno_color_data): Add member
12023 conflict_allocno_hard_prefs.
12024 (update_conflict_allocno_hard_prefs): New.
12025 (bucket_allocno_compare_func): Add a preference based on
12026 conflict_allocno_hard_prefs.
12027 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12028 (color_allocnos): Remove a dead code. Initiate
12029 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12031 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12034 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12035 constraint from =wa to wa. Avoid a subreg on the output operand,
12036 instead use a pseudo and subreg it in a move.
12037 (p9_xxbrd_<mode>): Changed to ...
12038 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12039 (p9_xxbrd_v2df): New expander.
12040 (p9_xxbrw_<mode>): Changed to ...
12041 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12042 (p9_xxbrw_v4sf): New expander.
12044 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12046 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12048 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12051 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12052 multiply in 32-bit mode.
12053 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12054 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12057 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12059 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12060 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12061 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12062 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12064 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12067 * dwarf2out.c (is_trivial_indirect_ref): New function.
12068 (dwarf2out_late_global_decl): Do not generate a location
12069 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12070 and that are not defined in the current unit.
12072 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12074 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12075 instead of a libcall for UNORDERED.
12077 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12080 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12081 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12083 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12086 * config/s390/s390.c (s390_set_current_function): Invoke
12087 s390_indirect_branch_settings also if fndecl didn't change.
12089 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12091 * config/rs6000/rs6000.md (blockage): Set length to zero.
12093 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12095 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12097 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12100 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12101 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12105 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12106 PARALLEL incoming that failed vt_get_decl_and_offset check.
12108 PR middle-end/84237
12109 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12110 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12112 (get_variable_section): For decls in named .bss* sections pass true as
12113 second argument to bss_initializer_p.
12115 2018-02-09 Marek Polacek <polacek@redhat.com>
12116 Jakub Jelinek <jakub@redhat.com>
12119 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12120 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12121 Sync some changes from cxx_fold_indirect_ref.
12123 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12125 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12127 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12128 (BLOCK_INLINE_ENTRY_LABEL): New.
12129 (dwarf2out_var_location): Disregard inline entry markers.
12130 (inline_entry_data): New struct.
12131 (inline_entry_data_hasher): New hashtable type.
12132 (inline_entry_data_hasher::hash): New.
12133 (inline_entry_data_hasher::equal): New.
12134 (inline_entry_data_table): New variable.
12135 (add_high_low_attributes): Add DW_AT_entry_pc and
12136 DW_AT_GNU_entry_view attributes if a pending entry is found
12137 in inline_entry_data_table. Add old entry_pc attribute only
12138 if debug nonbinding markers are disabled.
12139 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12140 markers are enabled.
12141 (block_within_block_p, dwarf2out_inline_entry): New.
12142 (dwarf2out_finish): Check that no entries remained in
12143 inline_entry_data_table.
12144 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12145 (final_scan_insn, notice_source_line): Likewise.
12146 (rest_of_clean_state): Skip inline entry markers.
12147 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12149 * gimple.c (gimple_build_debug_inline_entry): New.
12150 * gimple.h (enum gimple_debug_subcode): Add
12151 GIMPLE_DEBUG_INLINE_ENTRY.
12152 (gimple_build_debug_inline_entry): Declare.
12153 (gimple_debug_inline_entry_p): New.
12154 (gimple_debug_nonbind_marker_p): Adjust.
12155 * insn-notes.def (INLINE_ENTRY): New.
12156 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12157 inline entry marker notes.
12158 (print_insn): Likewise.
12159 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12160 (INSN_DEBUG_MARKER_KIND): Likewise.
12161 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12162 * tree-inline.c (expand_call_inline): Build and insert
12163 debug_inline_entry stmt.
12164 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12165 inline entry blocks early, if nonbind markers are enabled.
12166 (dump_scope_block): Dump fragment info.
12167 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12168 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12169 (gimple_build_debug_inline_entry): New.
12170 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12171 Enable/disable inline entry points too.
12172 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12173 (DEBUG_INSN): Describe inline entry markers.
12175 * common.opt (gvariable-location-views): New.
12176 (gvariable-location-views=incompat5): New.
12177 * config.in: Rebuilt.
12178 * configure: Rebuilt.
12179 * configure.ac: Test assembler for view support.
12180 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12181 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12182 * dwarf2out.c (var_loc_view): New typedef.
12183 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12184 (dwarf2out_locviews_in_attribute): New.
12185 (dwarf2out_locviews_in_loclist): New.
12186 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12187 (enum dw_line_info_opcode): Add LI_adv_address.
12188 (struct dw_line_info_table): Add view.
12189 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12190 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12191 (zero_view_p): New variable.
12192 (ZERO_VIEW_P): New macro.
12193 (output_asm_line_debug_info): New.
12194 (struct var_loc_node): Add view.
12195 (add_AT_view_list, AT_loc_list): New.
12196 (add_var_loc_to_decl): Add view param. Test it against last.
12197 (new_loc_list): Add view params. Record them.
12198 (AT_loc_list_ptr): Handle loc and view lists.
12199 (view_list_to_loc_list_val_node): New.
12200 (print_dw_val): Handle dw_val_class_view_list.
12201 (size_of_die): Likewise.
12202 (value_format): Likewise.
12203 (loc_list_has_views): New.
12204 (gen_llsym): Set vl_symbol too.
12205 (maybe_gen_llsym, skip_loc_list_entry): New.
12206 (dwarf2out_maybe_output_loclist_view_pair): New.
12207 (output_loc_list): Output view list or entries too.
12208 (output_view_list_offset): New.
12209 (output_die): Handle dw_val_class_view_list.
12210 (output_dwarf_version): New.
12211 (output_compilation_unit_header): Use it.
12212 (output_skeleton_debug_sections): Likewise.
12213 (output_rnglists, output_line_info): Likewise.
12214 (output_pubnames, output_aranges): Update version comments.
12215 (output_one_line_info_table): Output view numbers in asm comments.
12216 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12217 Call maybe_gen_llsym.
12218 (loc_list_from_tree_1): Adjust.
12219 (add_AT_location_description): Create view list attribute if
12220 needed, check it's absent otherwise.
12221 (convert_cfa_to_fb_loc_list): Adjust.
12222 (maybe_emit_file): Call output_asm_line_debug_info for test.
12223 (dwarf2out_var_location): Reset views as needed. Precompute
12224 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12225 attribute. Set view.
12226 (new_line_info_table): Reset next view.
12227 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12228 (dwarf2out_source_line): Likewise. Output view resets and labels to
12229 the assembler, or select appropriate line info opcodes.
12230 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12231 (optimize_string_length): Catch it. Adjust.
12232 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12233 dw_val_class_view_list, and remove it if no longer needed.
12234 (hash_loc_list): Hash view numbers.
12235 (loc_list_hasher::equal): Compare them.
12236 (optimize_location_lists): Check whether a view list symbol is
12237 needed, and whether the locview attribute is present, and
12238 whether they match. Remove the locview attribute if no longer
12240 (index_location_lists): Call skip_loc_list_entry for test.
12241 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12242 Use output_dwarf_version.
12243 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12244 (struct dw_val_node): Add val_view_list.
12245 * final.c (SEEN_NEXT_VIEW): New.
12246 (set_next_view_needed): New.
12247 (clear_next_view_needed): New.
12248 (maybe_output_next_view): New.
12249 (final_start_function): Rename to...
12250 (final_start_function_1): ... this. Take pointer to FIRST,
12251 add SEEN parameter. Emit param bindings in the initial view.
12252 (final_start_function): Reintroduce SEEN-less interface.
12253 (final): Rename to...
12254 (final_1): ... this. Take SEEN parameter. Output final pending
12255 next view at the end.
12256 (final): Reintroduce seen-less interface.
12257 (final_scan_insn): Output pending next view before switching
12258 sections or ending a block. Mark the next view as needed when
12259 outputting variable locations. Notify debug backend of section
12260 changes, and of location view changes.
12261 (rest_of_handle_final): Adjust.
12262 * toplev.c (process_options): Autodetect value for debug variable
12263 location views option. Warn on incompat5 without -gdwarf-5.
12264 * doc/invoke.texi (gvariable-location-views): New.
12265 (gvariable-location-views=incompat5): New.
12266 (gno-variable-location-views): New.
12268 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12270 PR tree-optimization/84136
12271 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12272 that the result of find_edge is non-NULL.
12274 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12277 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12278 storing integer register in SImode. Fix cost of 256 and 512
12279 byte aligned SSE register store.
12281 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12283 * config/i386/i386.c (ix86_multiplication_cost): Fix
12284 multiplication cost for TARGET_AVX512DQ.
12286 2018-02-08 Marek Polacek <polacek@redhat.com>
12288 PR tree-optimization/84238
12289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12294 PR tree-optimization/84265
12295 * tree-vect-stmts.c (vectorizable_store): Don't treat
12296 VMAT_CONTIGUOUS accesses as grouped.
12297 (vectorizable_load): Likewise.
12299 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12301 PR tree-optimization/81635
12302 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12303 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12304 (test_round_for_mask): New functions.
12305 (wide_int_cc_tests): Call test_round_for_mask.
12306 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12307 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12308 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12309 range returned by get_range_info.
12311 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12314 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12315 * symtab.c: Include builtins.h
12316 (symtab_node::output_to_lto_symbol_table_p): Move here
12317 from lto-streamer-out.c:output_symbol_p.
12318 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12319 (output_symbol_p): Move all logic to symtab.c
12320 (produce_symtab): Update.
12322 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12324 * config/s390/s390-opts.h (enum indirect_branch): Define.
12325 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12326 (s390_indirect_branch_via_thunk)
12327 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12328 (enum s390_indirect_branch_type): Define.
12329 * config/s390/s390.c (struct s390_frame_layout, struct
12330 machine_function): Remove.
12331 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12332 (indirect_branch_table_label_no, indirect_branch_table_name):
12334 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12335 (enum s390_indirect_branch_option): Define.
12336 (s390_return_addr_from_memory): New function.
12337 (s390_handle_string_attribute): New function.
12338 (s390_attribute_table): Add new attribute handler.
12339 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12340 (s390_indirect_branch_via_thunk): New function.
12341 (s390_indirect_branch_via_inline_thunk): New function.
12342 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12343 sibling call optimization for non z10 compiles.
12344 (s390_emit_call): Force indirect branch target to be a single
12345 register. Add r1 clobber for non-z10 compiles.
12346 (s390_emit_epilogue): Emit return jump via return_use expander.
12347 (s390_reorg): Handle JUMP_INSNs as execute targets.
12348 (s390_option_override_internal): Perform validity checks for the
12349 new command line options.
12350 (s390_indirect_branch_attrvalue): New function.
12351 (s390_indirect_branch_settings): New function.
12352 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12353 (s390_output_indirect_thunk_function): New function.
12354 (s390_code_end): Implement target hook.
12355 (s390_case_values_threshold): Implement target hook.
12356 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12358 * config/s390/s390.h (struct s390_frame_layout)
12359 (struct machine_function): Move here from s390.c.
12360 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12361 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12362 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12363 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12364 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12365 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12366 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12367 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12368 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12369 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12370 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12371 (mnemonic attribute): Add values which aren't recognized
12373 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12374 pattern for branch conversion. Fix mnemonic attribute.
12375 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12376 indirect branch via thunk if requested.
12377 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12378 ("*indirect_jump"): Disable for branch conversion using out of
12380 ("indirect_jump_via_thunk<mode>_z10")
12381 ("indirect_jump_via_thunk<mode>")
12382 ("indirect_jump_via_inlinethunk<mode>_z10")
12383 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12384 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12385 ("casesi_jump_via_inlinethunk<mode>_z10")
12386 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12387 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12388 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12389 ("*indirect2_jump"): Disable for branch conversion.
12390 ("casesi_jump"): Turn into expander and expand patterns for branch
12392 ("return_use"): New expander.
12393 ("*return"): Emit return via thunk and rename it to ...
12394 ("*return<mode>"): ... this one.
12395 * config/s390/s390.opt: Add new options and and enum for the
12398 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12400 * lra-constraints.c (match_reload): Unconditionally use
12401 gen_lowpart_SUBREG, rather than selecting between that
12402 and equivalent gen_rtx_SUBREG code.
12404 2018-02-08 Richard Biener <rguenther@suse.de>
12406 PR tree-optimization/84233
12407 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12408 changed flag instead of boguously re-using phi_inserted.
12410 2018-02-08 Martin Jambor <mjambor@suse.cz>
12412 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12413 static local variables.
12415 2018-02-08 Richard Biener <rguenther@suse.de>
12417 PR tree-optimization/84278
12418 * tree-vect-stmts.c (vectorizable_store): When looking for
12419 smaller vector types to perform grouped strided loads/stores
12420 make sure the mode is supported by the target.
12421 (vectorizable_load): Likewise.
12423 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12425 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12426 Increase LDP/STP opportunities by adding adjacent callee-saves.
12428 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12430 PR rtl-optimization/84068
12431 PR rtl-optimization/83459
12432 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12434 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12436 PR tree-optimization/84224
12437 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12438 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12439 non-zero arguments.
12441 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12444 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12445 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12446 count, remove one USE.
12448 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12450 * doc/install.texi (Configuration): Document the
12451 --with-long-double-format={ibm,ieee} PowerPC configuration
12455 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12456 Convert from define_expand to be define_insn_and_split. Rework
12457 float/double/_Float128 conversions to QI/HI/SImode to work with
12458 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12459 conversions to QI/HImode types did a store and then a load to
12460 truncate the value. For conversions to VSX registers, don't split
12461 the insn, instead emit the code directly. Use the code iterator
12462 any_fix to combine signed and unsigned conversions.
12463 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12464 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12465 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12466 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12467 (fix_<mode>di2_hw): Likewise.
12468 (fixuns_<mode>di2_hw): Likewise.
12469 (fix_<mode>si2_hw): Likewise.
12470 (fixuns_<mode>si2_hw): Likewise.
12471 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12472 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12473 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12474 fix<uns>_trunc<SFDF:mode>si2_p8.
12475 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12476 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12477 (fix<uns>_<mode>_mem): Likewise.
12478 (fctiw<u>z_<mode>_mem): Likewise.
12479 (fix<uns>_<mode>_mem): Likewise.
12480 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12481 the register allocator from doing a direct move to the GPRs to do
12482 a store, and instead use the ISA 3.0 store byte/half-word from
12483 vector register instruction. For IEEE 128-bit floating point,
12484 also optimize stores of 32-bit ints.
12485 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12487 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12489 * genextract.c (push_pathstr_operand): New function to support
12491 (walk_rtx): Call push_pathstr_operand.
12492 (print_path): Support [a-zA-Z].
12494 2018-02-07 Richard Biener <rguenther@suse.de>
12496 PR tree-optimization/84037
12497 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12498 (cse_and_gimplify_to_preheader): Declare.
12499 (vect_get_place_in_interleaving_chain): Likewise.
12500 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12502 (_loop_vec_info::~_loop_vec_info): Delete it.
12503 (cse_and_gimplify_to_preheader): New function.
12504 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12505 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12506 (vectorizable_load): Likewise. For grouped stores always base
12507 the IV on the first element.
12508 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12509 condition before gimplifying.
12511 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12513 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12514 *DIV_EXPR and *MOD_EXPR.
12516 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12519 * config/i386/i386.c (ix86_option_override_internal): Mask out
12520 the CF_SET bit when checking -fcf-protection.
12522 2018-02-07 Tom de Vries <tom@codesourcery.com>
12525 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12528 2018-02-07 Richard Biener <rguenther@suse.de>
12530 PR tree-optimization/84204
12531 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12534 PR tree-optimization/84205
12535 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12536 special-case isl_ast_op_zdiv_r.
12538 PR tree-optimization/84223
12539 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12540 Only add conditions from within the region.
12541 (gather_bbs::after_dom_children): Adjust.
12543 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12546 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12547 * config/avr/avr.md: Only post-reload split REG-REG moves if
12548 either register is GENERAL_REG_P.
12550 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12552 PR tree-optimization/84235
12553 * tree-ssa-scopedtables.c
12554 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12555 if the subtraction is performed in floating point type where NaNs are
12556 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12557 build 1. Formatting fix.
12559 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12562 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12563 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12564 and skip it regardless of bb boundaries. Use CALL_P macro,
12565 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12566 unnecessarily, formatting fix.
12568 2018-02-06 Michael Collison <michael.collison@arm.com>
12570 * config/arm/thumb2.md:
12571 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12572 (*thumb_mov_notscc): Ditto.
12574 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12577 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12578 unsigned_fix, not "s".
12580 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12582 * configure.ac (gcc_fn_eh_frame_ro): New function.
12583 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12584 correct .eh_frame permissions.
12585 * configure: Regenerate.
12587 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12589 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12590 irrelevant options.
12592 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12594 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12595 Display warning message for -mno-speculate-indirect-jumps.
12597 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12599 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12601 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12603 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12605 PR tree-optimization/84225
12606 * tree-eh.c (find_trapping_overflow): Only call
12607 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12609 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12612 * config/i386/i386.c: Reimplement the check of possible options
12613 -mibt/-mshstk conbination. Change error messages.
12614 * doc/invoke.texi: Fix a typo: remove extra '='.
12616 2018-02-06 Marek Polacek <polacek@redhat.com>
12618 PR tree-optimization/84228
12619 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12621 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12624 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12625 emitted arch directives.
12626 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12627 __ARM_FEATURE_COPROC before changing architectures.
12629 2018-02-06 Richard Biener <rguenther@suse.de>
12631 * config/i386/i386.c (print_reg): Fix typo.
12632 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12634 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12636 * configure: Regenerate.
12638 2018-02-05 Martin Sebor <msebor@redhat.com>
12640 PR tree-optimization/83369
12641 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12644 2018-02-05 Martin Liska <mliska@suse.cz>
12646 * doc/invoke.texi: Cherry-pick upstream r323995.
12648 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12650 * ira.c (ira_init_register_move_cost): Adjust comment.
12652 2018-02-05 Martin Liska <mliska@suse.cz>
12654 PR gcov-profile/84137
12655 * doc/gcov.texi: Fix typo in documentation.
12657 2018-02-05 Martin Liska <mliska@suse.cz>
12659 PR gcov-profile/83879
12660 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12661 using dlopen functionality.
12663 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12665 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12666 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12667 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12668 _mm_maskz_range_round_ss): New intrinsics.
12669 (__builtin_ia32_rangesd128_round)
12670 (__builtin_ia32_rangess128_round): Remove.
12671 (__builtin_ia32_rangesd128_mask_round,
12672 __builtin_ia32_rangess128_mask_round): New builtins.
12673 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12674 __builtin_ia32_rangess128_round): Remove.
12675 (__builtin_ia32_rangesd128_mask_round,
12676 __builtin_ia32_rangess128_mask_round): New builtins.
12677 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12678 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12679 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12680 "<round_saeonly_constraint>")): Changed to ...
12681 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12682 "<round_saeonly_scalar_constraint>")): ... this.
12683 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12684 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12685 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12686 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12687 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12689 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12691 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12693 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12694 Remove all values except native, 8540 and 8548.
12696 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12698 * config/i386/i386.c (ix86_output_function_return): Pass
12699 INVALID_REGNUM, instead of -1, as invalid register number to
12700 indirect_thunk_name and output_indirect_thunk.
12702 2018-02-02 Julia Koval <julia.koval@intel.com>
12704 * config.gcc: Add -march=icelake.
12705 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12706 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12707 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12708 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12709 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12710 (processor_target_table): Add icelake.
12711 (ix86_option_override_internal): Handle new PTAs.
12712 (get_builtin_code_for_version): Handle icelake.
12713 (M_INTEL_COREI7_ICELAKE): New.
12714 (fold_builtin_cpu): Handle icelake.
12715 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12716 * doc/invoke.texi: Add -march=icelake.
12718 2018-02-02 Julia Koval <julia.koval@intel.com>
12720 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12721 to wide_int_bitmask.
12722 * wide-int-bitmask.h: New.
12724 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12727 * config/i386/i386.md: Replace Pmode with word_mode in
12728 builtin_setjmp_setup and builtin_longjmp to support x32.
12730 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12734 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12736 (rs6000_supported_cpu_names): New static variable.
12737 (linux_cpu_translation_table): Likewise.
12738 (elf_platform) <cpu>: Define new static variable and use it.
12739 Translate kernel AT_PLATFORM name to canonical name if needed.
12740 Error if platform name is unknown.
12742 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12745 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12747 2018-02-01 Jeff Law <law@redhat.com>
12750 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12751 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12752 the scratch if RELEASE_VIA_POP is false.
12753 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12754 If we have to save a temporary register, decrement SIZE appropriately.
12755 Pass new arguments to release_scratch_register_on_entry.
12756 (ix86_adjust_stack_and_probe): Likewise.
12757 (ix86_emit_probe_stack_range): Pass new arguments to
12758 release_scratch_register_on_entry.
12760 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12762 PR rtl-optimization/84157
12763 * combine.c (change_zero_ext): Use REG_P predicate in
12764 front of HARD_REGISTER_P predicate.
12766 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12768 * config/avr/avr.c (avr_option_override): Move disabling of
12769 -fdelete-null-pointer-checks to...
12770 * common/config/avr/avr-common.c (avr_option_optimization_table):
12773 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12775 PR tree-optimization/81635
12776 * tree-data-ref.c (split_constant_offset_1): For types that
12777 wrap on overflow, try to use range info to prove that wrapping
12780 2018-02-01 Renlin Li <renlin.li@arm.com>
12783 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12784 TAILCALL_ADDR_REGS.
12785 (aarch64_register_move_cost): Likewise.
12786 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12787 TAILCALL_ADDR_REGS.
12788 (REG_CLASS_NAMES): Likewise.
12789 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12790 TAILCALL_ADDR_REGS. Remove IP registers.
12791 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12793 2018-02-01 Richard Biener <rguenther@suse.de>
12795 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12796 for specifying RPO order and allow NULL for that.
12797 * domwalk.c (dom_walker::dom_walker): Likewise.
12798 (dom_walker::walk): Handle NULL RPO order.
12799 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12801 (rewrite_update_dom_walker): Likewise.
12802 (mark_def_dom_walker): Likewise.
12804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12806 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12807 (aarch64_maybe_expand_sve_subreg_move): Declare.
12808 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12809 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12811 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12812 that are semantically a reverse operation.
12813 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12814 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12815 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12817 (aarch64_can_change_mode_class): For big-endian, forbid changes
12818 between two SVE modes if they have different element sizes.
12820 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12822 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12823 the TImode handling for big-endian targets.
12825 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12827 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12828 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12830 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12831 Remove BSWAP handing for big-endian targets and use the form of
12832 LD1RQ appropariate for the mode.
12834 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12836 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12837 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12838 duplicated element.
12840 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12843 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12844 check for operands that need to go through aarch64_sve_reload_be.
12846 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12848 PR tree-optimization/81661
12849 PR tree-optimization/84117
12850 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12851 * tree-eh.c: Include gimplify.h.
12852 (find_trapping_overflow, replace_trapping_overflow,
12853 rewrite_to_non_trapping_overflow): New functions.
12854 * tree-vect-loop.c: Include tree-eh.h.
12855 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12856 * tree-data-ref.c: Include tree-eh.h.
12857 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12859 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12861 PR rtl-optimization/84123
12862 * combine.c (change_zero_ext): Check if hard register satisfies
12863 can_change_dest_mode before calling gen_lowpart_SUBREG.
12865 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12868 * ira.c (ira_init_register_move_cost): Remove assert.
12870 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12872 PR rtl-optimization/84071
12873 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12874 * doc/tm.texi: Regenerate.
12876 2018-01-31 Richard Biener <rguenther@suse.de>
12878 PR tree-optimization/84132
12879 * tree-data-ref.c (analyze_miv_subscript): Properly
12880 check whether evolution_function_is_affine_multivariate_p
12881 before calling gcd_of_steps_may_divide_p.
12883 2018-01-31 Julia Koval <julia.koval@intel.com>
12886 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12887 * config/i386/i386.md (rdpid_rex64) New.
12888 (rdpid): Make 32bit only.
12890 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12893 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12894 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12896 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12899 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12901 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12903 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12905 PR rtl-optimization/84071
12906 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12907 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12909 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12911 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12912 (arc_attribute_table): Add 'aux' attribute.
12913 (arc_in_small_data_p): Consider aux like variables.
12914 (arc_is_aux_reg_p): New function.
12915 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12916 (arc_get_aux_arg): New function.
12917 (prepare_move_operands): Handle aux-register access.
12918 (arc_handle_aux_attribute): New function.
12919 * doc/extend.texi (ARC Variable attributes): Add subsection.
12921 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12923 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12924 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12925 (arc_attribute_table): Add 'uncached' attribute.
12926 (arc_print_operand): Print '.di' flag for uncached memory
12928 (arc_in_small_data_p): Do not consider for small data the uncached
12930 (arc_is_uncached_mem_p): New function.
12931 * config/arc/predicates.md (compact_store_memory_operand): Check
12932 for uncached memory accesses.
12933 (nonvol_nonimm_operand): Likewise.
12934 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12936 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12939 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12940 falign-loops=): Add Optimization flag.
12942 2018-01-30 Jeff Law <law@redhat.com>
12945 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12946 INT_REGISTERS_SAVED. Check it prior to calling
12947 get_scratch_register_on_entry.
12948 (ix86_adjust_stack_and_probe): Similarly.
12949 (ix86_emit_probe_stack_range): Similarly.
12950 (ix86_expand_prologue): Corresponding changes.
12952 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12955 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12956 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12958 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12961 * lra-constraints.c (curr_insn_transform): Process AND in the
12964 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12966 PR rtl-optimization/83986
12967 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12968 dependence against last_pending_memory_flush in addition to
12969 pending_jump_insns.
12971 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12973 PR tree-optimization/81611
12974 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12977 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12980 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12983 2018-01-30 Richard Biener <rguenther@suse.de>
12984 Jakub Jelinek <jakub@redhat.com>
12986 PR tree-optimization/84111
12987 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12988 inner loops added during recursion, as they don't have up-to-date
12991 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12994 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12995 (can_inline_edge_by_limits_p): ... here.
12996 (can_early_inline_edge_p, check_callers,
12997 update_caller_keys, update_callee_keys, recursive_inlining,
12998 add_new_edges_to_heap, speculation_useful_p,
12999 inline_small_functions,
13000 inline_small_functions, flatten_function,
13001 inline_to_all_callers_1): Update.
13003 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13005 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13008 2018-01-30 Richard Biener <rguenther@suse.de>
13010 PR tree-optimization/83008
13011 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13012 invariant and constant vector uses in stmts when they need
13013 more than one stmt.
13015 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13018 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13019 * configure: Regenerate.
13021 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13023 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13025 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13026 Use gen_rtx_REG rather than gen_lowpart.
13028 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13030 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13031 rather than 0 when creating partial subregs.
13033 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13035 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13038 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13041 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13042 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13043 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13044 flags. This restores the settings used before the 2017-07-24.
13045 Turning off pre increment/decrement/modify allows IVOPTS to
13046 optimize DF/SF loops where the index is an int.
13048 2018-01-29 Richard Biener <rguenther@suse.de>
13049 Kelvin Nilsen <kelvin@gcc.gnu.org>
13052 * tree-vect-stmts.c (vectorizable_call): Don't call
13053 targetm.vectorize_builtin_md_vectorized_function if callee is
13056 2018-01-22 Carl Love <cel@us.ibm.com>
13058 * doc/extend.tex: Fix typo in second arg in
13059 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13061 2018-01-29 Richard Biener <rguenther@suse.de>
13063 PR tree-optimization/84086
13064 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13065 (flush_ssaname_freelist): When SSA names were released reset
13066 the SCEV hash table.
13068 2018-01-29 Richard Biener <rguenther@suse.de>
13070 PR tree-optimization/84057
13071 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13072 removed paths when removing edges.
13074 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13076 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13077 -mfunction-return=@var{choice}.
13079 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13081 PR diagnostic/84034
13082 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13083 Handle CR like TAB.
13084 (layout::print_source_line): Likewise.
13085 (test_get_line_width_without_trailing_whitespace): Add test cases.
13087 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13089 PR middle-end/84040
13090 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13093 2018-01-26 Jim Wilson <jimw@sifive.com>
13095 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13097 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13100 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13102 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13103 and CMP + SUB-immediate -> SUBS.
13105 2018-01-26 Martin Sebor <msebor@redhat.com>
13107 PR tree-optimization/83896
13108 * tree-ssa-strlen.c (get_string_len): Rename...
13109 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13110 Avoid assuming length is constant.
13111 (handle_char_store): Use HOST_WIDE_INT for string length.
13113 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13116 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13117 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13119 2018-01-26 Richard Biener <rguenther@suse.de>
13121 PR rtl-optimization/84003
13122 * dse.c (record_store): Only record redundant stores when
13123 the earlier store aliases at least all accesses the later one does.
13125 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13127 PR rtl-optimization/83985
13128 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13129 REG_CFA_RESTORE insns.
13130 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13131 REG_CFA_RESTORE insns here.
13134 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13135 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13137 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13139 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13140 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13141 (arc_init): Likewise.
13142 (arc_override_options): Likewise.
13143 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13145 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13147 * config/arc/arc.h (TARGET_DBNZ): Define.
13148 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13149 properly set the tune attribute.
13150 (dbnz): Use TARGET_DBNZ guard.
13151 * config/arc/arc.opt (mtune): Add core3 option.
13153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13155 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13156 recognize new pic like addresses.
13157 (arc_delegitimize_address): Clean up.
13159 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13161 * config/arc/arc-arches.def: Option mrf16 valid for all
13163 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13164 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13165 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13166 * config/arc/arc-tables.opt: Regenerate.
13167 * config/arc/arc.c (arc_conditional_register_usage): Handle
13168 reduced register file case.
13169 (arc_file_start): Set must have build attributes.
13170 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13171 mrf16 option value.
13172 * config/arc/arc.opt (mrf16): Add new option.
13173 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13174 * config/arc/genmultilib.awk: Handle new mrf16 option.
13175 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13176 * config/arc/t-multilib: Regenerate.
13177 * doc/invoke.texi (ARC Options): Document mrf16 option.
13179 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13181 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13182 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13183 (arc_attribute_table): Add 'secure_call' attribute.
13184 (arc_print_operand): Print secure call operand.
13185 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13187 (arc_is_secure_call_p): New function. * config/arc/arc.md
13188 (call_i): Add support for sjli instruction.
13189 (call_value_i): Likewise.
13190 * config/arc/constraints.md (Csc): New constraint.
13192 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13193 John Eric Martin <John.Martin@emmicro-us.com>
13195 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13196 * config/arc/arc.c (_arc_jli_section): New struct.
13197 (arc_jli_section): New type.
13198 (rc_jli_sections): New static variable.
13199 (arc_handle_jli_attribute): New function.
13200 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13201 (arc_file_end): New function.
13202 (TARGET_ASM_FILE_END): Define.
13203 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13204 (arc_add_jli_section): New function.
13205 (jli_call_scan): Likewise.
13206 (arc_reorg): Call jli_call_scan.
13207 (arc_output_addsi): Remove 'S' from printing asm operand.
13208 (arc_is_jli_call_p): New function.
13209 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13211 (movhi_insn): Likewise.
13212 (movsi_insn): Likewise.
13213 (movsi_set_cc_insn): Likewise.
13214 (loadqi_update): Likewise.
13215 (load_zeroextendqisi_update): Likewise.
13216 (load_signextendqisi_update): Likewise.
13217 (loadhi_update): Likewise.
13218 (load_zeroextendhisi_update): Likewise.
13219 (load_signextendhisi_update): Likewise.
13220 (loadsi_update): Likewise.
13221 (loadsf_update): Likewise.
13222 (movsicc_insn): Likewise.
13223 (bset_insn): Likewise.
13224 (bxor_insn): Likewise.
13225 (bclr_insn): Likewise.
13226 (bmsk_insn): Likewise.
13227 (bicsi3_insn): Likewise.
13228 (cmpsi_cc_c_insn): Likewise.
13229 (movsi_ne): Likewise.
13230 (movsi_cond_exec): Likewise.
13231 (clrsbsi2): Likewise.
13232 (norm_f): Likewise.
13242 (call_i): Remove 'S' asm letter, add jli instruction.
13243 (call_value_i): Likewise.
13244 * config/arc/arc.op (mjli-always): New option.
13245 * config/arc/constraints.md (Cji): New constraint.
13246 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13248 (subsf3_fpx): Likewise.
13249 (mulsf3_fpx): Likewise.
13250 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13252 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13253 function attrbutes.
13254 * doc/invoke.texi (ARC): Document mjli-always option.
13256 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13258 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13259 avoid addition with 0 and use incw and decw where possible.
13261 2018-01-26 Richard Biener <rguenther@suse.de>
13263 PR tree-optimization/81082
13264 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13265 association if it requires casting to unsigned.
13266 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13267 from fold_plusminus_mult_expr to catch important cases late when
13268 range info is available.
13270 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13272 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13273 * configure.ac (hidden_linkonce): New test.
13274 * configure: Regenerate.
13275 * config.in: Regenerate.
13277 2018-01-26 Julia Koval <julia.koval@intel.com>
13279 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13280 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13281 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13282 _mm_mask_bitshuffle_epi64_mask): Fix type.
13283 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13284 USI_FTYPE_V4DI_V4DI_USI): Remove.
13285 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13286 __builtin_ia32_vpshufbitqmb256_mask,
13287 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13288 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13289 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13291 2018-01-26 Alan Modra <amodra@gmail.com>
13294 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13295 UNSPEC_VBPERMQ. Sort other unspecs.
13297 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13299 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13301 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13303 PR middle-end/83055
13304 * predict.c (drop_profile): Do not push/pop cfun; update also
13306 (handle_missing_profiles): Fix logic looking for zero profiles.
13308 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13310 PR middle-end/83977
13311 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13312 on functions with #pragma omp declare simd or functions with simd
13314 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13315 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13316 Remove trailing \n from warning_at calls.
13318 2018-01-25 Tom de Vries <tom@codesourcery.com>
13321 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13322 for neutered workers.
13324 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13327 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13328 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13330 2018-01-24 Jeff Law <law@redhat.com>
13333 * i386.c (get_probe_interval): Move to earlier point.
13334 (ix86_compute_frame_layout): If -fstack-clash-protection and
13335 the frame is larger than the probe interval, then use pushes
13336 to save registers rather than reg->mem moves.
13337 (ix86_expand_prologue): Remove conditional for int_registers_saved
13340 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13343 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13344 min/max for never referenced object.
13346 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13348 PR middle-end/83977
13349 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13351 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13352 attributes from DECL_ATTRIBUTES (decl) without affecting
13353 DECL_ATTRIBUTES (current_function_decl).
13354 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13355 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13357 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13359 PR tree-optimization/83979
13360 * fold-const.c (fold_comparison): Use constant_boolean_node
13361 instead of boolean_{true,false}_node.
13363 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13365 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13368 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13370 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13371 Simplify the clause that sets the length attribute.
13372 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13373 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13374 clause that sets the length attribute.
13375 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13377 2018-01-24 Tom de Vries <tom@codesourcery.com>
13380 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13381 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13382 Add strict parameter.
13383 (prevent_branch_around_nothing): Insert dummy insn between branch to
13384 label and label with no ptx insn inbetween.
13385 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13387 2018-01-24 Tom de Vries <tom@codesourcery.com>
13390 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13391 for neutered threads in warp.
13392 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13394 2018-01-24 Richard Biener <rguenther@suse.de>
13396 PR tree-optimization/83176
13397 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13400 2018-01-24 Richard Biener <rguenther@suse.de>
13402 PR tree-optimization/82819
13403 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13404 code generating pluses that are no-ops in the target precision.
13406 2018-01-24 Richard Biener <rguenther@suse.de>
13408 PR middle-end/84000
13409 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13411 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13413 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13414 to merge probabilities.
13415 * predict.c (probably_never_executed): Also mark as cold functions
13416 with global 0 profile and guessed local profile.
13417 * profile-count.c (profile_probability::combine_with_count): New
13419 * profile-count.h (profile_probability::operator*,
13420 profile_probability::operator*=, profile_probability::operator/,
13421 profile_probability::operator/=): Reduce precision to adjusted
13422 and set value to guessed on contradictory divisions.
13423 (profile_probability::combine_with_freq): Remove.
13424 (profile_probability::combine_wiht_count): Declare.
13425 (profile_count::force_nonzero):: Set to adjusted.
13426 (profile_count::probability_in):: Set quality to adjusted.
13427 * tree-ssa-tail-merge.c (replace_block_by): Use
13428 combine_with_count.
13430 2018-01-23 Andrew Waterman <andrew@sifive.com>
13431 Jim Wilson <jimw@sifive.com>
13433 * config/riscv/riscv.c (riscv_stack_boundary): New.
13434 (riscv_option_override): Set riscv_stack_boundary. Handle
13435 riscv_preferred_stack_boundary_arg.
13436 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13437 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13438 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13439 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13440 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13441 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13443 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13446 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13447 of struct ix86_frame.
13448 (ix86_expand_epilogue): Likewise. Add a local variable for
13449 the reg_save_offset field in struct ix86_frame.
13451 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13453 PR tree-optimization/82604
13454 * tree-loop-distribution.c (enum partition_kind): New enum item
13455 PKIND_PARTIAL_MEMSET.
13456 (partition_builtin_p): Support above new enum item.
13457 (generate_code_for_partition): Ditto.
13458 (compute_access_range): Differentiate cases that equality can be
13459 proven at all loops, the innermost loops or no loops.
13460 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13461 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13462 (finalize_partitions, distribute_loop): Don't fuse partition of
13463 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13464 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13465 parloop is enabled.
13467 2018-01-23 Martin Liska <mliska@suse.cz>
13469 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13470 order to ignore the predictor.
13471 (PRED_POLYMORPHIC_CALL): Likewise.
13472 (PRED_RECURSIVE_CALL): Likewise.
13474 2018-01-23 Martin Liska <mliska@suse.cz>
13476 * tree-profile.c (tree_profiling): Print function header to
13477 aware reader which function we are working on.
13478 * value-prof.c (gimple_find_values_to_profile): Do not print
13479 not interesting value histograms.
13481 2018-01-23 Martin Liska <mliska@suse.cz>
13483 * profile-count.h (enum profile_quality): Add
13484 profile_uninitialized as the first value. Do not number values
13485 as they are zero based.
13486 (profile_count::verify): Update sanity check.
13487 (profile_probability::verify): Likewise.
13489 2018-01-23 Nathan Sidwell <nathan@acm.org>
13491 * doc/invoke.texi (ffor-scope): Deprecate.
13493 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13495 PR tree-optimization/83510
13496 * domwalk.c (set_all_edges_as_executable): New function.
13497 (dom_walker::dom_walker): Convert bool param
13498 "skip_unreachable_blocks" to enum reachability. Move setup of
13499 edge flags to set_all_edges_as_executable and only do it when
13500 reachability is REACHABLE_BLOCKS.
13501 * domwalk.h (enum dom_walker::reachability): New enum.
13502 (dom_walker::dom_walker): Convert bool param
13503 "skip_unreachable_blocks" to enum reachability.
13504 (set_all_edges_as_executable): New decl.
13505 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13506 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13508 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13509 but converting true to REACHABLE_BLOCKS.
13510 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13512 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13513 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13514 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13516 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13517 if check_all_array_refs will be called.
13519 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13521 * tree.c (selftest::test_location_wrappers): Add more test
13524 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13526 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13527 (selftest::test_bit_in_range): Likewise.
13529 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13532 * doc/sourcebuild.texi (vect_float): Say that the selector
13533 only describes the situation when -funsafe-math-optimizations is on.
13534 (vect_float_strict): Document.
13536 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13538 PR tree-optimization/83965
13539 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13540 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13541 instead of checking only for a reduction.
13542 (vect_recog_widen_sum_pattern): Likewise.
13544 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13546 * predict.c (probably_never_executed): Only use precise profile info.
13547 (compute_function_frequency): Skip after inlining hack since we now
13548 have quality checking.
13550 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13552 * profile-count.h (profile_probability::very_unlikely,
13553 profile_probability::unlikely, profile_probability::even): Set
13554 precision to guessed.
13556 2018-01-23 Richard Biener <rguenther@suse.de>
13558 PR tree-optimization/83963
13559 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13560 Properly terminate dominator walk when crossing the exit edge not
13561 when visiting its source block.
13563 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13566 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13567 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13569 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13571 PR tree-optimization/83957
13572 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13573 semicolon after for body surrounded by braces.
13575 PR tree-optimization/83081
13576 * profile-count.h (profile_probability::split): New method.
13577 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13578 Use profile_probability::split.
13579 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13580 when splitting a single conditional jump into 2.
13582 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13584 PR tree-optimization/69452
13585 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13590 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13591 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13592 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13594 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13596 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13598 * config/rl78/rl78.md (movdi): New define_expand.
13599 * config/rl78/rl78.c (rl78_split_movdi): New function.
13601 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13604 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13606 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13607 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13608 128-bit to produce an UNSPEC move to get the double word with the
13609 signbit and then a shift directly to do signbit.
13610 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13611 implementation with a new version that just does either a direct
13612 move or a regular move. Move memory interface to separate insns.
13613 Move insns so they are next to the expander.
13614 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13615 with signbit move. Split big and little endian case.
13616 (signbit<mode>2_dm_mem_le): Likewise.
13617 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13618 (signbit<mode>2_dm2): Likewise.
13620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13622 * config/rl78/rl78.md (anddi3): New define_expand.
13624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13626 * config/rl78/rl78.md (umindi3): New define_expand.
13628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13630 * config/rl78/rl78.md (smindi3): New define_expand.
13632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13634 * config/rl78/rl78.md (smaxdi3): New define_expand.
13636 2018-01-22 Carl Love <cel@us.ibm.com>
13638 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13639 LVX_V1TI): Add macro expansion.
13640 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13641 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13642 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13643 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13644 Change check to determine if the instruction is a byte reversing
13645 entry. Fix typo in comment.
13646 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13647 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13648 Add def_builtin calls for new builtins.
13649 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13650 Add define_insn expansion.
13652 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13654 * config/rl78/rl78.md (umaxdi3): New define_expand.
13656 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13658 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13659 for non-QImode registers.
13661 2018-01-22 Richard Biener <rguenther@suse.de>
13663 PR tree-optimization/83963
13664 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13665 including the loop exit block.
13666 (scop_detection::merge_sese): Likewise.
13667 (scop_detection::add_scop): Do it here instead.
13669 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13671 * doc/sourcebuild.texi (arm_softfloat): Document.
13673 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13676 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13677 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13678 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13680 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13681 David Edelsohn <dje.gcc@gmail.com>
13684 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13685 Change "crset eq" to "crset 2".
13686 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13687 (*call_indirect_aix<mode>_nospec): Likewise.
13688 (*call_value_indirect_aix<mode>_nospec): Likewise.
13689 (*call_indirect_elfv2<mode>_nospec): Likewise.
13690 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13691 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13692 change assembly output from . to $.
13693 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13694 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13695 (*tablejump<mode>_internal1_nospec): Likewise.
13697 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13700 * config/sh/sh_optimize_sett_clrt.cc:
13701 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13703 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13705 PR tree-optimization/83940
13706 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13707 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13708 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13709 use it to pass back the definition type.
13710 (vect_check_store_rhs): Likewise rhs_dt_out.
13711 (vect_build_gather_load_calls): Add a mask_dt argument and use
13712 it instead of a call to vect_is_simple_use.
13713 (vectorizable_store): Update calls to vect_check_load_store_mask
13714 and vect_check_store_rhs. Use the dt returned by the latter instead
13715 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13716 instead of calls to vect_is_simple_use. Pass the scalar rather
13717 than the vector operand to vect_is_simple_use when handling
13718 second and subsequent copies of an rhs value.
13719 (vectorizable_load): Update calls to vect_check_load_store_mask
13720 and vect_build_gather_load_calls. Use the cached mask_dt and
13721 gs_info.offset_dt instead of calls to vect_is_simple_use.
13723 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13725 PR middle-end/83945
13726 * tree-emutls.c: Include gimplify.h.
13727 (lower_emutls_2): New function.
13728 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13729 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13730 it before further processing.
13733 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13734 UINTVAL (trueop1) instead of INTVAL (op1).
13736 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13740 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13741 INCOMING_FRAME_SP_OFFSET if not defined.
13742 (scan_trace): Add ENTRY argument. If true and
13743 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13744 emit a note to adjust the CFA offset.
13745 (create_cfi_notes): Adjust scan_trace callers.
13746 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13747 INCOMING_FRAME_SP_OFFSET in the CIE.
13748 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13749 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13751 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13752 * doc/tm.texi: Regenerated.
13754 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13756 PR rtl-optimization/83147
13757 * lra-constraints.c (remove_inheritance_pseudos): Use
13758 lra_substitute_pseudo_within_insn.
13760 2018-01-19 Tom de Vries <tom@codesourcery.com>
13761 Cesar Philippidis <cesar@codesourcery.com>
13764 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13766 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13769 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13770 spaces for function labels.
13772 2018-01-19 Martin Liska <mliska@suse.cz>
13774 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13775 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13776 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13777 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13778 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13779 (PRED_CONST_RETURN): Change from 69 to 65.
13780 (PRED_NULL_RETURN): Change from 91 to 71.
13781 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13782 (PRED_LOOP_GUARD): Change from 66 to 73.
13784 2018-01-19 Martin Liska <mliska@suse.cz>
13786 * predict.c (predict_insn_def): Add new assert.
13787 (struct branch_predictor): Change type to signed integer.
13788 (test_prediction_value_range): Amend test to cover
13789 PROB_UNINITIALIZED.
13790 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13791 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13792 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13793 (PRED_LOOP_IV_COMPARE): Likewise.
13794 * predict.h (PROB_UNINITIALIZED): Define new constant.
13796 2018-01-19 Martin Liska <mliska@suse.cz>
13798 * predict.c (dump_prediction): Add new format for
13799 analyze_brprob.py script which is enabled with -details
13801 * profile-count.h (precise_p): New function.
13803 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13805 PR tree-optimization/83922
13806 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13807 there are no statements that need masking.
13808 (vect_active_double_reduction_p): New function.
13809 (vect_analyze_loop_operations): Use it when handling phis that
13810 are not in the loop header.
13812 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13814 PR tree-optimization/83914
13815 * tree-vect-loop.c (vectorizable_induction): Don't convert
13816 init_expr or apply the peeling adjustment for inductions
13817 that are nested within the vectorized loop.
13819 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13821 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13824 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13828 * function.h (gimplify_parameters): Add gimple_seq * argument.
13829 * function.c: Include gimple.h and options.h.
13830 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13831 for the added local temporaries if needed.
13832 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13833 if there are any parameter cleanups, wrap whole body into a
13834 try/finally with the cleanups.
13836 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13839 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13840 Use GET_MODE_CLASS for scalar floating point.
13842 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13846 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13847 Fix call of call_cgraph_insertion_hooks.
13849 2018-01-18 Martin Sebor <msebor@redhat.com>
13851 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13853 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13856 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13859 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13862 * common.opt: (-ffile-prefix-map): New option.
13863 * opts.c (common_handle_option): Defer it.
13864 * opts-global.c (handle_common_deferred_options): Handle it.
13865 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13866 * file-prefix-map.h: New file.
13867 (remap_debug_filename, add_debug_prefix_map): ...here.
13868 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13869 * final.c (debug_prefix_map, add_debug_prefix_map
13870 remap_debug_filename): Move to...
13871 * file-prefix-map.c: New file.
13872 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13873 generalize, get rid of alloca(), use strrchr() instead of strchr().
13874 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13875 Implement in terms of add_prefix_map().
13876 (remap_macro_filename, remap_debug_filename): Implement in term of
13878 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13879 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13880 * dbxout.c: Include file-prefix-map.h.
13881 * varasm.c: Likewise.
13882 * vmsdbgout.c: Likewise.
13883 * xcoffout.c: Likewise.
13884 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13885 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13886 * doc/invoke.texi (-ffile-prefix-map): Document.
13887 (-fdebug-prefix-map): Update description.
13889 2018-01-18 Martin Liska <mliska@suse.cz>
13891 * config/i386/i386.c (indirect_thunk_name): Document that also
13893 (output_indirect_thunk): Document why both instructions
13894 (pause and lfence) are generated.
13896 2018-01-18 Richard Biener <rguenther@suse.de>
13898 PR tree-optimization/83887
13899 * graphite-scop-detection.c
13900 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13901 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13902 (scop_detection::merge_sese): Re-implement with a flood-fill
13903 algorithm that properly finds a SESE region if it exists.
13905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13908 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13909 pointer_diff optimizations use view_convert instead of convert.
13911 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13913 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13914 Generate different code for -mno-speculate-indirect-jumps.
13915 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13916 (*call_indirect_aix<mode>): Disable for
13917 -mno-speculate-indirect-jumps.
13918 (*call_indirect_aix<mode>_nospec): New define_insn.
13919 (*call_value_indirect_aix<mode>): Disable for
13920 -mno-speculate-indirect-jumps.
13921 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13922 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13923 -mno-speculate-indirect-jumps.
13924 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13926 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13928 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13929 long double type, set the flags for noting the default long double
13930 type, even if we don't pass or return a long double type.
13932 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13935 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13938 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13940 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13941 support for merge[hl].
13942 (fold_mergehl_helper): New helper function.
13943 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13944 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13945 (altivec_vmrglw_direct): Add xxmrglw insn.
13947 2018-01-17 Andrew Waterman <andrew@sifive.com>
13949 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13950 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13952 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13955 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13956 call the lto_location_cache before reading the
13957 DECL_SOURCE_LOCATION of the types.
13959 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13960 Richard Sandiford <richard.sandiford@linaro.org>
13962 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13963 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13964 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13965 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13966 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13968 * config/aarch64/constraints.md (aarch64_movti_operand):
13970 * config/aarch64/predicates.md (Uti): Add new constraint.
13972 2018-01-17 Carl Love <cel@us.ibm.com>
13974 * config/rs6000/vsx.md (define_expand xl_len_r,
13975 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13976 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13978 (define_expand, define_insn): Move the shift left from the
13979 define_insn to the define_expand for lxvl and stxvl instructions.
13980 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13981 and XL_LEN_R definitions to PURE.
13983 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13985 * config/i386/i386.c (indirect_thunk_name): Declare regno
13986 as unsigned int. Compare regno with INVALID_REGNUM.
13987 (output_indirect_thunk): Ditto.
13988 (output_indirect_thunk_function): Ditto.
13989 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13990 in the call to output_indirect_thunk_function.
13992 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13994 PR middle-end/83884
13995 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13996 rather than the size of inner_type to determine the stack slot size
13997 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13999 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14002 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14005 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14007 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14008 endian Linux systems to optionally enable multilibs for selecting
14009 the long double type if the user configured an explicit type.
14010 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14011 have no long double multilibs if not defined.
14012 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14013 warn if the user used -mabi={ieee,ibm}longdouble and we built
14014 multilibs for long double.
14015 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14016 appropriate multilib option.
14017 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14019 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14020 for building long double multilibs.
14021 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14023 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14025 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14028 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14030 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14033 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14036 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14039 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14041 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14042 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14044 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14046 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14047 different rtl trees depending on TARGET_64BIT.
14048 (rs6000_gen_lvx): Likewise.
14050 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14052 * config/visium/visium.md (nop): Tweak comment.
14053 (hazard_nop): Likewise.
14055 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14057 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14058 -mspeculate-indirect-jumps.
14059 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14060 for -mno-speculate-indirect-jumps.
14061 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14062 (*call_value_indirect_elfv2<mode>): Disable for
14063 -mno-speculate-indirect-jumps.
14064 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14065 (indirect_jump): Emit different RTL for
14066 -mno-speculate-indirect-jumps.
14067 (*indirect_jump<mode>): Disable for
14068 -mno-speculate-indirect-jumps.
14069 (*indirect_jump<mode>_nospec): New define_insn.
14070 (tablejump): Emit different RTL for
14071 -mno-speculate-indirect-jumps.
14072 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14073 (tablejumpsi_nospec): New define_expand.
14074 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14075 (tablejumpdi_nospec): New define_expand.
14076 (*tablejump<mode>_internal1): Disable for
14077 -mno-speculate-indirect-jumps.
14078 (*tablejump<mode>_internal1_nospec): New define_insn.
14079 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14082 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14084 * caller-save.c (insert_save): Drop unnecessary parameter. All
14087 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14088 Richard Biener <rguenth@suse.de>
14091 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14092 return early, inline manually is_gimple_sizepos. Make sure if we
14093 call gimplify_expr we don't end up with a gimple constant.
14094 * tree.c (variably_modified_type_p): Don't return true for
14095 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14096 * gimplify.h (is_gimple_sizepos): Remove.
14098 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14100 PR tree-optimization/83857
14101 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14102 vectorizable_live_operation for pure SLP statements.
14103 (vectorizable_live_operation): Handle PHIs.
14105 2018-01-16 Richard Biener <rguenther@suse.de>
14107 PR tree-optimization/83867
14108 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14109 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14111 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14114 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14115 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14116 If off is not INTEGER_CST, issue a may not be aligned warning
14117 rather than isn't aligned. Use isn%'t rather than isn't.
14118 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14120 <case MULT_EXPR>: Improve the case when bottom and one of the
14121 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14122 operand, in that case check if the other operand is multiple of
14123 bottom divided by the INTEGER_CST operand.
14125 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14128 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14129 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14130 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14131 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14132 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14133 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14134 (pa_function_arg_size): New function.
14136 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14138 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14139 in a separate statement.
14141 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14143 PR tree-optimization/83847
14144 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14145 group gathers and scatters.
14147 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14149 PR rtl-optimization/86620
14150 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14152 PR rtl-optimization/83213
14153 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14154 to last if both are JUMP_INSNs.
14156 PR tree-optimization/83843
14157 * gimple-ssa-store-merging.c
14158 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14159 store_immediate_info for bswap/nop orig_stores.
14161 2018-01-15 Andrew Waterman <andrew@sifive.com>
14163 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14165 <UDIV>: Increase cost if !TARGET_DIV.
14167 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14169 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14170 (define_attr "cr_logical_3op"): New.
14171 (cceq_ior_compare): Adjust.
14172 (cceq_ior_compare_complement): Adjust.
14173 (*cceq_rev_compare): Adjust.
14174 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14175 (is_cracked_insn): Adjust.
14176 (insn_must_be_first_in_group): Adjust.
14177 * config/rs6000/40x.md: Adjust.
14178 * config/rs6000/440.md: Adjust.
14179 * config/rs6000/476.md: Adjust.
14180 * config/rs6000/601.md: Adjust.
14181 * config/rs6000/603.md: Adjust.
14182 * config/rs6000/6xx.md: Adjust.
14183 * config/rs6000/7450.md: Adjust.
14184 * config/rs6000/7xx.md: Adjust.
14185 * config/rs6000/8540.md: Adjust.
14186 * config/rs6000/cell.md: Adjust.
14187 * config/rs6000/e300c2c3.md: Adjust.
14188 * config/rs6000/e500mc.md: Adjust.
14189 * config/rs6000/e500mc64.md: Adjust.
14190 * config/rs6000/e5500.md: Adjust.
14191 * config/rs6000/e6500.md: Adjust.
14192 * config/rs6000/mpc.md: Adjust.
14193 * config/rs6000/power4.md: Adjust.
14194 * config/rs6000/power5.md: Adjust.
14195 * config/rs6000/power6.md: Adjust.
14196 * config/rs6000/power7.md: Adjust.
14197 * config/rs6000/power8.md: Adjust.
14198 * config/rs6000/power9.md: Adjust.
14199 * config/rs6000/rs64.md: Adjust.
14200 * config/rs6000/titan.md: Adjust.
14202 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14204 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14205 ix86_indirect_branch_register logic.
14207 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14209 * config/i386/constraints.md (Bs): Update
14210 ix86_indirect_branch_register check. Don't check
14211 ix86_indirect_branch_register with GOT_memory_operand.
14213 * config/i386/predicates.md (GOT_memory_operand): Don't check
14214 ix86_indirect_branch_register here.
14215 (GOT32_symbol_operand): Likewise.
14217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14219 * config/i386/predicates.md (constant_call_address_operand):
14220 Rewrite ix86_indirect_branch_register logic.
14221 (sibcall_insn_operand): Likewise.
14223 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14225 * config/i386/constraints.md (Bs): Replace
14226 ix86_indirect_branch_thunk_register with
14227 ix86_indirect_branch_register.
14229 * config/i386/i386.md (indirect_jump): Likewise.
14230 (tablejump): Likewise.
14231 (*sibcall_memory): Likewise.
14232 (*sibcall_value_memory): Likewise.
14233 Peepholes of indirect call and jump via memory: Likewise.
14234 * config/i386/i386.opt: Likewise.
14235 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14236 (GOT_memory_operand): Likewise.
14237 (call_insn_operand): Likewise.
14238 (sibcall_insn_operand): Likewise.
14239 (GOT32_symbol_operand): Likewise.
14241 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14243 PR middle-end/83837
14244 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14245 type rather than type addr's type points to.
14246 (expand_omp_atomic_mutex): Likewise.
14247 (expand_omp_atomic): Likewise.
14249 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14252 * config/i386/i386.c (output_indirect_thunk_function): Use
14253 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14254 for __x86_return_thunk.
14256 2018-01-15 Richard Biener <rguenther@suse.de>
14258 PR middle-end/83850
14259 * expmed.c (extract_bit_field_1): Fix typo.
14261 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14264 * config/arm/iterators.md (VF): New mode iterator.
14265 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14266 Remove integer-related logic from pattern.
14267 (neon_vabd<mode>_3): Likewise.
14269 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14271 PR middle-end/82694
14272 * common.opt (fstrict-overflow): No longer an alias.
14273 (fwrapv-pointer): New option.
14274 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14275 also for pointer types based on flag_wrapv_pointer.
14276 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14277 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14278 opts->x_flag_wrapv got set.
14279 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14280 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14281 POINTER_TYPE_OVERFLOW_UNDEFINED.
14282 * match.pd: Likewise in address comparison pattern.
14283 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14285 2018-01-15 Richard Biener <rguenther@suse.de>
14288 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14289 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14290 Reset type names to their identifier if their TYPE_DECL doesn't
14291 have linkage (and thus is used for ODR and devirt).
14292 (save_debug_info_for_decl): Remove.
14293 (save_debug_info_for_type): Likewise.
14294 (add_tree_to_fld_list): Adjust.
14295 * tree-pretty-print.c (dump_generic_node): Make dumping of
14296 type names more robust.
14298 2018-01-15 Richard Biener <rguenther@suse.de>
14300 * BASE-VER: Bump to 8.0.1.
14302 2018-01-14 Martin Sebor <msebor@redhat.com>
14305 * builtins.c (check_access): Avoid warning when the no-warning bit
14308 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14310 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14311 * ira-color (allocno_hard_regs_compare): Likewise.
14313 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14316 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14317 Use .pushsection/.popsection.
14319 2018-01-14 Martin Sebor <msebor@redhat.com>
14322 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14324 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14326 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14327 entry from extra_headers.
14328 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14329 extra_headers, make the list bitwise identical to the i?86-*-* one.
14331 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14333 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14334 -mcmodel=large with -mindirect-branch=thunk,
14335 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14336 -mfunction-return=thunk-extern.
14337 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14338 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14339 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14343 * config/i386/i386.c (print_reg): Print the name of the full
14344 integer register without '%'.
14345 (ix86_print_operand): Handle 'V'.
14346 * doc/extend.texi: Document 'V' modifier.
14348 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14350 * config/i386/constraints.md (Bs): Disallow memory operand for
14351 -mindirect-branch-register.
14353 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14354 (GOT_memory_operand): Likewise.
14355 (call_insn_operand): Likewise.
14356 (sibcall_insn_operand): Likewise.
14357 (GOT32_symbol_operand): Likewise.
14358 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14359 for -mindirect-branch-register.
14360 (tablejump): Likewise.
14361 (*sibcall_memory): Likewise.
14362 (*sibcall_value_memory): Likewise.
14363 Disallow peepholes of indirect call and jump via memory for
14364 -mindirect-branch-register.
14365 (*call_pop): Replace m with Bw.
14366 (*call_value_pop): Likewise.
14367 (*sibcall_pop_memory): Replace m with Bs.
14368 * config/i386/i386.opt (mindirect-branch-register): New option.
14369 * doc/invoke.texi: Document -mindirect-branch-register option.
14371 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14373 * config/i386/i386-protos.h (ix86_output_function_return): New.
14374 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14375 set function_return_type.
14376 (indirect_thunk_name): Add ret_p to indicate thunk for function
14378 (output_indirect_thunk_function): Pass false to
14379 indirect_thunk_name.
14380 (ix86_output_indirect_branch_via_reg): Likewise.
14381 (ix86_output_indirect_branch_via_push): Likewise.
14382 (output_indirect_thunk_function): Create alias for function
14383 return thunk if regno < 0.
14384 (ix86_output_function_return): New function.
14385 (ix86_handle_fndecl_attribute): Handle function_return.
14386 (ix86_attribute_table): Add function_return.
14387 * config/i386/i386.h (machine_function): Add
14388 function_return_type.
14389 * config/i386/i386.md (simple_return_internal): Use
14390 ix86_output_function_return.
14391 (simple_return_internal_long): Likewise.
14392 * config/i386/i386.opt (mfunction-return=): New option.
14393 (indirect_branch): Mention -mfunction-return=.
14394 * doc/extend.texi: Document function_return function attribute.
14395 * doc/invoke.texi: Document -mfunction-return= option.
14397 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14399 * config/i386/i386-opts.h (indirect_branch): New.
14400 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14401 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14402 with local indirect jump when converting indirect call and jump.
14403 (ix86_set_indirect_branch_type): New.
14404 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14405 (indirectlabelno): New.
14406 (indirect_thunk_needed): Likewise.
14407 (indirect_thunk_bnd_needed): Likewise.
14408 (indirect_thunks_used): Likewise.
14409 (indirect_thunks_bnd_used): Likewise.
14410 (INDIRECT_LABEL): Likewise.
14411 (indirect_thunk_name): Likewise.
14412 (output_indirect_thunk): Likewise.
14413 (output_indirect_thunk_function): Likewise.
14414 (ix86_output_indirect_branch_via_reg): Likewise.
14415 (ix86_output_indirect_branch_via_push): Likewise.
14416 (ix86_output_indirect_branch): Likewise.
14417 (ix86_output_indirect_jmp): Likewise.
14418 (ix86_code_end): Call output_indirect_thunk_function if needed.
14419 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14421 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14422 (ix86_attribute_table): Add indirect_branch.
14423 * config/i386/i386.h (machine_function): Add indirect_branch_type
14424 and has_local_indirect_jump.
14425 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14427 (tablejump): Likewise.
14428 (*indirect_jump): Use ix86_output_indirect_jmp.
14429 (*tablejump_1): Likewise.
14430 (simple_return_indirect_internal): Likewise.
14431 * config/i386/i386.opt (mindirect-branch=): New option.
14432 (indirect_branch): New.
14435 (thunk-inline): Likewise.
14436 (thunk-extern): Likewise.
14437 * doc/extend.texi: Document indirect_branch function attribute.
14438 * doc/invoke.texi: Document -mindirect-branch= option.
14440 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14443 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14445 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14447 * ipa-inline.c (want_inline_small_function_p): Return false if
14448 inlining has already failed with CIF_FINAL_ERROR.
14449 (update_caller_keys): Call want_inline_small_function_p before
14451 (update_callee_keys): Likewise.
14453 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14455 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14457 (rs6000_quadword_masked_address_p): Likewise.
14458 (quad_aligned_load_p): Likewise.
14459 (quad_aligned_store_p): Likewise.
14460 (const_load_sequence_p): Add comment to describe the outer-most loop.
14461 (mimic_memory_attributes_and_flags): New function.
14462 (rs6000_gen_stvx): Likewise.
14463 (replace_swapped_aligned_store): Likewise.
14464 (rs6000_gen_lvx): Likewise.
14465 (replace_swapped_aligned_load): Likewise.
14466 (replace_swapped_load_constant): Capitalize argument name in
14467 comment describing this function.
14468 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14469 and stores that access quad-word aligned addresses and replace
14470 with stvx or lvx instructions when appropriate.
14471 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14472 New function prototype.
14473 (rs6000_quadword_masked_address_p): Likewise.
14474 (rs6000_gen_lvx): Likewise.
14475 (rs6000_gen_stvx): Likewise.
14476 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14477 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14478 when memory address is aligned.
14479 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14480 this split to select lvx instruction when memory address is aligned.
14481 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14482 instruction when memory address is aligned.
14483 (*vsx_le_perm_load_v16qi): Likewise.
14484 (four unnamed splitters): Modify to select the stvx instruction
14485 when memory is aligned.
14487 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14489 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14490 which appears in the queue multiple times.
14492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14493 Alan Hayward <alan.hayward@arm.com>
14494 David Sherwood <david.sherwood@arm.com>
14496 * tree-vectorizer.h (vec_lower_bound): New structure.
14497 (_loop_vec_info): Add check_nonzero and lower_bounds.
14498 (LOOP_VINFO_CHECK_NONZERO): New macro.
14499 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14500 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14501 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14502 fields. Make seg_len the distance travelled, not including the
14504 (dr_direction_indicator): Declare.
14505 (dr_zero_step_indicator): Likewise.
14506 (dr_known_forward_stride_p): Likewise.
14507 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14509 (runtime_alias_check_p): Allow runtime alias checks with
14511 (operator ==): Compare access_size and align.
14512 (prune_runtime_alias_test_list): Rework for new distinction between
14513 the access_size and seg_len.
14514 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14516 (get_segment_min_max): New function.
14517 (create_intersect_range_checks): Use it.
14518 (dr_step_indicator): New function.
14519 (dr_direction_indicator): Likewise.
14520 (dr_zero_step_indicator): Likewise.
14521 (dr_known_forward_stride_p): Likewise.
14522 * tree-loop-distribution.c (data_ref_segment_size): Return
14523 DR_STEP * (niters - 1).
14524 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14526 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14527 (vect_preserves_scalar_order_p): New function, split out from...
14528 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14529 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14530 (vect_vfa_access_size): New function.
14531 (vect_vfa_align): Likewise.
14532 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14533 (dump_lower_bound): New function.
14534 (vect_check_lower_bound): Likewise.
14535 (vect_small_gap_p): Likewise.
14536 (vectorizable_with_step_bound_p): Likewise.
14537 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14538 depencies if the vectorization factor is 1. Convert the checks
14539 for nonzero steps into checks on the bounds of DR_STEP. Try using
14540 a bunds check for variable steps if the minimum required step is
14541 relatively small. Update calls to the dr_with_seg_len
14542 constructor and to vect_compile_time_alias.
14543 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14545 (vect_loop_versioning): Call it.
14546 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14548 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14551 Alan Hayward <alan.hayward@arm.com>
14552 David Sherwood <david.sherwood@arm.com>
14554 * doc/sourcebuild.texi (vect_scatter_store): Document.
14555 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14557 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14559 * genopinit.c (main): Add supports_vec_scatter_store and
14560 supports_vec_scatter_store_cached to target_optabs.
14561 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14562 IFN_MASK_SCATTER_STORE.
14563 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14565 * internal-fn.h (internal_store_fn_p): Declare.
14566 (internal_fn_stored_value_index): Likewise.
14567 * internal-fn.c (scatter_store_direct): New macro.
14568 (expand_scatter_store_optab_fn): New function.
14569 (direct_scatter_store_optab_supported_p): New macro.
14570 (internal_store_fn_p): New function.
14571 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14572 IFN_MASK_SCATTER_STORE.
14573 (internal_fn_mask_index): Likewise.
14574 (internal_fn_stored_value_index): New function.
14575 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14576 for scatter stores.
14577 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14578 * optabs-query.c (supports_vec_scatter_store_p): New function.
14579 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14580 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14581 true for scatter stores.
14582 (vect_gather_scatter_fn_p): Handle scatter stores too.
14583 (vect_check_gather_scatter): Consider using scatter stores if
14584 supports_vec_scatter_store_p.
14585 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14586 scatter stores too.
14587 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14588 internal_fn_stored_value_index.
14589 (check_load_store_masking): Handle scatter stores too.
14590 (vect_get_store_rhs): Make public.
14591 (vectorizable_call): Use internal_store_fn_p.
14592 (vectorizable_store): Handle scatter store internal functions.
14593 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14594 when deciding whether the end of the group has been reached.
14595 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14596 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14597 (mask_scatter_store<mode>): New insns.
14599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14600 Alan Hayward <alan.hayward@arm.com>
14601 David Sherwood <david.sherwood@arm.com>
14603 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14604 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14605 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14607 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14608 Use vect_truncate_gather_scatter_offset if we can't treat the
14609 operation as a normal gather load or scatter store.
14610 (get_group_load_store_type): Take the gather_scatter_info
14611 as argument. Try using a gather load or scatter store for
14612 single-element groups.
14613 (get_load_store_type): Update calls to get_group_load_store_type
14614 and vect_use_strided_gather_scatters_p.
14616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14617 Alan Hayward <alan.hayward@arm.com>
14618 David Sherwood <david.sherwood@arm.com>
14620 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14621 optional tree argument.
14622 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14624 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14625 but continue to use the current value as a fallback.
14626 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14627 to compare the updates.
14628 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14629 (get_load_store_type): Use it when handling a strided access.
14630 (vect_get_strided_load_store_ops): New function.
14631 (vect_get_data_ptr_increment): Likewise.
14632 (vectorizable_load): Handle strided gather loads. Always pass
14633 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14636 Alan Hayward <alan.hayward@arm.com>
14637 David Sherwood <david.sherwood@arm.com>
14639 * doc/md.texi (gather_load@var{m}): Document.
14640 (mask_gather_load@var{m}): Likewise.
14641 * genopinit.c (main): Add supports_vec_gather_load and
14642 supports_vec_gather_load_cached to target_optabs.
14643 * optabs-tree.c (init_tree_optimization_optabs): Use
14644 ggc_cleared_alloc to allocate target_optabs.
14645 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14646 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14648 * internal-fn.h (internal_load_fn_p): Declare.
14649 (internal_gather_scatter_fn_p): Likewise.
14650 (internal_fn_mask_index): Likewise.
14651 (internal_gather_scatter_fn_supported_p): Likewise.
14652 * internal-fn.c (gather_load_direct): New macro.
14653 (expand_gather_load_optab_fn): New function.
14654 (direct_gather_load_optab_supported_p): New macro.
14655 (direct_internal_fn_optab): New function.
14656 (internal_load_fn_p): Likewise.
14657 (internal_gather_scatter_fn_p): Likewise.
14658 (internal_fn_mask_index): Likewise.
14659 (internal_gather_scatter_fn_supported_p): Likewise.
14660 * optabs-query.c (supports_at_least_one_mode_p): New function.
14661 (supports_vec_gather_load_p): Likewise.
14662 * optabs-query.h (supports_vec_gather_load_p): Declare.
14663 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14664 and memory_type field.
14665 (NUM_PATTERNS): Bump to 15.
14666 * tree-vect-data-refs.c: Include internal-fn.h.
14667 (vect_gather_scatter_fn_p): New function.
14668 (vect_describe_gather_scatter_call): Likewise.
14669 (vect_check_gather_scatter): Try using internal functions for
14670 gather loads. Recognize existing calls to a gather load function.
14671 (vect_analyze_data_refs): Consider using gather loads if
14672 supports_vec_gather_load_p.
14673 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14674 (vect_get_gather_scatter_offset_type): Likewise.
14675 (vect_convert_mask_for_vectype): Likewise.
14676 (vect_add_conversion_to_patterm): Likewise.
14677 (vect_try_gather_scatter_pattern): Likewise.
14678 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14679 (vect_vect_recog_func_ptrs): Add it.
14680 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14681 internal_fn_mask_index and internal_gather_scatter_fn_p.
14682 (check_load_store_masking): Take the gather_scatter_info as an
14683 argument and handle gather loads.
14684 (vect_get_gather_scatter_ops): New function.
14685 (vectorizable_call): Check internal_load_fn_p.
14686 (vectorizable_load): Likewise. Handle gather load internal
14688 (vectorizable_store): Update call to check_load_store_masking.
14689 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14690 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14691 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14692 (aarch64_gather_scale_operand_d): New predicates.
14693 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14694 (mask_gather_load<mode>): New insns.
14696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14697 Alan Hayward <alan.hayward@arm.com>
14698 David Sherwood <david.sherwood@arm.com>
14700 * optabs.def (fold_left_plus_optab): New optab.
14701 * doc/md.texi (fold_left_plus_@var{m}): Document.
14702 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14703 * internal-fn.c (fold_left_direct): Define.
14704 (expand_fold_left_optab_fn): Likewise.
14705 (direct_fold_left_optab_supported_p): Likewise.
14706 * fold-const-call.c (fold_const_fold_left): New function.
14707 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14708 * tree-parloops.c (valid_reduction_p): New function.
14709 (gather_scalar_reductions): Use it.
14710 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14711 (vect_finish_replace_stmt): Declare.
14712 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14713 (needs_fold_left_reduction_p): New function, split out from...
14714 (vect_is_simple_reduction): ...here. Accept reductions that
14715 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14716 (vect_force_simple_reduction): Also store the reduction type in
14717 the assignment's STMT_VINFO_REDUC_TYPE.
14718 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14719 (merge_with_identity): New function.
14720 (vect_expand_fold_left): Likewise.
14721 (vectorize_fold_left_reduction): Likewise.
14722 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14723 scalar phi in place for it. Check for target support and reject
14724 cases that would reassociate the operation. Defer the transform
14725 phase to vectorize_fold_left_reduction.
14726 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14727 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14728 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14732 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14733 call to ifc_temp_var.
14735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14736 Alan Hayward <alan.hayward@arm.com>
14737 David Sherwood <david.sherwood@arm.com>
14739 * target.def (legitimize_address_displacement): Take the original
14740 offset as a poly_int.
14741 * targhooks.h (default_legitimize_address_displacement): Update
14743 * targhooks.c (default_legitimize_address_displacement): Likewise.
14744 * doc/tm.texi: Regenerate.
14745 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14746 as an argument, moving assert of ad->disp == ad->disp_term to...
14747 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14748 Try calling targetm.legitimize_address_displacement before expanding
14749 the address rather than afterwards, and adjust for the new interface.
14750 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14751 Match the new hook interface. Handle SVE addresses.
14752 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14753 new hook interface.
14755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14757 * Makefile.in (OBJS): Add early-remat.o.
14758 * target.def (select_early_remat_modes): New hook.
14759 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14760 * doc/tm.texi: Regenerate.
14761 * targhooks.h (default_select_early_remat_modes): Declare.
14762 * targhooks.c (default_select_early_remat_modes): New function.
14763 * timevar.def (TV_EARLY_REMAT): New timevar.
14764 * passes.def (pass_early_remat): New pass.
14765 * tree-pass.h (make_pass_early_remat): Declare.
14766 * early-remat.c: New file.
14767 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14769 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14772 Alan Hayward <alan.hayward@arm.com>
14773 David Sherwood <david.sherwood@arm.com>
14775 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14776 vfm1 with a bound_epilog parameter.
14777 (vect_do_peeling): Update calls accordingly, and move the prologue
14778 call earlier in the function. Treat the base bound_epilog as 0 for
14779 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14780 this base when peeling for gaps.
14781 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14782 with fully-masked loops.
14783 (vect_estimate_min_profitable_iters): Handle the single peeled
14784 iteration in that case.
14786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14787 Alan Hayward <alan.hayward@arm.com>
14788 David Sherwood <david.sherwood@arm.com>
14790 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14791 single-element interleaving even if the size is not a power of 2.
14792 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14793 accesses for single-element interleaving if the group size is
14796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14797 Alan Hayward <alan.hayward@arm.com>
14798 David Sherwood <david.sherwood@arm.com>
14800 * doc/md.texi (fold_extract_last_@var{m}): Document.
14801 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14802 * optabs.def (fold_extract_last_optab): New optab.
14803 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14804 * internal-fn.c (fold_extract_direct): New macro.
14805 (expand_fold_extract_optab_fn): Likewise.
14806 (direct_fold_extract_optab_supported_p): Likewise.
14807 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14808 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14809 EXTRACT_LAST_REDUCTION.
14810 (get_initial_def_for_reduction): Do not create an initial vector
14811 for EXTRACT_LAST_REDUCTION reductions.
14812 (vectorizable_reduction): Leave the scalar phi in place for
14813 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14814 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14815 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14816 transform phase to vectorizable_condition.
14817 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14819 (vect_finish_stmt_generation): ...here.
14820 (vect_finish_replace_stmt): New function.
14821 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14822 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14824 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14827 Alan Hayward <alan.hayward@arm.com>
14828 David Sherwood <david.sherwood@arm.com>
14830 * doc/md.texi (extract_last_@var{m}): Document.
14831 * optabs.def (extract_last_optab): New optab.
14832 * internal-fn.def (EXTRACT_LAST): New internal function.
14833 * internal-fn.c (cond_unary_direct): New macro.
14834 (expand_cond_unary_optab_fn): Likewise.
14835 (direct_cond_unary_optab_supported_p): Likewise.
14836 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14837 loops using EXTRACT_LAST.
14838 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14839 (extract_last_<mode>): ...this optab.
14840 (vec_extract<mode><Vel>): Update accordingly.
14842 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14843 Alan Hayward <alan.hayward@arm.com>
14844 David Sherwood <david.sherwood@arm.com>
14846 * target.def (empty_mask_is_expensive): New hook.
14847 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14848 * doc/tm.texi: Regenerate.
14849 * targhooks.h (default_empty_mask_is_expensive): Declare.
14850 * targhooks.c (default_empty_mask_is_expensive): New function.
14851 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14852 if the target says that empty masks are expensive.
14853 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14855 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14858 Alan Hayward <alan.hayward@arm.com>
14859 David Sherwood <david.sherwood@arm.com>
14861 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14862 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14863 (vect_use_loop_mask_for_alignment_p): New function.
14864 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14865 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14866 niters_skip argument. Make sure that the first niters_skip elements
14867 of the first iteration are inactive.
14868 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14869 Update call to vect_set_loop_masks_directly.
14870 (get_misalign_in_elems): New function, split out from...
14871 (vect_gen_prolog_loop_niters): ...here.
14872 (vect_update_init_of_dr): Take a code argument that specifies whether
14873 the adjustment should be added or subtracted.
14874 (vect_update_init_of_drs): Likewise.
14875 (vect_prepare_for_masked_peels): New function.
14876 (vect_do_peeling): Skip prologue peeling if we're using a mask
14877 instead. Update call to vect_update_inits_of_drs.
14878 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14880 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14881 alignment. Do not include the number of peeled iterations in
14882 the minimum threshold in that case.
14883 (vectorizable_induction): Adjust the start value down by
14884 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14885 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14886 Take the number of skipped iterations into account when calculating
14888 * tree-vect-stmts.c (vect_gen_while_not): New function.
14890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14891 Alan Hayward <alan.hayward@arm.com>
14892 David Sherwood <david.sherwood@arm.com>
14894 * doc/sourcebuild.texi (vect_fully_masked): Document.
14895 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14896 default value to 0.
14897 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14899 (vect_analyze_loop_2): ...here. Don't check the vectorization
14900 factor against the number of loop iterations if the loop is
14903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14904 Alan Hayward <alan.hayward@arm.com>
14905 David Sherwood <david.sherwood@arm.com>
14907 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14908 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14909 (dump_groups): Update accordingly.
14910 (iv_use::mem_type): New member variable.
14911 (address_p): New function.
14912 (record_use): Add a mem_type argument and initialize the new
14914 (record_group_use): Add a mem_type argument. Use address_p.
14915 Remove obsolete null checks of base_object. Update call to record_use.
14916 (find_interesting_uses_op): Update call to record_group_use.
14917 (find_interesting_uses_cond): Likewise.
14918 (find_interesting_uses_address): Likewise.
14919 (get_mem_type_for_internal_fn): New function.
14920 (find_address_like_use): Likewise.
14921 (find_interesting_uses_stmt): Try find_address_like_use before
14922 calling find_interesting_uses_op.
14923 (addr_offset_valid_p): Use the iv mem_type field as the type
14924 of the addressed memory.
14925 (add_autoinc_candidates): Likewise.
14926 (get_address_cost): Likewise.
14927 (split_small_address_groups_p): Use address_p.
14928 (split_address_groups): Likewise.
14929 (add_iv_candidate_for_use): Likewise.
14930 (autoinc_possible_for_pair): Likewise.
14931 (rewrite_groups): Likewise.
14932 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14933 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14934 (get_alias_ptr_type_for_ptr_address): New function.
14935 (rewrite_use_address): Rewrite address uses in calls that were
14936 identified by find_address_like_use.
14938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14939 Alan Hayward <alan.hayward@arm.com>
14940 David Sherwood <david.sherwood@arm.com>
14942 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14944 * gimple-expr.h (is_gimple_addressable: Likewise.
14945 * gimple-expr.c (is_gimple_address): Likewise.
14946 * internal-fn.c (expand_call_mem_ref): New function.
14947 (expand_mask_load_optab_fn): Use it.
14948 (expand_mask_store_optab_fn): Likewise.
14950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14951 Alan Hayward <alan.hayward@arm.com>
14952 David Sherwood <david.sherwood@arm.com>
14954 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14955 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14956 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14957 (cond_umax@var{mode}): Document.
14958 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14959 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14960 (cond_umin_optab, cond_umax_optab): New optabs.
14961 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14962 (COND_IOR, COND_XOR): New internal functions.
14963 * internal-fn.h (get_conditional_internal_fn): Declare.
14964 * internal-fn.c (cond_binary_direct): New macro.
14965 (expand_cond_binary_optab_fn): Likewise.
14966 (direct_cond_binary_optab_supported_p): Likewise.
14967 (get_conditional_internal_fn): New function.
14968 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14969 Cope with reduction statements that are vectorized as calls rather
14971 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14972 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14973 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14974 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14975 (UNSPEC_COND_EOR): New unspecs.
14976 (optab): Add mappings for them.
14977 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14978 (sve_int_op, sve_fp_op): New int attributes.
14980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14981 Alan Hayward <alan.hayward@arm.com>
14982 David Sherwood <david.sherwood@arm.com>
14984 * optabs.def (while_ult_optab): New optab.
14985 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14986 * internal-fn.def (WHILE_ULT): New internal function.
14987 * internal-fn.h (direct_internal_fn_supported_p): New override
14988 that takes two types as argument.
14989 * internal-fn.c (while_direct): New macro.
14990 (expand_while_optab_fn): New function.
14991 (convert_optab_supported_p): Likewise.
14992 (direct_while_optab_supported_p): New macro.
14993 * wide-int.h (wi::udiv_ceil): New function.
14994 * tree-vectorizer.h (rgroup_masks): New structure.
14995 (vec_loop_masks): New typedef.
14996 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14997 and fully_masked_p.
14998 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14999 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15000 (vect_max_vf): New function.
15001 (slpeel_make_loop_iterate_ntimes): Delete.
15002 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15003 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15004 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15005 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15006 internal-fn.h, stor-layout.h and optabs-query.h.
15007 (vect_set_loop_mask): New function.
15008 (add_preheader_seq): Likewise.
15009 (add_header_seq): Likewise.
15010 (interleave_supported_p): Likewise.
15011 (vect_maybe_permute_loop_masks): Likewise.
15012 (vect_set_loop_masks_directly): Likewise.
15013 (vect_set_loop_condition_masked): Likewise.
15014 (vect_set_loop_condition_unmasked): New function, split out from
15015 slpeel_make_loop_iterate_ntimes.
15016 (slpeel_make_loop_iterate_ntimes): Rename to..
15017 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15018 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15019 (vect_do_peeling): Update call accordingly.
15020 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15022 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15023 mask_compare_type, can_fully_mask_p and fully_masked_p.
15024 (release_vec_loop_masks): New function.
15025 (_loop_vec_info): Use it to free the loop masks.
15026 (can_produce_all_loop_masks_p): New function.
15027 (vect_get_max_nscalars_per_iter): Likewise.
15028 (vect_verify_full_masking): Likewise.
15029 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15030 retries, and free the mask rgroups before retrying. Check loop-wide
15031 reasons for disallowing fully-masked loops. Make the final decision
15032 about whether use a fully-masked loop or not.
15033 (vect_estimate_min_profitable_iters): Do not assume that peeling
15034 for the number of iterations will be needed for fully-masked loops.
15035 (vectorizable_reduction): Disable fully-masked loops.
15036 (vectorizable_live_operation): Likewise.
15037 (vect_halve_mask_nunits): New function.
15038 (vect_double_mask_nunits): Likewise.
15039 (vect_record_loop_mask): Likewise.
15040 (vect_get_loop_mask): Likewise.
15041 (vect_transform_loop): Handle the case in which the final loop
15042 iteration might handle a partial vector. Call vect_set_loop_condition
15043 instead of slpeel_make_loop_iterate_ntimes.
15044 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15045 (check_load_store_masking): New function.
15046 (prepare_load_store_mask): Likewise.
15047 (vectorizable_store): Handle fully-masked loops.
15048 (vectorizable_load): Likewise.
15049 (supportable_widening_operation): Use vect_halve_mask_nunits for
15051 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15052 (vect_gen_while): New function.
15053 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15054 (aarch64_uqdec<mode>): New insn.
15056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15057 Alan Hayward <alan.hayward@arm.com>
15058 David Sherwood <david.sherwood@arm.com>
15060 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15061 (reduc_xor_scal_optab): New optabs.
15062 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15063 (reduc_xor_scal_@var{m}): Document.
15064 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15065 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15066 internal functions.
15067 * fold-const-call.c (fold_const_call): Handle them.
15068 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15069 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15070 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15071 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15072 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15073 (UNSPEC_XORV): New unspecs.
15074 (optab): Add entries for them.
15075 (BITWISEV): New int iterator.
15076 (bit_reduc_op): New int attributes.
15078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15079 Alan Hayward <alan.hayward@arm.com>
15080 David Sherwood <david.sherwood@arm.com>
15082 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15083 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15084 * optabs.def (vec_shl_insert_optab): New optab.
15085 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15086 (duplicate_and_interleave): Likewise.
15087 * tree-vect-loop.c: Include internal-fn.h.
15088 (neutral_op_for_slp_reduction): New function, split out from
15089 get_initial_defs_for_reduction.
15090 (get_initial_def_for_reduction): Handle option 2 for variable-length
15091 vectors by loading the neutral value into a vector and then shifting
15092 the initial value into element 0.
15093 (get_initial_defs_for_reduction): Replace the code argument with
15094 the neutral value calculated by neutral_op_for_slp_reduction.
15095 Use gimple_build_vector for constant-length vectors.
15096 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15097 but the first group_size elements have a neutral value.
15098 Use duplicate_and_interleave otherwise.
15099 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15100 Update call to get_initial_defs_for_reduction. Handle SLP
15101 reductions for variable-length vectors by creating one vector
15102 result for each scalar result, with the elements associated
15103 with other scalar results stubbed out with the neutral value.
15104 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15105 Require IFN_VEC_SHL_INSERT for double reductions on
15106 variable-length vectors, or SLP reductions that have
15107 a neutral value. Require can_duplicate_and_interleave_p
15108 support for variable-length unchained SLP reductions if there
15109 is no neutral value, such as for MIN/MAX reductions. Also require
15110 the number of vector elements to be a multiple of the number of
15111 SLP statements when doing variable-length unchained SLP reductions.
15112 Update call to vect_create_epilog_for_reduction.
15113 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15114 and remove initial values.
15115 (duplicate_and_interleave): Make public.
15116 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15117 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15120 Alan Hayward <alan.hayward@arm.com>
15121 David Sherwood <david.sherwood@arm.com>
15123 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15124 (can_duplicate_and_interleave_p): New function.
15125 (vect_get_and_check_slp_defs): Take the vector of statements
15126 rather than just the current one. Remove excess parentheses.
15127 Restriction rejectinon of vect_constant_def and vect_external_def
15128 for variable-length vectors to boolean types, or types for which
15129 can_duplicate_and_interleave_p is false.
15130 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15131 (duplicate_and_interleave): New function.
15132 (vect_get_constant_vectors): Use gimple_build_vector for
15133 constant-length vectors and suitable variable-length constant
15134 vectors. Use duplicate_and_interleave for other variable-length
15135 vectors. Don't defer the update when inserting new statements.
15137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15138 Alan Hayward <alan.hayward@arm.com>
15139 David Sherwood <david.sherwood@arm.com>
15141 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15142 min_profitable_iters doesn't go negative.
15144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15145 Alan Hayward <alan.hayward@arm.com>
15146 David Sherwood <david.sherwood@arm.com>
15148 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15149 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15150 * optabs.def (vec_mask_load_lanes_optab): New optab.
15151 (vec_mask_store_lanes_optab): Likewise.
15152 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15153 (MASK_STORE_LANES): Likewise.
15154 * internal-fn.c (mask_load_lanes_direct): New macro.
15155 (mask_store_lanes_direct): Likewise.
15156 (expand_mask_load_optab_fn): Handle masked operations.
15157 (expand_mask_load_lanes_optab_fn): New macro.
15158 (expand_mask_store_optab_fn): Handle masked operations.
15159 (expand_mask_store_lanes_optab_fn): New macro.
15160 (direct_mask_load_lanes_optab_supported_p): Likewise.
15161 (direct_mask_store_lanes_optab_supported_p): Likewise.
15162 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15164 (vect_load_lanes_supported): Likewise.
15165 * tree-vect-data-refs.c (strip_conversion): New function.
15166 (can_group_stmts_p): Likewise.
15167 (vect_analyze_data_ref_accesses): Use it instead of checking
15168 for a pair of assignments.
15169 (vect_store_lanes_supported): Take a masked_p parameter.
15170 (vect_load_lanes_supported): Likewise.
15171 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15172 vect_store_lanes_supported and vect_load_lanes_supported.
15173 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15174 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15175 parameter. Don't allow gaps for masked accesses.
15176 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15177 and vect_load_lanes_supported.
15178 (get_load_store_type): Take a masked_p parameter and update
15179 call to get_group_load_store_type.
15180 (vectorizable_store): Update call to get_load_store_type.
15181 Handle IFN_MASK_STORE_LANES.
15182 (vectorizable_load): Update call to get_load_store_type.
15183 Handle IFN_MASK_LOAD_LANES.
15185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15186 Alan Hayward <alan.hayward@arm.com>
15187 David Sherwood <david.sherwood@arm.com>
15189 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15191 * config/aarch64/aarch64-protos.h
15192 (aarch64_sve_struct_memory_operand_p): Declare.
15193 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15194 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15195 (VPRED, vpred): Handle SVE structure modes.
15196 * config/aarch64/constraints.md (Utx): New constraint.
15197 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15198 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15199 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15200 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15201 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15202 structure modes. Split into pieces after RA.
15203 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15204 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15206 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15207 SVE structure modes.
15208 (aarch64_classify_address): Likewise.
15209 (sizetochar): Move earlier in file.
15210 (aarch64_print_operand): Handle SVE register lists.
15211 (aarch64_array_mode): New function.
15212 (aarch64_sve_struct_memory_operand_p): Likewise.
15213 (TARGET_ARRAY_MODE): Redefine.
15215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15216 Alan Hayward <alan.hayward@arm.com>
15217 David Sherwood <david.sherwood@arm.com>
15219 * target.def (array_mode): New target hook.
15220 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15221 * doc/tm.texi: Regenerate.
15222 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15223 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15224 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15225 targetm.array_mode.
15226 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15230 Alan Hayward <alan.hayward@arm.com>
15231 David Sherwood <david.sherwood@arm.com>
15233 * fold-const.c (fold_binary_loc): Check the argument types
15234 rather than the result type when testing for a vector operation.
15236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15238 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15239 * doc/tm.texi: Regenerate.
15241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15242 Alan Hayward <alan.hayward@arm.com>
15243 David Sherwood <david.sherwood@arm.com>
15245 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15246 (sve): Document new AArch64 extension.
15247 * doc/md.texi (w): Extend the description of the AArch64
15248 constraint to include SVE vectors.
15249 (Upl, Upa): Document new AArch64 predicate constraints.
15250 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15252 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15253 (msve-vector-bits=): New option.
15254 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15255 SVE when these are disabled.
15256 (sve): New extension.
15257 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15258 modes. Adjust their number of units based on aarch64_sve_vg.
15259 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15260 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15261 aarch64_addr_query_type.
15262 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15263 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15264 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15265 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15266 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15267 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15268 (aarch64_simd_imm_zero_p): Delete.
15269 (aarch64_check_zero_based_sve_index_immediate): Declare.
15270 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15271 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15272 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15273 (aarch64_sve_float_mul_immediate_p): Likewise.
15274 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15275 rather than an rtx.
15276 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15277 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15278 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15279 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15280 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15281 (aarch64_regmode_natural_size): Likewise.
15282 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15283 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15285 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15286 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15287 for VG and the SVE predicate registers.
15288 (V_ALIASES): Add a "z"-prefixed alias.
15289 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15290 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15291 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15292 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15293 (REG_CLASS_NAMES): Add entries for them.
15294 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15295 and the predicate registers.
15296 (aarch64_sve_vg): Declare.
15297 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15298 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15299 (REGMODE_NATURAL_SIZE): Define.
15300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15302 * config/aarch64/aarch64.c: Include cfgrtl.h.
15303 (simd_immediate_info): Add a constructor for series vectors,
15304 and an associated step field.
15305 (aarch64_sve_vg): New variable.
15306 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15307 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15308 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15309 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15310 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15311 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15312 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15313 (aarch64_get_mask_mode): New functions.
15314 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15315 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15316 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15317 predicate modes and predicate registers. Explicitly restrict
15318 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15319 to store a vector mode if it is recognized by
15320 aarch64_classify_vector_mode.
15321 (aarch64_regmode_natural_size): New function.
15322 (aarch64_hard_regno_caller_save_mode): Return the original mode
15324 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15325 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15326 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15327 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15329 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15330 does not overlap dest if the function is frame-related. Handle
15332 (aarch64_split_add_offset): New function.
15333 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15334 them aarch64_add_offset.
15335 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15336 and update call to aarch64_sub_sp.
15337 (aarch64_add_cfa_expression): New function.
15338 (aarch64_expand_prologue): Pass extra temporary registers to the
15339 functions above. Handle the case in which we need to emit new
15340 DW_CFA_expressions for registers that were originally saved
15341 relative to the stack pointer, but now have to be expressed
15342 relative to the frame pointer.
15343 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15345 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15346 IP0 and IP1 values for SVE frames.
15347 (aarch64_expand_vec_series): New function.
15348 (aarch64_expand_sve_widened_duplicate): Likewise.
15349 (aarch64_expand_sve_const_vector): Likewise.
15350 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15351 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15352 into the register, rather than emitting a SET directly.
15353 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15354 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15355 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15356 (offset_9bit_signed_scaled_p): New functions.
15357 (aarch64_replicate_bitmask_imm): New function.
15358 (aarch64_bitmask_imm): Use it.
15359 (aarch64_cannot_force_const_mem): Reject expressions involving
15360 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15361 (aarch64_classify_index): Handle SVE indices, by requiring
15362 a plain register index with a scale that matches the element size.
15363 (aarch64_classify_address): Handle SVE addresses. Assert that
15364 the mode of the address is VOIDmode or an integer mode.
15365 Update call to aarch64_classify_symbol.
15366 (aarch64_classify_symbolic_expression): Update call to
15367 aarch64_classify_symbol.
15368 (aarch64_const_vec_all_in_range_p): New function.
15369 (aarch64_print_vector_float_operand): Likewise.
15370 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15371 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15372 and the FP immediates 1.0 and 0.5.
15373 (aarch64_print_address_internal): Handle SVE addresses.
15374 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15375 (aarch64_regno_regclass): Handle predicate registers.
15376 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15378 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15379 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15380 (aarch64_convert_sve_vector_bits): New function.
15381 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15382 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15383 rather than an rtx.
15384 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15385 Handle SVE vector and predicate modes. Accept VL-based constants
15386 that need only one temporary register, and VL offsets that require
15387 no temporary registers.
15388 (aarch64_conditional_register_usage): Mark the predicate registers
15389 as fixed if SVE isn't available.
15390 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15391 Return true for SVE vector and predicate modes.
15392 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15393 rather than an unsigned int. Handle SVE modes.
15394 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15396 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15398 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15399 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15400 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15401 (aarch64_sve_float_mul_immediate_p): New functions.
15402 (aarch64_sve_valid_immediate): New function.
15403 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15404 Explicitly reject structure modes. Check for INDEX constants.
15405 Handle PTRUE and PFALSE constants.
15406 (aarch64_check_zero_based_sve_index_immediate): New function.
15407 (aarch64_simd_imm_zero_p): Delete.
15408 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15409 vector modes. Accept constants in the range of CNT[BHWD].
15410 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15411 ask for an Advanced SIMD mode.
15412 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15413 (aarch64_simd_vector_alignment): Handle SVE predicates.
15414 (aarch64_vectorize_preferred_vector_alignment): New function.
15415 (aarch64_simd_vector_alignment_reachable): Use it instead of
15417 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15418 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15420 (MAX_VECT_LEN): Delete.
15421 (expand_vec_perm_d): Add a vec_flags field.
15422 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15423 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15424 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15426 (aarch64_evpc_rev): Rename to...
15427 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15428 (aarch64_evpc_rev_global): New function.
15429 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15430 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15432 (aarch64_evpc_sve_tbl): New function.
15433 (aarch64_expand_vec_perm_const_1): Update after rename of
15434 aarch64_evpc_rev. Handle SVE permutes too, trying
15435 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15436 than aarch64_evpc_tbl.
15437 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15438 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15439 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15440 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15441 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15442 (aarch64_expand_sve_vcond): New functions.
15443 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15444 of aarch64_vector_mode_p.
15445 (aarch64_dwarf_poly_indeterminate_value): New function.
15446 (aarch64_compute_pressure_classes): Likewise.
15447 (aarch64_can_change_mode_class): Likewise.
15448 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15449 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15450 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15451 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15452 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15453 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15454 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15455 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15457 (Dn, Dl, Dr): Accept const as well as const_vector.
15458 (Dz): Likewise. Compare against CONST0_RTX.
15459 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15460 of "vector" where appropriate.
15461 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15462 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15463 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15464 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15465 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15466 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15467 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15468 (v_int_equiv): Extend to SVE modes.
15469 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15471 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15472 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15473 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15474 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15475 (SVE_COND_FP_CMP): New int iterators.
15476 (perm_hilo): Handle the new unpack unspecs.
15477 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15479 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15480 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15481 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15482 (aarch64_equality_operator, aarch64_constant_vector_operand)
15483 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15484 (aarch64_sve_nonimmediate_operand): Likewise.
15485 (aarch64_sve_general_operand): Likewise.
15486 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15487 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15488 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15489 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15490 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15491 (aarch64_sve_float_arith_immediate): Likewise.
15492 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15493 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15494 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15495 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15496 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15497 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15498 (aarch64_sve_float_arith_operand): Likewise.
15499 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15500 (aarch64_sve_float_mul_operand): Likewise.
15501 (aarch64_sve_vec_perm_operand): Likewise.
15502 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15503 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15504 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15505 as well as const_vector.
15506 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15507 in file. Use CONST0_RTX and CONSTM1_RTX.
15508 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15509 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15510 Use aarch64_simd_imm_zero.
15511 * config/aarch64/aarch64-sve.md: New file.
15512 * config/aarch64/aarch64.md: Include it.
15513 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15514 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15515 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15516 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15517 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15518 (sve): New attribute.
15519 (enabled): Disable instructions with the sve attribute unless
15521 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15522 aarch64_expand_mov_immediate.
15523 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15524 CNT[BHSD] immediates.
15525 (movti): Split CONST_POLY_INT moves into two halves.
15526 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15527 Split additions that need a temporary here if the destination
15528 is the stack pointer.
15529 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15530 (*add<mode>3_poly_1): New instruction.
15531 (set_clobber_cc): New expander.
15533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15535 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15536 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15537 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15538 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15539 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15540 Change innermode from fixed_mode_size to machine_mode.
15541 (simplify_subreg): Update call accordingly. Handle a constant-sized
15542 subreg of a variable-length CONST_VECTOR.
15544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15545 Alan Hayward <alan.hayward@arm.com>
15546 David Sherwood <david.sherwood@arm.com>
15548 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15549 (add_offset_to_base): New function, split out from...
15550 (create_mem_ref): ...here. When handling a scale other than 1,
15551 check first whether the address is valid without the offset.
15552 Add it into the base if so, leaving the index and scale as-is.
15554 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15557 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15558 fold_for_warn before checking if arg2 is INTEGER_CST.
15560 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15562 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15563 (store_multiple_operation): Delete.
15564 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15565 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15566 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15567 guarded by TARGET_STRING.
15568 (rs6000_output_load_multiple): Delete.
15569 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15570 OPTION_MASK_STRING / TARGET_STRING handling.
15571 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15572 (const rs6000_opt_masks) <"string">: Change mask to 0.
15573 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15574 (MASK_STRING): Delete.
15575 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15577 (load_multiple): Delete.
15584 (store_multiple): Delete.
15591 (movmemsi_8reg): Delete.
15592 (corresponding unnamed define_insn): Delete.
15593 (movmemsi_6reg): Delete.
15594 (corresponding unnamed define_insn): Delete.
15595 (movmemsi_4reg): Delete.
15596 (corresponding unnamed define_insn): Delete.
15597 (movmemsi_2reg): Delete.
15598 (corresponding unnamed define_insn): Delete.
15599 (movmemsi_1reg): Delete.
15600 (corresponding unnamed define_insn): Delete.
15601 * config/rs6000/rs6000.opt (mno-string): New.
15602 (mstring): Replace by deprecation warning stub.
15603 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15605 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15607 * regrename.c (regrename_do_replace): If replacing the same
15608 reg multiple times, try to reuse last created gen_raw_REG.
15611 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15612 main to workaround a bug in GDB.
15614 2018-01-12 Tom de Vries <tom@codesourcery.com>
15617 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15619 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15621 PR rtl-optimization/80481
15622 * ira-color.c (get_cap_member): New function.
15623 (allocnos_conflict_by_live_ranges_p): Use it.
15624 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15625 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15627 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15630 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15631 (*saddl_se_1): Ditto.
15632 (*ssubsi_1): Ditto.
15633 (*ssubl_se_1): Ditto.
15635 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15637 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15638 rather than wi::to_widest for DR_INITs.
15639 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15640 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15641 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15643 (vect_analyze_group_access_1): Note that here.
15645 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15647 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15648 polynomial type sizes.
15650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15652 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15653 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15654 (gimple_add_tmp_var): Likewise.
15656 2018-01-12 Martin Liska <mliska@suse.cz>
15658 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15659 (gimple_alloc_sizes): Likewise.
15660 (dump_gimple_statistics): Use PRIu64 in printf format.
15661 * gimple.h: Change uint64_t to int.
15663 2018-01-12 Martin Liska <mliska@suse.cz>
15665 * tree-core.h: Use uint64_t instead of int.
15666 * tree.c (tree_node_counts): Likewise.
15667 (tree_node_sizes): Likewise.
15668 (dump_tree_statistics): Use PRIu64 in printf format.
15670 2018-01-12 Martin Liska <mliska@suse.cz>
15672 * Makefile.in: As qsort_chk is implemented in vec.c, add
15673 vec.o to linkage of gencfn-macros.
15674 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15675 passing the info to record_node_allocation_statistics.
15676 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15678 * ggc-common.c (struct ggc_usage): Add operator== and use
15679 it in operator< and compare function.
15680 * mem-stats.h (struct mem_usage): Likewise.
15681 * vec.c (struct vec_usage): Remove operator< and compare
15682 function. Can be simply inherited.
15684 2018-01-12 Martin Jambor <mjambor@suse.cz>
15687 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15688 * tree-ssa-math-opts.c: Include domwalk.h.
15689 (convert_mult_to_fma_1): New function.
15690 (fma_transformation_info): New type.
15691 (fma_deferring_state): Likewise.
15692 (cancel_fma_deferring): New function.
15693 (result_of_phi): Likewise.
15694 (last_fma_candidate_feeds_initial_phi): Likewise.
15695 (convert_mult_to_fma): Added deferring logic, split actual
15696 transformation to convert_mult_to_fma_1.
15697 (math_opts_dom_walker): New type.
15698 (math_opts_dom_walker::after_dom_children): New method, body moved
15699 here from pass_optimize_widening_mul::execute, added deferring logic
15701 (pass_optimize_widening_mul::execute): Moved most of code to
15702 math_opts_dom_walker::after_dom_children.
15703 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15704 * config/i386/i386.c (ix86_option_override_internal): Added
15705 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15707 2018-01-12 Richard Biener <rguenther@suse.de>
15710 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15711 inline instance vars.
15713 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15716 * config/rx/rx.c (rx_is_restricted_memory_address):
15717 Handle SUBREG case.
15719 2018-01-12 Richard Biener <rguenther@suse.de>
15721 PR tree-optimization/80846
15722 * target.def (split_reduction): New target hook.
15723 * targhooks.c (default_split_reduction): New function.
15724 * targhooks.h (default_split_reduction): Declare.
15725 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15726 target requests first reduce vectors by combining low and high
15728 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15729 (get_vectype_for_scalar_type_and_size): Export.
15730 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15731 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15732 * doc/tm.texi: Regenerate.
15733 * config/i386/i386.c (ix86_split_reduction): Implement
15734 TARGET_VECTORIZE_SPLIT_REDUCTION.
15736 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15739 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15740 in PIC mode except for TARGET_VXWORKS_RTP.
15741 * config/sparc/sparc.c: Include cfgrtl.h.
15742 (TARGET_INIT_PIC_REG): Define.
15743 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15744 (sparc_pic_register_p): New predicate.
15745 (sparc_legitimate_address_p): Use it.
15746 (sparc_legitimize_pic_address): Likewise.
15747 (sparc_delegitimize_address): Likewise.
15748 (sparc_mode_dependent_address_p): Likewise.
15749 (gen_load_pcrel_sym): Remove 4th parameter.
15750 (load_got_register): Adjust call to above. Remove obsolete stuff.
15751 (sparc_expand_prologue): Do not call load_got_register here.
15752 (sparc_flat_expand_prologue): Likewise.
15753 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15754 (sparc_use_pseudo_pic_reg): New function.
15755 (sparc_init_pic_reg): Likewise.
15756 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15757 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15759 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15761 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15762 Add item for branch_cost.
15764 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15766 PR rtl-optimization/83565
15767 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15768 not extend the result to a larger mode for rotate operations.
15769 (num_sign_bit_copies1): Likewise.
15771 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15774 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15776 Use values-Xc.o for -pedantic.
15777 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15779 2018-01-12 Martin Liska <mliska@suse.cz>
15782 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15784 (possible_polymorphic_call_targets): Use it.
15785 (ipa_devirt): Likewise.
15787 2018-01-12 Martin Liska <mliska@suse.cz>
15789 * profile-count.h (enum profile_quality): Use 0 as invalid
15790 enum value of profile_quality.
15792 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15794 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15795 -mext-string options.
15797 2018-01-12 Richard Biener <rguenther@suse.de>
15799 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15800 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15801 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15803 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15805 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15807 * configure.ac (--with-long-double-format): Add support for the
15808 configuration option to change the default long double format on
15810 * config.gcc (powerpc*-linux*-*): Likewise.
15811 * configure: Regenerate.
15812 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15813 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15814 used without modification.
15816 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15818 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15819 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15820 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15821 MISC_BUILTIN_SPEC_BARRIER.
15822 (rs6000_init_builtins): Likewise.
15823 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15825 (speculation_barrier): New define_insn.
15826 * doc/extend.texi: Document __builtin_speculation_barrier.
15828 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15831 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15832 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15833 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15835 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15836 integral modes instead of "ss" and "sd".
15837 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15838 vectors with 32-bit and 64-bit elements.
15839 (vecdupssescalarmodesuffix): New mode attribute.
15840 (vec_dup<mode>): Use it.
15842 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15845 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15846 frame if argument is passed on stack.
15848 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15851 * ree.c (combine_reaching_defs): Optimize also
15852 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15853 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15855 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15857 PR middle-end/83189
15858 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15860 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15862 PR middle-end/83718
15863 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15864 after they are computed.
15866 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15868 PR tree-optimization/83695
15869 * gimple-loop-linterchange.cc
15870 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15871 reset cached scev information after interchange.
15872 (pass_linterchange::execute): Remove call to scev_reset_htab.
15874 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15876 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15877 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15878 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15879 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15880 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15881 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15882 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15883 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15884 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15885 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15886 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15887 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15888 (V_lane_reg): Likewise.
15889 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15891 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15892 (vfmal_lane_low<mode>_intrinsic,
15893 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15894 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15895 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15896 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15897 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15898 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15900 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15902 * config/arm/arm-cpus.in (fp16fml): New feature.
15903 (ALL_SIMD): Add fp16fml.
15904 (armv8.2-a): Add fp16fml as an option.
15905 (armv8.3-a): Likewise.
15906 (armv8.4-a): Add fp16fml as part of fp16.
15907 * config/arm/arm.h (TARGET_FP16FML): Define.
15908 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15910 * config/arm/arm-modes.def (V2HF): Define.
15911 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15912 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15913 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15914 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15915 vfmsl_low, vfmsl_high): New set of builtins.
15916 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15917 (vfml_op): New code attribute.
15918 (VFMLHALVES): New int iterator.
15919 (VFML, VFMLSEL): New mode attributes.
15920 (V_reg): Define mapping for V2HF.
15921 (V_hi, V_lo): New mode attributes.
15922 (VF_constraint): Likewise.
15923 (vfml_half, vfml_half_selector): New int attributes.
15924 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15926 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15927 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15929 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15930 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15931 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15932 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15934 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15935 Document new effective target and option set.
15937 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15939 * config/arm/arm-cpus.in (armv8_4): New feature.
15940 (ARMv8_4a): New fgroup.
15941 (armv8.4-a): New arch.
15942 * config/arm/arm-tables.opt: Regenerate.
15943 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15944 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15945 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15946 Add matching rules for -march=armv8.4-a and extensions.
15947 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15949 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15952 * config/rx/rx.md (BW): New mode attribute.
15953 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15955 2018-01-11 Richard Biener <rguenther@suse.de>
15957 PR tree-optimization/83435
15958 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15959 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15960 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15962 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15963 Alan Hayward <alan.hayward@arm.com>
15964 David Sherwood <david.sherwood@arm.com>
15966 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15968 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15969 (aarch64_print_address_internal): Use it to check for a zero offset.
15971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15972 Alan Hayward <alan.hayward@arm.com>
15973 David Sherwood <david.sherwood@arm.com>
15975 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15976 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15977 Return a poly_int64 rather than a HOST_WIDE_INT.
15978 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15979 rather than a HOST_WIDE_INT.
15980 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15981 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15982 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15983 final_offset from HOST_WIDE_INT to poly_int64.
15984 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15985 to_constant when getting the number of units in an Advanced SIMD
15987 (aarch64_builtin_vectorized_function): Check for a constant number
15989 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15991 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15992 attribute instead of GET_MODE_NUNITS.
15993 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15994 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15995 GET_MODE_SIZE for fixed-size registers.
15996 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15997 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15998 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15999 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16000 (aarch64_print_operand, aarch64_print_address_internal)
16001 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16002 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16003 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16004 Handle polynomial GET_MODE_SIZE.
16005 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16006 wider than SImode without modification.
16007 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16008 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16009 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16010 passing and returning SVE modes.
16011 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16012 rather than GEN_INT.
16013 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16014 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16015 (aarch64_allocate_and_probe_stack_space): Likewise.
16016 (aarch64_layout_frame): Cope with polynomial offsets.
16017 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16018 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16019 polynomial offsets.
16020 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16021 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16022 poly_int64 rather than a HOST_WIDE_INT.
16023 (aarch64_get_separate_components, aarch64_process_components)
16024 (aarch64_expand_prologue, aarch64_expand_epilogue)
16025 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16026 (aarch64_anchor_offset): New function, split out from...
16027 (aarch64_legitimize_address): ...here.
16028 (aarch64_builtin_vectorization_cost): Handle polynomial
16029 TYPE_VECTOR_SUBPARTS.
16030 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16032 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16033 number of elements from the PARALLEL rather than the mode.
16034 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16035 rather than GET_MODE_BITSIZE.
16036 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16037 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16038 (aarch64_expand_vec_perm_const_1): Handle polynomial
16039 d->perm.length () and d->perm elements.
16040 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16041 Apply to_constant to d->perm elements.
16042 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16043 polynomial CONST_VECTOR_NUNITS.
16044 (aarch64_move_pointer): Take amount as a poly_int64 rather
16046 (aarch64_progress_pointer): Avoid temporary variable.
16047 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16048 the mode attribute instead of GET_MODE.
16050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16051 Alan Hayward <alan.hayward@arm.com>
16052 David Sherwood <david.sherwood@arm.com>
16054 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16055 x exists before using it.
16056 (aarch64_add_constant_internal): Rename to...
16057 (aarch64_add_offset_1): ...this. Replace regnum with separate
16058 src and dest rtxes. Handle the case in which they're different,
16059 including when the offset is zero. Replace scratchreg with an rtx.
16060 Use 2 additions if there is no spare register into which we can
16061 move a 16-bit constant.
16062 (aarch64_add_constant): Delete.
16063 (aarch64_add_offset): Replace reg with separate src and dest
16064 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16065 Use aarch64_add_offset_1.
16066 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16067 an rtx rather than an int. Take the delta as a poly_int64
16068 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16069 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16070 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16071 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16072 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16073 and aarch64_add_sp.
16074 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16075 aarch64_add_constant.
16077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16079 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16080 Use scalar_float_mode.
16082 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16084 * config/aarch64/aarch64-simd.md
16085 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16086 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16087 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16088 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16089 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16090 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16091 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16092 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16093 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16094 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16096 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16099 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16100 targ_options->x_arm_arch_string is non NULL.
16102 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16104 * config/aarch64/aarch64.h
16105 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16107 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16110 * expmed.c (emit_store_flag_force): Swap if const op0
16111 and change VOIDmode to mode of op0.
16113 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16115 PR rtl-optimization/83761
16116 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16117 than bytes to mode_for_size.
16119 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16121 PR middle-end/83189
16122 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16123 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16126 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16128 PR middle-end/83575
16129 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16130 when in layout mode.
16131 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16132 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16135 2018-01-10 Michael Collison <michael.collison@arm.com>
16137 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16138 * config/aarch64/aarch64-option-extension.def: Add
16139 AARCH64_OPT_EXTENSION of 'fp16fml'.
16140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16141 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16142 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16143 * config/aarch64/constraints.md (Ui7): New constraint.
16144 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16145 (VFMLA_SEL_W): Ditto.
16148 (VFMLA16_LOW): New int iterator.
16149 (VFMLA16_HIGH): Ditto.
16150 (UNSPEC_FMLAL): New unspec.
16151 (UNSPEC_FMLSL): Ditto.
16152 (UNSPEC_FMLAL2): Ditto.
16153 (UNSPEC_FMLSL2): Ditto.
16154 (f16mac): New code attribute.
16155 * config/aarch64/aarch64-simd-builtins.def
16156 (aarch64_fmlal_lowv2sf): Ditto.
16157 (aarch64_fmlsl_lowv2sf): Ditto.
16158 (aarch64_fmlalq_lowv4sf): Ditto.
16159 (aarch64_fmlslq_lowv4sf): Ditto.
16160 (aarch64_fmlal_highv2sf): Ditto.
16161 (aarch64_fmlsl_highv2sf): Ditto.
16162 (aarch64_fmlalq_highv4sf): Ditto.
16163 (aarch64_fmlslq_highv4sf): Ditto.
16164 (aarch64_fmlal_lane_lowv2sf): Ditto.
16165 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16166 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16167 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16168 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16169 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16170 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16171 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16172 (aarch64_fmlal_lane_highv2sf): Ditto.
16173 (aarch64_fmlsl_lane_highv2sf): Ditto.
16174 (aarch64_fmlal_laneq_highv2sf): Ditto.
16175 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16176 (aarch64_fmlalq_lane_highv4sf): Ditto.
16177 (aarch64_fmlsl_lane_highv4sf): Ditto.
16178 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16179 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16180 * config/aarch64/aarch64-simd.md:
16181 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16182 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16183 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16184 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16185 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16186 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16187 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16188 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16189 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16190 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16191 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16192 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16193 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16194 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16195 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16196 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16197 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16198 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16199 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16200 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16201 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16202 (vfmlsl_low_u32): Ditto.
16203 (vfmlalq_low_u32): Ditto.
16204 (vfmlslq_low_u32): Ditto.
16205 (vfmlal_high_u32): Ditto.
16206 (vfmlsl_high_u32): Ditto.
16207 (vfmlalq_high_u32): Ditto.
16208 (vfmlslq_high_u32): Ditto.
16209 (vfmlal_lane_low_u32): Ditto.
16210 (vfmlsl_lane_low_u32): Ditto.
16211 (vfmlal_laneq_low_u32): Ditto.
16212 (vfmlsl_laneq_low_u32): Ditto.
16213 (vfmlalq_lane_low_u32): Ditto.
16214 (vfmlslq_lane_low_u32): Ditto.
16215 (vfmlalq_laneq_low_u32): Ditto.
16216 (vfmlslq_laneq_low_u32): Ditto.
16217 (vfmlal_lane_high_u32): Ditto.
16218 (vfmlsl_lane_high_u32): Ditto.
16219 (vfmlal_laneq_high_u32): Ditto.
16220 (vfmlsl_laneq_high_u32): Ditto.
16221 (vfmlalq_lane_high_u32): Ditto.
16222 (vfmlslq_lane_high_u32): Ditto.
16223 (vfmlalq_laneq_high_u32): Ditto.
16224 (vfmlslq_laneq_high_u32): Ditto.
16225 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16226 (AARCH64_FL_FOR_ARCH8_4): New.
16227 (AARCH64_ISA_F16FML): New ISA flag.
16228 (TARGET_F16FML): New feature flag for fp16fml.
16229 (doc/invoke.texi): Document new fp16fml option.
16231 2018-01-10 Michael Collison <michael.collison@arm.com>
16233 * config/aarch64/aarch64-builtins.c:
16234 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16235 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16236 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16237 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16238 (AARCH64_ISA_SHA3): New ISA flag.
16239 (TARGET_SHA3): New feature flag for sha3.
16240 * config/aarch64/iterators.md (sha512_op): New int attribute.
16241 (CRYPTO_SHA512): New int iterator.
16242 (UNSPEC_SHA512H): New unspec.
16243 (UNSPEC_SHA512H2): Ditto.
16244 (UNSPEC_SHA512SU0): Ditto.
16245 (UNSPEC_SHA512SU1): Ditto.
16246 * config/aarch64/aarch64-simd-builtins.def
16247 (aarch64_crypto_sha512hqv2di): New builtin.
16248 (aarch64_crypto_sha512h2qv2di): Ditto.
16249 (aarch64_crypto_sha512su0qv2di): Ditto.
16250 (aarch64_crypto_sha512su1qv2di): Ditto.
16251 (aarch64_eor3qv8hi): Ditto.
16252 (aarch64_rax1qv2di): Ditto.
16253 (aarch64_xarqv2di): Ditto.
16254 (aarch64_bcaxqv8hi): Ditto.
16255 * config/aarch64/aarch64-simd.md:
16256 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16257 (aarch64_crypto_sha512su0qv2di): Ditto.
16258 (aarch64_crypto_sha512su1qv2di): Ditto.
16259 (aarch64_eor3qv8hi): Ditto.
16260 (aarch64_rax1qv2di): Ditto.
16261 (aarch64_xarqv2di): Ditto.
16262 (aarch64_bcaxqv8hi): Ditto.
16263 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16264 (vsha512h2q_u64): Ditto.
16265 (vsha512su0q_u64): Ditto.
16266 (vsha512su1q_u64): Ditto.
16267 (veor3q_u16): Ditto.
16268 (vrax1q_u64): Ditto.
16269 (vxarq_u64): Ditto.
16270 (vbcaxq_u16): Ditto.
16271 * config/arm/types.md (crypto_sha512): New type attribute.
16272 (crypto_sha3): Ditto.
16273 (doc/invoke.texi): Document new sha3 option.
16275 2018-01-10 Michael Collison <michael.collison@arm.com>
16277 * config/aarch64/aarch64-builtins.c:
16278 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16279 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16280 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16281 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16282 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16283 (AARCH64_ISA_SM4): New ISA flag.
16284 (TARGET_SM4): New feature flag for sm4.
16285 * config/aarch64/aarch64-simd-builtins.def
16286 (aarch64_sm3ss1qv4si): Ditto.
16287 (aarch64_sm3tt1aq4si): Ditto.
16288 (aarch64_sm3tt1bq4si): Ditto.
16289 (aarch64_sm3tt2aq4si): Ditto.
16290 (aarch64_sm3tt2bq4si): Ditto.
16291 (aarch64_sm3partw1qv4si): Ditto.
16292 (aarch64_sm3partw2qv4si): Ditto.
16293 (aarch64_sm4eqv4si): Ditto.
16294 (aarch64_sm4ekeyqv4si): Ditto.
16295 * config/aarch64/aarch64-simd.md:
16296 (aarch64_sm3ss1qv4si): Ditto.
16297 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16298 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16299 (aarch64_sm4eqv4si): Ditto.
16300 (aarch64_sm4ekeyqv4si): Ditto.
16301 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16302 (sm3part_op): Ditto.
16303 (CRYPTO_SM3TT): Ditto.
16304 (CRYPTO_SM3PART): Ditto.
16305 (UNSPEC_SM3SS1): New unspec.
16306 (UNSPEC_SM3TT1A): Ditto.
16307 (UNSPEC_SM3TT1B): Ditto.
16308 (UNSPEC_SM3TT2A): Ditto.
16309 (UNSPEC_SM3TT2B): Ditto.
16310 (UNSPEC_SM3PARTW1): Ditto.
16311 (UNSPEC_SM3PARTW2): Ditto.
16312 (UNSPEC_SM4E): Ditto.
16313 (UNSPEC_SM4EKEY): Ditto.
16314 * config/aarch64/constraints.md (Ui2): New constraint.
16315 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16316 * config/arm/types.md (crypto_sm3): New type attribute.
16317 (crypto_sm4): Ditto.
16318 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16319 (vsm3tt1aq_u32): Ditto.
16320 (vsm3tt1bq_u32): Ditto.
16321 (vsm3tt2aq_u32): Ditto.
16322 (vsm3tt2bq_u32): Ditto.
16323 (vsm3partw1q_u32): Ditto.
16324 (vsm3partw2q_u32): Ditto.
16325 (vsm4eq_u32): Ditto.
16326 (vsm4ekeyq_u32): Ditto.
16327 (doc/invoke.texi): Document new sm4 option.
16329 2018-01-10 Michael Collison <michael.collison@arm.com>
16331 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16332 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16333 (AARCH64_FL_FOR_ARCH8_4): New.
16334 (AARCH64_FL_V8_4): New flag.
16335 (doc/invoke.texi): Document new armv8.4-a option.
16337 2018-01-10 Michael Collison <michael.collison@arm.com>
16339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16340 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16341 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16342 * config/aarch64/aarch64-option-extension.def: Add
16343 AARCH64_OPT_EXTENSION of 'sha2'.
16344 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16345 (crypto): Disable sha2 and aes if crypto disabled.
16346 (crypto): Enable aes and sha2 if enabled.
16347 (simd): Disable sha2 and aes if simd disabled.
16348 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16350 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16351 (TARGET_SHA2): New feature flag for sha2.
16352 (TARGET_AES): New feature flag for aes.
16353 * config/aarch64/aarch64-simd.md:
16354 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16355 conditional on TARGET_AES.
16356 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16357 (aarch64_crypto_sha1hsi): Make pattern conditional
16359 (aarch64_crypto_sha1hv4si): Ditto.
16360 (aarch64_be_crypto_sha1hv4si): Ditto.
16361 (aarch64_crypto_sha1su1v4si): Ditto.
16362 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16363 (aarch64_crypto_sha1su0v4si): Ditto.
16364 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16365 (aarch64_crypto_sha256su0v4si): Ditto.
16366 (aarch64_crypto_sha256su1v4si): Ditto.
16367 (doc/invoke.texi): Document new aes and sha2 options.
16369 2018-01-10 Martin Sebor <msebor@redhat.com>
16371 PR tree-optimization/83781
16372 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16375 2018-01-11 Martin Sebor <msebor@gmail.com>
16376 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16378 PR tree-optimization/83501
16379 PR tree-optimization/81703
16381 * tree-ssa-strlen.c (get_string_cst): Rename...
16382 (get_string_len): ...to this. Handle global constants.
16383 (handle_char_store): Adjust.
16385 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16386 Jim Wilson <jimw@sifive.com>
16388 * config/riscv/riscv-protos.h (riscv_output_return): New.
16389 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16390 (riscv_attribute_table, riscv_output_return),
16391 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16392 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16393 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16395 (riscv_expand_prologue): Add early return for naked function.
16396 (riscv_expand_epilogue): Likewise.
16397 (riscv_function_ok_for_sibcall): Return false for naked function.
16398 (riscv_set_current_function): New.
16399 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16400 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16401 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16402 * doc/extend.texi (RISC-V Function Attributes): New.
16404 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16406 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16407 check for 128-bit long double before checking TCmode.
16408 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16409 128-bit long doubles before checking TFmode or TCmode.
16410 (FLOAT128_IBM_P): Likewise.
16412 2018-01-10 Martin Sebor <msebor@redhat.com>
16414 PR tree-optimization/83671
16415 * builtins.c (c_strlen): Unconditionally return zero for the empty
16417 Use -Warray-bounds for warnings.
16418 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16419 for non-constant array indices with COMPONENT_REF, arrays of
16420 arrays, and pointers to arrays.
16421 (gimple_fold_builtin_strlen): Determine and set length range for
16422 non-constant character arrays.
16424 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16426 PR middle-end/81897
16427 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16430 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16432 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16434 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16437 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16438 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16439 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16440 indexed_or_indirect_operand predicate.
16441 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16442 (*vsx_le_perm_load_v8hi): Likewise.
16443 (*vsx_le_perm_load_v16qi): Likewise.
16444 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16445 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16446 (*vsx_le_perm_store_v8hi): Likewise.
16447 (*vsx_le_perm_store_v16qi): Likewise.
16448 (eight unnamed splitters): Likewise.
16450 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16452 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16453 * config/rs6000/emmintrin.h: Likewise.
16454 * config/rs6000/mmintrin.h: Likewise.
16455 * config/rs6000/xmmintrin.h: Likewise.
16457 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16460 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16462 * tree.c (tree_nop_conversion): Return true for location wrapper
16464 (maybe_wrap_with_location): New function.
16465 (selftest::check_strip_nops): New function.
16466 (selftest::test_location_wrappers): New function.
16467 (selftest::tree_c_tests): Call it.
16468 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16469 (maybe_wrap_with_location): New decl.
16470 (EXPR_LOCATION_WRAPPER_P): New macro.
16471 (location_wrapper_p): New inline function.
16472 (tree_strip_any_location_wrapper): New inline function.
16474 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16477 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16478 stack_realign_offset for the largest alignment of stack slot
16480 (ix86_find_max_used_stack_alignment): New function.
16481 (ix86_finalize_stack_frame_flags): Use it. Set
16482 max_used_stack_alignment if we don't realign stack.
16483 * config/i386/i386.h (machine_function): Add
16484 max_used_stack_alignment.
16486 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16488 * config/arm/arm.opt (-mbranch-cost): New option.
16489 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16492 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16495 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16496 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16498 2018-01-10 Richard Biener <rguenther@suse.de>
16501 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16502 early out so it also covers the case where we have a non-NULL
16505 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16507 PR tree-optimization/83753
16508 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16509 for non-strided grouped accesses if the number of elements is 1.
16511 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16514 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16515 * i386.h (TARGET_USE_GATHER): Define.
16516 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16518 2018-01-10 Martin Liska <mliska@suse.cz>
16521 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16522 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16524 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16525 CLEANUP_NO_PARTITIONING is not set.
16527 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16529 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16530 for vectors, as a partial revert of r254296.
16531 * rtl.h (const_vec_p): Delete.
16532 (const_vec_duplicate_p): Don't test for vector CONSTs.
16533 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16534 * expmed.c (make_tree): Likewise.
16537 * common.md (E, F): Use CONSTANT_P instead of checking for
16539 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16540 checking for CONST_VECTOR.
16542 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16544 PR middle-end/83575
16545 * predict.c (force_edge_cold): Handle in more sane way edges
16546 with no prediction.
16548 2018-01-09 Carl Love <cel@us.ibm.com>
16550 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16552 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16553 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16554 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16555 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16556 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16557 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16558 * config/rs6000/rs6000-protos.h: Add extern defition for
16559 rs6000_generate_float2_double_code.
16560 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16562 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16563 (float2_v2df): Add define_expand.
16565 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16568 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16569 op_mode in the force_to_mode call.
16571 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16573 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16574 instead of checking each element individually.
16575 (aarch64_evpc_uzp): Likewise.
16576 (aarch64_evpc_zip): Likewise.
16577 (aarch64_evpc_ext): Likewise.
16578 (aarch64_evpc_rev): Likewise.
16579 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16580 instead of checking each element individually. Return true without
16582 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16583 whether all selected elements come from the same input, instead of
16584 checking each element individually. Remove calls to gen_rtx_REG,
16585 start_sequence and end_sequence and instead assert that no rtl is
16588 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16590 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16591 order of HIGH and CONST checks.
16593 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16595 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16596 if the destination isn't an SSA_NAME.
16598 2018-01-09 Richard Biener <rguenther@suse.de>
16600 PR tree-optimization/83668
16601 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16603 (canonicalize_loop_form): ... here, renamed from ...
16604 (canonicalize_loop_closed_ssa_form): ... this and amended to
16605 swap successor edges for loop exit blocks to make us use
16606 the RPO order we need for initial schedule generation.
16608 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16610 PR tree-optimization/64811
16611 * match.pd: When optimizing comparisons with Inf, avoid
16612 introducing or losing exceptions from comparisons with NaN.
16614 2018-01-09 Martin Liska <mliska@suse.cz>
16617 * asan.c (shadow_mem_size): Add gcc_assert.
16619 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16621 Don't save registers in main().
16624 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16625 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16626 * config/avr/avr.c (avr_set_current_function): Don't error if
16627 naked, OS_task or OS_main are specified at the same time.
16628 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16630 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16632 * common/config/avr/avr-common.c (avr_option_optimization_table):
16633 Switch on -mmain-is-OS_task for optimizing compilations.
16635 2018-01-09 Richard Biener <rguenther@suse.de>
16637 PR tree-optimization/83572
16638 * graphite.c: Include cfganal.h.
16639 (graphite_transform_loops): Connect infinite loops to exit
16640 and remove fake edges at the end.
16642 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16644 * ipa-inline.c (edge_badness): Revert accidental checkin.
16646 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16649 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16650 symbols; not inline clones.
16652 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16655 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16656 hard registers. Formatting fixes.
16658 PR preprocessor/83722
16659 * gcc.c (try_generate_repro): Pass
16660 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16661 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16664 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16665 Kito Cheng <kito.cheng@gmail.com>
16667 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16668 (riscv_leaf_function_p): Delete.
16669 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16671 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16673 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16675 (do_ifelse): New function.
16676 (do_isel): New function.
16677 (do_sub3): New function.
16678 (do_add3): New function.
16679 (do_load_mask_compare): New function.
16680 (do_overlap_load_compare): New function.
16681 (expand_compare_loop): New function.
16682 (expand_block_compare): Call expand_compare_loop() when appropriate.
16683 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16684 option description.
16685 (-mblock-compare-inline-loop-limit): New option.
16687 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16690 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16691 Reverse order of second and third operands in first alternative.
16692 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16693 of first and second elements in UNSPEC_VPERMR vector.
16694 (altivec_expand_vec_perm_le): Likewise.
16696 2018-01-08 Jeff Law <law@redhat.com>
16698 PR rtl-optimizatin/81308
16699 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16700 (process_switch): If group_case_labels makes a change, then set
16702 (pass_convert_switch::execute): If a switch is converted, then
16703 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16705 PR rtl-optimization/81308
16706 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16709 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16711 PR target/83663 - Revert r255946
16712 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16713 generation for cases where splatting a value is not useful.
16714 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16715 across a vec_duplicate and a paradoxical subreg forming a vector
16716 mode to a vec_concat.
16718 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16720 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16721 -march=armv8.3-a variants.
16722 * config/arm/t-multilib: Likewise.
16723 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16725 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16727 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16729 (cceq_ior_compare_complement): Give it a name so I can use it, and
16730 change boolean_or_operator predicate to boolean_operator so it can
16731 be used to generate a crand.
16732 (eqne): New code iterator.
16733 (bd/bd_neg): New code_attrs.
16734 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16735 a single define_insn.
16736 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16737 decrement (bdnzt/bdnzf/bdzt/bdzf).
16738 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16739 with the new names of the branch decrement patterns, and added the
16740 names of the branch decrement conditional patterns.
16742 2018-01-08 Richard Biener <rguenther@suse.de>
16744 PR tree-optimization/83563
16745 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16748 2018-01-08 Richard Biener <rguenther@suse.de>
16750 PR middle-end/83713
16751 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16753 2018-01-08 Richard Biener <rguenther@suse.de>
16755 PR tree-optimization/83685
16756 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16757 references to abnormals.
16759 2018-01-08 Richard Biener <rguenther@suse.de>
16762 * dwarf2out.c (output_indirect_strings): Handle empty
16763 skeleton_debug_str_hash.
16764 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16766 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16768 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16769 (emit_store_direct): Likewise.
16770 (arc_trampoline_adjust_address): Likewise.
16771 (arc_asm_trampoline_template): New function.
16772 (arc_initialize_trampoline): Use asm_trampoline_template.
16773 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16774 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16775 * config/arc/arc.md (flush_icache): Delete pattern.
16777 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16779 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16780 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16783 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16786 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16787 by not USED_FOR_TARGET.
16788 (make_pass_resolve_sw_modes): Likewise.
16790 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16792 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16795 2018-01-08 Richard Biener <rguenther@suse.de>
16797 PR middle-end/83580
16798 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16800 2018-01-08 Richard Biener <rguenther@suse.de>
16802 PR middle-end/83517
16803 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16805 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16807 PR middle-end/81897
16808 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16809 basic blocks with a small number of successors.
16810 (convert_control_dep_chain_into_preds): Improve handling of
16812 (dump_predicates): Split apart into...
16813 (dump_pred_chain): ...here...
16814 (dump_pred_info): ...and here.
16815 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16816 (can_chain_union_be_invalidated_p): Improve check for invalidation
16818 (uninit_uses_cannot_happen): Avoid unnecessary if
16819 convert_control_dep_chain_into_preds yielded nothing.
16821 2018-01-06 Martin Sebor <msebor@redhat.com>
16823 PR tree-optimization/83640
16824 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16825 subtracting negative offset from size.
16826 (builtin_access::overlap): Adjust offset bounds of the access to fall
16827 within the size of the object if possible.
16829 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16831 PR rtl-optimization/83699
16832 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16833 extract_bit_field_as_subreg to cases in which the extracted
16834 value is also a vector.
16836 * lra-constraints.c (process_alt_operands): Test for the equivalence
16837 substitutions when detecting a possible reload cycle.
16839 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16842 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16843 by default if flag_selective_schedling{,2}. Formatting fixes.
16845 PR rtl-optimization/83682
16846 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16847 if it has non-VECTOR_MODE element mode.
16848 (vec_duplicate_p): Likewise.
16850 PR middle-end/83694
16851 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16852 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16854 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16857 * config/i386/i386-builtin.def
16858 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16859 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16860 Require also OPTION_MASK_ISA_AVX512F in addition to
16861 OPTION_MASK_ISA_GFNI.
16862 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16863 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16864 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16865 to OPTION_MASK_ISA_GFNI.
16866 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16867 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16868 OPTION_MASK_ISA_AVX512BW.
16869 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16870 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16871 addition to OPTION_MASK_ISA_GFNI.
16872 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16873 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16874 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16875 to OPTION_MASK_ISA_GFNI.
16876 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16877 a requirement for all ISAs rather than any of them with a few
16879 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16881 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16882 bitmasks to be enabled with 3 exceptions, instead of requiring any
16883 enabled ISA with lots of exceptions.
16884 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16885 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16886 Change avx512bw in isa attribute to avx512f.
16887 * config/i386/sgxintrin.h: Add license boilerplate.
16888 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16889 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16890 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16891 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16893 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16894 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16895 temporarily sse2 rather than sse if not enabled already.
16898 * config/i386/sse.md (VI248_VLBW): Rename to ...
16899 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16900 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16901 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16902 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16903 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16904 mode iterator instead of VI248_VLBW.
16906 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16908 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16909 (record_modified): Skip clobbers; add debug output.
16910 (param_change_prob): Use sreal frequencies.
16912 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16914 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16915 punt for user-aligned variables.
16917 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16919 * tree-chrec.c (chrec_contains_symbols): Return true for
16922 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16925 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16926 of (x|y) == x for BICS pattern.
16928 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16930 PR tree-optimization/83605
16931 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16932 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16935 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16937 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16938 * config/epiphany/rtems.h: New file.
16940 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16941 Uros Bizjak <ubizjak@gmail.com>
16944 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16945 QIreg_operand instead of register_operand predicate.
16946 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16947 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16948 comments instead of -fmitigate[-_]rop.
16950 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16953 * cgraphunit.c (symbol_table::compile): Switch to text_section
16954 before calling assembly_start debug hook.
16955 * run-rtl-passes.c (run_rtl_passes): Likewise.
16958 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16960 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16961 range_int_cst_p rather than !symbolic_range_p before calling
16962 extract_range_from_multiplicative_op_1.
16964 2018-01-04 Jeff Law <law@redhat.com>
16966 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16967 redundant test in assertion.
16969 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16971 * doc/rtl.texi: Document machine_mode wrapper classes.
16973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16975 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16976 using tree_to_uhwi.
16978 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16980 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16981 the VEC_PERM_EXPR fold to fail.
16983 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16986 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16987 to switched_sections.
16989 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16992 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16993 test for d.testing.
16995 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16998 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16999 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17001 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17004 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17005 is BLKmode and bitpos not zero or mode change is needed.
17007 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17010 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17013 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17016 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17017 instead of MULT rtx. Update all corresponding splitters.
17018 (*saddl_se): Ditto.
17019 (*ssub<modesuffix>): Ditto.
17020 (*ssubl_se): Ditto.
17021 (*cmp_sadd_di): Update split patterns.
17022 (*cmp_sadd_si): Ditto.
17023 (*cmp_sadd_sidi): Ditto.
17024 (*cmp_ssub_di): Ditto.
17025 (*cmp_ssub_si): Ditto.
17026 (*cmp_ssub_sidi): Ditto.
17027 * config/alpha/predicates.md (const23_operand): New predicate.
17028 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17029 Look for ASHIFT, not MULT inner operand.
17030 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17032 2018-01-04 Martin Liska <mliska@suse.cz>
17034 PR gcov-profile/83669
17035 * gcov.c (output_intermediate_file): Add version to intermediate
17037 * doc/gcov.texi: Document new field 'version' in intermediate
17038 file format. Fix location of '-k' option of gcov command.
17040 2018-01-04 Martin Liska <mliska@suse.cz>
17043 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17045 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17047 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17049 2018-01-03 Martin Sebor <msebor@redhat.com>
17051 PR tree-optimization/83655
17052 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17053 checking calls with invalid arguments.
17055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17057 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17058 (vectorizable_mask_load_store): Delete.
17059 (vectorizable_call): Return false for masked loads and stores.
17060 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17061 instead of gimple_assign_rhs1.
17062 (vectorizable_load): Handle IFN_MASK_LOAD.
17063 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17067 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17069 (vectorizable_mask_load_store): ...here.
17070 (vectorizable_load): ...and here.
17072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17074 * tree-vect-stmts.c (vect_build_all_ones_mask)
17075 (vect_build_zero_merge_argument): New functions, split out from...
17076 (vectorizable_load): ...here.
17078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17080 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17082 (vectorizable_mask_load_store): ...here.
17083 (vectorizable_store): ...and here.
17085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17087 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17089 (vectorizable_mask_load_store): ...here.
17091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17093 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17094 (vect_model_store_cost): Take a vec_load_store_type instead of a
17096 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17097 (vect_model_store_cost): Take a vec_load_store_type instead of a
17099 (vectorizable_mask_load_store): Update accordingly.
17100 (vectorizable_store): Likewise.
17101 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17105 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17106 IFN_MASK_LOAD calls here rather than...
17107 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17110 Alan Hayward <alan.hayward@arm.com>
17111 David Sherwood <david.sherwood@arm.com>
17113 * expmed.c (extract_bit_field_1): For vector extracts,
17114 fall back to extract_bit_field_as_subreg if vec_extract
17117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17118 Alan Hayward <alan.hayward@arm.com>
17119 David Sherwood <david.sherwood@arm.com>
17121 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17122 they are variable or constant sized.
17123 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17124 slots for constant-sized data.
17126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17127 Alan Hayward <alan.hayward@arm.com>
17128 David Sherwood <david.sherwood@arm.com>
17130 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17131 handling COND_EXPRs with boolean comparisons, try to find a better
17132 basis for the mask type than the boolean itself.
17134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17136 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17137 is calculated and how it can be overridden.
17138 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17139 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17141 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17145 Alan Hayward <alan.hayward@arm.com>
17146 David Sherwood <david.sherwood@arm.com>
17148 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17149 Remove the mode argument.
17150 (aarch64_simd_valid_immediate): Remove the mode and inverse
17152 * config/aarch64/iterators.md (bitsize): New iterator.
17153 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17154 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17155 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17156 aarch64_simd_valid_immediate.
17157 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17158 (aarch64_reg_or_bic_imm): Likewise.
17159 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17160 with an insn_type enum and msl with a modifier_type enum.
17161 Replace element_width with a scalar_mode. Change the shift
17162 to unsigned int. Add constructors for scalar_float_mode and
17163 scalar_int_mode elements.
17164 (aarch64_vect_float_const_representable_p): Delete.
17165 (aarch64_can_const_movi_rtx_p)
17166 (aarch64_simd_scalar_immediate_valid_for_move)
17167 (aarch64_simd_make_constant): Update call to
17168 aarch64_simd_valid_immediate.
17169 (aarch64_advsimd_valid_immediate_hs): New function.
17170 (aarch64_advsimd_valid_immediate): Likewise.
17171 (aarch64_simd_valid_immediate): Remove mode and inverse
17172 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17173 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17174 and aarch64_float_const_representable_p on the result.
17175 (aarch64_output_simd_mov_immediate): Remove mode argument.
17176 Update call to aarch64_simd_valid_immediate and use of
17177 simd_immediate_info.
17178 (aarch64_output_scalar_simd_mov_immediate): Update call
17181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17182 Alan Hayward <alan.hayward@arm.com>
17183 David Sherwood <david.sherwood@arm.com>
17185 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17186 (mode_nunits): Likewise CONST_MODE_NUNITS.
17187 * machmode.def (ADJUST_NUNITS): Document.
17188 * genmodes.c (mode_data::need_nunits_adj): New field.
17189 (blank_mode): Update accordingly.
17190 (adj_nunits): New variable.
17191 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17193 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17194 listed in adj_nunits.
17195 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17196 listed in adj_nunits. Don't emit case statements for such modes.
17197 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17198 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17199 nothing if adj_nunits is nonnull.
17200 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17201 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17202 (emit_mode_fbit): Update use of print_maybe_const_decl.
17203 (emit_move_size): Likewise. Treat the array as non-const
17205 (emit_mode_adjustments): Handle adj_nunits.
17207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17209 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17210 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17211 (VECTOR_MODES): Use it.
17212 (make_vector_modes): Take the prefix as an argument.
17214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17215 Alan Hayward <alan.hayward@arm.com>
17216 David Sherwood <david.sherwood@arm.com>
17218 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17219 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17220 for MODE_VECTOR_BOOL.
17221 * machmode.def (VECTOR_BOOL_MODE): Document.
17222 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17223 (make_vector_bool_mode): New function.
17224 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17226 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17227 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17229 * stor-layout.c (int_mode_for_mode): Likewise.
17230 * tree.c (build_vector_type_for_mode): Likewise.
17231 * varasm.c (output_constant_pool_2): Likewise.
17232 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17233 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17234 for MODE_VECTOR_BOOL.
17235 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17236 of mode class checks.
17237 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17238 instead of a list of mode class checks.
17239 (expand_vector_scalar_condition): Likewise.
17240 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17243 Alan Hayward <alan.hayward@arm.com>
17244 David Sherwood <david.sherwood@arm.com>
17246 * machmode.h (mode_size): Change from unsigned short to
17248 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17249 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17250 or if measurement_type is not polynomial.
17251 (fixed_size_mode::includes_p): Check for constant-sized modes.
17252 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17253 return a poly_uint16 rather than an unsigned short.
17254 (emit_mode_size): Change the type of mode_size from unsigned short
17255 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17256 (emit_mode_adjustments): Cope with polynomial vector sizes.
17257 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17259 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17261 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17262 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17263 * caller-save.c (setup_save_areas): Likewise.
17264 (replace_reg_with_saved_mem): Likewise.
17265 * calls.c (emit_library_call_value_1): Likewise.
17266 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17267 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17268 (gen_lowpart_for_combine): Likewise.
17269 * convert.c (convert_to_integer_1): Likewise.
17270 * cse.c (equiv_constant, cse_insn): Likewise.
17271 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17272 (cselib_subst_to_values): Likewise.
17273 * dce.c (word_dce_process_block): Likewise.
17274 * df-problems.c (df_word_lr_mark_ref): Likewise.
17275 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17276 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17277 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17278 (rtl_for_decl_location): Likewise.
17279 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17280 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17281 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17282 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17283 (expand_expr_real_1): Likewise.
17284 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17285 (pad_below): Likewise.
17286 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17287 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17288 * ira.c (get_subreg_tracking_sizes): Likewise.
17289 * ira-build.c (ira_create_allocno_objects): Likewise.
17290 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17291 (ira_sort_regnos_for_alter_reg): Likewise.
17292 * ira-costs.c (record_operand_costs): Likewise.
17293 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17294 (resolve_simple_move): Likewise.
17295 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17296 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17297 (lra_constraints): Likewise.
17298 (CONST_POOL_OK_P): Reject variable-sized modes.
17299 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17300 (add_pseudo_to_slot, lra_spill): Likewise.
17301 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17302 * optabs-query.c (get_best_extraction_insn): Likewise.
17303 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17304 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17305 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17306 * recog.c (offsettable_address_addr_space_p): Likewise.
17307 * regcprop.c (maybe_mode_change): Likewise.
17308 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17309 * regrename.c (build_def_use): Likewise.
17310 * regstat.c (dump_reg_info): Likewise.
17311 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17312 (find_reloads, find_reloads_subreg_address): Likewise.
17313 * reload1.c (eliminate_regs_1): Likewise.
17314 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17315 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17316 (simplify_binary_operation_1, simplify_subreg): Likewise.
17317 * targhooks.c (default_function_arg_padding): Likewise.
17318 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17319 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17320 (verify_gimple_assign_ternary): Likewise.
17321 * tree-inline.c (estimate_move_cost): Likewise.
17322 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17323 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17324 (get_address_cost_ainc): Likewise.
17325 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17326 (vect_supportable_dr_alignment): Likewise.
17327 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17328 (vectorizable_reduction): Likewise.
17329 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17330 (vectorizable_operation, vectorizable_load): Likewise.
17331 * tree.c (build_same_sized_truth_vector_type): Likewise.
17332 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17333 * var-tracking.c (emit_note_insn_var_location): Likewise.
17334 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17335 (ADDR_VEC_ALIGN): Likewise.
17337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17338 Alan Hayward <alan.hayward@arm.com>
17339 David Sherwood <david.sherwood@arm.com>
17341 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17343 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17344 or if measurement_type is polynomial.
17345 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17346 * combine.c (make_extraction): Likewise.
17347 * dse.c (find_shift_sequence): Likewise.
17348 * dwarf2out.c (mem_loc_descriptor): Likewise.
17349 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17350 (extract_bit_field, extract_low_bits): Likewise.
17351 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17352 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17353 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17354 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17355 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17356 * reload.c (find_reloads): Likewise.
17357 * reload1.c (alter_reg): Likewise.
17358 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17359 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17360 * tree-if-conv.c (predicate_mem_writes): Likewise.
17361 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17362 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17363 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17364 * valtrack.c (dead_debug_insert_temp): Likewise.
17365 * varasm.c (mergeable_constant_section): Likewise.
17366 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17369 Alan Hayward <alan.hayward@arm.com>
17370 David Sherwood <david.sherwood@arm.com>
17372 * expr.c (expand_assignment): Cope with polynomial mode sizes
17373 when assigning to a CONCAT.
17375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17376 Alan Hayward <alan.hayward@arm.com>
17377 David Sherwood <david.sherwood@arm.com>
17379 * machmode.h (mode_precision): Change from unsigned short to
17381 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17383 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17384 or if measurement_type is not polynomial.
17385 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17386 in which the mode is already known to be a scalar_int_mode.
17387 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17388 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17390 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17391 for GET_MODE_PRECISION.
17392 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17393 for GET_MODE_PRECISION.
17394 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17396 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17397 (expand_field_assignment, make_extraction): Likewise.
17398 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17399 (get_last_value): Likewise.
17400 * convert.c (convert_to_integer_1): Likewise.
17401 * cse.c (cse_insn): Likewise.
17402 * expr.c (expand_expr_real_1): Likewise.
17403 * lra-constraints.c (simplify_operand_subreg): Likewise.
17404 * optabs-query.c (can_atomic_load_p): Likewise.
17405 * optabs.c (expand_atomic_load): Likewise.
17406 (expand_atomic_store): Likewise.
17407 * ree.c (combine_reaching_defs): Likewise.
17408 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17409 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17410 * tree.h (type_has_mode_precision_p): Likewise.
17411 * ubsan.c (instrument_si_overflow): Likewise.
17413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17414 Alan Hayward <alan.hayward@arm.com>
17415 David Sherwood <david.sherwood@arm.com>
17417 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17418 polynomial numbers of units.
17419 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17420 (valid_vector_subparts_p): New function.
17421 (build_vector_type): Remove temporary shim and take the number
17422 of units as a poly_uint64 rather than an int.
17423 (build_opaque_vector_type): Take the number of units as a
17424 poly_uint64 rather than an int.
17425 * tree.c (build_vector_from_ctor): Handle polynomial
17426 TYPE_VECTOR_SUBPARTS.
17427 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17428 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17429 (build_vector_from_val): If the number of units is variable,
17430 use build_vec_duplicate_cst for constant operands and
17431 VEC_DUPLICATE_EXPR otherwise.
17432 (make_vector_type): Remove temporary is_constant ().
17433 (build_vector_type, build_opaque_vector_type): Take the number of
17434 units as a poly_uint64 rather than an int.
17435 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17437 * cfgexpand.c (expand_debug_expr): Likewise.
17438 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17439 (store_constructor, expand_expr_real_1): Likewise.
17440 (const_scalar_mask_from_tree): Likewise.
17441 * fold-const-call.c (fold_const_reduction): Likewise.
17442 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17443 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17444 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17445 (fold_relational_const): Likewise.
17446 (native_interpret_vector): Likewise. Change the size from an
17447 int to an unsigned int.
17448 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17449 TYPE_VECTOR_SUBPARTS.
17450 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17451 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17452 duplicating a non-constant operand into a variable-length vector.
17453 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17454 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17455 * ipa-icf.c (sem_variable::equals): Likewise.
17456 * match.pd: Likewise.
17457 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17458 * print-tree.c (print_node): Likewise.
17459 * stor-layout.c (layout_type): Likewise.
17460 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17461 * tree-cfg.c (verify_gimple_comparison): Likewise.
17462 (verify_gimple_assign_binary): Likewise.
17463 (verify_gimple_assign_ternary): Likewise.
17464 (verify_gimple_assign_single): Likewise.
17465 * tree-pretty-print.c (dump_generic_node): Likewise.
17466 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17467 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17468 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17469 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17470 (vect_shift_permute_load_chain): Likewise.
17471 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17472 (expand_vector_condition, optimize_vector_constructor): Likewise.
17473 (lower_vec_perm, get_compute_type): Likewise.
17474 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17475 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17476 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17477 (vect_recog_mask_conversion_pattern): Likewise.
17478 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17479 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17480 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17481 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17482 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17483 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17484 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17485 (supportable_widening_operation): Likewise.
17486 (supportable_narrowing_operation): Likewise.
17487 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17489 * varasm.c (output_constant): Likewise.
17491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17492 Alan Hayward <alan.hayward@arm.com>
17493 David Sherwood <david.sherwood@arm.com>
17495 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17496 so that both the length == 3 and length != 3 cases set up their
17497 own permute vectors. Add comments explaining why we know the
17498 number of elements is constant.
17499 (vect_permute_load_chain): Likewise.
17501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17502 Alan Hayward <alan.hayward@arm.com>
17503 David Sherwood <david.sherwood@arm.com>
17505 * machmode.h (mode_nunits): Change from unsigned char to
17507 (ONLY_FIXED_SIZE_MODES): New macro.
17508 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17509 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17510 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17512 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17513 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17514 or if measurement_type is not polynomial.
17515 * genmodes.c (ZERO_COEFFS): New macro.
17516 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17518 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17519 Use ZERO_COEFFS when emitting initializers.
17520 * data-streamer.h (bp_pack_poly_value): New function.
17521 (bp_unpack_poly_value): Likewise.
17522 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17523 for GET_MODE_NUNITS.
17524 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17525 for GET_MODE_NUNITS.
17526 * tree.c (make_vector_type): Remove temporary shim and make
17527 the real function take the number of units as a poly_uint64
17528 rather than an int.
17529 (build_vector_type_for_mode): Handle polynomial nunits.
17530 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17531 * emit-rtl.c (const_vec_series_p_1): Likewise.
17532 (gen_rtx_CONST_VECTOR): Likewise.
17533 * fold-const.c (test_vec_duplicate_folding): Likewise.
17534 * genrecog.c (validate_pattern): Likewise.
17535 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17536 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17537 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17538 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17539 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17540 * rtlanal.c (subreg_get_info): Likewise.
17541 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17542 (vect_grouped_load_supported): Likewise.
17543 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17544 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17545 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17546 (simplify_const_unary_operation, simplify_binary_operation_1)
17547 (simplify_const_binary_operation, simplify_ternary_operation)
17548 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17549 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17550 instead of CONST_VECTOR_NUNITS.
17551 * varasm.c (output_constant_pool_2): Likewise.
17552 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17553 explicit-encoded elements in the XVEC for variable-length vectors.
17555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17557 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17560 Alan Hayward <alan.hayward@arm.com>
17561 David Sherwood <david.sherwood@arm.com>
17563 * coretypes.h (fixed_size_mode): Declare.
17564 (fixed_size_mode_pod): New typedef.
17565 * builtins.h (target_builtins::x_apply_args_mode)
17566 (target_builtins::x_apply_result_mode): Change type to
17567 fixed_size_mode_pod.
17568 * builtins.c (apply_args_size, apply_result_size, result_vector)
17569 (expand_builtin_apply_args_1, expand_builtin_apply)
17570 (expand_builtin_return): Update accordingly.
17572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17574 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17575 * cselib.c (cselib_hash_rtx): Likewise.
17576 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17577 CONST_VECTOR encoding.
17579 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17580 Jeff Law <law@redhat.com>
17583 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17584 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17585 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17586 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17589 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17590 explicitly probe *sp in a noreturn function if there were any callee
17591 register saves or frame pointer is needed.
17593 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17596 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17597 BLKmode for ternary, binary or unary expressions.
17600 * var-tracking.c (delete_vta_debug_insn): New inline function.
17601 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17602 insns from get_insns () to NULL instead of each bb separately.
17603 Use delete_vta_debug_insn. No longer static.
17604 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17605 delete_vta_debug_insns callers.
17606 * rtl.h (delete_vta_debug_insns): Declare.
17607 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17608 instead of variable_tracking_main.
17610 2018-01-03 Martin Sebor <msebor@redhat.com>
17612 PR tree-optimization/83603
17613 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17614 arguments past the endof the argument list in functions declared
17615 without a prototype.
17616 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17617 Avoid checking when arguments are null.
17619 2018-01-03 Martin Sebor <msebor@redhat.com>
17622 * doc/extend.texi (attribute const): Fix a typo.
17623 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17624 issuing -Wsuggest-attribute for void functions.
17626 2018-01-03 Martin Sebor <msebor@redhat.com>
17628 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17629 offset_int::from instead of wide_int::to_shwi.
17630 (maybe_diag_overlap): Remove assertion.
17631 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17632 * gimple-ssa-sprintf.c (format_directive): Same.
17633 (parse_directive): Same.
17634 (sprintf_dom_walker::compute_format_length): Same.
17635 (try_substitute_return_value): Same.
17637 2018-01-03 Jeff Law <law@redhat.com>
17639 PR middle-end/83654
17640 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17641 non-constant residual for zero at runtime and avoid probing in
17642 that case. Reorganize code for trailing problem to mirror handling
17645 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17647 PR tree-optimization/83501
17648 * tree-ssa-strlen.c (get_string_cst): New.
17649 (handle_char_store): Call get_string_cst.
17651 2018-01-03 Martin Liska <mliska@suse.cz>
17653 PR tree-optimization/83593
17654 * tree-ssa-strlen.c: Include tree-cfg.h.
17655 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17656 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17657 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17659 (strlen_dom_walker::before_dom_children): Call
17660 gimple_purge_dead_eh_edges. Dump tranformation with details
17662 (strlen_dom_walker::before_dom_children): Update call by adding
17663 new argument cleanup_eh.
17664 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17666 2018-01-03 Martin Liska <mliska@suse.cz>
17669 * cif-code.def (VARIADIC_THUNK): New enum value.
17670 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17673 2018-01-03 Jan Beulich <jbeulich@suse.com>
17675 * sse.md (mov<mode>_internal): Tighten condition for when to use
17676 vmovdqu<ssescalarsize> for TI and OI modes.
17678 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17680 Update copyright years.
17682 2018-01-03 Martin Liska <mliska@suse.cz>
17685 * ipa-visibility.c (function_and_variable_visibility): Skip
17686 functions with noipa attribure.
17688 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17690 * gcc.c (process_command): Update copyright notice dates.
17691 * gcov-dump.c (print_version): Ditto.
17692 * gcov.c (print_version): Ditto.
17693 * gcov-tool.c (print_version): Ditto.
17694 * gengtype.c (create_file): Ditto.
17695 * doc/cpp.texi: Bump @copying's copyright year.
17696 * doc/cppinternals.texi: Ditto.
17697 * doc/gcc.texi: Ditto.
17698 * doc/gccint.texi: Ditto.
17699 * doc/gcov.texi: Ditto.
17700 * doc/install.texi: Ditto.
17701 * doc/invoke.texi: Ditto.
17703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17705 * vector-builder.h (vector_builder::m_full_nelts): Change from
17706 unsigned int to poly_uint64.
17707 (vector_builder::full_nelts): Update prototype accordingly.
17708 (vector_builder::new_vector): Likewise.
17709 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17710 (vector_builder::operator ==): Likewise.
17711 (vector_builder::finalize): Likewise.
17712 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17713 Take the number of elements as a poly_uint64 rather than an
17715 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17716 from unsigned int to poly_uint64.
17717 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17718 (vec_perm_indices::new_vector): Likewise.
17719 (vec_perm_indices::length): Likewise.
17720 (vec_perm_indices::nelts_per_input): Likewise.
17721 (vec_perm_indices::input_nelts): Likewise.
17722 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17723 number of elements per input as a poly_uint64 rather than an
17724 unsigned int. Use the original encoding for variable-length
17725 vectors, rather than clamping each individual element.
17726 For the second and subsequent elements in each pattern,
17727 clamp the step and base before clamping their sum.
17728 (vec_perm_indices::series_p): Handle polynomial element counts.
17729 (vec_perm_indices::all_in_range_p): Likewise.
17730 (vec_perm_indices_to_tree): Likewise.
17731 (vec_perm_indices_to_rtx): Likewise.
17732 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17733 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17734 (tree_vector_builder::new_binary_operation): Handle polynomial
17735 element counts. Return false if we need to know the number
17736 of elements at compile time.
17737 * fold-const.c (fold_vec_perm): Punt if the number of elements
17738 isn't known at compile time.
17740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17742 * vec-perm-indices.h (vec_perm_builder): Change element type
17743 from HOST_WIDE_INT to poly_int64.
17744 (vec_perm_indices::element_type): Update accordingly.
17745 (vec_perm_indices::clamp): Handle polynomial element_types.
17746 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17747 (vec_perm_indices::all_in_range_p): Likewise.
17748 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17750 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17751 polynomial vec_perm_indices element types.
17752 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17753 * fold-const.c (fold_vec_perm): Likewise.
17754 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17755 * tree-vect-generic.c (lower_vec_perm): Likewise.
17756 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17757 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17758 element type to HOST_WIDE_INT.
17760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17761 Alan Hayward <alan.hayward@arm.com>
17762 David Sherwood <david.sherwood@arm.com>
17764 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17765 rather than an int. Use plus_constant.
17766 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17767 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17770 Alan Hayward <alan.hayward@arm.com>
17771 David Sherwood <david.sherwood@arm.com>
17773 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17774 a HOST_WIDE_INT to a poly_int64.
17776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17777 Alan Hayward <alan.hayward@arm.com>
17778 David Sherwood <david.sherwood@arm.com>
17780 * calls.c (load_register_parameters): Cope with polynomial
17781 mode sizes. Require a constant size for BLKmode parameters
17782 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17783 forces a parameter to be padded at the lsb end in order to
17784 fill a complete number of words, require the parameter size
17785 to be ordered wrt UNITS_PER_WORD.
17787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17788 Alan Hayward <alan.hayward@arm.com>
17789 David Sherwood <david.sherwood@arm.com>
17791 * reload1.c (spill_stack_slot_width): Change element type
17792 from unsigned int to poly_uint64_pod.
17793 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
17800 (reload_inner_reg_of_subreg, push_reload): Use it.
17802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17803 Alan Hayward <alan.hayward@arm.com>
17804 David Sherwood <david.sherwood@arm.com>
17806 * lra-constraints.c (process_alt_operands): Reject matched
17807 operands whose sizes aren't ordered.
17808 (match_reload): Refer to this check here.
17810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17811 Alan Hayward <alan.hayward@arm.com>
17812 David Sherwood <david.sherwood@arm.com>
17814 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17815 that the mode size is in the set {1, 2, 4, 8, 16}.
17817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17818 Alan Hayward <alan.hayward@arm.com>
17819 David Sherwood <david.sherwood@arm.com>
17821 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17822 Use plus_constant instead of gen_rtx_PLUS.
17824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17825 Alan Hayward <alan.hayward@arm.com>
17826 David Sherwood <david.sherwood@arm.com>
17828 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17829 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17830 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17831 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17832 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17833 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17834 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17835 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17836 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17837 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17839 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17840 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17841 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17842 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17843 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17844 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17845 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17846 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17847 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17848 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17850 * expr.c (emit_move_resolve_push): Treat the input and result
17851 of PUSH_ROUNDING as a poly_int64.
17852 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17853 (emit_push_insn): Likewise.
17854 * lra-eliminations.c (mark_not_eliminable): Likewise.
17855 * recog.c (push_operand): Likewise.
17856 * reload1.c (elimination_effects): Likewise.
17857 * rtlanal.c (nonzero_bits1): Likewise.
17858 * calls.c (store_one_arg): Likewise. Require the padding to be
17859 known at compile time.
17861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17862 Alan Hayward <alan.hayward@arm.com>
17863 David Sherwood <david.sherwood@arm.com>
17865 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17866 Use plus_constant instead of gen_rtx_PLUS.
17868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17869 Alan Hayward <alan.hayward@arm.com>
17870 David Sherwood <david.sherwood@arm.com>
17872 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17873 rather than an int.
17875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17876 Alan Hayward <alan.hayward@arm.com>
17877 David Sherwood <david.sherwood@arm.com>
17879 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17880 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17881 via stack temporaries. Treat the mode size as polynomial too.
17883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17884 Alan Hayward <alan.hayward@arm.com>
17885 David Sherwood <david.sherwood@arm.com>
17887 * expr.c (expand_expr_real_2): When handling conversions involving
17888 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17889 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17890 as a poly_uint64 too.
17892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17893 Alan Hayward <alan.hayward@arm.com>
17894 David Sherwood <david.sherwood@arm.com>
17896 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17899 Alan Hayward <alan.hayward@arm.com>
17900 David Sherwood <david.sherwood@arm.com>
17902 * combine.c (can_change_dest_mode): Handle polynomial
17903 REGMODE_NATURAL_SIZE.
17904 * expmed.c (store_bit_field_1): Likewise.
17905 * expr.c (store_constructor): Likewise.
17906 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17907 and polynomial REGMODE_NATURAL_SIZE.
17908 (gen_lowpart_common): Likewise.
17909 * reginfo.c (record_subregs_of_mode): Likewise.
17910 * rtlanal.c (read_modify_subreg_p): Likewise.
17912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17913 Alan Hayward <alan.hayward@arm.com>
17914 David Sherwood <david.sherwood@arm.com>
17916 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17917 numbers of elements.
17919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17920 Alan Hayward <alan.hayward@arm.com>
17921 David Sherwood <david.sherwood@arm.com>
17923 * match.pd: Cope with polynomial numbers of vector elements.
17925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17926 Alan Hayward <alan.hayward@arm.com>
17927 David Sherwood <david.sherwood@arm.com>
17929 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17930 in a POINTER_PLUS_EXPR.
17932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17933 Alan Hayward <alan.hayward@arm.com>
17934 David Sherwood <david.sherwood@arm.com>
17936 * omp-simd-clone.c (simd_clone_subparts): New function.
17937 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17938 (ipa_simd_modify_function_body): Likewise.
17940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17941 Alan Hayward <alan.hayward@arm.com>
17942 David Sherwood <david.sherwood@arm.com>
17944 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17945 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17946 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17947 (expand_vector_condition, vector_element): Likewise.
17948 (subparts_gt): New function.
17949 (get_compute_type): Use subparts_gt.
17950 (count_type_subparts): Delete.
17951 (expand_vector_operations_1): Use subparts_gt instead of
17952 count_type_subparts.
17954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17955 Alan Hayward <alan.hayward@arm.com>
17956 David Sherwood <david.sherwood@arm.com>
17958 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17959 (vect_compile_time_alias): ...this new function. Do the calculation
17960 on poly_ints rather than trees.
17961 (vect_prune_runtime_alias_test_list): Update call accordingly.
17963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17964 Alan Hayward <alan.hayward@arm.com>
17965 David Sherwood <david.sherwood@arm.com>
17967 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17969 (vect_schedule_slp_instance): Likewise.
17971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17972 Alan Hayward <alan.hayward@arm.com>
17973 David Sherwood <david.sherwood@arm.com>
17975 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17976 constant and extern definitions for variable-length vectors.
17977 (vect_get_constant_vectors): Note that the number of units
17978 is known to be constant.
17980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17981 Alan Hayward <alan.hayward@arm.com>
17982 David Sherwood <david.sherwood@arm.com>
17984 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17985 of units as polynomial. Choose between WIDE and NARROW based
17988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17989 Alan Hayward <alan.hayward@arm.com>
17990 David Sherwood <david.sherwood@arm.com>
17992 * tree-vect-stmts.c (simd_clone_subparts): New function.
17993 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17996 Alan Hayward <alan.hayward@arm.com>
17997 David Sherwood <david.sherwood@arm.com>
17999 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18000 vectors as polynomial. Use build_index_vector for
18001 IFN_GOMP_SIMD_LANE.
18003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18004 Alan Hayward <alan.hayward@arm.com>
18005 David Sherwood <david.sherwood@arm.com>
18007 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18008 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18009 for variable-length vectors.
18010 (vectorizable_mask_load_store): Treat the number of units as
18011 polynomial, asserting that it is constant if the condition has
18012 already been enforced.
18013 (vectorizable_store, vectorizable_load): Likewise.
18015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18016 Alan Hayward <alan.hayward@arm.com>
18017 David Sherwood <david.sherwood@arm.com>
18019 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18020 of units as polynomial. Punt if we can't tell at compile time
18021 which vector contains the final result.
18023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18024 Alan Hayward <alan.hayward@arm.com>
18025 David Sherwood <david.sherwood@arm.com>
18027 * tree-vect-loop.c (vectorizable_induction): Treat the number
18028 of units as polynomial. Punt on SLP inductions. Use an integer
18029 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18030 cast of such a series for variable-length floating-point
18033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18034 Alan Hayward <alan.hayward@arm.com>
18035 David Sherwood <david.sherwood@arm.com>
18037 * tree.h (build_index_vector): Declare.
18038 * tree.c (build_index_vector): New function.
18039 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18040 of units as polynomial, forcibly converting it to a constant if
18041 vectorizable_reduction has already enforced the condition.
18042 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18043 to create a {1,2,3,...} vector.
18044 (vectorizable_reduction): Treat the number of units as polynomial.
18045 Choose vectype_in based on the largest scalar element size rather
18046 than the smallest number of units. Enforce the restrictions
18049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18050 Alan Hayward <alan.hayward@arm.com>
18051 David Sherwood <david.sherwood@arm.com>
18053 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18054 number of units as polynomial.
18056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18057 Alan Hayward <alan.hayward@arm.com>
18058 David Sherwood <david.sherwood@arm.com>
18060 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18061 * target.def (autovectorize_vector_sizes): Return the vector sizes
18062 by pointer, using vector_sizes rather than a bitmask.
18063 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18064 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18065 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18067 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18068 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18069 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18070 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18071 * omp-general.c (omp_max_vf): Likewise.
18072 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18073 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18074 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18075 * tree-vect-slp.c (vect_slp_bb): Likewise.
18076 * doc/tm.texi: Regenerate.
18077 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18079 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18080 the vector size as a poly_uint64 rather than an unsigned int.
18081 (current_vector_size): Change from an unsigned int to a poly_uint64.
18082 (get_vectype_for_scalar_type): Update accordingly.
18083 * tree.h (build_truth_vector_type): Take the size and number of
18084 units as a poly_uint64 rather than an unsigned int.
18085 (build_vector_type): Add a temporary overload that takes
18086 the number of units as a poly_uint64 rather than an unsigned int.
18087 * tree.c (make_vector_type): Likewise.
18088 (build_truth_vector_type): Take the number of units as a poly_uint64
18089 rather than an unsigned int.
18091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18092 Alan Hayward <alan.hayward@arm.com>
18093 David Sherwood <david.sherwood@arm.com>
18095 * target.def (get_mask_mode): Take the number of units and length
18096 as poly_uint64s rather than unsigned ints.
18097 * targhooks.h (default_get_mask_mode): Update accordingly.
18098 * targhooks.c (default_get_mask_mode): Likewise.
18099 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18100 * doc/tm.texi: Regenerate.
18102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18103 Alan Hayward <alan.hayward@arm.com>
18104 David Sherwood <david.sherwood@arm.com>
18106 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18107 * omp-general.c (omp_max_vf): Likewise.
18108 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18109 (expand_omp_simd): Handle polynomial safelen.
18110 * omp-low.c (omplow_simd_context): Add a default constructor.
18111 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18112 (lower_rec_simd_input_clauses): Update accordingly.
18113 (lower_rec_input_clauses): Likewise.
18115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18116 Alan Hayward <alan.hayward@arm.com>
18117 David Sherwood <david.sherwood@arm.com>
18119 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18120 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18121 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18122 (vect_analyze_slp_cost): Likewise.
18123 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18124 (vect_model_load_cost): Likewise.
18126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18127 Alan Hayward <alan.hayward@arm.com>
18128 David Sherwood <david.sherwood@arm.com>
18130 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18131 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18132 from an unsigned int * to a poly_uint64_pod *.
18133 (calculate_unrolling_factor): New function.
18134 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18137 Alan Hayward <alan.hayward@arm.com>
18138 David Sherwood <david.sherwood@arm.com>
18140 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18141 from an unsigned int to a poly_uint64.
18142 (_loop_vec_info::slp_unrolling_factor): Likewise.
18143 (_loop_vec_info::vectorization_factor): Change from an int
18145 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18146 (vect_get_num_vectors): New function.
18147 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18148 (vect_get_num_copies): Use vect_get_num_vectors.
18149 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18150 to an unsigned int *.
18151 (vect_analyze_data_refs): Change min_vf from an int * to a
18153 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18154 than an unsigned HOST_WIDE_INT.
18155 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18156 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18157 to an unsigned int *.
18158 (vect_analyze_data_ref_dependences): Likewise.
18159 (vect_compute_data_ref_alignment): Handle polynomial vf.
18160 (vect_enhance_data_refs_alignment): Likewise.
18161 (vect_prune_runtime_alias_test_list): Likewise.
18162 (vect_shift_permute_load_chain): Likewise.
18163 (vect_supportable_dr_alignment): Likewise.
18164 (dependence_distance_ge_vf): Take the vectorization factor as a
18165 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18166 (vect_analyze_data_refs): Change min_vf from an int * to a
18168 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18169 vfm1 as a poly_uint64 rather than an int. Make the same change
18170 for the returned bound_scalar.
18171 (vect_gen_vector_loop_niters): Handle polynomial vf.
18172 (vect_do_peeling): Likewise. Update call to
18173 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18174 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18176 * tree-vect-loop.c (vect_determine_vectorization_factor)
18177 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18178 (vect_get_known_peeling_cost): Likewise.
18179 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18180 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18181 (vect_transform_loop): Likewise. Use the lowest possible VF when
18182 updating the upper bounds of the loop.
18183 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18184 rather than an int.
18185 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18186 polynomial unroll factors.
18187 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18188 (vect_make_slp_decision): Likewise.
18189 (vect_supported_load_permutation_p): Likewise, and polynomial
18191 (vect_analyze_slp_cost): Handle polynomial vf.
18192 (vect_slp_analyze_node_operations): Likewise.
18193 (vect_slp_analyze_bb_1): Likewise.
18194 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18195 than an unsigned HOST_WIDE_INT.
18196 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18197 (vectorizable_load): Handle polynomial vf.
18198 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18200 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18203 Alan Hayward <alan.hayward@arm.com>
18204 David Sherwood <david.sherwood@arm.com>
18206 * match.pd: Handle bit operations involving three constants
18207 and try to fold one pair.
18209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18211 * tree-vect-loop-manip.c: Include gimple-fold.h.
18212 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18213 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18214 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18215 Add a path that uses a step of VF instead of 1, but disable it
18217 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18218 and niters_no_overflow parameters. Update calls to
18219 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18220 Create a new SSA name if the latter choses to use a ste other
18221 than zero, and return it via niters_vector_mult_vf_var.
18222 * tree-vect-loop.c (vect_transform_loop): Update calls to
18223 vect_do_peeling, vect_gen_vector_loop_niters and
18224 slpeel_make_loop_iterate_ntimes.
18225 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18226 (vect_gen_vector_loop_niters): Update declarations after above changes.
18228 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18230 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18231 128-bit round to integer instructions.
18232 (ceil<mode>2): Likewise.
18233 (btrunc<mode>2): Likewise.
18234 (round<mode>2): Likewise.
18236 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18238 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18239 unaligned VSX load/store on P8/P9.
18240 (expand_block_clear): Allow the use of unaligned VSX
18241 load/store on P8/P9.
18243 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18245 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18247 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18248 swap associated with both a load and a store.
18250 2018-01-02 Andrew Waterman <andrew@sifive.com>
18252 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18253 * config/riscv/riscv.md (clear_cache): Use it.
18255 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18257 * web.c: Remove out-of-date comment.
18259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18261 * expr.c (fixup_args_size_notes): Check that any existing
18262 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18263 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18264 (emit_single_push_insn): ...here.
18266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18268 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18269 (const_vector_encoded_nelts): New function.
18270 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18271 (const_vector_int_elt, const_vector_elt): Declare.
18272 * emit-rtl.c (const_vector_int_elt_1): New function.
18273 (const_vector_elt): Likewise.
18274 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18275 of CONST_VECTOR_ELT.
18277 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18279 * expr.c: Include rtx-vector-builder.h.
18280 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18281 directly on the tree encoding.
18282 (const_vector_from_tree): Likewise.
18283 * optabs.c: Include rtx-vector-builder.h.
18284 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18285 sequence of "u" values.
18286 * vec-perm-indices.c: Include rtx-vector-builder.h.
18287 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18288 directly on the vec_perm_indices encoding.
18290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18292 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18293 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18294 * rtx-vector-builder.h: New file.
18295 * rtx-vector-builder.c: Likewise.
18296 * rtl.h (rtx_def::u2): Add a const_vector field.
18297 (CONST_VECTOR_NPATTERNS): New macro.
18298 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18299 (CONST_VECTOR_DUPLICATE_P): Likewise.
18300 (CONST_VECTOR_STEPPED_P): Likewise.
18301 (CONST_VECTOR_ENCODED_ELT): Likewise.
18302 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18303 (unwrap_const_vec_duplicate): Likewise.
18304 (const_vec_series_p): Check for a non-duplicated vector encoding.
18305 Say that the function only returns true for integer vectors.
18306 * emit-rtl.c: Include rtx-vector-builder.h.
18307 (gen_const_vec_duplicate_1): Delete.
18308 (gen_const_vector): Call gen_const_vec_duplicate instead of
18309 gen_const_vec_duplicate_1.
18310 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18311 (gen_const_vec_duplicate): Use rtx_vector_builder.
18312 (gen_const_vec_series): Likewise.
18313 (gen_rtx_CONST_VECTOR): Likewise.
18314 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18315 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18316 Build a new vector rather than modifying a CONST_VECTOR in-place.
18317 (handle_special_swappables): Update call accordingly.
18318 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18319 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18320 Build a new vector rather than modifying a CONST_VECTOR in-place.
18321 (handle_special_swappables): Update call accordingly.
18323 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18325 * simplify-rtx.c (simplify_const_binary_operation): Use
18326 CONST_VECTOR_ELT instead of XVECEXP.
18328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18330 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18331 the selector elements to be different from the data elements
18332 if the selector is a VECTOR_CST.
18333 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18334 ssizetype for the selector.
18336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18338 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18339 before testing each element individually.
18340 * tree-vect-generic.c (lower_vec_perm): Likewise.
18342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18344 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18345 * selftest-run-tests.c (selftest::run_tests): Call it.
18346 * vector-builder.h (vector_builder::operator ==): New function.
18347 (vector_builder::operator !=): Likewise.
18348 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18349 (vec_perm_indices::all_from_input_p): New function.
18350 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18351 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18352 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18353 instead of reading the VECTOR_CST directly. Detect whether both
18354 vector inputs are the same before constructing the vec_perm_indices,
18355 and update the number of inputs argument accordingly. Use the
18356 utility functions added above. Only construct sel2 if we need to.
18358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18360 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18361 the broadcast of the low byte.
18362 (expand_mult_highpart): Use an explicit encoding for the permutes.
18363 * optabs-query.c (can_mult_highpart_p): Likewise.
18364 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18365 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18366 (vectorizable_bswap): Likewise.
18367 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18368 explicit encoding for the power-of-2 permutes.
18369 (vect_permute_store_chain): Likewise.
18370 (vect_grouped_load_supported): Likewise.
18371 (vect_permute_load_chain): Likewise.
18373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18375 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18376 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18377 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18378 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18379 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18380 (vect_gen_perm_mask_any): Likewise.
18382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18384 * int-vector-builder.h: New file.
18385 * vec-perm-indices.h: Include int-vector-builder.h.
18386 (vec_perm_indices): Redefine as an int_vector_builder.
18387 (auto_vec_perm_indices): Delete.
18388 (vec_perm_builder): Redefine as a stand-alone class.
18389 (vec_perm_indices::vec_perm_indices): New function.
18390 (vec_perm_indices::clamp): Likewise.
18391 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18392 (vec_perm_indices::new_vector): New function.
18393 (vec_perm_indices::new_expanded_vector): Update for new
18394 vec_perm_indices class.
18395 (vec_perm_indices::rotate_inputs): New function.
18396 (vec_perm_indices::all_in_range_p): Operate directly on the
18397 encoded form, without computing elided elements.
18398 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18399 encoding. Update for new vec_perm_indices class.
18400 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18401 the given vec_perm_builder.
18402 (expand_vec_perm_var): Update vec_perm_builder constructor.
18403 (expand_mult_highpart): Use vec_perm_builder instead of
18404 auto_vec_perm_indices.
18405 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18406 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18407 or double series encoding as appropriate.
18408 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18409 vec_perm_indices instead of auto_vec_perm_indices.
18410 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18411 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18412 (vect_permute_store_chain): Likewise.
18413 (vect_grouped_load_supported): Likewise.
18414 (vect_permute_load_chain): Likewise.
18415 (vect_shift_permute_load_chain): Likewise.
18416 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18417 (vect_transform_slp_perm_load): Likewise.
18418 (vect_schedule_slp_instance): Likewise.
18419 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18420 (vectorizable_mask_load_store): Likewise.
18421 (vectorizable_bswap): Likewise.
18422 (vectorizable_store): Likewise.
18423 (vectorizable_load): Likewise.
18424 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18425 vec_perm_indices instead of auto_vec_perm_indices. Use
18426 tree_to_vec_perm_builder to read the vector from a tree.
18427 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18428 vec_perm_builder instead of a vec_perm_indices.
18429 (have_whole_vector_shift): Use vec_perm_builder and
18430 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18431 truncation to calc_vec_perm_mask_for_shift.
18432 (vect_create_epilog_for_reduction): Likewise.
18433 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18434 from auto_vec_perm_indices to vec_perm_indices.
18435 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18436 instead of changing individual elements.
18437 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18438 the vector in d.perm.
18439 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18440 from auto_vec_perm_indices to vec_perm_indices.
18441 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18442 instead of changing individual elements.
18443 (arm_vectorize_vec_perm_const): Use new_vector to install
18444 the vector in d.perm.
18445 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18446 Update vec_perm_builder constructor.
18447 (rs6000_expand_interleave): Likewise.
18448 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18449 (rs6000_expand_interleave): Likewise.
18451 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18453 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18454 to qimode could truncate the indices.
18455 * optabs.c (expand_vec_perm_var): Likewise.
18457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18459 * Makefile.in (OBJS): Add vec-perm-indices.o.
18460 * vec-perm-indices.h: New file.
18461 * vec-perm-indices.c: Likewise.
18462 * target.h (vec_perm_indices): Replace with a forward class
18464 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18465 * optabs.h: Include vec-perm-indices.h.
18466 (expand_vec_perm): Delete.
18467 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18468 (expand_vec_perm_const): Declare.
18469 * target.def (vec_perm_const_ok): Replace with...
18470 (vec_perm_const): ...this new hook.
18471 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18472 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18473 * doc/tm.texi: Regenerate.
18474 * optabs.def (vec_perm_const): Delete.
18475 * doc/md.texi (vec_perm_const): Likewise.
18476 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18477 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18478 expand_vec_perm for constant permutation vectors. Assert that
18479 the mode of variable permutation vectors is the integer equivalent
18480 of the mode that is being permuted.
18481 * optabs-query.h (selector_fits_mode_p): Declare.
18482 * optabs-query.c: Include vec-perm-indices.h.
18483 (selector_fits_mode_p): New function.
18484 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18485 is defined, instead of checking whether the vec_perm_const_optab
18486 exists. Use targetm.vectorize.vec_perm_const instead of
18487 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18488 fit in the vector mode before using a variable permute.
18489 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18490 vec_perm_indices instead of an rtx.
18491 (expand_vec_perm): Replace with...
18492 (expand_vec_perm_const): ...this new function. Take the selector
18493 as a vec_perm_indices rather than an rtx. Also take the mode of
18494 the selector. Update call to shift_amt_for_vec_perm_mask.
18495 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18496 Use vec_perm_indices::new_expanded_vector to expand the original
18497 selector into bytes. Check whether the indices fit in the vector
18498 mode before using a variable permute.
18499 (expand_vec_perm_var): Make global.
18500 (expand_mult_highpart): Use expand_vec_perm_const.
18501 * fold-const.c: Includes vec-perm-indices.h.
18502 * tree-ssa-forwprop.c: Likewise.
18503 * tree-vect-data-refs.c: Likewise.
18504 * tree-vect-generic.c: Likewise.
18505 * tree-vect-loop.c: Likewise.
18506 * tree-vect-slp.c: Likewise.
18507 * tree-vect-stmts.c: Likewise.
18508 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18510 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18511 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18512 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18513 (aarch64_vectorize_vec_perm_const): ...this new function.
18514 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18515 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18516 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18517 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18518 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18519 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18520 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18522 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18523 check for NEON modes.
18524 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18525 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18526 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18527 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18529 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18530 the old VEC_PERM_CONST conditions.
18531 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18532 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18533 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18534 (ia64_vectorize_vec_perm_const_ok): Merge into...
18535 (ia64_vectorize_vec_perm_const): ...this new function.
18536 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18537 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18538 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18539 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18540 * config/mips/mips.c (mips_expand_vec_perm_const)
18541 (mips_vectorize_vec_perm_const_ok): Merge into...
18542 (mips_vectorize_vec_perm_const): ...this new function.
18543 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18544 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18545 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18546 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18547 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18548 (rs6000_expand_vec_perm_const): Delete.
18549 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18551 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18552 (altivec_expand_vec_perm_const_le): Take each operand individually.
18553 Operate on constant selectors rather than rtxes.
18554 (altivec_expand_vec_perm_const): Likewise. Update call to
18555 altivec_expand_vec_perm_const_le.
18556 (rs6000_expand_vec_perm_const): Delete.
18557 (rs6000_vectorize_vec_perm_const_ok): Delete.
18558 (rs6000_vectorize_vec_perm_const): New function.
18559 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18560 an element count and rtx array.
18561 (rs6000_expand_extract_even): Update call accordingly.
18562 (rs6000_expand_interleave): Likewise.
18563 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18564 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18565 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18566 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18567 (rs6000_expand_vec_perm_const): Delete.
18568 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18569 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18570 (altivec_expand_vec_perm_const_le): Take each operand individually.
18571 Operate on constant selectors rather than rtxes.
18572 (altivec_expand_vec_perm_const): Likewise. Update call to
18573 altivec_expand_vec_perm_const_le.
18574 (rs6000_expand_vec_perm_const): Delete.
18575 (rs6000_vectorize_vec_perm_const_ok): Delete.
18576 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18577 reference to the SPE evmerge intructions.
18578 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18579 an element count and rtx array.
18580 (rs6000_expand_extract_even): Update call accordingly.
18581 (rs6000_expand_interleave): Likewise.
18582 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18583 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18585 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18587 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18589 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18590 vector mode and that that mode matches the mode of the data
18592 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18593 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18594 directly using expand_vec_perm_1 when forcing selectors into
18596 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18600 * optabs-query.h (can_vec_perm_p): Delete.
18601 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18602 * optabs-query.c (can_vec_perm_p): Split into...
18603 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18604 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18605 particular selector is valid.
18606 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18607 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18608 (vect_grouped_load_supported): Likewise.
18609 (vect_shift_permute_load_chain): Likewise.
18610 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18611 (vect_transform_slp_perm_load): Likewise.
18612 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18613 (vectorizable_bswap): Likewise.
18614 (vect_gen_perm_mask_checked): Likewise.
18615 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18616 implementations of variable permutation vectors into account
18617 when deciding which selector to use.
18618 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18619 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18620 with a false third argument.
18621 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18622 to test whether the constant selector is valid and can_vec_perm_var_p
18623 to test whether a variable selector is valid.
18625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18627 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18628 * optabs-query.c (can_vec_perm_p): Likewise.
18629 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18630 instead of vec_perm_indices.
18631 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18632 (vect_gen_perm_mask_checked): Likewise,
18633 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18634 (vect_gen_perm_mask_checked): Likewise,
18636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18638 * optabs-query.h (qimode_for_vec_perm): Declare.
18639 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18640 (qimode_for_vec_perm): ...this new function.
18641 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18643 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18645 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18646 does not have a conditional at the top.
18648 2018-01-02 Richard Biener <rguenther@suse.de>
18650 * ipa-inline.c (big_speedup_p): Fix expression.
18652 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18655 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18658 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18662 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18663 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18664 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18665 cond_taken_branch_cost 3->4.
18667 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18669 PR tree-optimization/83581
18670 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18671 TODO_cleanup_cfg if any changes have been made.
18673 PR middle-end/83608
18674 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18675 convert_modes if target mode has the right side, but different mode
18678 PR middle-end/83609
18679 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18680 last argument when extracting from CONCAT. If either from_real or
18681 from_imag is NULL, use expansion through memory. If result is not
18682 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18683 the parts directly to inner mode, if even that fails, use expansion
18686 PR middle-end/83623
18687 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18688 check for bswap in mode rather than HImode and use that in expand_unop
18691 Copyright (C) 2018 Free Software Foundation, Inc.
18693 Copying and distribution of this file, with or without modification,
18694 are permitted in any medium without royalty provided the copyright
18695 notice and this notice are preserved.