1 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (vect_split_statement): New function.
4 (vect_convert_input): Use it to try to split an existing cast.
6 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8 * poly-int.h (print_hex): New function.
9 * dumpfile.h (dump_dec, dump_hex): Declare.
10 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
12 min_input_precision, operation_precision and operation_sign.
13 * tree-vect-patterns.c (vect_get_range_info): New function.
14 (vect_same_loop_or_bb_p, vect_single_imm_use)
15 (vect_operation_fits_smaller_type): Delete.
16 (vect_look_through_possible_promotion): Add an optional
17 single_use_p parameter.
18 (vect_recog_over_widening_pattern): Rewrite to use new
19 stmt_vec_info infomration. Handle one operation at a time.
20 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
21 (vect_truncatable_operation_p, vect_set_operation_type)
22 (vect_set_min_input_precision): New functions.
23 (vect_determine_min_output_precision_1): Likewise.
24 (vect_determine_min_output_precision): Likewise.
25 (vect_determine_precisions_from_range): Likewise.
26 (vect_determine_precisions_from_users): Likewise.
27 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
28 (vect_vect_recog_func_ptrs): Put over_widening first.
30 (vect_pattern_recog): Call vect_determine_precisions.
32 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
34 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
35 statements that have been replaced by further pattern statements.
36 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
38 2018-07-03 Richard Biener <rguenther@suse.de>
40 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
41 always set *dt. Dump vectype in vectype overload.
42 * dumpfile.h (dump_gimple_expr): New function.
43 (dump_gimple_expr_loc): Likewise.
44 * dumpfile.c (dump_gimple_expr): New function.
45 (dump_gimple_expr_loc): Likewise.
47 2018-07-02 Jeff Law <law@redhat.com>
49 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
50 the H8/300, H8/300H and H8/S variants into a single pattern.
51 (movhi_h8300, movqi_h8300hs): Similarly.
52 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
53 (QHI mode iterator): New.
55 * config/h8300/h8300.md: Remove trailing whitespace.
57 2018-07-02 Jim Wilson <jimw@sifive.com>
59 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
60 instead of emit_insn for interrupt returns.
61 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
62 (riscv_sret, riscv_uret): Likewise.
64 2018-07-02 David Malcolm <dmalcolm@redhat.com>
66 * pretty-print.c (selftest::test_pp_format): Move save and restore
67 of quotes to class auto_fix_quotes, and add an instance.
68 * selftest.c: Include "intl.h".
69 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
70 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
71 * selftest.h (selftest::auto_fix_quotes): New class.
73 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
75 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
76 (aarch64_sve_prepare_conditional_op): Remove.
77 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
78 Allow aarch64_simd_reg_or_zero as select operand; remove
79 the aarch64_sve_prepare_conditional_op call.
80 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
81 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
82 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
83 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
84 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
85 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
86 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
87 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
88 and a splitters to match all of the *_any patterns.
89 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
91 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
92 (SVE_COND_FP_BINARY_REV): Remove.
93 (sve_int_op_rev, sve_fp_op_rev): New.
94 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
95 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
96 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
97 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
98 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
99 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
100 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
101 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
102 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
104 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
105 Remove match_dup 1 from the inner unspec.
106 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
108 * config/aarch64/aarch64.md (movprfx): New attr.
109 (length): Default movprfx to 8.
110 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
111 (*madd<SVE_I>, *msub<SVE_I): Likewise.
112 (*<su>mul<SVE_I>3_highpart): Likewise.
113 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
114 (*v<ASHIFT><SVE_I>3): Likewise.
115 (*<su><MAXMIN><SVE_I>3): Likewise.
116 (*<su><MAXMIN><SVE_F>3): Likewise.
117 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
118 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
119 (*div<SVE_F>4): Likewise.
121 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
123 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
126 2018-07-02 Richard Biener <rguenther@suse.de>
128 PR tree-optimization/86363
129 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
130 memset argument refers to a non-variable address.
132 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
134 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
135 {PLUS,MINUS}_EXPR code to...
136 (adjust_symbolic_bound): ...here,
137 (combine_bound): ...here,
138 (set_value_range_with_overflow): ...and here.
140 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
142 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
144 (extract_range_from_abs_expr): ...here.
146 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
148 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
149 -fno-omit-frame-pointer when not optimizing.
151 2018-07-02 Martin Liska <mliska@suse.cz>
154 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
155 (propagate_nothrow): Likewise.
157 2018-07-02 Martin Liska <mliska@suse.cz>
160 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
162 2018-07-02 David Malcolm <dmalcolm@redhat.com>
164 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
165 function in r262149, changing "loc" param from source_location to
166 const dump_location_t &.
167 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
168 declaration, as above.
170 2018-07-01 Paul Koning <ni1d@arrl.net>
172 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
173 -munit-asm, -mgnu-asm, -mdec-asm.
174 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
175 (pdp11_output_labelref): New.
176 (pdp11_output_def): New.
177 (pdp11_output_addr_vec_elt): New.
178 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
179 %# and %@ format codes.
180 (pdp11_option_override): New.
181 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
182 (pdp11_output_ident): New.
183 (pdp11_asm_named_section): New.
184 (pdp11_asm_init_sections): New.
185 (pdp11_file_start): New.
186 (pdp11_file_end): New.
187 (output_ascii): Use .ascii/.asciz for -mdec-asm.
188 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
189 %o, like %c but octal.
190 (pdp11_option_override): New.
191 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
193 (DATA_SECTION_ASM_OP): Ditto.
194 (READONLY_DATA_SECTION_ASM_OP): New.
195 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
196 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
197 (ASM_OUTPUT_LABELREF): Ditto.
198 (ASM_OUTPUT_DEF): Ditto.
199 (ASM_OUTPUT_EXTERNAL): New.
200 (ASM_OUTPUT_SOURCE_FILENAME): New.
201 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
202 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
203 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
204 %# and %@ format codes.
205 * config/pdp11/pdp11.opt (mgnu-asm): New.
206 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
207 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
208 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
210 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
212 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
213 dereferencing path[] beyond its length.
215 (debug_all_paths): New.
216 (rewire_first_differing_edge): New.
217 (adjust_paths_after_duplication): New.
218 (duplicate_thread_path): Call adjust_paths_after_duplication.
220 (thread_through_all_blocks): Add new argument to
221 duplicate_thread_path.
223 2018-06-30 Jim Wilson <jimw@sifive.com>
225 * config/riscv/predicates.md (p2m1_shift_operand): New.
226 (high_mask_shift_operand): New.
227 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
228 pattern using p2m1_shift_operand.
229 (lshsi3_zero_extend_3+2): New combiner pattern using
230 high_mask_shift_operand.
232 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
234 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
236 (vect_recog_rotate_pattern): ...here.
237 (vect_convert_input): Try to insert casts of invariants in the
239 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
240 preheader to be empty.
242 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
244 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
245 vector type. If given, install it in the new statement's
247 (vect_element_precision): New function.
248 (vect_unpromoted_value): New struct.
249 (vect_unpromoted_value::vect_unpromoted_value): New function.
250 (vect_unpromoted_value::set_op): Likewise.
251 (vect_look_through_possible_promotion): Likewise.
252 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
253 (vect_widened_op_tree, vect_convert_input): Likewise.
254 (vect_convert_inputs, vect_convert_output): Likewise.
255 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
256 to handle the optional cast of the multiplication result and
257 vect_widened_op_tree to detect the widened multiplication itself.
258 Do not require the input and output of promotion casts to have
259 the same sign, but base the signedness of the operation on the
260 input rather than the result. If the pattern includes two
261 promotions, check that those promotions have the same sign.
262 Do not restrict the MULT_EXPR handling to a double-width result;
263 handle quadruple-width results and wider. Use vect_convert_inputs
264 to convert the inputs to the common type.
265 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
266 to handle the optional cast of the ABS result. Also allow a sign
267 change or a sign extension between the ABS and MINUS.
268 Use vect_widened_op_tree to detect the widened subtraction and use
269 vect_convert_inputs to convert the inputs to the common type.
270 (vect_handle_widen_op_by_const): Delete.
271 (vect_recog_widen_op_pattern): New function.
272 (vect_recog_widen_mult_pattern): Use it.
273 (vect_recog_widen_shift_pattern): Likewise.
274 (vect_recog_widen_sum_pattern): Use
275 vect_look_through_possible_promotion to handle the promoted
278 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
280 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
281 the containing gimple_seq *.
282 * gimple-iterator.h (gsi_for_stmt): Declare it.
283 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
284 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
285 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
286 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
287 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
289 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
291 (vect_mark_pattern_stmts): ...here. Handle cases in which the
292 statement being replaced is part of an existing pattern
293 definition sequence, inserting the new pattern statements before
295 (vect_pattern_recog_1): Don't return a bool. If the statement
296 is already part of a pattern, instead apply pattern matching
297 to the pattern definition statements. Don't clear the
298 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
299 (vect_pattern_recog): Don't break after the first match;
300 continue processing the pattern definition statements instead.
301 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
303 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
305 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
306 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
307 (vect_recog_widen_sum_pattern): Use it.
309 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
311 * tree-vect-loop.c (vectorizable_reduction): Assert that the
312 phi is not a pattern statement and has not been replaced by
314 * tree-vect-patterns.c (type_conversion_p): Don't check
315 STMT_VINFO_IN_PATTERN_P.
316 (vect_recog_vector_vector_shift_pattern): Likewise.
317 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
318 the pattern statement rather than the original statement; check
319 directly for a WIDEN_MULT_EXPR here.
320 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
321 vect_is_simple_use to return the pattern statement rather
322 than the original statement; use is_pattern_stmt_p to check
323 for such a pattern statement.
324 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
325 to return the pattern statement rather than the original statement;
326 don't do the same transformation here.
327 (vect_is_simple_use): If the defining statement has been replaced
328 by a pattern statement, return the pattern statement instead.
329 Remove the corresponding (local) transformation from the vectype
332 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
334 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
335 end and default to null.
336 * tree-vect-loop.c (vect_create_epilog_for_reduction)
337 (vectorizable_reduction): Update calls accordingly, dropping the
338 gimple ** argument if the passed-back statement isn't needed.
339 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
340 (vect_recog_rotate_pattern): Likewise.
341 (vect_recog_mask_conversion_pattern): Likewise.
342 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
343 (vect_mask_constant_operand_p): Likewise.
344 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
345 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
346 (get_group_load_store_type, get_load_store_type): Likewise.
347 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
348 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
349 (vectorizable_conversion, vectorizable_assignment): Likewise.
350 (vectorizable_shift, vectorizable_operation): Likewise.
351 (vectorizable_store, vect_is_simple_cond): Likewise.
352 (vectorizable_condition, vectorizable_comparison): Likewise.
353 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
354 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
355 and move it to the end. Cope with null def_stmt_outs.
357 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
359 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
361 2018-06-29 Jeff Law <law@redhat.com>
363 * config/v850/v850.c (v850_legitimate_address_p): Handle large
364 displacements for TARGET_V850E2V3 and newer.
365 (TARGET_LRA_P): Remove. Defaults to LRA now.
366 * config/v850/v850.md (sign23byte_load): Remove.
367 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
368 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
370 2018-06-29 Martin Liska <mliska@suse.cz>
373 * coverage.c (coverage_init): Mangle full path name.
374 * doc/invoke.texi: Document the change.
375 * gcov-io.c (mangle_path): New.
376 * gcov-io.h (mangle_path): Likewise.
377 * gcov.c (mangle_name): Use mangle_path for path mangling.
379 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
381 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
382 if starting source register is not even.
384 2018-06-29 Martin Liska <mliska@suse.cz>
386 PR tree-optimization/86263
387 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
388 Make edge redirection.
390 2018-06-29 David Malcolm <dmalcolm@redhat.com>
392 * dumpfile.c (dump_loc): Add indentation based on scope depth.
393 (dump_scope_depth): New variable.
394 (get_dump_scope_depth): New function.
395 (dump_begin_scope): New function.
396 (dump_end_scope): New function.
397 * dumpfile.h (get_dump_scope_depth): New declaration.
398 (dump_begin_scope): New declaration.
399 (dump_end_scope): New declaration.
400 (class auto_dump_scope): New class.
401 (AUTO_DUMP_SCOPE): New macro.
402 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
405 2018-06-29 Richard Biener <rguenther@suse.de>
407 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
408 compute_all_dependences succeeds.
409 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
410 exceed --param loop-max-datarefs-for-datadeps.
412 2018-06-29 Jakub Jelinek <jakub@redhat.com>
414 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
416 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
419 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
420 alternative 0 in preferred_for_speed attribute.
422 2018-06-28 Paul Koning <ni1d@arrl.net>
424 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
425 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
426 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
427 * config/pdp11/pdp11.md: Correct "length" attribute calculation
428 for shift insn patterns.
430 2018-06-28 David Malcolm <dmalcolm@redhat.com>
432 * cgraph.c (cgraph_node::get_body): Replace assignments to
433 "dump_file" with calls to set_dump_file.
434 * dumpfile.c (alt_dump_file): Make static, and group with...
435 (alt_flags): ...this definition.
436 (dumps_are_enabled): New variable.
437 (refresh_dumps_are_enabled): New function.
438 (set_dump_file): New function.
439 (set_alt_dump_file): New function.
440 (gcc::dump_manager::dump_start): Replace assignments to
441 "dump_file" and "alt_dump_file" with calls to set_dump_file and
443 (gcc::dump_manager::dump_finish): Likewise.
444 * dumpfile.h (alt_dump_file): Delete decl.
445 (dumps_are_enabled): New variable decl.
446 (set_dump_file): New function decl.
447 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
449 * tree-nested.c (lower_nested_functions): Replace assignments to
450 "dump_file" with calls to set_dump_file.
452 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
454 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
455 goto_locus of each outgoing edge of each basic block.
457 2018-06-28 Richard Biener <rguenther@suse.de>
459 * dwarf2out.c (decl_scope_table): Remove.
460 (push_decl_scope): Likewise.
461 (pop_decl_scope): Likewise.
462 (gen_type_die_for_member): Do not call push/pop_decl_scope.
463 (gen_struct_or_union_type_die): Likewise.
464 (gen_tagged_type_die): Likewise.
465 (dwarf2out_init): Do not initialize decl_scope_table.
466 (dwarf2out_c_finalize): Do not free it.
468 2018-06-28 Richard Biener <rguenther@suse.de>
470 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
471 deciding whether to not re-use a DIE.
473 2018-06-28 Richard Biener <rguenther@suse.de>
475 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
476 DW_AT_abstract_origin attribute.
478 2018-06-28 Martin Liska <mliska@suse.cz>
480 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
481 Use newly introduced constants.
482 * tree-switch-conversion.h (struct jump_table_cluster):
483 Define max_ratio_for_size and max_ratio_for_speed.
485 2018-06-28 Martin Liska <mliska@suse.cz>
487 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
488 Add new checking assert to catch invalid state.
489 (jump_table_cluster::can_be_handled): Handle single case
491 (jump_table_cluster::is_beneficial): Bail out for such case.
492 (bit_test_cluster::find_bit_tests):
493 Add new checking assert to catch invalid state.
494 (bit_test_cluster::can_be_handled): Handle single case
496 (bit_test_cluster::is_beneficial): Bail out for such case.
497 (switch_decision_tree::analyze_switch_statement):
500 2018-06-28 Martin Liska <mliska@suse.cz>
502 * common.opt: Introduce -completion option.
503 * gcc.c (driver_handle_option): Handle it.
504 (driver::main): Print completions if completion
506 * opt-suggestions.c (option_proposer::get_completions):
508 (option_proposer::suggest_completion): Likewise.
509 (option_proposer::find_param_completions): Likewise.
510 (verify_autocompletions): Likewise.
511 (test_completion_valid_options): Likewise.
512 (test_completion_valid_params): Likewise.
513 (in_completion_p): Likewise.
514 (empty_completion_p): Likewise.
515 (test_completion_partial_match): Likewise.
516 (test_completion_garbage): Likewise.
517 (opt_proposer_c_tests): Likewise.
518 * opt-suggestions.h: Declare new functions.
519 * opts.c (common_handle_option): Handle OPT__completion_.
520 * selftest-run-tests.c (selftest::run_tests): Add
521 opt_proposer_c_tests.
522 * selftest.c (assert_str_startswith): New.
523 * selftest.h (assert_str_startswith): Likewise.
524 (opt_proposer_c_tests): New.
525 (ASSERT_STR_STARTSWITH): Likewise.
527 2018-06-28 Martin Liska <mliska@suse.cz>
529 * Makefile.in: Add opt-suggestions.o.
530 * gcc-main.c: Include opt-suggestions.h.
531 * gcc.c (driver::driver): Likewise.
532 (driver::~driver): Remove m_option_suggestions.
533 (driver::build_option_suggestions): Moved to option_proposer.
534 (driver::suggest_option): Likewise.
535 (driver::handle_unrecognized_options): Use option_proposer.
536 * gcc.h (class driver): Add new memver m_option_proposer.
537 * opt-suggestions.c: New file.
538 * opt-suggestions.h: New file.
540 2018-06-28 Martin Liska <mliska@suse.cz>
542 * vec.h (class auto_string_vec): New (moved from auto_argvec).
543 (auto_string_vec::~auto_string_vec): Likewise.
545 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
547 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
548 prevent_decl_creation_for_types fields up and add reset_location field.
549 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
550 statement if id->reset_location is true.
551 (copy_edges_for_bb): Do not set goto_locus on the new edges if
552 id->reset_location is true.
553 (copy_phis_for_bb): Force input_location on the arguments if
554 id->reset_location is true.
555 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
556 is set on the function to be inlined.
558 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
560 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
562 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
564 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
566 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
567 hard registers for the clobbered pseudo.
569 2018-06-27 Paul Koning <ni1d@arrl.net>
571 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
572 mutually exclusive options.
573 * config/pdp11/constraints.md (h): New constraint.
574 (O): Update definition to match shift code generation.
576 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
578 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
580 (output_jump): Change arguments.
581 (pdp11_fixed_cc_regs): New function.
582 (pdp11_cc_mode): Ditto.
583 (pdp11_expand_shift): Ditto.
584 (pdp11_assemble_shift): Ditto.
585 (pdp11_small_shift): Ditto.
586 (pdp11_branch_cost): Remove.
587 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
589 (pdp11_register_move_cost): Update for CC registers.
590 (pdp11_rtx_costs): Add case for LSHIFTRT.
591 (pdp11_output_jump): Add CCNZ mode conditional branches.
592 (notice_update_cc_on_set): Remove.
593 (pdp11_cc_mode): New function.
594 (simple_memory_operand): Correct pre/post decrement case.
595 (no_side_effect_operand): New function.
596 (pdp11_regno_reg_class): Add CC_REGS class.
597 (pdp11_fixed_cc_regs): New function.
598 (pdp11_small_shift): New function.
599 (pdp11_expand_shift): New function to expand shift insns.
600 (pdp11_assemble_shift): New function to output shifts.
601 (pdp11_branch_cost): Remove.
602 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
603 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
605 (PTRDIFF_TYPE): Ditto.
606 (ADJUST_INSN_LENGTH): New macro.
607 (FIXED_REGISTERS): Add CC registers.
608 (CALL_USED_REGISTERS): Ditto.
610 (REG_CLASS_NAMES): Ditto.
611 (REG_CLASS_CONTENTS): Ditto.
612 (SELECT_CC_MODE): Use new function.
613 (TARGET_FLAGS_REGNUM): New macro.
614 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
615 (cc0_reg_rtx): Remove.
616 (CC_STATUS_MDEP): Remove.
617 (CC_STATUS_MDEFP_INIT): Remove.
619 (NOTICE_UPDATE_CC): Remove.
620 (REGISTER_NAMES): Add CC registers.
621 (BRANCH_COST): Change to constant 1.
622 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
624 * config/pdp11/pdp11.opt (mbcopy): Remove.
625 (mbcopy-builtin): Remove.
626 (mbranch-cheap): Remove.
627 (mbranch-expensive): Remove.
628 * config/pdp11/predicates.md (expand_shift_operand): Update to
629 match shift code generation.
630 (ccnz_operator): New predicate.
631 * doc/invoke.texi (PDP-11 Options): Remove deleted options
632 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
633 Remove non-existent option -mabshi, -mno-abshi. Document mutually
635 * doc/md.texi (PDP-11): Document new D and h constraints. Update
636 description of O constraint.
638 2018-06-27 Jeff Law <law@redhat.com>
639 Austin Law <austinklaw@gmail.com>
641 * config/v850/v850.md (addsi3_set_flags): New pattern.
642 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
643 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
644 (zero_extendhisi2_v850_set_flags): Likewise.
645 (zero_extendqisi2_v850_set_flags): Likewise.
646 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
647 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
648 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
650 * config/v850/v850-protos.h (notice_update_cc): Remove.
651 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
652 (v850_print_operand): Handle 'D' and "d".
653 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
654 Add handling of arithmetic/logical operations compared against zero.
655 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
656 Do not look at v850_compare_op, instead get mode from last argument.
657 (v850_gen_compare): Remove
658 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
659 after reload for prologue insns.
660 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
662 (construct_save_jarl): Likewise.
663 (TARGET_FLAGS_REGNUM): Define.
664 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
665 (NOTICE_UPDATE_CC): Remove.
666 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
667 than cc0. Conditionalize on reload_completed.
668 (cmpsi_insn, setfcc_insn): Likewise.
669 (tst1 splitter): Turn into define_and_split which sets the flags
671 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
672 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
673 (cstoresf4, cstoredf4): Clobber the flags.
674 (cmpsi, cmpsf, cmpdf): Remove expanders.
675 (setf_insn): Remove pattern.
676 (addsi3): Turn into define_and_split which clobbers the flags after
677 reload and a suitable pattern (addsi3_clobber_flags) for use after
679 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
680 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
681 (ashrsi3, ashrsi3_v850e2): Likewise.
682 (bins): Clobber the flags.
683 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
684 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
685 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
686 (call_value_internal_short, call_value_internal_long): Likewise.
687 (callt_save_interrupt, callt_return_interrupt): Likewise.
688 (save_interrupt, return_interrupt): Likewise.
689 (callt_save_all_interrupt, save_all_interrupt): Likewise.
690 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
691 (restore_all_interrupt, _restore_all_interrupt): Likewise.
692 (All FP comparisons): Only allow after reload has completed.
694 (divh, divhu): Tweak output template.
695 (branch_z_normal, branch_z_invert): Remove
696 (branch_nz_normal, branch_nz_invert): Likewise.
697 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
699 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
700 * config/v850/v850.c (notice_update_cc): Remove.
701 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
702 (CC_NO_CARRY): Likewise.
703 (NOTICE_UPDATE_CC): Define to nothing.
704 * config/v850/v850.md: Remove block comment on cc0 handling
705 Remove "cc" attribute from all patterns. Remove cc_status handling
706 from all patterns. Minor formatting fixes.
708 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
710 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
711 (cortex-a76.cortex-a55): Likewise.
712 * config/aarch64/aarch64-tune.md: Regenerate.
713 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
714 cortex-a76.cortex-a55.
716 2018-06-27 Jeff Law <law@redhat.com>
718 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
719 (MULTILIB_DIRNAMES): Similarly.
721 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
723 * gimple.h (gimple_return_retbnd): Delete.
724 (gimple_return_set_retbnd): Likewise.
725 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
726 gimple_return_set_retbnd.
727 * gimple-pretty-print.c (dump_gimple_return): Remove call to
728 gimple_return_retbnd and adjust.
729 * tree-inline.h (struct copy_body_data): Remove retbnd field.
730 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
731 Explicitly return NULL in a couple more cases. Move assertion
732 on debug statements and remove unreachable code.
733 (reset_debug_binding): Do not test id->retbnd.
734 (expand_call_inline): Do not set it.
736 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
738 * configure.ac: Add --disable-gcov option.
739 * configure: Regenerate.
740 * Makefile.in: Honour @enable_gcov@.
741 * doc/install.texi: Document --disable-gcov.
743 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
745 * config/arm/arm-cpus.in (cortex-a76): New entry.
746 (cortex-a76.cortex-a55): Likewise.
747 * config/arm/arm-tables.opt: Regenerate.
748 * config/arm/arm-tune.md: Likewise.
749 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
750 * doc/invoke.texi (ARM Options): Document cortex-a76 and
751 cortex-a76.cortex-a55.
753 2018-06-27 Tamar Christina <tamar.christina@arm.com>
756 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
758 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
760 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
762 (EPILOGUE_USES): Likewise.
764 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
766 * tree-inline.c (remap_location): New function extracted from...
767 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
768 (copy_phis_for_bb): ...here. Call remap_location.
769 (copy_cfg_body): Adjust call to copy_edges_for_bb.
771 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
773 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
774 unaligned vsx for 16B memset.
776 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
779 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
780 ieee128_float_type_node to long_double_type_node unless
781 TARGET_LONG_DOUBLE_128 is set.
783 2018-06-26 David Malcolm <dmalcolm@redhat.com>
785 * cfgloop.c (get_loop_location): Convert return type from
786 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
787 by implicit construction from rtx_insn *, and using
788 dump_user_location_t::from_function_decl for the fallback case.
789 * cfgloop.h (get_loop_location): Convert return type from
790 location_t to dump_user_location_t.
791 * cgraphunit.c (walk_polymorphic_call_targets): 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 * coverage.c (get_coverage_counts): Update calls to
795 dump_printf_loc to pass in dump_location_t rather than a
797 * doc/optinfo.texi (Dump types): Convert example of
798 dump_printf_loc from taking "locus" to taking "insn". Update
799 description of the "_loc" calls to cover dump_location_t.
800 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
802 (dump_user_location_t::dump_user_location_t): New constructors,
803 from gimple *stmt and rtx_insn *.
804 (dump_user_location_t::from_function_decl): New function.
805 (dump_loc): Make static.
806 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
807 const dump_location_t &.
808 (dump_generic_expr_loc): Delete.
809 (dump_printf_loc): Convert param "loc" from location_t to
810 const dump_location_t &.
811 (selftest::test_impl_location): New function.
812 (selftest::dumpfile_c_tests): New function.
813 * dumpfile.h: Include "profile-count.h".
814 (class dump_user_location_t): New class.
815 (struct dump_impl_location_t): New struct.
816 (class dump_location_t): New class.
817 (dump_printf_loc): Convert 2nd param from source_location to
818 const dump_location_t &.
819 (dump_generic_expr_loc): Delete.
820 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
821 const dump_location_t &.
822 * gimple-fold.c (fold_gimple_assign): Update call to
823 dump_printf_loc to pass in a dump_location_t rather than a
824 location_t, via the gimple stmt.
825 (gimple_fold_call): Likewise.
826 * gimple-loop-interchange.cc
827 (loop_cand::analyze_iloop_reduction_var): Update for change to
828 check_reduction_path.
829 (tree_loop_interchange::interchange): Update for change to
831 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
832 change in return-type of find_loop_location.
833 (graphite_regenerate_ast_isl): Likewise.
834 * graphite-optimize-isl.c (optimize_isl): Likewise.
835 * graphite.c (graphite_transform_loops): Likewise.
836 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
837 pass in a dump_location_t rather than a location_t, via the
839 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
840 * ipa.c (walk_polymorphic_call_targets): Likewise.
841 * loop-unroll.c (report_unroll): Convert "locus" param from
842 location_t to dump_location_t.
843 (decide_unrolling): Update for change to get_loop_location's
845 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
846 location_t to dump_user_location_t.
847 (grid_find_single_omp_among_assignments_1): Updates calls to
848 dump_printf_loc to pass in a dump_location_t rather than a
849 location_t, via the gimple stmt.
850 (grid_parallel_clauses_gridifiable): Convert "tloc" from
851 location_t to dump_location_t. Updates calls to dump_printf_loc
852 to pass in a dump_location_t rather than a location_t, via the
854 (grid_inner_loop_gridifiable_p): Likewise.
855 (grid_dist_follows_simple_pattern): Likewise.
856 (grid_gfor_follows_tiling_pattern): Likewise.
857 (grid_target_follows_gridifiable_pattern): Likewise.
858 (grid_attempt_target_gridification): Convert initialization
859 of local "grid" from memset to zero-initialization; FIXME: does
860 this require C++11? Update call to dump_printf_loc to pass in a
861 optinfo_location rather than a location_t, via the gimple stmt.
862 * profile.c (read_profile_edge_counts): Updates call to
863 dump_printf_loc to pass in a dump_location_t rather than a
865 (compute_branch_probabilities): Likewise.
866 * selftest-run-tests.c (selftest::run_tests): Call
868 * selftest.h (dumpfile_c_tests): New decl.
869 * tree-loop-distribution.c (pass_loop_distribution::execute):
870 Update for change in return type of find_loop_location.
871 * tree-parloops.c (parallelize_loops): Likewise.
872 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
873 "locus" from location_t to dump_user_location_t.
874 (canonicalize_loop_induction_variables): Likewise.
875 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
876 for change in return type of find_loop_location.
877 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
878 to dump_printf_loc to pass in a dump_location_t rather than a
879 location_t, via the stmt.
880 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
882 * tree-vect-loop-manip.c (find_loop_location): Convert return
883 type from source_location to dump_user_location_t.
884 (vect_do_peeling): Update for above change.
885 (vect_loop_versioning): Update for change in type of
887 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
888 from location_t to dump_user_location_t.
889 (vect_estimate_min_profitable_iters): Update for change in type
891 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
892 location_t to dump_location_t.
893 (vect_slp_bb): Update for change in type of vect_location.
894 * tree-vectorizer.c (vect_location): Convert from source_location
895 to dump_user_location_t.
896 (try_vectorize_loop_1): Update for change in vect_location's type.
897 (vectorize_loops): Likewise.
898 (increase_alignment): Likewise.
899 * tree-vectorizer.h (vect_location): Convert from source_location
900 to dump_user_location_t.
901 (find_loop_location): Convert return type from source_location to
902 dump_user_location_t.
903 (check_reduction_path): Convert 1st param from location_t to
904 dump_user_location_t.
905 * value-prof.c (check_counter): Update call to dump_printf_loc to
906 pass in a dump_user_location_t rather than a location_t; update
907 call to error_at for change in type of "locus".
908 (check_ic_target): Update call to dump_printf_loc to
909 pass in a dump_user_location_t rather than a location_t, via the
912 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
914 * config/s390/s390.h (enum processor_flags): Do not use
915 default tune parameter when -march was specified.
917 2018-06-26 Jakub Jelinek <jakub@redhat.com>
920 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
921 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
924 2018-06-26 Richard Biener <rguenther@suse.de>
926 PR tree-optimization/86287
928 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
929 (vect_analyze_loop): Initialize n_stmts.
931 2018-06-26 Richard Biener <rguenther@suse.de>
934 * fold-const.c (fold_convertible_p): Pointer extension
937 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
940 * dwarf2out.c (loc_list_has_views): Adjust comments.
941 (dw_loc_list): Split single cross-partition range with
944 2018-06-25 Jeff Law <law@redhat.com>
946 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
947 on -mbig-switch by default.
949 * config/v850/predicates.md (const_float_1_operand): Fix match_code
951 (const_float_0_operand): Remove unused predicate.
952 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
953 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
954 (recipsf2): New expander. Original pattern now called
956 (recipdf2, recipdf2_insn): Similarly.
957 (rsqrtsf2, rsqrtsf2_insn): Similarly
958 (rsqrtdf2, rsqrtdf2_insn): Similarly
960 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
962 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
963 Simplify logic for FreeBSD (twice).
965 2018-06-25 Martin Sebor <msebor@redhat.com>
967 PR tree-optimization/86204
968 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
969 a strnlen result if it's less than the length of the string.
971 2018-06-25 Martin Sebor <msebor@redhat.com>
973 PR tree-optimization/85700
974 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
975 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
976 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
978 2018-06-25 Martin Sebor <msebor@redhat.com>
980 * doc/extend.texi (Zero-length arrays): Update and clarify.
982 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
984 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
985 added IEEE/IBM long double multilib support on PowerPC little
986 endian Linux systems.
987 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
988 (MULTILIB_DEFAULTS): Likewise.
989 * config/rs6000/rs6000.c (rs6000_option_override_internal):
991 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
992 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
993 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
995 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
998 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
999 (REORDER_45): Likewise.
1001 2018-06-25 Jeff Law <law@redhat.com>
1003 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1004 dividend to 32 bits. Adjust length.
1005 (udivmodhi4): Cleanup output template. Fix length.
1007 2018-06-25 Carl Love <cel@us.ibm.com>
1009 * config/rs6000/vsx.md: Change word selector to prefered location.
1011 2018-06-25 Richard Biener <rguenther@suse.de>
1013 PR tree-optimization/86304
1014 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1015 epilogue-if-converted loops as well.
1017 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1019 * lto-section-out.c (lto_begin_section): Do not print section
1020 name for noaddr and unnumbered dumps.
1022 2018-06-25 Richard Biener <rguenther@suse.de>
1024 * tree-vectorizer.h (struct vec_info_shared): New structure
1025 with parts split out from struct vec_info and loop_nest from
1026 struct _loop_vec_info.
1027 (struct vec_info): Adjust accordingly.
1028 (struct _loop_vec_info): Likewise.
1029 (LOOP_VINFO_LOOP_NEST): Adjust.
1030 (LOOP_VINFO_DATAREFS): Likewise.
1031 (LOOP_VINFO_DDRS): Likewise.
1032 (struct _bb_vec_info): Likewise.
1033 (BB_VINFO_DATAREFS): Likewise.
1034 (BB_VINFO_DDRS): Likewise.
1035 (struct _stmt_vec_info): Add dr_aux member.
1036 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1037 (DR_MISALIGNMENT_UNINITIALIZED): New.
1038 (set_dr_misalignment): Adjust.
1039 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1040 (vect_analyze_loop): Adjust prototype.
1041 (vect_analyze_loop_form): Likewise.
1042 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1043 Compute dependences lazily.
1044 (vect_record_base_alignments): Use shared datarefs/ddrs.
1045 (vect_verify_datarefs_alignment): Likewise.
1046 (vect_analyze_data_refs_alignment): Likewise.
1047 (vect_analyze_data_ref_accesses): Likewise.
1048 (vect_analyze_data_refs): Likewise.
1049 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1050 constructor parameter for shared part.
1051 (vect_analyze_loop_form): Pass in shared part and adjust.
1052 (vect_analyze_loop_2): Pass in storage for the number of
1053 stmts. Move loop nest finding to the caller. Compute
1055 (vect_analyze_loop): Pass in shared part.
1056 (vect_transform_loop): Verify shared datarefs are unchanged.
1057 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1058 constructor parameter for shared part.
1059 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1060 (vect_slp_bb): Verify shared datarefs are unchanged before
1062 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1064 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1065 DR_MISALIGNMENT_UNINITIALIZED.
1066 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1067 parameter for shared part.
1068 (vec_info::~vec_info): Adjust.
1069 (vec_info_shared::vec_info_shared): New.
1070 (vec_info_shared::~vec_info_shared): Likewise.
1071 (vec_info_shared::save_datarefs): Likewise.
1072 (vec_info_shared::check_datarefs): Likewise.
1073 (try_vectorize_loop_1): Construct shared part live for analyses
1074 of a single loop for multiple vector sizes.
1075 * tree-parloops.c (gather_scalar_reductions): Adjust.
1077 2018-06-25 Richard Biener <rguenther@suse.de>
1079 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1080 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1081 (vect_analyze_data_refs): Remove similar code from here and
1082 simplify accordingly.
1084 2018-06-25 Richard Biener <rguenther@suse.de>
1086 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1087 for reverse storage order accesses rather than asserting
1088 they cannot happen here.
1090 2018-06-25 Tom de Vries <tdevries@suse.de>
1093 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1094 Use data16 instead of .byte for insn prefix.
1096 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1099 * parser.c (make_char_string_pack): Pass this literal chars
1100 through cpp_interpret_string.
1101 (cp_parser_userdef_numeric_literal): Check the result of
1102 make_char_string_pack.
1104 2018-06-24 Maya Rashish <coypu@sdf.org>
1106 * ginclude/stddef.h: Simplify conditions around avoiding
1107 re-definition of __size_t.
1109 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1111 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1112 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1114 2018-06-22 Maya Rashish <coypu@sdf.org>
1116 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1117 made to OpenBSD and NetBSD through the years.
1119 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1121 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1122 behavior of vec_pack (vector double, vector double) to match
1123 behavior of vec_float2 (vector double, vector double).
1125 2018-06-22 Olivier Hainque <hainque@adacore.com>
1127 * gimplify.c (gimplify_function_tree): Prevent creation
1128 of a trampoline for the address of the current function
1129 passed to entry/exit instrumentation hooks.
1131 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1134 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1137 2018-06-22 Martin Liska <mliska@suse.cz>
1139 PR tree-optimization/86263
1140 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1141 Bail out if is_enabled is false.
1142 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1144 (jump_table_cluster::is_enabled): New function.
1146 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1148 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1149 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1150 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1151 (lto_input_ts_binfo_tree_pointers): Likewise.
1152 * tree-streamer-out.c (streamer_write_tree_bitfields,
1153 write_ts_binfo_tree_pointers): Likewise.
1154 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1156 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1158 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1160 2018-06-22 Martin Liska <mliska@suse.cz>
1162 * symbol-summary.h (get): Make it pure and inline move
1163 functionality from ::get function.
1164 (get): Remove and inline into ::get and ::get_create.
1165 (get_create): Move code from ::get function.
1167 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1170 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1171 -x assembler-with-cpp.
1173 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1175 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1176 _FILE_OFFSET_BITS=64 for C++.
1178 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1180 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1181 conversion insn that shows up when pr85657-3.c is compiled using
1182 IEEE 128-bit long double.
1183 (neg<mode>2_internal): Use the correct mode to check whether the
1184 mode is IBM extended.
1185 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1186 multiply and divide external functions from being created more
1189 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1191 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1193 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1194 the edge can be forwarded.
1195 (cfg_layout_merge_blocks): Likewise.
1197 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1199 * except.c (finish_eh_generation): Commit edge insertions only after
1200 the EH edges have been redirected from post-landing to landing pads.
1202 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1204 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1205 create_tmp_var_for to create the FRAME decl.
1206 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1208 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1210 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1211 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1212 debug statement when resetting its value.
1213 (expand_call_inline): Copy the locus of the call onto the assignment
1214 of the return value, if any. Use local variable in more cases.
1216 2018-06-21 Martin Liska <mliska@suse.cz>
1218 * ipa-pure-const.c (propagate_nothrow): Use
1219 funct_state_summaries->get.
1220 (dump_malloc_lattice): Likewise.
1221 (propagate_malloc): Likewise.
1223 2018-06-21 Richard Biener <rguenther@suse.de>
1225 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1226 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1227 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1229 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1230 BLOCK_ABSTRACT_ORIGIN unconditionally.
1232 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1234 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1236 * ipa-reference.c (ipa_reference_c_finalize): Delete
1237 ipa_ref_opt_sum_summaries and set it to NULL.
1239 2018-06-21 Tom de Vries <tdevries@suse.de>
1241 PR tree-optimization/85859
1242 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1243 test with comment from bb_no_side_effects_p.
1245 2018-06-21 Richard Biener <rguenther@suse.de>
1247 PR tree-optimization/86232
1248 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1249 max for constant niter.
1251 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1253 * config/aarch64/aarch64-simd.md
1254 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1256 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1258 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1259 Make opernads of the unspec commutative.
1261 2018-06-21 Richard Biener <rguenther@suse.de>
1263 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1264 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1265 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1266 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1268 (vect_record_base_alignments): Do not record base alignment
1269 for gathers or scatters.
1270 (vect_compute_data_ref_alignment): Drop return value that is always
1271 true. Bail out early for gathers or scatters.
1272 (vect_enhance_data_refs_alignment): Bail out early for gathers
1274 (vect_find_same_alignment_drs): Likewise.
1275 (vect_analyze_data_refs_alignment): Remove dead code.
1276 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1277 (vect_analyze_data_refs): For possible gathers or scatters do
1278 not create an alternate DR, just check their possible validity
1279 and mark them. Adjust DECL_NONALIASED handling to not rely
1281 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1282 update inits of gathers or scatters.
1283 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1284 Also copy gather/scatter flag to pattern vinfo.
1286 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1288 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1289 behavior of vec_packsu (vector unsigned long long, vector unsigned
1290 long long) to match behavior of vec_packs with same signature.
1292 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1293 Thomas Schwinge <thomas@codesourcery.com>
1294 Cesar Philippidis <cesar@codesourcery.com>
1296 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1297 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1298 (gimplify_adjust_omp_clauses): Likewise.
1299 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1300 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1301 (gimplify_omp_target_update): Update handling of acc update and
1303 * omp-low.c (install_var_field): Remove unused parameter
1304 base_pointers_restrict.
1305 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1306 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1308 (omp_target_base_pointers_restrict_p): Delete.
1309 (scan_omp_target): Update call to scan_sharing_clauses.
1310 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1312 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1313 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1314 (convert_local_omp_clauses): Likewise.
1315 * tree-pretty-print.c (dump_omp_clause): Likewise.
1316 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1318 (omp_clause_code_name): Likewise.
1320 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1323 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1326 PR tree-optimization/86231
1327 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1328 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1330 2018-06-20 Tom de Vries <tdevries@suse.de>
1332 PR tree-optimization/86097
1333 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1334 iv type if signedness of iv type is not the same as that of *nit.
1336 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1338 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1339 EDGE_EH edges, verify they are all EDGE_EH.
1341 2018-06-20 Maya Rashish <coypu@sdf.org>
1343 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1345 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1347 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1348 * config/aarch64/aarch64.c (xgene1_tunings): Add
1349 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1350 (aarch64_mode_valid_for_sched_fusion_p):
1351 Allow 16-byte modes.
1352 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1353 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1355 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1357 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1358 * config/aarch64/iterators.md (VQ2): New mode iterator.
1360 2018-06-20 Martin Liska <mliska@suse.cz>
1362 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1363 Change default ratio from 10 to 8.
1365 2018-06-20 Martin Liska <mliska@suse.cz>
1367 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1369 (bit_test_cluster::find_bit_tests): Likewise.
1370 (switch_decision_tree::analyze_switch_statement): Find clusters.
1371 * tree-switch-conversion.h (struct jump_table_cluster): Document
1374 2018-06-20 Martin Liska <mliska@suse.cz>
1376 * tree-switch-conversion.c (switch_conversion::collect):
1377 Record m_uniq property.
1378 (switch_conversion::expand): Bail out for special conditions.
1379 (group_cluster::~group_cluster): New.
1380 (group_cluster::group_cluster): Likewise.
1381 (group_cluster::dump): Likewise.
1382 (jump_table_cluster::emit): New.
1383 (switch_decision_tree::fix_phi_operands_for_edges): New.
1384 (struct case_node): Remove struct.
1385 (jump_table_cluster::can_be_handled): New.
1386 (case_values_threshold): Moved to header.
1387 (reset_out_edges_aux): Likewise.
1388 (jump_table_cluster::is_beneficial): New.
1389 (bit_test_cluster::can_be_handled): Likewise.
1390 (add_case_node): Remove.
1391 (bit_test_cluster::is_beneficial): New.
1392 (case_bit_test::cmp): New.
1393 (bit_test_cluster::emit): New.
1394 (expand_switch_as_decision_tree_p): Remove.
1395 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1396 (fix_phi_operands_for_edge): Likewise.
1397 (switch_decision_tree::analyze_switch_statement): New.
1398 (compute_cases_per_edge): Move ...
1399 (switch_decision_tree::compute_cases_per_edge): ... here.
1400 (try_switch_expansion): Likewise.
1401 (switch_decision_tree::try_switch_expansion): Likewise.
1402 (record_phi_operand_mapping): Likewise.
1403 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1404 (emit_case_decision_tree): Likewise.
1405 (switch_decision_tree::emit): Likewise.
1406 (balance_case_nodes): Likewise.
1407 (switch_decision_tree::balance_case_nodes): Likewise.
1408 (dump_case_nodes): Likewise.
1409 (switch_decision_tree::dump_case_nodes): Likewise.
1410 (emit_jump): Likewise.
1411 (switch_decision_tree::emit_jump): Likewise.
1412 (emit_cmp_and_jump_insns): Likewise.
1413 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1414 (emit_case_nodes): Likewise.
1415 (switch_decision_tree::emit_case_nodes): Likewise.
1416 (conditional_probability): Remove.
1417 * tree-switch-conversion.h (enum cluster_type): New.
1419 (struct cluster): Likewise.
1420 (cluster::cluster): Likewise.
1421 (struct simple_cluster): Likewise.
1422 (simple_cluster::simple_cluster): Likewise.
1423 (struct group_cluster): Likewise.
1424 (struct jump_table_cluster): Likewise.
1425 (struct bit_test_cluster): Likewise.
1426 (struct min_cluster_item): Likewise.
1427 (struct case_tree_node): Likewise.
1428 (case_tree_node::case_tree_node): Likewise.
1429 (jump_table_cluster::case_values_threshold): Likewise.
1430 (struct case_bit_test): Likewise.
1431 (struct switch_decision_tree): Likewise.
1432 (struct switch_conversion): Likewise.
1433 (switch_decision_tree::reset_out_edges_aux): Likewise.
1435 2018-06-20 Martin Liska <mliska@suse.cz>
1437 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1438 (hoist_edge_and_branch_if_true): Likewise.
1439 (expand_switch_using_bit_tests_p): Likewise.
1440 (struct case_bit_test): Likewise.
1441 (case_bit_test_cmp): Likewise.
1442 (emit_case_bit_tests): Likewise.
1443 (switch_conversion::switch_conversion): New class.
1444 (struct switch_conv_info): Remove old struct.
1445 (collect_switch_conv_info): More to ...
1446 (switch_conversion::collect): ... this.
1447 (check_range): Likewise.
1448 (switch_conversion::check_range): Likewise.
1449 (check_all_empty_except_final): Likewise.
1450 (switch_conversion::check_all_empty_except_final): Likewise.
1451 (check_final_bb): Likewise.
1452 (switch_conversion::check_final_bb): Likewise.
1453 (create_temp_arrays): Likewise.
1454 (switch_conversion::create_temp_arrays): Likewise.
1455 (free_temp_arrays): Likewise.
1456 (gather_default_values): Likewise.
1457 (switch_conversion::gather_default_values): Likewise.
1458 (build_constructors): Likewise.
1459 (switch_conversion::build_constructors): Likewise.
1460 (constructor_contains_same_values_p): Likewise.
1461 (switch_conversion::contains_same_values_p): Likewise.
1462 (array_value_type): Likewise.
1463 (switch_conversion::array_value_type): Likewise.
1464 (build_one_array): Likewise.
1465 (switch_conversion::build_one_array): Likewise.
1466 (build_arrays): Likewise.
1467 (switch_conversion::build_arrays): Likewise.
1468 (gen_def_assigns): Likewise.
1469 (switch_conversion::gen_def_assigns): Likewise.
1470 (prune_bbs): Likewise.
1471 (switch_conversion::prune_bbs): Likewise.
1472 (fix_phi_nodes): Likewise.
1473 (switch_conversion::fix_phi_nodes): Likewise.
1474 (gen_inbound_check): Likewise.
1475 (switch_conversion::gen_inbound_check): Likewise.
1476 (process_switch): Use the newly created class.
1477 (switch_conversion::expand): New.
1478 (switch_conversion::~switch_conversion): New.
1479 * tree-switch-conversion.h: New file.
1481 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1483 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1484 tree-vect-patterns.c.
1485 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1486 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1487 (vect_recog_sad_pattern): Likewise.
1488 (vect_recog_widen_sum_pattern): Likewise.
1489 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1490 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1491 (vect_recog_rotate_pattern): Likewise.
1492 (vect_recog_mult_pattern): Likewise.
1493 (vect_recog_vector_vector_shift_pattern): Likewise.
1494 (vect_recog_divmod_pattern): Likewise.
1495 (vect_recog_mixed_size_cond_pattern): Likewise.
1496 (vect_recog_bool_pattern): Likewise.
1497 (vect_recog_mask_conversion_pattern): Likewise.
1498 (vect_try_gather_scatter_pattern): Likewise.
1499 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1500 (vect_recog_over_widening_pattern): Likewise.
1501 (vect_recog_gather_scatter_pattern): Likewise.
1502 (vect_recog_func_ptr): Move from tree-vectorizer.h
1503 (vect_vect_recog_func_ptrs): Move further down the file.
1504 (vect_recog_func): Likewise. Remove the third argument.
1505 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1506 (vect_pattern_recog_1): Expect the pattern function to do any
1507 necessary target tests. Also expect it to provide a vector type.
1508 Remove the type_in handling.
1510 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1512 * tree-vect-patterns.c (vect_pattern_detected): New function.
1513 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1514 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1515 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1516 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1517 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1518 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1519 (vect_recog_mask_conversion_pattern)
1520 (vect_try_gather_scatter_pattern): Likewise.
1522 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1524 * tree-vect-patterns.c (vect_get_internal_def): New function.
1525 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1526 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1527 (search_type_for_mask_1): Use it.
1529 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1531 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1532 redundant WIDEN_SUM_EXPR handling.
1533 (vect_recog_sad_pattern): Likewise.
1535 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1537 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1538 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1539 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1540 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1541 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1543 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1545 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1546 use the stmt_vec_info of the original bb statement for the
1547 new zero assignment, even if the call is part of a pattern.
1549 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1551 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1552 that the sequence is attached to the original statement rather
1553 than the pattern statement.
1554 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1555 PATTERN_DEF_SEQ from the original statement rather than
1556 the main pattern statement.
1557 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1558 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1559 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1561 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1563 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1564 definition statements before the early exit for statements that aren't
1566 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1568 (vect_transform_loop): ...here. Process pattern definition
1569 statements without first checking whether the main pattern
1570 statement is live or relevant.
1572 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1574 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1575 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1577 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1579 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1580 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1581 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1582 (expand_block_compare): Change select_block_compare_mode call.
1583 (expand_strncmp_align_check): Use new functions, fix comment.
1584 (emit_final_str_compare_gpr): New function.
1585 (expand_strn_compare): Refactor and clean up code.
1586 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1588 2018-06-19 Tony Reix <tony.reix@atos.com>
1589 Damien Bergamini <damien.bergamini@atos.com>
1590 David Edelsohn <dje.gcc@gmail.com>
1592 * collect2.c (static_obj): New variable.
1593 (static_libs): New variable.
1594 (is_in_list): Uncomment declaration.
1595 (main): Track AIX libraries linked statically.
1596 (is_in_list): Uncomment definition.
1597 (scan_prog_file): Don't add AIX shared libraries initializer
1598 to constructor list if linking statically.
1600 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1602 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1604 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1606 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1608 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1611 2018-06-19 Martin Liska <mliska@suse.cz>
1613 * config/i386/i386.c (ix86_can_inline_p): Do not use
1614 ipa_fn_summaries::get_create.
1615 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1617 (devirtualization_time_bonus): Likewise.
1618 (ipcp_propagate_stage): Likewise.
1619 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1620 (edge_set_predicate): Likewise.
1621 (evaluate_conditions_for_known_args): Likewise.
1622 (evaluate_properties_for_edge): Likewise.
1623 (ipa_call_summary::reset): Tranform to ...
1624 (ipa_call_summary::~ipa_call_summary): ... this.
1625 (ipa_fn_summary::reset): Transform to ...
1626 (ipa_fn_summary::~ipa_fn_summary): ... this.
1627 (ipa_fn_summary_t::remove): Rename to ...
1628 (ipa_fn_summary_t::remove_callees): ... this.
1629 (ipa_fn_summary_t::duplicate): Use placement new
1630 instead of memory copy.
1631 (ipa_call_summary_t::duplicate): Likewise.
1632 (ipa_call_summary_t::remove): Remove.
1633 (dump_ipa_call_summary): Change get_create to get.
1634 (ipa_dump_fn_summary): Dump only when summary exists.
1635 (analyze_function_body): Use symbol_summary::get instead
1637 (compute_fn_summary): Likewise.
1638 (estimate_edge_devirt_benefit): Likewise.
1639 (estimate_edge_size_and_time): Likewise.
1640 (inline_update_callee_summaries): Likewise.
1641 (remap_edge_change_prob): Likewise.
1642 (remap_edge_summaries): Likewise.
1643 (ipa_merge_fn_summary_after_inlining): Likewise.
1644 (write_ipa_call_summary): Likewise.
1645 (ipa_fn_summary_write): Likewise.
1646 (ipa_free_fn_summary): Likewise.
1647 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1648 (struct ipa_call_summary): Likewise.
1649 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1651 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1652 (estimate_size_after_inlining): Likewise.
1653 (estimate_growth): Likewise.
1654 (growth_likely_positive): Likewise.
1655 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1656 (inline_call): Likewise.
1657 * ipa-inline.c (caller_growth_limits): Likewise.
1658 (can_inline_edge_p): Likewise.
1659 (can_inline_edge_by_limits_p): Likewise.
1660 (compute_uninlined_call_time): Likewise.
1661 (compute_inlined_call_time): Likewise.
1662 (want_inline_small_function_p): Likewise.
1663 (edge_badness): Likewise.
1664 (update_caller_keys): Likewise.
1665 (update_callee_keys): Likewise.
1666 (inline_small_functions): Likewise.
1667 (inline_to_all_callers_1): Likewise.
1668 (dump_overall_stats): Likewise.
1669 (early_inline_small_functions): Likewise.
1670 (early_inliner): Likewise.
1671 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1672 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1673 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1674 * ipa-split.c (execute_split_functions): Likewise.
1675 * symbol-summary.h: Likewise.
1676 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1678 2018-06-19 Richard Biener <rguenther@suse.de>
1680 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1681 (vectorize_loops): ... here. Fix dbgcnt handling.
1682 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1684 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1687 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1688 ieee128 argument takes up only one (vector) register, not two (floating
1691 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1693 * gimplify.c (gimplify_init_constructor): Really never clear for an
1694 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1696 2018-06-19 Richard Biener <rguenther@suse.de>
1698 PR tree-optimization/86179
1699 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1700 after failed recognition.
1702 2018-06-18 Martin Sebor <msebor@redhat.com>
1705 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1706 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1707 Handle integer subtraction.
1708 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1709 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1711 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1713 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1714 param from rtx to rtx_insn *.
1715 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1717 (frv_ifcvt_modify_insn): Likwise.
1718 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1719 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1720 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1721 as_a <rtx_insn *> cast to local "unprotected_region" once
1722 it's been established that it's not NULL or pc_rtx.
1723 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1724 param "sethi" from rtx to rtx_insn *.
1725 (nds32_group_float_insns): Likewise for param "insn".
1726 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1728 (vax_output_int_subtract): Likewise.
1729 * config/vax/vax.c (vax_output_int_add): Likewise for param
1731 (vax_output_int_subtract): Likewise.
1732 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1733 (emit_pattern_after): Likewise for param "after".
1734 (emit_insn_after): Likewise.
1735 (emit_jump_insn_after): Likewise.
1736 (emit_call_insn_after): Likewise.
1737 (emit_debug_insn_after): Likewise.
1738 (emit_pattern_before): Likewise for param "before".
1739 (emit_insn_before): Likewise.
1740 (emit_jump_insn_before): Likewise.
1741 * final.c (get_insn_template): Likewise for param "insn", removing
1743 * output.h (get_insn_template): Likewise for 2nd param.
1744 * rtl.h (emit_insn_before): Likewise.
1745 (emit_jump_insn_before): Likewise.
1746 (emit_debug_insn_before_noloc): Likewise.
1747 (emit_insn_after): Likewise.
1748 (emit_jump_insn_after): Likewise.
1749 (emit_call_insn_after): Likewise.
1750 (emit_debug_insn_after): Likewise.
1751 (set_insn_deleted): Likewise for param.
1753 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1756 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1757 floating point modes, so that IFmode is numerically greater than
1758 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1759 to declare the ordering. This prevents IFmode from being
1760 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1761 machine. Include rs6000-modes.h to share the fractional values
1762 between genmodes* and the rest of the compiler.
1766 * config/rs6000/rs6000-modes.h: New file.
1767 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1768 meaning of rs6000_long_double_size so that 126..128 selects an
1769 appropriate 128-bit floating point type.
1770 (rs6000_option_override_internal): Likewise.
1771 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1772 (TARGET_LONG_DOUBLE_128): Change the meaning of
1773 rs6000_long_double_size so that 126..128 selects an appropriate
1774 128-bit floating point type.
1775 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1776 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1777 source and destination to match the standard usage.
1778 (truncifkf2): Likewise.
1779 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1780 ISA 2.07 to use an explicit clobber, instead of passing in a
1782 (copysign<mode>3_soft): Likewise.
1784 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1786 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1787 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1788 (vect_slp_analyze_instance_dependence): Likewise.
1789 (vect_enhance_data_refs_alignment): Likewise.
1790 (vect_analyze_data_refs_alignment): Likewise.
1791 (vect_slp_analyze_and_verify_instance_alignment
1792 (vect_analyze_data_ref_accesses): Likewise.
1793 (vect_prune_runtime_alias_test_list): Likewise.
1794 (vect_analyze_data_refs): Likewise.
1795 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1796 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1797 (vect_analyze_scalar_cycles_1): Likewise.
1798 (vect_get_loop_niters): Likewise.
1799 (vect_analyze_loop_form_1): Likewise.
1800 (vect_update_vf_for_slp): Likewise.
1801 (vect_analyze_loop_operations): Likewise.
1802 (vect_analyze_loop): Likewise.
1803 (vectorizable_induction): Likewise.
1804 (vect_transform_loop): Likewise.
1805 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1806 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1807 (vect_make_slp_decision): Likewise.
1808 (vect_detect_hybrid_slp): Likewise.
1809 (vect_slp_analyze_operations): Likewise.
1810 (vect_slp_bb): Likewise.
1811 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1812 (vectorizable_bswap): Likewise.
1813 (vectorizable_call): Likewise.
1814 (vectorizable_simd_clone_call): Likewise.
1815 (vectorizable_conversion): Likewise.
1816 (vectorizable_assignment): Likewise.
1817 (vectorizable_shift): Likewise.
1818 (vectorizable_operation): Likewise.
1819 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1821 2018-06-18 Martin Sebor <msebor@redhat.com>
1823 PR tree-optimization/81384
1824 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1825 * builtins.c (expand_builtin_strnlen): New function.
1826 (expand_builtin): Call it.
1827 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1828 * builtins.def (BUILT_IN_STRNLEN): New.
1829 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1830 Warn for bounds in excess of maximum object size.
1831 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1832 single-value ranges. Handle strnlen.
1833 (handle_builtin_strlen): Handle strnlen.
1834 (strlen_check_and_optimize_stmt): Same.
1835 * doc/extend.texi (Other Builtins): Document strnlen.
1837 2018-06-18 Maya Rashish <coypu@sdf.org>
1839 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1840 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1841 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1843 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1845 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1847 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1849 * tree.c (escaped_string::escape): Replace cast to char * by
1850 const_cast<char *> (unescaped).
1852 2018-06-18 Nick Clifton <nickc@redhat.com>
1855 * tree.c (escaped_string): New class. Converts an unescaped
1856 string into its escaped equivalent.
1857 (warn_deprecated_use): Use the new class to convert the
1858 deprecation message, if present.
1859 (test_escaped_strings): New self test.
1860 (test_c_tests): Add test_escaped_strings.
1861 * doc/extend.texi (deprecated): Add a note that the
1862 deprecation message is affected by the -fmessage-length
1863 option, and that control characters will be escaped.
1864 (#pragma GCC error): Document this pragma.
1865 (#pragma GCC warning): Likewise.
1866 * doc/invoke.texi (-fmessage-length): Document this option's
1867 effect on the #warning and #error preprocessor directives and
1868 the deprecated attribute.
1870 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1872 * tree.c (decl_value_expr_lookup): Revert latest change.
1873 (decl_value_expr_insert): Likewise.
1875 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1877 * gimplify.c (nonlocal_vlas): Delete.
1878 (nonlocal_vla_vars): Likewise.
1879 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1881 (gimplify_body): Do not create and destroy nonlocal_vlas.
1882 * tree-nested.c: Include diagnostic.h.
1883 (use_pointer_in_frame): Tweak.
1884 (lookup_field_for_decl): Add assertion and declare the transformation.
1885 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1886 internal error when the reference is in a wrong context. Do not
1887 create a debug decl by default.
1888 (note_nonlocal_block_vlas): Delete.
1889 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1890 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1891 create a debug decl by default.
1892 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1893 call is in a wrong context.
1894 (fixup_vla_decls): New function.
1895 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1896 debug variables were created.
1897 * tree.c (decl_value_expr_lookup): Add checking assertion.
1898 (decl_value_expr_insert): Likewise.
1900 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1903 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1904 * tree-scalar-evolution.c (interpret_expr): Likewise.
1905 (expression_expensive_p): Likewise.
1906 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1907 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1908 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1909 (ssa_defined_by_minus_one_stmt_p): New.
1911 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1914 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1915 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1916 * dojump.c (do_jump): Likewise.
1917 * expr.c (expand_expr_real_2): Check operand type's sign.
1918 * fold-const.c (const_unop): Handle ABSU_EXPR.
1919 (fold_abs_const): Likewise.
1920 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1921 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1922 (strip_sign_op_1): Likesise.
1923 * match.pd: Add new pattern to generate ABSU_EXPR.
1924 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1925 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1926 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1927 * tree-inline.c (estimate_operator_cost): Likewise.
1928 * tree-pretty-print.c (dump_generic_node): Likewise.
1929 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1930 * tree.def (ABSU_EXPR): New.
1932 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1935 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1936 documented as preserved for backward compatibility only.
1937 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1939 PR rtl-optimization/86108
1940 * bb-reorder.c (create_forwarder_block): Renamed to ...
1941 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1942 jump from new landing pad to the second part.
1943 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1946 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1949 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1950 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1951 Only call store_expr for halves if the mode is the same.
1954 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1955 Fix up comment formatting.
1957 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1959 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1960 (typed_splay_tree::closure,
1961 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1962 (typed_splay_tree::typed_splay_tree,
1963 typed_splay_tree::operator =): Declared private.
1964 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1965 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1966 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1967 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1968 typed_splay_tree::splay_tree_splay,
1969 typed_splay_tree::splay_tree_foreach_helper,
1970 typed_splay_tree::splay_tree_insert,
1971 typed_splay_tree::splay_tree_remove,
1972 typed_splay_tree::splay_tree_lookup,
1973 typed_splay_tree::splay_tree_predecessor,
1974 typed_splay_tree::splay_tree_successor,
1975 typed_splay_tree::splay_tree_min,
1976 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1977 (typed_splay_tree::root, typed_splay_tree::comp,
1978 typed_splay_tree::delete_key,
1979 typed_splay_tree::delete_value): New data members.
1980 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1981 typed_splay_tree::remove.
1983 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1985 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1986 -mginv and -mno-ginv to the assembler.
1987 * config/mips/mips.opt (-mcrc): New option.
1989 * doc/invoke.text (-mcrc): Document.
1992 2018-06-15 Nick Clifton <nickc@redhat.com>
1995 * tree.c (escaped_string): New class. Converts an unescaped
1996 string into its escaped equivalent.
1997 (warn_deprecated_use): Use the new class to convert the
1998 deprecation message, if present.
1999 (test_escaped_strings): New self test.
2000 (test_c_tests): Add test_escaped_strings.
2001 * doc/extend.texi (deprecated): Add a note that the
2002 deprecation message is affected by the -fmessage-length
2003 option, and that control characters will be escaped.
2004 (#pragma GCC error): Document this pragma.
2005 (#pragma GCC warning): Likewise.
2006 * doc/invoke.texi (-fmessage-length): Document this option's
2007 effect on the #warning and #error preprocessor directives and
2008 the deprecated attribute.
2010 2018-06-15 Richard Biener <rguenther@suse.de>
2012 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2013 here, also noting vector size used.
2014 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2015 size used in MSG_OPTIMIZED_LOCATIONS dump.
2016 (pass_slp_vectorize::execute): Adjust.
2018 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2021 * config/arc/arc.c (arc_return_address_register): Fix
2024 2018-06-15 Richard Biener <rguenther@suse.de>
2027 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2028 leave useless conversion stripping to force_gimple_operand_gsi.
2029 (gimplify_build2): Likewise.
2030 (gimplify_build1): Likewise.
2032 2018-06-15 Richard Biener <rguenther@suse.de>
2035 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2036 before adjusting their block.
2038 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2040 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2041 multilibs for *-*-rtems*.
2042 * config/riscv/t-rtems: New file.
2044 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2047 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2048 unsigned_type_for returns NULL.
2051 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2052 subregs of multi-word pseudos unless the float mode has word size.
2054 2018-06-14 Richard Biener <rguenther@suse.de>
2057 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2058 duplicate and harmful type_hash_canon.
2059 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2061 2018-06-14 Richard Biener <rguenther@suse.de>
2064 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2067 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2069 * config/rtems.h (STDINT_LONG32): Define.
2071 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2072 Prachi Godbole <prachi.godbole@imgtec.com>
2074 * config/mips/mips-cpus.def: Define P6600.
2075 * config/mips/mips-tables.opt: Regenerate.
2076 * config/mips/mips.c (mips_ucbranch_type): New enum.
2077 (mips_rtx_cost_data): Add support for P6600.
2078 (mips_issue_rate): Likewise.
2079 (mips_multipass_dfa_lookahead): Likewise.
2080 (mips_avoid_hazard): Likewise.
2081 (mips_reorg_process_insns): Likewise.
2082 (mips_classify_branch_p6600): New function.
2083 * config/mips/mips.h (TUNE_P6600): New define.
2084 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2085 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2086 * config/mips/mips.md: Include p6600.md.
2087 (processor): Add p6600.
2088 * config/mips/p6600.md: New file.
2089 * doc/invoke.texi: Add p6600 to supported architectures.
2091 2018-06-13 Martin Sebor <msebor@redhat.com>
2093 PR tree-optimization/86114
2094 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2096 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2098 2018-06-13 Richard Biener <rguenther@suse.de>
2100 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2101 Properly set vector type of the intermediate stmt.
2102 * tree-vect-stmts.c (vectorizable_operation): The destination
2103 var always has vectype_out type.
2105 2018-06-13 Jeff Law <law@redhat.com>
2107 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2108 integer 0 for argument to print_rtl_with_bb.
2109 (rl78_reorg): Likewise.
2111 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2113 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2114 from rtx to rtx_insn *.
2115 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2117 (add_sched_insns_for_speculation): Likewise for local "target",
2118 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2119 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2120 from rtx_insn ** to rtx_code_label **.
2121 (reorg_emit_nops): Likewise.
2122 (c6x_reorg): Likewise for local "call_labels".
2123 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2125 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2126 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2127 the loops over LABEL_REFS.
2128 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2130 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2131 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2132 (split_branches): Strengthen local "olabel" from rtx to
2133 rtx_insn *, adding a safe_as_a cast.
2134 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2136 (add_insn_after): Likewise for first two params.
2137 (add_insn_before): Likewise.
2138 (remove_insn): Likewise for param.
2139 (emit_pattern_before_noloc): Likewise for second and third params.
2140 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2141 (emit_call_insn_before_noloc): Likewise.
2142 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2144 (emit_barrier_before): Likewise.
2145 (emit_label_before): Strengthen "label" param from "rtx" to
2146 "rtx_code_label *". Strengthen "before" param from "rtx" to
2148 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2150 (emit_pattern_after_noloc): Likewise.
2151 (emit_insn_after_noloc): Likewise.
2152 (emit_jump_insn_after_noloc): Likewise.
2153 (emit_call_insn_after_noloc): Likewise.
2154 (emit_debug_insn_after_noloc): Likewise.
2155 (emit_barrier_after): Likewise.
2156 (emit_label_after): Likewise for both params.
2157 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2158 "loc" param from "int" to "location_t".
2159 (emit_insn_after_setloc): Likewise.
2160 (emit_jump_insn_after_setloc): Likewise.
2161 (emit_call_insn_after_setloc): Likewise.
2162 (emit_debug_insn_after_setloc): Likewise.
2163 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2164 "loc" param from "int" to "location_t".
2165 (emit_pattern_before): Convert NULL_RTX to NULL.
2166 (emit_insn_before_setloc): Convert "loc" param from "int" to
2168 (emit_jump_insn_before_setloc): Likewise.
2169 (emit_call_insn_before_setloc): Likewise.
2170 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2171 rtx_insn *. Convert "loc" param from "int" to "location_t".
2172 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2173 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2174 Convert 3rd param from "int" to "location_t".
2175 (emit_barrier_before, emit_barrier_after, next_real_insn):
2176 Strengthen param from rtx to rtx_insn *.
2177 (emit_label_before): Strengthen 1st param from "rtx" to
2178 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2180 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2181 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2182 Strengthen 2nd param from "rtx" to "rtx_insn *".
2183 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2184 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2185 Likewise. Convert 3rd param from "int" to "location_t".
2186 (emit_label_after): Strengthen 1st param from "rtx" to
2188 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2190 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2191 from "rtx" to "rtx_insn *".
2193 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2195 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2196 bodies streamed in with -Q.
2197 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2198 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2199 * gimple-streamer-out.c: Include gimple-pretty-print.h
2200 (output_bb): Dump stmts streamed.
2201 * lto-section-out.c: Include print-tree.h
2202 (lto_begin_section): Dump sections created.
2203 (lto_output_decl_index): Dump decl encoded.
2204 * lto-streamer-out.c: Include print-tree.h
2205 (create_output_block): Dump output block created.
2206 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2207 (output_function): Dump function output.
2208 (output_constructor): Dump constructor streamed.
2209 (write_global_stream): Output indexes encoded.
2210 (produce_asm_for_decls): Dump streams encoded.
2211 * lto-streamer.c (streamer_dump_file): New global var.
2212 * lto-streamer.h (streamer_dump_file): Declare.
2213 * passes.c (ipa_write_summaries): Initialize streamer dump.
2214 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2217 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2220 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2221 offsets for register save directives. Emit a second batch of save
2222 directives, if need be, when the function accesses prior frames.
2224 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2226 * config/arc/fpu.md (fmasf4): Force operand to register.
2227 (fnmasf4): Likewise.
2229 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2231 * config/arc/arc-protos.h (arc_pad_return): Remove.
2232 * config/arc/arc.c (machine_function): Remove force_short_suffix
2234 (arc_print_operand): Adjust printing of '&'.
2235 (arc_verify_short): Remove conditional printing of short suffix.
2236 (arc_final_prescan_insn): Remove reference to size_reason.
2237 (pad_return): New function.
2238 (arc_reorg): Call pad_return.
2239 (arc_pad_return): Remove.
2240 (arc_init_machine_status): Remove reference to force_short_suffix.
2241 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2242 (attr length): When attribute iscompact is true force to 2
2243 regardless; in the case of maybe check if we want to force the
2244 instruction to have 4 bytes length.
2245 (nopv): Change it to generate 4 byte long nop as well.
2246 (blockage): New pattern.
2247 (simple_return): Remove call to arc_pad_return.
2248 (p_return_i): Likewise.
2250 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2252 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2254 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2256 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2259 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2261 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2262 for ARC700 and ARCv2.
2264 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2267 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2268 operands[2] instead of operands[1].
2271 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2273 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2274 case, check whether the outer register overlaps an unallocatable
2275 register, not just whether it fits the required class.
2277 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2279 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2281 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2282 (memrefs_conflict_p): Likewise.
2283 (init_alias_analysis): Likewise.
2284 * cfgexpand.c (expand_debug_expr): Likewise.
2285 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2286 * cse.c (fold_rtx): Likewise.
2287 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2288 * expr.c (emit_block_move_hints): Likewise.
2289 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2290 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2291 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2292 (emit_group_store): Likewise.
2293 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2294 to read the PRE/POST_MODIFY increment.
2295 * calls.c (store_one_arg): Use strip_offset.
2296 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2298 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2300 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2301 (simplify_binary_operation_1): Extend CONST_INT handling to
2303 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2304 than a HOST_WIDE_INT.
2305 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2307 (adjust_mems, add_stores): Update accodingly.
2308 (vt_canonicalize_addr): Track polynomial offsets.
2309 (emit_note_insn_var_location): Likewise.
2310 (vt_add_function_parameter): Likewise.
2311 (vt_initialize): Likewise.
2313 2018-06-12 Jeff Law <law@redhat.com>
2315 * config.gcc (alpha*-*-freebsd*): Remove.
2316 * config/alpha/freebsd.h: Remove.
2318 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2321 * spellcheck-tree.c (levenshtein_distance): Rename to...
2322 (get_edit_distance): ...this, and update for underlying renaming.
2323 * spellcheck-tree.h (levenshtein_distance): Rename to...
2324 (get_edit_distance): ...this.
2325 * spellcheck.c (levenshtein_distance): Rename to...
2326 (get_edit_distance): ...this. Convert from Levenshtein distance
2327 to Damerau-Levenshtein distance by supporting transpositions of
2328 adjacent characters. Rename "v1" to "v_next" and "v0" to
2330 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2331 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2332 update for underlying renaming.
2333 (selftest::levenshtein_distance_unit_test): Rename to...
2334 (selftest::test_get_edit_distance_unit): ...this, and update for
2335 underlying renaming.
2336 (selftest::test_find_closest_string): Add example from PR 69968
2337 where transposition helps
2338 (selftest::test_metric_conditions): Update for renaming.
2339 (selftest::test_metric_conditions): Likewise.
2340 (selftest::spellcheck_c_tests): Likewise.
2341 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2342 (get_edit_distance): ...this.
2343 (best_match::consider): Update for renaming.
2345 2018-06-12 Martin Sebor <msebor@redhat.com>
2347 PR tree-optimization/85259
2348 * builtins.c (compute_objsize): Handle constant offsets.
2349 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2350 true iff a warning has been issued.
2351 * gimple.h (gimple_nonartificial_location): New function.
2352 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2353 gimple_nonartificial_location and handle -Wno-system-headers.
2354 (handle_builtin_stxncpy): Same.
2356 2018-06-12 Martin Sebor <msebor@redhat.com>
2359 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2361 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2363 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2364 BUILTIN_VEC_XST entries for pointer to double and long long.
2366 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2369 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2371 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2374 2018-06-12 Martin Liska <mliska@suse.cz>
2376 * doc/options.texi: Document IntegerRange.
2378 2018-06-12 Martin Liska <mliska@suse.cz>
2380 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2381 * opt-functions.awk: Handle Deprecated flag.
2382 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2384 (read_cmdline_option): Report warning for a deprecated option.
2385 * opts.h (struct cl_option): Add new field cl_deprecated.
2386 (CL_ERR_DEPRECATED): New.
2388 2018-06-12 Martin Liska <mliska@suse.cz>
2390 * doc/options.texi: Document Deprecated option flag.
2392 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2394 * config/arc/arc-arch.h (arc_extras): New enum.
2395 (arc_cpu_t):Add field extra.
2396 (arc_cpu_types): Consider the extras.
2397 * config/arc/arc-cpus.def: Add extras info.
2398 * config/arc/arc-opts.h (processor_type): Consider extra field.
2399 * config/arc/arc.c (arc_override_options): Handle extra field.
2401 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2403 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2404 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2406 * config/arc/arc.c (arc_init): Update pic warning.
2407 (irq_range): Update irq range parsing warnings.
2408 (arc_override_options): Update various warning messages.
2409 (arc_handle_aux_attribute): Likewise.
2411 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2413 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2415 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2417 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2418 compiler messages directives.
2420 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2422 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2423 * config/mips/mips-tables.opt: Regenerate.
2424 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2426 * doc/invoke.texi: Document -march=i6500.
2428 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2430 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2431 (i6400_gpmul): Add cpu_unit.
2432 (i6400_gpdiv): Likewise.
2433 (i6400_msa_add_d): Update reservations.
2434 (i6400_msa_int_add) Likewise.
2435 (i6400_msa_short_logic3) Likewise.
2436 (i6400_msa_short_logic2) Likewise.
2437 (i6400_msa_short_logic) Likewise.
2438 (i6400_msa_move) Likewise.
2439 (i6400_msa_cmp) Likewise.
2440 (i6400_msa_short_float2) Likewise.
2441 (i6400_msa_div_d) Likewise.
2442 (i6400_msa_long_logic1) Likewise.
2443 (i6400_msa_long_logic2) Likewise.
2444 (i6400_msa_mult) Likewise.
2445 (i6400_msa_long_float2) Likewise.
2446 (i6400_msa_long_float4) Likewise.
2447 (i6400_msa_long_float5) Likewise.
2448 (i6400_msa_long_float8) Likewise.
2449 (i6400_fpu_fadd): Include frint type.
2450 (i6400_fpu_store): New define_insn_reservation.
2451 (i6400_fpu_load): Likewise.
2452 (i6400_fpu_move): Likewise.
2453 (i6400_fpu_fcmp): Likewise.
2454 (i6400_fpu_fmadd): Likewise.
2455 (i6400_int_mult): Include imul3nc type and update reservation.
2456 (i6400_int_div): Include idiv3 type and update reservation.
2457 (i6400_int_load): Update to check type not move_type.
2458 (i6400_int_store): Likewise.
2459 (i6400_int_prefetch): Set zero latency.
2461 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2463 * gcc.c: Document new %@{...} sequence.
2464 (LINK_COMMAND_SPEC): Use it for the -L switches.
2465 (cpp_unique_options): Use it for the -I switches.
2466 (at_file_argbuf): New global variable.
2467 (in_at_file): Likewise.
2468 (alloc_args): Create at_file_argbuf.
2469 (clear_args): Truncate at_file_argbuf.
2470 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2471 (open_at_file): New function.
2472 (close_at_file): Likewise.
2473 (create_at_file): Delete.
2474 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2477 (validate_switches_from_spec): Deal with %@{...} sequence.
2478 (validate_switches): Likewise.
2479 (driver::finalize): Call clear_args.
2481 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2483 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2485 2018-06-11 Martin Sebor <msebor@redhat.com>
2487 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2488 (Wno-class-memaccess): Add @opindex.
2489 (Wno-templates, Wno-multiple-inheritance): Same.
2490 (Wno-virtual-inheritance, Wno-namespaces): Same.
2491 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2492 (Wno-format-overflow, Wno-format-truncation): Same.
2493 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2494 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2495 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2496 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2497 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2498 (Wno-misspelled-isr): Same.
2500 2018-06-11 Martin Sebor <msebor@redhat.com>
2502 * PR tree-optimization/86083
2503 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2505 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2507 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2509 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2512 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2513 on the correct operand.
2514 (*movdi_internal64): Ditto.
2516 2018-06-11 Martin Liska <mliska@suse.cz>
2518 PR tree-optimization/86089
2519 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2521 2018-06-11 Julia Koval <julia.koval@intel.com>
2523 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2524 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2525 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2527 2018-06-11 Olivier Hainque <hainque@adacore.com>
2529 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2530 for Ada with strict dwarf2.
2532 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2535 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2538 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2540 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2542 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2544 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2547 2018-06-08 Martin Liska <mliska@suse.cz>
2549 * tree-cfg.h (debug_function): Fix argument type to match
2552 2018-06-08 Martin Liska <mliska@suse.cz>
2554 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2555 Remove usage of MPX-related (and removed) fields.
2556 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2558 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2560 * cfg.c (debug): Use TDF_NONE rather than 0.
2561 * cfghooks.c (debug): Likewise.
2562 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2563 (struct dump_option_value_info): Convert to...
2564 (struct kv_pair): ...this template type.
2565 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2567 (optinfo_verbosity_options): Likewise.
2568 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2570 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2571 than int for "optgroup_flags" param.
2572 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2574 (dump_dec): Likewise.
2575 (dump_finish): Use TDF_NONE rather than 0.
2576 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2577 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2578 than 0. Update for change to option_ptr.
2579 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2580 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2581 0. Update for changes to optinfo_verbosity_options and
2583 (opt_info_switch_p): Convert optgroup_flags from int to
2585 (dump_basic_block): Use dump_flags_t rather than int
2586 for "dump_kind" param.
2587 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2588 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2589 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2590 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2591 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2592 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2593 TDF_NONE): Convert from macros to...
2594 (enum dump_flag): ...this new enum.
2595 (dump_flags_t): Update to use enum.
2596 (operator|, operator&, operator~, operator|=, operator&=):
2597 Implement for dump_flags_t.
2598 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2599 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2600 Convert from macros to...
2601 (enum optgroup_flag): ...this new enum.
2602 (optgroup_flags_t): New typedef.
2603 (operator|, operator|=): Implement for optgroup_flags_t.
2604 (struct dump_file_info): Convert field "alt_flags" to
2605 dump_flags_t. Convert field "optgroup_flags" to
2607 (dump_basic_block): Use dump_flags_t rather than int for param.
2608 (dump_generic_expr_loc): Likewise.
2609 (dump_dec): Likewise.
2610 (dump_register): Convert param "optgroup_flags" to
2612 (opt_info_enable_passes): Likewise.
2613 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2615 * gimple-pretty-print.c (debug): Likewise.
2616 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2617 (merged_store_group::apply_stores): Likewise.
2618 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2619 * gimple.c (verify_gimple_pp): Likewise.
2620 * graphite-poly.c (print_pbb_body): Likewise.
2621 * passes.c (pass_manager::register_one_dump_file): Convert
2622 local "optgroup_flags" to optgroup_flags_t.
2623 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2625 (debug_body): Likewise.
2626 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2627 to optgroup_flags_t.
2628 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2630 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2631 (convert_mult_to_fma): Likewise.
2632 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2633 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2634 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2635 "dump_kind" to dump_flags_t.
2637 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2639 * config/rs6000/rs6000.c (min, max): Delete.
2641 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2643 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2644 -mabi=spe and -mabi=no-spe.
2646 2018-06-08 Martin Liska <mliska@suse.cz>
2648 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2649 where we expect an existing summary.
2651 2018-06-08 Martin Liska <mliska@suse.cz>
2653 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2654 * ipa-inline.h (estimate_edge_growth): Likewise.
2656 2018-06-08 Martin Liska <mliska@suse.cz>
2658 * cgraph.c (function_version_hasher::hash): Use
2659 cgraph_node::get_uid ().
2660 (function_version_hasher::equal):
2661 * cgraph.h (cgraph_node::get_uid): New method.
2662 * ipa-inline.c (update_caller_keys): Use
2663 cgraph_node::get_uid ().
2664 (update_callee_keys): Likewise.
2665 * ipa-utils.c (searchc): Likewise.
2666 (ipa_reduced_postorder): Likewise.
2667 * lto-cgraph.c (input_node): Likewise.
2668 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2669 * symbol-summary.h (symtab_insertion): Likewise.
2670 (symtab_removal): Likewise.
2671 (symtab_duplication): Likewise.
2672 * tree-pretty-print.c (dump_function_header): Likewise.
2673 * tree-sra.c (convert_callers_for_node): Likewise.
2675 2018-06-08 Martin Liska <mliska@suse.cz>
2677 * cgraph.c (symbol_table::create_edge): Always assign a new
2679 (symbol_table::free_edge): Do not recycle numbers.
2680 * cgraph.h (cgraph_edge::get): New method.
2681 * symbol-summary.h (symtab_removal): Use it.
2682 (symtab_duplication): Likewise.
2683 (call_summary::hashable_uid): Remove.
2685 2018-06-08 Martin Liska <mliska@suse.cz>
2687 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2688 (initialize_growth_caches): Remove.
2689 (free_growth_caches): Likewise.
2690 (do_estimate_edge_time): Use edge_growth_cache.
2691 (do_estimate_edge_size): Likewise.
2692 (do_estimate_edge_hints): Likewise.
2693 * ipa-inline.c (reset_edge_caches): Likewise.
2694 (recursive_inlining): Likewise.
2695 (inline_small_functions): Likewise.
2696 * ipa-inline.h (initialize_growth_caches): Remove.
2697 (estimate_edge_size): Likewise.
2698 (estimate_edge_time): Likewise.
2699 (estimate_edge_hints): Likewise.
2700 (reset_edge_growth_cache): Likewise.
2701 * symbol-summary.h (call_summary::remove): New method.
2703 2018-06-08 Martin Liska <mliska@suse.cz>
2705 * ipa-cp.c (class edge_clone_summary): New summary.
2706 (grow_edge_clone_vectors): Remove.
2707 (ipcp_edge_duplication_hook): Remove.
2708 (class edge_clone_summary_t): New call_summary class.
2709 (ipcp_edge_removal_hook): Remove.
2710 (edge_clone_summary_t::duplicate): New function.
2711 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2712 (create_specialized_node): Likewise.
2713 (ipcp_driver): Initialize edge_clone_summaries and do not
2716 2018-06-08 Martin Liska <mliska@suse.cz>
2718 * symbol-summary.h (get): New function.
2719 (call_summary::m_initialize_when_cloning): New class member.
2721 2018-06-08 Martin Liska <mliska@suse.cz>
2723 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2724 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2725 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2726 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2727 (remove_cgraph_node_from_order): Use the removed_nodes set.
2728 (do_per_function_toporder): Likwise.
2729 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2730 instead of summary_uid.
2731 (symtab_removal): Likewise.
2732 (symtab_duplication): Likewise.
2734 2018-06-08 Martin Liska <mliska@suse.cz>
2736 * ipa-cp.c (ipcp_store_bits_results): Use
2737 ipcp_transformation_sum.
2738 (ipcp_store_vr_results): Likewise.
2739 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2741 (ipcp_transformation_initialize): ... this.
2742 (ipa_set_node_agg_value_chain):
2743 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2744 (write_ipcp_transformation_info): Likewise.
2745 (read_ipcp_transformation_info): Likewise.
2746 (ipcp_update_bits): Likewise.
2747 (ipcp_update_vr): Likewise.
2748 (ipcp_transform_function): Likewise.
2749 * ipa-prop.h: Rename ipcp_transformation_summary to
2750 ipcp_transformation.
2751 (class ipcp_transformation_t): New function summary.
2752 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2753 (ipa_get_agg_replacements_for_node): Likewise.
2755 2018-06-08 Martin Liska <mliska@suse.cz>
2757 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2759 (class funct_state_summary_t): New function_summary class.
2760 (has_function_state): Remove.
2761 (get_function_state): Likewise.
2762 (set_function_state): Likewise.
2763 (add_new_function): Likewise.
2764 (funct_state_summary_t::insert): New function.
2765 (duplicate_node_data): Remove.
2766 (remove_node_data): Remove.
2767 (funct_state_summary_t::duplicate): New function.
2768 (register_hooks): Create new funct_state_summaries.
2769 (pure_const_generate_summary): Use it.
2770 (pure_const_write_summary): Likewise.
2771 (pure_const_read_summary): Likewise.
2772 (propagate_pure_const): Likewise.
2773 (propagate_nothrow): Likewise.
2774 (dump_malloc_lattice): Likewise.
2775 (propagate_malloc): Likewise.
2776 (execute): Do not register hooks, just remove summary
2778 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2781 2018-06-08 Martin Liska <mliska@suse.cz>
2783 * ipa-reference.c (remove_node_data): Remove.
2784 (duplicate_node_data): Likewise.
2785 (class ipa_ref_var_info_summary_t): New class.
2786 (class ipa_ref_opt_summary_t): Likewise.
2787 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2788 (get_reference_optimization_summary): Use
2789 ipa_ref_opt_sum_summaries.
2790 (set_reference_vars_info): Remove.
2791 (set_reference_optimization_summary): Likewise.
2792 (ipa_init): Create summaries.
2793 (init_function_info): Use function summary.
2794 (ipa_ref_opt_summary_t::duplicate): New function.
2795 (ipa_ref_opt_summary_t::remove): New function.
2796 (get_read_write_all_from_node): Fix GNU coding style.
2797 (propagate): Use function summary.
2798 (write_node_summary_p): Fix GNU coding style.
2799 (stream_out_bitmap): Likewise.
2800 (ipa_reference_read_optimization_summary): Use function summary.
2801 (ipa_reference_c_finalize): Do not release hooks.
2803 2018-06-08 Martin Liska <mliska@suse.cz>
2805 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2806 (analyze_function_body): Extract multiple calls of get_create.
2807 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2808 * ipa-inline.c (recursive_inlining): Use ::get method.
2809 * ipa-inline.h (estimate_edge_growth): Likewise.
2811 2018-06-08 Martin Liska <mliska@suse.cz>
2813 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2815 (hsa_function_summary::hsa_function_summary): Use the new enum
2817 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2818 * hsa-gen.c (hsa_get_host_function): Likewise.
2819 (get_brig_function_name): Likewise.
2820 * ipa-hsa.c (process_hsa_functions): Likewise.
2821 (ipa_hsa_write_summary): Likewise.
2822 * symbol-summary.h (symtab_duplication): Use ::get function/
2823 (get): New function.
2825 2018-06-08 Martin Liska <mliska@suse.cz>
2827 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2829 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2830 (hsa_register_kernel): Likewise.
2831 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2832 * hsa-gen.c (hsa_get_host_function): Likewise.
2833 (get_brig_function_name): Likewise.
2834 (generate_hsa): Likewise.
2835 (pass_gen_hsail::execute): Likewise.
2836 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2837 (devirtualization_time_bonus): Likewise.
2838 (ipcp_propagate_stage): Likewise.
2839 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2840 (edge_set_predicate): Likewise.
2841 (evaluate_conditions_for_known_args): Likewise.
2842 (evaluate_properties_for_edge): Likewise.
2843 (ipa_fn_summary::reset): Likewise.
2844 (ipa_fn_summary_t::duplicate): Likewise.
2845 (dump_ipa_call_summary): Likewise.
2846 (ipa_dump_fn_summary): Likewise.
2847 (analyze_function_body): Likewise.
2848 (compute_fn_summary): Likewise.
2849 (estimate_edge_devirt_benefit): Likewise.
2850 (estimate_edge_size_and_time): Likewise.
2851 (estimate_calls_size_and_time): Likewise.
2852 (estimate_node_size_and_time): Likewise.
2853 (inline_update_callee_summaries): Likewise.
2854 (remap_edge_change_prob): Likewise.
2855 (remap_edge_summaries): Likewise.
2856 (ipa_merge_fn_summary_after_inlining): Likewise.
2857 (ipa_update_overall_fn_summary): Likewise.
2858 (read_ipa_call_summary): Likewise.
2859 (inline_read_section): Likewise.
2860 (write_ipa_call_summary): Likewise.
2861 (ipa_fn_summary_write): Likewise.
2862 (ipa_free_fn_summary): Likewise.
2863 * ipa-hsa.c (process_hsa_functions): Likewise.
2864 (ipa_hsa_write_summary): Likewise.
2865 (ipa_hsa_read_section): Likewise.
2866 * ipa-icf.c (sem_function::merge): Likewise.
2867 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2868 (do_estimate_edge_time): Likewise.
2869 (estimate_size_after_inlining): Likewise.
2870 (estimate_growth): Likewise.
2871 (growth_likely_positive): Likewise.
2872 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2873 (inline_call): Likewise.
2874 * ipa-inline.c (caller_growth_limits): Likewise.
2875 (can_inline_edge_p): Likewise.
2876 (can_inline_edge_by_limits_p): Likewise.
2877 (compute_uninlined_call_time): Likewise.
2878 (compute_inlined_call_time): Likewise.
2879 (want_inline_small_function_p): Likewise.
2880 (edge_badness): Likewise.
2881 (update_caller_keys): Likewise.
2882 (update_callee_keys): Likewise.
2883 (recursive_inlining): Likewise.
2884 (inline_small_functions): Likewise.
2885 (inline_to_all_callers_1): Likewise.
2886 (dump_overall_stats): Likewise.
2887 (early_inline_small_functions): Likewise.
2888 (early_inliner): Likewise.
2889 * ipa-inline.h (estimate_edge_growth): Likewise.
2890 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2891 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2892 * ipa-prop.h (IPA_NODE_REF): Likewise.
2893 (IPA_EDGE_REF): Likewise.
2894 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2895 (propagate_malloc): Likewise.
2896 * ipa-split.c (execute_split_functions): Likewise.
2897 * symbol-summary.h: Rename get to get_create.
2899 (get_create): Likewise.
2900 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2902 2018-06-08 Martin Liska <mliska@suse.cz>
2904 * symbol-summary.h (release): Move definition out of class
2906 (symtab_removal): Likewise.
2907 (symtab_duplication): Likewise.
2909 2018-06-08 Martin Liska <mliska@suse.cz>
2911 * symbol-summary.h (function_summary): Move constructor
2912 implementation out of class declaration.
2913 (release): Likewise.
2914 (symtab_insertion): Likewise.
2915 (symtab_removal): Likewise.
2916 (symtab_duplication): Likewise.
2919 2018-06-08 Martin Liska <mliska@suse.cz>
2921 * Makefile.in: Remove support for MPX (macros, related functions,
2922 fields in cgraph_node, ...).
2923 * builtin-types.def (BT_BND): Likewise.
2924 (BT_FN_BND_CONST_PTR): Likewise.
2925 (BT_FN_CONST_PTR_BND): Likewise.
2926 (BT_FN_VOID_PTR_BND): Likewise.
2927 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2928 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2929 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2930 (expand_builtin_mempcpy_with_bounds): Likewise.
2931 (expand_builtin_memset_with_bounds): Likewise.
2932 (expand_builtin_memset_args): Likewise.
2933 (std_expand_builtin_va_start): Likewise.
2934 (expand_builtin): Likewise.
2935 (expand_builtin_with_bounds): Likewise.
2936 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2937 (DEF_LIB_BUILTIN_CHKP): Likewise.
2938 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2939 (DEF_CHKP_BUILTIN): Likewise.
2940 (BUILT_IN_MEMCPY): Likewise.
2941 (BUILT_IN_MEMMOVE): Likewise.
2942 (BUILT_IN_MEMPCPY): Likewise.
2943 (BUILT_IN_MEMSET): Likewise.
2944 (BUILT_IN_STPCPY): Likewise.
2945 (BUILT_IN_STRCAT): Likewise.
2946 (BUILT_IN_STRCHR): Likewise.
2947 (BUILT_IN_STRCPY): Likewise.
2948 (BUILT_IN_STRLEN): Likewise.
2949 (BUILT_IN_MEMCPY_CHK): Likewise.
2950 (BUILT_IN_MEMMOVE_CHK): Likewise.
2951 (BUILT_IN_MEMPCPY_CHK): Likewise.
2952 (BUILT_IN_MEMSET_CHK): Likewise.
2953 (BUILT_IN_STPCPY_CHK): Likewise.
2954 (BUILT_IN_STRCAT_CHK): Likewise.
2955 (BUILT_IN_STRCPY_CHK): Likewise.
2956 * calls.c (store_bounds): Likewise.
2957 (emit_call_1): Likewise.
2958 (special_function_p): Likewise.
2959 (maybe_warn_nonstring_arg): Likewise.
2960 (initialize_argument_information): Likewise.
2961 (finalize_must_preallocate): Likewise.
2962 (compute_argument_addresses): Likewise.
2963 (expand_call): Likewise.
2964 * cfgexpand.c (expand_call_stmt): Likewise.
2965 (expand_return): Likewise.
2966 (expand_gimple_stmt_1): Likewise.
2967 (pass_expand::execute): Likewise.
2968 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2969 (cgraph_node::remove): Likewise.
2970 (cgraph_node::dump): Likewise.
2971 (cgraph_node::verify_node): Likewise.
2972 * cgraph.h (chkp_function_instrumented_p): Likewise.
2973 (symtab_node::get_alias_target): Likewise.
2974 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2975 (cgraph_local_p): Likewise.
2976 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2977 (cgraph_edge::rebuild_references): Likewise.
2978 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2979 (walk_polymorphic_call_targets): Likewise.
2980 (cgraph_node::expand_thunk): Likewise.
2981 (symbol_table::output_weakrefs): Likewise.
2982 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2983 (ix86_handle_option): Likewise.
2984 * config/i386/constraints.md: Likewise.
2985 * config/i386/i386-builtin-types.def (BND): Likewise.
2989 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2990 (BDESC_FIRST): Likewise.
2992 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2993 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2994 * config/i386/i386.c (enum reg_class): Likewise.
2995 (ix86_target_string): Likewise.
2996 (ix86_option_override_internal): Likewise.
2997 (ix86_conditional_register_usage): Likewise.
2998 (ix86_valid_target_attribute_inner_p): Likewise.
2999 (ix86_set_indirect_branch_type): Likewise.
3000 (ix86_set_current_function): Likewise.
3001 (ix86_function_arg_regno_p): Likewise.
3002 (init_cumulative_args): Likewise.
3003 (ix86_function_arg_advance): Likewise.
3004 (ix86_function_arg): Likewise.
3005 (ix86_pass_by_reference): Likewise.
3006 (ix86_function_value_regno_p): Likewise.
3007 (ix86_function_value_1): Likewise.
3008 (ix86_function_value_bounds): Likewise.
3009 (ix86_return_in_memory): Likewise.
3010 (ix86_setup_incoming_vararg_bounds): Likewise.
3011 (ix86_va_start): Likewise.
3012 (indirect_thunk_need_prefix): Likewise.
3013 (print_reg): Likewise.
3014 (ix86_print_operand): Likewise.
3015 (ix86_expand_call): Likewise.
3016 (ix86_output_function_return): Likewise.
3017 (reg_encoded_number): Likewise.
3018 (BDESC_VERIFYS): Likewise.
3019 (ix86_init_mpx_builtins): Likewise.
3020 (ix86_init_builtins): Likewise.
3021 (ix86_emit_cmove): Likewise.
3022 (ix86_emit_move_max): Likewise.
3023 (ix86_expand_builtin): Likewise.
3024 (ix86_builtin_mpx_function): Likewise.
3025 (ix86_get_arg_address_for_bt): Likewise.
3026 (ix86_load_bounds): Likewise.
3027 (ix86_store_bounds): Likewise.
3028 (ix86_load_returned_bounds): Likewise.
3029 (ix86_store_returned_bounds): Likewise.
3030 (ix86_class_likely_spilled_p): Likewise.
3031 (ix86_hard_regno_mode_ok): Likewise.
3032 (x86_order_regs_for_local_alloc): Likewise.
3033 (ix86_mitigate_rop): Likewise.
3034 (ix86_bnd_prefixed_insn_p): Likewise.
3035 (ix86_mpx_bound_mode): Likewise.
3036 (ix86_make_bounds_constant): Likewise.
3037 (ix86_initialize_bounds): Likewise.
3038 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3039 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3040 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3041 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3042 (TARGET_CHKP_BOUND_MODE): Likewise.
3043 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3044 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3045 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3046 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3047 * config/i386/i386.h (TARGET_MPX): Likewise.
3048 (TARGET_MPX_P): Likewise.
3049 (VALID_BND_REG_MODE): Likewise.
3050 (FIRST_BND_REG): Likewise.
3051 (LAST_BND_REG): Likewise.
3052 (enum reg_class): Likewise.
3053 (BND_REG_P): Likewise.
3054 (BND_REGNO_P): Likewise.
3055 (BNDmode): Likewise.
3056 (ADJUST_INSN_LENGTH): Likewise.
3057 * config/i386/i386.md: Likewise.
3058 * config/i386/i386.opt: Likewise.
3059 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3060 (defined): Likewise.
3061 (LINK_MPX): Likewise.
3062 (MPX_SPEC): Likewise.
3063 (LIBMPX_SPEC): Likewise.
3064 (LIBMPXWRAPPERS_SPEC): Likewise.
3065 (CHKP_SPEC): Likewise.
3066 * config/i386/predicates.md: Likewise.
3067 * dbxout.c (dbxout_type): Likewise.
3068 * doc/extend.texi: Likewise.
3069 * doc/invoke.texi: Likewise.
3070 * doc/md.texi: Likewise.
3071 * doc/tm.texi: Likewise.
3072 * doc/tm.texi.in: Likewise.
3073 * dwarf2out.c (is_base_type): Likewise.
3074 (gen_formal_types_die): Likewise.
3075 (gen_subprogram_die): Likewise.
3076 (gen_type_die_with_usage): Likewise.
3077 (gen_decl_die): Likewise.
3078 (dwarf2out_late_global_decl): Likewise.
3079 * expr.c (expand_assignment): Likewise.
3080 (emit_storent_insn): Likewise.
3081 (store_expr_with_bounds): Likewise.
3082 (store_expr): Likewise.
3083 (expand_expr_real_1): Likewise.
3084 * expr.h (store_expr_with_bounds): Likewise.
3085 * function.c (use_register_for_decl): Likewise.
3086 (struct bounds_parm_data): Likewise.
3087 (assign_parms_augmented_arg_list): Likewise.
3088 (assign_parm_find_entry_rtl): Likewise.
3089 (assign_parm_is_stack_parm): Likewise.
3090 (assign_parm_load_bounds): Likewise.
3091 (assign_bounds): Likewise.
3092 (assign_parms): Likewise.
3093 (expand_function_start): Likewise.
3094 * gcc.c (CHKP_SPEC): Likewise.
3095 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3096 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3097 (wrestrict_dom_walker::check_call): Likewise.
3098 * gimple.c (gimple_build_call_from_tree): Likewise.
3099 * gimple.h (enum gf_mask): Likewise.
3100 (gimple_call_with_bounds_p): Likewise.
3101 (gimple_call_set_with_bounds): Likewise.
3102 * gimplify.c (gimplify_init_constructor): Likewise.
3103 * ipa-cp.c (initialize_node_lattices): Likewise.
3104 (propagate_constants_across_call): Likewise.
3105 (find_more_scalar_values_for_callers_subset): Likewise.
3106 * ipa-hsa.c (process_hsa_functions): Likewise.
3107 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3108 * ipa-icf.c (sem_function::merge): Likewise.
3109 * ipa-inline.c (early_inliner): Likewise.
3110 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3111 (warn_function_cold): Likewise.
3112 (propagate_pure_const): Likewise.
3113 * ipa-ref.h (enum GTY): Likewise.
3114 * ipa-split.c (find_retbnd): Likewise.
3115 (consider_split): Likewise.
3116 (split_function): Likewise.
3117 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3118 * ipa.c (walk_polymorphic_call_targets): Likewise.
3119 (symbol_table::remove_unreachable_nodes): Likewise.
3120 (process_references): Likewise.
3121 (cgraph_build_static_cdtor_1): Likewise.
3122 * lto-cgraph.c (lto_output_node): Likewise.
3123 (output_refs): Likewise.
3124 (compute_ltrans_boundary): Likewise.
3125 (input_overwrite_node): Likewise.
3126 (input_node): Likewise.
3127 (input_cgraph_1): Likewise.
3128 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3129 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3130 (class pass_chkp_instrumentation_passes): Likewise.
3131 (make_pass_chkp_instrumentation_passes): Likewise.
3132 * passes.def: Likewise.
3133 * rtl.h (struct GTY): Likewise.
3134 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3135 * stor-layout.c (layout_type): Likewise.
3136 * symtab.c: Likewise.
3137 * target.def: Likewise.
3138 * targhooks.c (default_chkp_bound_type): Likewise.
3139 (default_chkp_bound_mode): Likewise.
3140 (default_builtin_chkp_function): Likewise.
3141 (default_chkp_function_value_bounds): Likewise.
3142 (default_chkp_make_bounds_constant): Likewise.
3143 (default_chkp_initialize_bounds): Likewise.
3144 * targhooks.h (default_chkp_bound_type): Likewise.
3145 (default_chkp_bound_mode): Likewise.
3146 (default_builtin_chkp_function): Likewise.
3147 (default_chkp_function_value_bounds): Likewise.
3148 (default_chkp_make_bounds_constant): Likewise.
3149 (default_chkp_initialize_bounds): Likewise.
3150 * toplev.c (compile_file): Likewise.
3151 (process_options): Likewise.
3152 * tree-core.h (DEF_BUILTIN): Likewise.
3153 (DEF_BUILTIN_CHKP): Likewise.
3154 * tree-inline.c (declare_return_variable): Likewise.
3155 (remap_gimple_stmt): Likewise.
3156 (copy_bb): Likewise.
3157 (initialize_inlined_parameters): Likewise.
3158 (expand_call_inline): Likewise.
3159 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3160 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3161 (make_pass_ipa_chkp_produce_thunks): Likewise.
3162 (make_pass_chkp): Likewise.
3163 (make_pass_chkp_opt): Likewise.
3164 (make_pass_chkp_instrumentation_passes): Likewise.
3165 * tree-pretty-print.c (dump_generic_node): Likewise.
3166 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3167 * tree-ssa-dce.c (propagate_necessity): Likewise.
3168 (eliminate_unnecessary_stmts): Likewise.
3169 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3170 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3171 * tree-ssa-sccvn.h: Likewise.
3172 * tree-ssa-strlen.c (get_string_length): Likewise.
3173 (valid_builtin_call): Likewise.
3174 (adjust_last_stmt): Likewise.
3175 (handle_builtin_strchr): Likewise.
3176 (handle_builtin_strcpy): Likewise.
3177 (handle_builtin_stxncpy): Likewise.
3178 (handle_builtin_memcpy): Likewise.
3179 (handle_builtin_strcat): Likewise.
3180 (strlen_check_and_optimize_stmt): Likewise.
3181 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3182 * tree-streamer-in.c: Likewise.
3183 * tree-streamer.c (record_common_node): Likewise.
3184 * tree.c (tree_code_size): Likewise.
3185 (wide_int_to_tree_1): Likewise.
3186 (type_contains_placeholder_1): Likewise.
3187 (build_common_tree_nodes): Likewise.
3188 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3189 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3190 (POINTER_BOUNDS_P): Likewise.
3191 (BOUNDED_TYPE_P): Likewise.
3192 (BOUNDED_P): Likewise.
3193 (CALL_WITH_BOUNDS_P): Likewise.
3194 (pointer_bounds_type_node): Likewise.
3195 * value-prof.c (gimple_ic): Likewise.
3196 * var-tracking.c (vt_add_function_parameters): Likewise.
3197 * varasm.c (make_decl_rtl): Likewise.
3198 (assemble_start_function): Likewise.
3199 (output_constant): Likewise.
3200 (maybe_assemble_visibility): Likewise.
3201 * varpool.c (ctor_for_folding): Likewise.
3202 * chkp-builtins.def: Remove.
3203 * ipa-chkp.c: Remove.
3204 * ipa-chkp.h: Remove.
3205 * rtl-chkp.c: Remove.
3206 * rtl-chkp.h: Remove.
3207 * tree-chkp-opt.c: Remove.
3208 * tree-chkp.c: Remove.
3209 * tree-chkp.h: Remove.
3211 2018-06-07 Carl Love <cel@us.ibm.com>
3213 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3214 vextract_fp_from_shortl): Add BE support.
3216 2018-06-07 Paul Koning <ni1d@arrl.net>
3218 * compare-elim.c (try_merge_compare): Don't merge compare if
3219 address contains a side effect.
3220 (try_eliminate_compare): Likewise.
3222 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3224 * config.gcc: Support "tremont".
3225 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3226 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3228 * config/i386/i386.c (m_TREMONT): Define.
3229 (processor_target_table): Add "tremont".
3230 (PTA_TREMONT): Define.
3231 (ix86_lea_outperforms): Add TARGET_TREMONT.
3232 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3233 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3234 and M_INTEL_GOLDMONT_PLUS.
3235 (fold_builtin_cpu): Add "tremont".
3236 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3237 (ix86_option_override_internal): Add "tremont".
3238 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3239 (processor_type): Add PROCESSOR_TREMONT.
3240 * config/i386/x86-tune.def: Add m_TREMONT.
3241 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3243 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3245 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3246 symbol defined for msp430i* devices to be lower case.
3248 gcc/testsuite/gcc.target/msp430/
3249 * msp430i-device-symbol.c: New test.
3250 * msp430f-device-symbol.c: New test.
3251 * msp430.h: New test header file.
3253 2018-06-07 Richard Biener <rguenther@suse.de>
3255 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3256 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3257 Properly wrap signed arithmetic if overflow wraps.
3259 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3261 PR tree-optimization/69615
3262 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3263 of a cast from a same precision integral SSA_NAME in a bb dominated
3264 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3265 cast to utype if rhs2 has already a compatible type.
3267 2018-06-07 Richard Biener <rguenther@suse.de>
3269 PR tree-optimization/85935
3270 * graphite-scop-detection.c (find_params_in_bb): Analyze
3271 condition operands with respect to the correct loop. Assert
3272 the analysis doesn't fail.
3274 2018-06-04 Carl Love <cel@us.ibm.com>
3276 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3277 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3278 as it is slightly cheaper.
3279 (first_match_or_eos_index_<mode>):
3280 Calculate index using natural element order.
3281 (first_match_index_<mode>):
3282 Calculate index using natural element order.
3283 (first_match_or_eos_index_<mode>):
3284 Calculate index using natural order.
3285 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3286 for BE and LE modes.
3287 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3288 P9V_BUILTIN_VCLZLSBB_V16QI.
3289 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3292 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3294 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3295 indentation and line wrap for many prototypes. Add missing
3296 @smallexample directives around block of prototypes for vec_xl and
3299 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3301 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3302 track if we pass or return IEEE 128-bit floating point.
3303 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3304 C++ mangling that is compatible with GCC 8.1.
3305 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3306 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3307 floating point types.
3308 (rs6000_function_arg_advance_1): Likewise.
3309 (rs6000_mangle_type): Optionally generate mangled names that match
3310 what GCC 8.1 generated for IEEE 128-bit floating point types.
3311 (rs6000_globalize_decl_name): If we have an external function that
3312 passes or returns IEEE 128-bit types, generate a weak reference
3313 from the mangled name used in GCC 8.1 to the current mangled
3315 (rs6000_init_builtins): Make __ibm128 use the long double type if
3316 long double is IBM extended double. Make __float128 use the long
3317 double type if long double is IEEE 128-bit.
3320 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3321 macro for __ibm128 built-in functions.
3322 (PACK_IF): Add __ibm128 pack/unpack functions.
3323 (UNPACK_IF): Likewise.
3324 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3325 enable long double built-in functions if long double is IEEE
3326 128-bit floating point.
3327 (rs6000_invalid_builtin): Update long double built-in function
3329 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3330 functions, adjust the built-in function to use the long double
3331 built-in function if __ibm128 and long double are the same type.
3332 * doc/extend.texi (PowerPC builtins): Update documention for
3333 __builtin_{,un}pack_longdouble. Add documentation for
3334 __builtin_{,un}pack_ibm128.
3336 2018-06-06 Jim Wilson <jimw@sifive.com>
3338 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3339 (struct machine_function): New field interrupt_mode.
3340 (riscv_handle_type_attribute): New function. Add forward declaration.
3341 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3342 (riscv_expand_epilogue): Check interrupt_mode field.
3343 (riscv_set_current_function): Check interrupt attribute args and
3344 set interrupt_mode field.
3345 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3346 (riscv_sret, riscv_uret): New.
3347 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3348 new arguments to interrupt attribute.
3350 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3353 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3354 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3356 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3358 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3359 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3360 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3361 several redundant entries.
3363 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3365 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3366 type from "rtx" to "rtx_insn *".
3367 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3368 for local "call_insn", removing cast.
3369 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3371 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3373 PR tree-optimization/86066
3374 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3375 for BIT_INSERT_EXPR stores.
3377 2018-06-06 Richard Biener <rguenther@suse.de>
3379 PR tree-optimization/86062
3380 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3381 component refs ontop
3382 of to be offsetted base.
3384 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3386 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3387 to be static and remove check on interrupt attribute name.
3389 gcc/testsuite/gcc.target/msp430/
3390 * function-attributes-4.c: New test.
3391 * static-interrupts.c: New test.
3393 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3395 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3396 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3398 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3401 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3403 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3404 Remove second argument, change how error is called.
3405 (aarch64_layout_arg): Remove second argument from
3406 aarch64_err_no_fpadvsimd call.
3407 (aarch64_init_cumulative_args): Ditto.
3408 (aarch64_gimplify_va_arg_expr): Ditto.
3409 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3411 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3413 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3414 (*simple_return_indirect_internal<mode>): Rename from
3415 simple_return_indirect_internal. Use W mode iterator.
3416 (rstorssp): New expander.
3417 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3418 (clrssbsy): New expander.
3419 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3421 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3423 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3425 (cmse_check_pointed_object): Likewise.
3427 2018-06-05 Martin Liska <mliska@suse.cz>
3429 PR gcov-profile/47618
3430 * doc/invoke.texi: Document how -fprofile-dir format
3433 2018-06-05 Richard Biener <rguenther@suse.de>
3435 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3436 removal pretend DOM info isn't available so we do not update
3437 it and only remove edges, not dominated blocks. Actually free
3438 DOM info in case we removed something. Remove unreachable blocks.
3439 (mfb_keep_latches): Work with either DOM info or marked backedges.
3440 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3441 first. Mark backedges if DOM info isn't available.
3442 (Re-)compute DOM info after cleanup_control_flow_pre.
3444 2018-06-05 Richard Biener <rguenther@suse.de>
3446 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3447 (locus_discrim_hasher::hash): Adjust.
3448 (locus_discrim_hasher::equal): Likewise.
3449 (next_discriminator_for_locus): Work on line directly.
3450 (same_line_p): Pass in expanded locus1 as well.
3451 (assign_discriminators): Avoid redundant location expansions.
3453 2018-06-05 Richard Biener <rguenther@suse.de>
3455 PR tree-optimization/86046
3456 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3457 if required after clearing TREE_ADDRESSABLE.
3459 2018-06-05 Richard Biener <rguenther@suse.de>
3461 PR tree-optimization/86047
3462 * tree-ssa-loop.c (for_each_index): Glob handling of all
3463 decls and constants and really handle all of them.
3465 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3468 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3469 qualifier_void_pointer and qualifier_const_void_pointer.
3470 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3471 (arm_init_builtins): Handle the above.
3472 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3473 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3474 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3477 2018-06-05 Martin Liska <mliska@suse.cz>
3479 * auto-profile.c (read_autofdo_file): Do not use
3480 gcov_ctr_summary struct.
3481 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3482 * coverage.c (struct counts_entry): Likewise.
3483 (read_counts_file): Read just single summary entry.
3484 (get_coverage_counts): Use gcov_summary struct.
3485 * coverage.h (get_coverage_counts): Likewise.
3486 * gcov-dump.c (dump_working_sets): Likewise.
3487 (tag_summary): Dump just single summary.
3488 * gcov-io.c (gcov_write_summary): Write just histogram
3490 (gcov_read_summary): Read just single summary.
3491 (compute_working_sets): Use gcov_summary struct.
3492 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3493 of GCOV_COUNTERS_SUMMABLE.
3494 (GCOV_COUNTERS_SUMMABLE): Remove.
3495 (GCOV_FIRST_VALUE_COUNTER): Replace with
3496 GCOV_COUNTER_V_INTERVAL.
3497 (struct gcov_ctr_summary): Remove.
3498 (struct gcov_summary): Directly use fields of former
3500 (compute_working_sets): Use gcov_summary struct.
3501 * gcov.c (read_count_file): Do not use ctrs fields.
3502 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3504 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3506 * profile.c: Likewise.
3507 * profile.h: Likewise.
3509 2018-06-05 Martin Liska <mliska@suse.cz>
3511 PR gcov-profile/84846
3512 * gcov.c (output_lines): Print working directory only
3513 in intermediate format.
3515 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3517 * config/s390/s390-builtin-types.def: Add void function type.
3518 * config/s390/s390-builtins.def: Use the function type for the
3521 2018-06-04 Jim Wilson <jimw@sifive.com>
3523 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3525 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3526 and maybe_eh_return. Change regno to unsigned int. Use new args to
3527 handle EH_RETURN_DATA_REGNO registers properly.
3528 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3529 (riscv_expand_epilogue): Update comment. Change argument name and
3530 type. Update code to use new name and type. Pass new args to
3531 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3533 * config/riscv/riscv.md (NORMAL_RETURN): New.
3534 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3535 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3536 (eh_return): Call gen_eh_return_internal and emit barrier.
3537 (eh_return_internal): Call riscv_expand_epilogue.
3539 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3541 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3542 bit_insertion field and declare can_be_merged_into method.
3543 (merged_store_group::can_be_merged_into): New method.
3544 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3545 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3546 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3548 2018-06-04 Richard Biener <rguenther@suse.de>
3550 PR tree-optimization/85955
3551 * builtins.c (fold_builtin_sincos): Convert pointers to
3552 destination to appropriate type before dereferencing.
3554 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3556 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3558 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3560 * expr.c (expand_expr_real_1): Force the operand into memory if
3561 its TYPE_MODE is BLKmode and if there is no integer mode for
3562 the number of bits being extracted.
3564 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3568 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3569 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3571 2018-06-04 Richard Biener <rguenther@suse.de>
3573 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3574 (cleanup_tree_cfg_noloop): ... single caller. Do
3575 start_recording_case_labels later.
3577 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3579 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3580 to _IMMINTRIN_H_INCLUDED.
3581 * config/i386/pconfigintrin.h: Ditto.
3582 * config/i386/waitpkgintrin.h: Ditto.
3583 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3584 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3585 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3586 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3587 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3588 waitpkgintrin.h and cldemoteintrin.h.
3590 2018-06-04 Richard Biener <rguenther@suse.de>
3592 PR tree-optimization/86038
3593 * tracer.c (find_best_successor): Check probability for
3594 being initialized, bail out if not.
3596 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3599 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3600 of bits to ignore when comparing architectures.
3602 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3604 PR tree-optimization/69615
3605 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3606 maximum or minimum of the type, try to merge it also as if
3607 range1 is + [-, x - 1] or + [x + 1, -].
3610 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3612 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3614 PR tree-optimization/86034
3615 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3616 the unsigned bitfield type in a bit insertion sequence if it does not
3617 have a larger precision than the bitfield size.
3618 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3620 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3622 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3624 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3626 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3627 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3628 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3629 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3631 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3634 Disable -fdelete-null-pointer-checks for ELF toolchain.
3636 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3637 Kito Cheng <kito.cheng@gmail.com>
3639 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3640 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3641 (nds32*-*-*): Add glibc and uclibc conditions.
3642 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3643 (TARGET_EXCEPT_UNWIND_INFO): Define.
3644 * config/nds32/elf.h: New file.
3645 * config/nds32/linux.h: New file.
3646 * config/nds32/nds32-elf.opt: New file.
3647 * config/nds32/nds32-linux.opt: New file.
3648 * config/nds32/nds32-fp-as-gp.c
3649 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3650 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3652 (nds32_asm_file_end): Ditto.
3653 (nds32_print_operand): Ditto.
3654 (nds32_insert_attributes): Ditto.
3655 (nds32_init_libfuncs): New function.
3656 (TARGET_HAVE_TLS): Define.
3657 (TARGET_INIT_LIBFUNCS): Define.
3658 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3660 (TARGET_ELF): Apply different mcmodel setting.
3661 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3662 been migrated into elf.h and linux.h files.
3663 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3664 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3665 (mcmodel): The content has been migrated into nds32-elf.opt and
3666 nds32-linux.opt files.
3667 * config/nds32/t-elf: New file.
3668 * config/nds32/t-linux: New file.
3670 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3671 Shiva Chen <shiva0217@gmail.com>
3673 * config/nds32/constants.md (unspec_volatile_element): Add
3674 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3675 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3677 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3678 (make_pass_nds32_fp_as_gp): Declare.
3679 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3681 (nds32_asm_function_end_prologue): Remove unused asm output.
3682 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3683 (nds32_asm_file_start): Output necessary fp_as_gp information.
3684 (nds32_option_override): Adjust register usage.
3685 (nds32_expand_prologue): Consider fp_as_gp situation.
3686 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3687 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3690 (simple_return): Ditto.
3691 (omit_fp_begin): Output special directive for fp_as_gp.
3692 (omit_fp_end): Output special directive for fp_as_gp.
3693 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3694 mforbid-fp-as-gp): New options.
3696 2018-06-01 Mark Wielaard <mark@klomp.org>
3698 * dwarf2out.c (dwarf2out_finish): Remove generation of
3699 DW_AT_loclists_base.
3701 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3703 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3704 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3705 (struct merged_store_group): Add bit_insertion field.
3706 (dump_char_array): Use standard hexadecimal format.
3707 (merged_store_group::merged_store_group): Set bit_insertion to false.
3708 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3709 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3710 also print the mask in the dump file.
3711 (pass_store_merging::gate): Minor tweak.
3712 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3713 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3714 stores with INTEGER_CST stores.
3715 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3716 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3717 and use it throughout. Generate bit insertion sequences if need be.
3718 (pass_store_merging::process_store): Remove redundant condition.
3719 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3721 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3723 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3724 the 128-bit floating point types. Fix function comment.
3726 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3728 * config/aarch64/aarch64-simd.md
3729 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3731 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3734 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3736 PR tree-optimization/85989
3737 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3739 (backprop::intersect_uses): Check it when deciding whether this
3740 is a backedge reference.
3741 (backprop::process_block): Add each phi to m_visited_phis
3742 after visiting it, then clear it at the end.
3744 2018-06-01 Richard Biener <rguenther@suse.de>
3746 * tree-vectorizer.h (vect_dr_stmt): New function.
3747 (vect_get_load_cost): Adjust.
3748 (vect_get_store_cost): Likewise.
3749 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3750 Use vect_dr_stmt instead of DR_SMTT.
3751 (vect_record_base_alignments): Likewise.
3752 (vect_calculate_target_alignment): Likewise.
3753 (vect_compute_data_ref_alignment): Likewise and make static.
3754 (vect_update_misalignment_for_peel): Likewise.
3755 (vect_verify_datarefs_alignment): Likewise.
3756 (vector_alignment_reachable_p): Likewise.
3757 (vect_get_data_access_cost): Likewise. Pass down
3758 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3759 (vect_get_peeling_costs_all_drs): Likewise.
3760 (vect_peeling_hash_get_lowest_cost): Likewise.
3761 (vect_enhance_data_refs_alignment): Likewise.
3762 (vect_find_same_alignment_drs): Likewise.
3763 (vect_analyze_data_refs_alignment): Likewise.
3764 (vect_analyze_group_access_1): Likewise.
3765 (vect_analyze_group_access): Likewise.
3766 (vect_analyze_data_ref_access): Likewise.
3767 (vect_analyze_data_ref_accesses): Likewise.
3768 (vect_vfa_segment_size): Likewise.
3769 (vect_small_gap_p): Likewise.
3770 (vectorizable_with_step_bound_p): Likewise.
3771 (vect_prune_runtime_alias_test_list): Likewise.
3772 (vect_analyze_data_refs): Likewise.
3773 (vect_supportable_dr_alignment): Likewise.
3774 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3775 (vect_gen_prolog_loop_niters): Likewise.
3776 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3777 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3779 (vect_recog_mask_conversion_pattern): Likewise.
3780 (vect_try_gather_scatter_pattern): Likewise.
3781 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3782 to vect_get_store_cost.
3783 (vect_get_store_cost): Get stmt_info instead of DR.
3784 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3785 (vect_get_load_cost): Get stmt_info instead of DR.
3787 2018-06-01 Richard Biener <rguenther@suse.de>
3790 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3791 vars wrapped in MEM_REFs.
3793 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3795 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3796 Fix subreg tests so that we only return a choice between
3797 GENERAL_REGS and FP_REGS if the original classes included both.
3799 2018-06-01 Richard Biener <rguenther@suse.de>
3802 * tree-ssa-structalias.c (get_function_part_constraint):
3803 Handle NULL fi->decl.
3804 (find_func_aliases_for_call): Properly handle indirect
3805 fi from direct call.
3806 (find_func_clobbers): Likewise.
3807 (ipa_pta_execute): Likewise.
3808 (create_variable_info_for): For functions that are ifunc_resolver
3809 resolve to a varinfo that contains the result of the resolver call.
3810 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3813 2018-05-31 Michael Collison <michael.collison@arm.com>
3815 * config/aarch64/aarch64.md:
3816 (*fix_to_zero_extenddfdi2): New pattern.
3817 * gcc.target/aarch64/fix_extend1.c: New testcase.
3819 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3823 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3824 and BUILT_IN_STRNCMP_EQ.
3825 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3826 BUILT_IN_STRNCMP_EQ.
3827 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3828 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3829 (gimple_fold_builtin): Likewise.
3830 * tree-ssa-strlen.c (compute_string_length): New function.
3831 (determine_min_obsize): New function.
3832 (handle_builtin_string_cmp): New function to handle calls to
3833 string compare functions.
3834 (strlen_optimize_stmt): Add handling to builtin string compare
3836 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3837 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3838 * tree.c (build_common_builtin_nodes): Add new defines of
3839 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3841 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3844 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3845 functions with naked attribute.
3847 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3849 * config/i386/sse.md (avx_vec_concat<mode>):
3850 Substitute concat_tg_mode mode attribute with xtg_mode.
3851 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3852 (concat_tg_mode): Remove mode attribute.
3854 2018-05-31 Martin Sebor <msebor@redhat.com>
3857 * calls.c (alloc_max_size): Correct a logic error/typo.
3858 Treat excessive arguments as infinite. Warn for invalid arguments.
3859 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3861 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3864 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3865 and movx for Haswell.
3867 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3868 Cesar Philippidis <cesar@codesourcery.com>
3871 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3872 when emitting error on private/firstprivate reductions.
3873 * omp-low.c (lower_omp_target): Avoid reference-type processing
3874 on pointers for firstprivate clause.
3876 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3878 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3881 * config/aarch64/aarch64-simd.md
3882 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3883 (aarch64_ld1_x3_<mode>): Likewise
3884 (aarch64_st1x2<VALLDIF:mode>): Likewise
3885 (aarch64_st1_x2_<mode>): Likewise
3886 (aarch64_st1x3<VALLDIF:mode>): Likewise
3887 (aarch64_st1_x3_<mode>): Likewise
3888 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3889 (vld1_s8_x3): Likewise.
3890 (vld1_u16_x3): Likewise.
3891 (vld1_s16_x3): Likewise.
3892 (vld1_u32_x3): Likewise.
3893 (vld1_s32_x3): Likewise.
3894 (vld1_u64_x3): Likewise.
3895 (vld1_s64_x3): Likewise.
3896 (vld1_f16_x3): Likewise.
3897 (vld1_f32_x3): Likewise.
3898 (vld1_f64_x3): Likewise.
3899 (vld1_p8_x3): Likewise.
3900 (vld1_p16_x3): Likewise.
3901 (vld1_p64_x3): Likewise.
3902 (vld1q_u8_x3): Likewise.
3903 (vld1q_s8_x3): Likewise.
3904 (vld1q_u16_x3): Likewise.
3905 (vld1q_s16_x3): Likewise.
3906 (vld1q_u32_x3): Likewise.
3907 (vld1q_s32_x3): Likewise.
3908 (vld1q_u64_x3): Likewise.
3909 (vld1q_s64_x3): Likewise.
3910 (vld1q_f16_x3): Likewise.
3911 (vld1q_f32_x3): Likewise.
3912 (vld1q_f64_x3): Likewise.
3913 (vld1q_p8_x3): Likewise.
3914 (vld1q_p16_x3): Likewise.
3915 (vld1q_p64_x3): Likewise.
3916 (vst1_s64_x2): Likewise.
3917 (vst1_u64_x2): Likewise.
3918 (vst1_f64_x2): Likewise.
3919 (vst1_s8_x2): Likewise.
3920 (vst1_p8_x2): Likewise.
3921 (vst1_s16_x2): Likewise.
3922 (vst1_p16_x2): Likewise.
3923 (vst1_s32_x2): Likewise.
3924 (vst1_u8_x2): Likewise.
3925 (vst1_u16_x2): Likewise.
3926 (vst1_u32_x2): Likewise.
3927 (vst1_f16_x2): Likewise.
3928 (vst1_f32_x2): Likewise.
3929 (vst1_p64_x2): Likewise.
3930 (vst1q_s8_x2): Likewise.
3931 (vst1q_p8_x2): Likewise.
3932 (vst1q_s16_x2): Likewise.
3933 (vst1q_p16_x2): Likewise.
3934 (vst1q_s32_x2): Likewise.
3935 (vst1q_s64_x2): Likewise.
3936 (vst1q_u8_x2): Likewise.
3937 (vst1q_u16_x2): Likewise.
3938 (vst1q_u32_x2): Likewise.
3939 (vst1q_u64_x2): Likewise.
3940 (vst1q_f16_x2): Likewise.
3941 (vst1q_f32_x2): Likewise.
3942 (vst1q_f64_x2): Likewise.
3943 (vst1q_p64_x2): Likewise.
3944 (vst1_s64_x3): Likewise.
3945 (vst1_u64_x3): Likewise.
3946 (vst1_f64_x3): Likewise.
3947 (vst1_s8_x3): Likewise.
3948 (vst1_p8_x3): Likewise.
3949 (vst1_s16_x3): Likewise.
3950 (vst1_p16_x3): Likewise.
3951 (vst1_s32_x3): Likewise.
3952 (vst1_u8_x3): Likewise.
3953 (vst1_u16_x3): Likewise.
3954 (vst1_u32_x3): Likewise.
3955 (vst1_f16_x3): Likewise.
3956 (vst1_f32_x3): Likewise.
3957 (vst1_p64_x3): Likewise.
3958 (vst1q_s8_x3): Likewise.
3959 (vst1q_p8_x3): Likewise.
3960 (vst1q_s16_x3): Likewise.
3961 (vst1q_p16_x3): Likewise.
3962 (vst1q_s32_x3): Likewise.
3963 (vst1q_s64_x3): Likewise.
3964 (vst1q_u8_x3): Likewise.
3965 (vst1q_u16_x3): Likewise.
3966 (vst1q_u32_x3): Likewise.
3967 (vst1q_u64_x3): Likewise.
3968 (vst1q_f16_x3): Likewise.
3969 (vst1q_f32_x3): Likewise.
3970 (vst1q_f64_x3): Likewise.
3971 (vst1q_p64_x3): Likewise.
3973 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3975 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3976 user_label_prefix to name.
3978 * tree-core.h: Update comment about the format of NAME string
3979 passed to handler in attribute_spec.
3981 * config/msp430/msp430.md: Remove erroneous subreg expression from
3982 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3983 zero_extend{q,h}isi2.
3985 2018-05-30 Borislav Petkov <bp@suse.de>
3987 * doc/extend.texi: Document some architecture specific
3988 constraints and sort entries.
3990 2018-05-30 Martin Sebor <msebor@redhat.com>
3993 * builtins.c (expand_builtin_stpcpy_1): New function.
3994 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3995 only if the former succeeds.
3997 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3999 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4002 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4004 * doc/invoke.texi (-flinker-output): Document
4006 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4008 * passes.c (ipa_write_summaries): Only modify statements if body
4010 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4011 incrementally linking.
4012 (ipa_passes): Likewise.
4013 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4014 pass down resolution info.
4015 * common.opt (flag_incremental_link): Update info.
4016 * gcc.c (plugin specs): Turn flinker-output=* to
4017 -plugin-opt=-linker-output-known
4018 * toplev.c (compile_file): Also cut compilation when doing incremental
4020 * flag-types. (enum lto_partition_model): Add
4021 LTO_LINKER_OUTPUT_NOLTOREL.
4022 (invoke.texi): Add -flinker-output docs.
4023 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4024 link same way as WPA; do not stream in dead initializers.
4026 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4027 darf2out_register_external_decl): Support incremental link.
4029 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4031 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4033 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4035 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4036 it down to simple_object_copy_lto_debug_sections.
4037 (run_gcc): Determine incremental LTO link time and configure
4038 lto1 into non-wpa mode, disable renaming of debug sections.
4040 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4042 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4043 descriptions of various incorrectly documented functions.
4045 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4048 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4049 address check not strict.
4051 2018-05-30 Richard Biener <rguenther@suse.de>
4053 PR tree-optimization/85964
4054 * tracer.c (better_p): Drop initialized count check, we only
4055 call the function with initialized counts now.
4056 (find_best_successor): Do find a best edge if one
4057 has uninitialized count.
4058 (find_best_predecessor): Likewise. Do BB frequency check only
4059 if count is initialized.
4061 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4063 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4064 (aarch64_ldrstr_offset_compare): New.
4065 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4066 load/store orderings.
4067 (aarch64_gen_adjusted_ldpstp): Likewise.
4069 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4071 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4072 Check for subset of GENERAL_REGS and FP_REGS.
4073 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4076 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4078 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4079 and wi::to_poly_offset. Add the current offset and then check
4080 whether the sum fits, rather than using an unchecked addition of
4081 a checked term. Check for a shwi rather than a uhwi.
4082 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4083 (store_constructor): Use poly_int_tree_p.
4084 (expand_expr_real_1): Likewise.
4085 * function.c (assign_temp): Likewise.
4086 * fold-const.c (const_binop): Use poly_int_tree_p and
4088 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4090 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4091 to_poly_offset for MEM offsets.
4092 * ipa-icf.c (sem_variable::equals): Likewise.
4093 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4094 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4095 wi::to_poly_offset for BIT_FIELD_REF offsets.
4096 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4098 * var-tracking.c (emit_note_insn_var_location): Use
4099 tree_to_poly_uint64.
4101 2018-05-29 Jim Wilson <jimw@sifive.com>
4103 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4105 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4108 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4109 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4111 (sse4_1_round<mode>2): Use nonimmediate_operand
4112 for operand 1 predicate.
4114 2018-05-29 Martin Sebor <msebor@redhat.com>
4115 Richard Biener <rguenther@suse.de>
4118 * calls.c (get_size_range): Call determine_value_range instead
4119 of get_value_range..
4120 * tree-vrp.h (determine_value_range): Declared new function.
4121 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4123 2018-05-29 Richard Biener <rguenther@suse.de>
4125 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4126 sure to use non-pattern stmts for get_earlier_stmt arguments.
4127 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4128 called on pattern stmts.
4129 (get_later_stmt): Likewise.
4131 2018-05-29 Martin Liska <mliska@suse.cz>
4133 PR gcov-profile/85759
4134 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4137 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4139 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4141 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4145 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4146 VEC_PACK_FLOAT_EXPR): New tree codes.
4147 * tree-pretty-print.c (op_code_prio): Handle
4148 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4149 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4150 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4151 * tree-inline.c (estimate_operator_cost): Likewise.
4152 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4153 * fold-const.c (const_binop): Likewise.
4154 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4155 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4156 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4157 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4158 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4159 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4160 * expr.c (expand_expr_real_2): Likewise.
4161 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4162 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4163 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4165 * optabs.c (expand_widen_pattern_expr): For
4166 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4167 sign from result type rather than operand's type.
4168 (expand_binop_directly): For vec_packu_float_optab and
4169 vec_packs_float_optab allow result type to be different from operand's
4171 * optabs-tree.c (optab_for_tree_code): Handle
4172 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4173 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4174 * tree-vect-generic.c (expand_vector_operations_1): Handle
4175 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4176 VEC_PACK_FLOAT_EXPR.
4177 * tree-vect-stmts.c (supportable_widening_operation): Handle
4179 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4180 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4181 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4182 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4183 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4185 (vec_pack<floatprefix>_float_<mode>): New expander.
4186 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4188 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4189 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4190 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4191 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4192 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4194 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4195 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4196 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4197 VEC_PACK_FLOAT_EXPR): Document.
4199 2018-05-29 Richard Biener <rguenther@suse.de>
4201 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4203 (stmt_vec_info_vec): Make pointer.
4204 (init_stmt_vec_info_vec): Remove.
4205 (free_stmt_vec_info_vec): Likewise.
4206 (set_stmt_vec_info_vec): New function.
4207 (free_stmt_vec_infos): Likewise.
4208 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4209 (set_vinfo_for_stmt): Likewise.
4210 (get_earlier_stmt): Likewise.
4211 (get_later_stmt): Likewise.
4212 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4213 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4214 (vec_info::~vec_info): Free stmt_vec_infos.
4215 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4216 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4217 (pass_slp_vectorize::execute): Likewise.
4218 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4219 (free_stmt_vec_info_vec): Likewise.
4220 (set_stmt_vec_info_vec): New function.
4221 (free_stmt_vec_infos): Likewise.
4222 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4223 the global stmt_vec_info_vec.
4224 * tree-parloops.c (gather_scalar_reductions): Use
4225 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4228 2018-05-29 Richard Biener <rguenther@suse.de>
4230 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4232 2018-05-29 Martin Liska <mliska@suse.cz>
4233 David Malcolm <dmalcolm@redhat.com>
4235 * vec.c (test_reverse): New.
4236 (vec_c_tests): Add new test.
4237 * vec.h (vl_ptr>::reverse): New function.
4239 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4241 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4243 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4246 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4248 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4250 2018-05-28 Richard Biener <rguenther@suse.de>
4252 PR tree-optimization/85933
4253 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4254 look at stmts marked as vectorizable.
4256 2018-05-28 Richard Biener <rguenther@suse.de>
4258 PR tree-optimization/85934
4259 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4260 vector boolean check before scalar optimization.
4262 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4264 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4267 2018-05-28 Mark Wielaard <mark@klomp.org>
4269 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4270 if it is an expression containing a minus sign.
4272 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4274 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4276 2018-05-27 Paul Koning <ni1d@arrl.net>
4278 * config/pdp11/pdp11.md (truncsihi2): Remove.
4280 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4281 Chung-Ju Wu <jasonwucj@gmail.com>
4283 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4285 (unaligned_store_dw): Ditto.
4286 * config/nds32/nds32-memory-manipulation.c
4287 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4288 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4289 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4290 (emit_setmem_word_loop): Rename to ...
4291 (emit_setmem_doubleword_loop): ... this.
4292 (nds32_gen_dup_4_byte_to_word_value): New function.
4293 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4294 (nds32_expand_setmem_loop): Refine implementation.
4295 (nds32_expand_setmem_loop_v3m): Ditto.
4296 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4299 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4301 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4303 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4305 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4306 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4307 machine->attr_no_prologue_p.
4308 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4309 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4310 (nds32_expand_epilogue): Consider attr_naked_p.
4311 (nds32_expand_epilogue_v3pop): Likewise.
4312 (nds32_can_use_return_insn): Likewise.
4313 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4314 attr_no_prologue_p fields.
4315 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4317 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4320 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4322 * config/i386/sse.md
4323 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4325 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4327 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4329 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4331 (*<floatsuffix>floatv2div2sf2): Rename to ...
4332 (*float<floatunssuffix>v2div2sf2): ... this.
4333 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4334 (float<floatunssuffix>v2div2sf2_mask): ... this.
4335 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4336 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4337 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4339 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4341 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4343 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4345 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4347 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4349 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4350 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4351 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4352 gen_ufix_truncv8dfv8si2.
4353 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4354 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4355 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4356 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4357 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4358 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4359 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4360 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4362 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4366 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4368 2018-05-25 Jim Wilson <jimw@sifive.com>
4370 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4371 * config/riscv/riscv.c (struct machine_function): Add
4372 interrupt_handler_p and attribute_checked_p fields.
4373 (riscv_attribute_table): Add interrupt.
4374 (riscv_interrupt_type_p): New.
4375 (riscv_save_reg_p): Save extra regs for interrupt handler.
4376 (riscv_use_save_libcall): Return false for interrupt handler.
4377 (riscv_first_stack_step): Add forward declaration.
4378 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4379 for interrupt handler with large frame. Use it for saved reg list.
4380 (riscv_expand_prologue): Move flag_stack_usage_info support to
4381 eliminate duplication.
4382 (riscv_expand_epilogue): Generate mret for interrupt handler.
4383 (riscv_epilogue_uses): New.
4384 (riscv_can_use_return_insn): Return false for interrupt handler.
4385 (riscv_function_ok_for_sibcall): Likewise.
4386 (riscv_set_current_function): Add interrupt handler support.
4387 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4388 * config/riscv/riscv.md (UNSPECV_MRET): New.
4390 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4392 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4394 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4396 PR tree-optimization/85712
4397 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4398 this candidate has already been replaced in-situ by a copy.
4400 2018-05-25 Jason Merrill <jason@redhat.com>
4402 PR c++/80485 - inline function non-zero address.
4403 * symtab.c (nonzero_address): Check DECL_COMDAT.
4405 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4408 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4409 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4410 extension of SImode operation. Use const123_operand predicate.
4411 (*saddsi_1): Remove.
4412 (*saddl_se_1): Ditto.
4414 (*ssubl_se_1): Ditto.
4415 * config/alpha/predicates.md (const123_operand): New predicate.
4416 * config/alpha/constraints.md (P): Use IN_RANGE.
4418 2018-05-25 Richard Biener <rguenther@suse.de>
4420 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4422 (ref_maybe_used_by_stmt_p): Likewise.
4423 (stmt_may_clobber_ref_p): Likewise.
4424 (stmt_may_clobber_ref_p_1): Likewise.
4425 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4427 (ref_maybe_used_by_stmt_p): Likewise.
4428 (stmt_may_clobber_ref_p): Likewise.
4429 (stmt_may_clobber_ref_p_1): Likewise.
4430 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4431 the alias oracle to disambiguate DRs with stmts DR analysis
4433 (vect_analyze_data_refs): Do not give up on not analyzable
4434 DRs for BB vectorization. Remove code truncating the dataref
4437 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4440 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4441 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4442 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4444 2018-05-25 Richard Biener <rguenther@suse.de>
4446 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4447 function, combining stmt data ref gathering and fatal analysis
4449 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4450 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4451 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4452 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4453 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4455 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4457 PR tree-optimization/85720
4458 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4459 SCC if all partitions are builtins.
4460 (version_loop_by_alias_check): New parameter. Generate cancelable
4461 runtime alias check if all partitions are builtins.
4462 (distribute_loop): Update call to above function.
4464 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4466 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4467 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4468 (parm_default_def_partition_arg): Ditto.
4469 (set_parm_default_def_partition): Ditto.
4470 (get_parm_default_def_partitions): Ditto and make it static.
4471 (get_undefined_value_partitions): Ditto and make it static.
4472 (remove_ssa_form): Refactor call to init_var_map here.
4473 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4474 computation for loop region.
4475 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4476 (register_default_def): Delete.
4477 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4478 (parm_default_def_partition_arg): Ditto.
4479 (set_parm_default_def_partition): Ditto.
4480 (get_parm_default_def_partitions): Ditto and make it static.
4481 (get_undefined_value_partitions): Ditto and make it static.
4482 (coalesce_with_default, coalesce_with_default): Update comment.
4483 (create_coalesce_list_for_region): New func factored out from
4484 create_outofssa_var_map.
4485 (populate_coalesce_list_for_outofssa): New func factored out from
4486 create_outofssa_var_map and coalesce_ssa_name.
4487 (create_outofssa_var_map): Delete.
4488 (coalesce_ssa_name): Refactor to support live range computation.
4489 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4490 (get_parm_default_def_partitions): Delete.
4491 (get_undefined_value_partitions): Ditto.
4492 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4493 computation for loop region.
4494 (new_tree_live_info, loe_visit_block): Ditto.
4495 (live_worklist, set_var_live_on_entry): Ditto.
4496 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4497 * tree-ssa-live.h (struct _var_map): New fields.
4498 (init_var_map): Change decl.
4499 (region_contains_p): New.
4501 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4503 * tree-ssa-live.h (live_merge_and_clear): Delete.
4505 2018-05-25 Richard Biener <rguenther@suse.de>
4508 * tree-dump.c (dequeue_and_dump): Remove access to removed
4509 operand 2 of a SWITCH_EXPR.
4511 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4513 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4514 multiplication and division.
4515 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4516 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4517 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4518 (cond_udiv_optab, cond_umod_optab): New optabs.
4519 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4520 (IFN_COND_RDIV): New internal functions.
4521 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4522 TRUNC_MOD_EXPR and RDIV_EXPR.
4523 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4524 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4526 (SVE_INT_BINARY): Include mult.
4527 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4528 (optab, sve_int_op): Handle mult.
4529 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4531 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4532 for SVE_INT_BINARY_SD.
4534 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4536 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4537 (optab, sve_int_op): Handle div and udiv.
4538 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4539 for SVE_INT_BINARY_SD.
4540 (*<optab><mode>3): New insn for the same.
4542 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4544 * tree-vect-patterns.c: Include predict.h.
4545 (vect_recog_divmod_pattern): Restrict check for division support
4546 to when optimizing for size.
4548 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4550 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4551 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4552 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4553 (gimple_match_op::set_op): Likewise.
4554 (gimple_resimplify4): Declare.
4555 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4556 (expr::gen_transform): Likewise.
4557 (decision_tree::gen): Generate a simplification routine for 4 operands.
4558 * gimple-match-head.c (gimple_simplify): Add an overload for
4559 4 operands. In the top-level function, handle up to 4 call
4560 arguments and call gimple_resimplify4.
4561 (gimple_resimplify4): New function.
4562 (build_call_internal): Pass a fourth operand.
4563 (maybe_push_to_seq): Likewise.
4564 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4565 Fold VEC_COND_EXPRs of an operation and a default value into
4566 an IFN_COND_* function if possible.
4567 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4569 (SVE_COND_FP_BINARY): Include them.
4570 (optab, sve_fp_op): Handle them.
4571 (SVE_INT_BINARY_REV): New code iterator.
4572 (SVE_COND_FP_BINARY_REV): New int iterator.
4573 (commutative): New int attribute.
4574 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4576 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4578 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4579 (*cond_<optab><mode>): New patterns for reversed operands.
4581 2018-05-25 Richard Biener <rguenther@suse.de>
4583 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4584 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4585 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4586 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4587 * tree-vect-data-refs.c (everywhere): Adjust users.
4588 * tree-vect-loop.c (everywhere): Likewise.
4589 * tree-vect-slp.c (everywhere): Likewise.
4590 * tree-vect-stmts.c (everywhere): Likewise.
4591 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4593 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4595 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4597 (gcc_cv_as_section_exclude): ... this.
4598 Try Solaris as #exclude syntax.
4599 * configure: Regenerate.
4600 * config.in: Regenerate.
4601 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4603 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4604 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4606 * varasm.c (default_elf_asm_named_section): Don't check if
4607 HAVE_GAS_SECTION_EXCLUDE is defined.
4609 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4611 * doc/md.texi: Update the documentation of the cond_* optabs
4612 to mention the new final operand. Fix GET_MODE_NUNITS call.
4613 Describe the scalar case too.
4614 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4615 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4617 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4618 (get_conditional_internal_fn): Update comment.
4619 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4620 accumulator value as a final argument to conditional functions.
4621 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4622 a define_expand and add an "else" operand. Assert for now that
4623 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4624 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4625 (*cond_<optab><mode>): New patterns.
4626 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4627 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4628 (UNSPEC_COND_EOR): Delete.
4629 (optab): Remove associated mappings.
4630 (SVE_INT_BINARY): New code iterator.
4631 (sve_int_op): Remove int attribute and add "minus" to the code
4633 (SVE_COND_INT_OP): Delete.
4634 (SVE_COND_FP_OP): Rename to...
4635 (SVE_COND_FP_BINARY): ...this.
4637 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4639 * optabs.c (can_reuse_operands_p): New function.
4640 (maybe_legitimize_operands): Try to reuse the results for
4643 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4645 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4646 Add {q} suffix to insn mnemonic.
4648 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4650 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4651 (msp430_warn_func_return): New.
4653 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4655 * fold-const.c (tree_nonzero_bits): New function.
4656 * fold-const.h (tree_nonzero_bits): Likewise.
4657 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4658 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4660 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4664 * varasm.c (assemble_alias): Check ifunc_resolver only on
4667 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4670 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4671 when memory input operand is handled.
4673 2018-05-24 Luis Machado <luis.machado@linaro.org>
4675 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4677 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4679 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4681 * match.pd: Delay FMA folds until after vectorization.
4683 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4686 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4687 address check not strict.
4689 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4691 * gimple-match.h (gimple_match_op): New class.
4692 (mprts_hook): Replace parameters with a gimple_match_op *.
4693 (maybe_build_generic_op): Likewise.
4694 (gimple_simplified_result_is_gimple_val): Replace parameters with
4695 a const gimple_match_op *.
4696 (gimple_simplify): Replace code_helper * and tree * parameters with
4697 a gimple_match_op * parameter.
4698 (gimple_resimplify1): Replace code_helper *, tree and tree *
4699 parameters with a gimple_match_op * parameter.
4700 (gimple_resimplify2): Likewise.
4701 (gimple_resimplify3): Likewise.
4702 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4703 parameters with a gimple_match_op * parameter.
4704 * gimple-match-head.c (gimple_simplify): Change prototypes of
4705 auto-generated functions to take a gimple_match_op * instead of
4706 separate code_helper * and tree * parameters. Make the same
4707 change in the top-level overload and update calls to the
4708 gimple_resimplify routines. Update calls to the auto-generated
4709 functions and to maybe_push_res_to_seq in the publicly-facing
4710 operation-specific gimple_simplify overloads.
4711 (gimple_match_op::MAX_NUM_OPS): Define.
4712 (gimple_resimplify1): Replace rcode and ops with a single res_op
4713 parameter. Update call to gimple_simplify.
4714 (gimple_resimplify2): Likewise.
4715 (gimple_resimplify3): Likewise.
4716 (mprts_hook): Replace parameters with a gimple_match_op *.
4717 (maybe_build_generic_op): Likewise.
4718 (build_call_internal): Replace type, nargs and ops with
4719 a gimple_match_op *.
4720 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4721 with a single gimple_match_op *. Update calls to mprts_hook,
4722 build_call_internal and gimple_simplified_result_is_gimple_val.
4723 Factor out code that is common to the tree_code and combined_fn cases.
4724 * genmatch.c (expr::gen_transform): Replace tem_code and
4725 tem_ops with a gimple_match_op called tem_op. Update calls
4726 to the gimple_resimplify functions and maybe_push_res_to_seq.
4727 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4728 res_ops. Update call to the gimple_resimplify functions.
4729 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4730 (decision_tree::gen): Make the functions take a gimple_match_op *
4731 called res_op instead of separate res_code and res_ops parameters.
4732 Update call accordingly.
4733 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4734 and ops with a single res_op parameter. Update calls to
4735 maybe_build_generic_op and maybe_push_res_to_seq.
4736 (fold_stmt_1): Update calls to gimple_simplify and
4737 replace_stmt_with_simplification.
4738 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4739 and gimple_simplified_result_is_gimple_val.
4740 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4742 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4743 with a gimple_match_op *.
4744 (vn_nary_build_or_lookup): Likewise. Update call to
4745 vn_nary_build_or_lookup_1.
4746 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4747 gimple_match_op *. Update calls to the gimple_resimplify routines
4748 and to gimple_simplified_result_is_gimple_val.
4749 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4750 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4751 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4752 (visit_nary_op): Likewise.
4753 (visit_reference_op_load): Likewise.
4755 2018-05-23 Luis Machado <luis.machado@linaro.org>
4757 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4758 modifier for printing the step amount.
4760 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4763 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4766 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4768 * doc/sourcebuild.texi (Endianness): New subsubsection.
4770 2018-05-23 Luis Machado <luis.machado@linaro.org>
4772 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4773 <prefetch_dynamic_strides>: New const bool field.
4774 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4775 prefetch_dynamic_strides.
4776 (exynosm1_prefetch_tune): Likewise.
4777 (thunderxt88_prefetch_tune): Likewise.
4778 (thunderx_prefetch_tune): Likewise.
4779 (thunderx2t99_prefetch_tune): Likewise.
4780 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4782 (aarch64_override_options_internal): Update to set
4783 PARAM_PREFETCH_DYNAMIC_STRIDES.
4784 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4785 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4786 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4787 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4788 prefetch-dynamic-strides setting.
4790 2018-05-23 Luis Machado <luis.machado@linaro.org>
4792 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4793 <minimum_stride>: New const int field.
4794 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4795 minimum_stride field defaulting to -1.
4796 (exynosm1_prefetch_tune): Likewise.
4797 (thunderxt88_prefetch_tune): Likewise.
4798 (thunderx_prefetch_tune): Likewise.
4799 (thunderx2t99_prefetch_tune): Likewise.
4800 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4801 <default_opt_level>: Set to 3.
4802 (aarch64_override_options_internal): Update to set
4803 PARAM_PREFETCH_MINIMUM_STRIDE.
4804 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4805 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4806 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4807 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4808 stride is constant and is below the minimum stride threshold.
4810 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4812 * config/arm/arm-cpus.in (mode26): Delete.
4813 (armv4): Delete mode26 reference.
4814 * config/arm/arm.c (arm_configure_build_target): Delete use of
4817 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4819 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4821 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4822 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4824 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4825 (floatunsdidf2): Ditto.
4827 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4829 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4830 (fixuns_trunc<mode>si2_avx512f): Ditto.
4831 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4832 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4833 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4835 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4837 PR rtl-optimization/79985
4838 * df-scan.c (df_insn_refs_collect): Remove special case for
4839 global registers and asm statements.
4841 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4843 * extend.texi (Global Register Variables): Rewrite the bullet list.
4844 Note that the register is available for allocation. Note that access
4845 via inline asm must use constraints. Add note about async-signal
4846 handlers. Remove paragraph about automagic register selection.
4848 2018-05-23 Richard Biener <rguenther@suse.de>
4850 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4851 of fixed offset from memset VN.
4853 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4855 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4857 (alloc_cand_and_find_basis): Initialize first_interp field.
4858 (slsr_process_mul): Modify first_interp field.
4859 (slsr_process_add): Likewise.
4860 (slsr_process_cast): Modify first_interp field for each new
4862 (slsr_process_copy): Likewise.
4863 (dump_candidate): Dump first_interp field.
4864 (replace_mult_candidate): Process all interpretations, not just
4866 (replace_rhs_if_not_dup): Likewise.
4867 (replace_one_candidate): Likewise.
4869 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4871 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4873 (aarch64_needs_frame_chain): New function.
4874 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4876 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4879 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4880 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4881 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4882 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4883 as true for strict-align.
4884 (aarch64_can_inline_p): Perform checks even when callee has no
4885 attributes to check for strict alignment.
4886 * doc/extend.texi (AArch64 Function Attributes): Document
4888 * doc/invoke.texi: (AArch64 Options): Likewise.
4890 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4892 PR tree-optimization/85853
4893 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4894 the handling of the root of the node to...
4895 (vect_slp_analyze_node_operations_1): ...this new function,
4896 and run the whole thing with the child nodes' def types
4897 set according to their SLP node's def type.
4899 2018-05-23 Richard Biener <rguenther@suse.de>
4902 * tree-data-ref.c (create_runtime_alias_checks): Defer
4903 and ignore overflow warnings.
4905 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4907 PR tree-optimization/85822
4908 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4911 2018-05-23 Richard Biener <rguenther@suse.de>
4913 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4914 memset constants via native_interpret_expr.
4916 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4919 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4921 (cgraph_node::create_alias): Likewise.
4922 (cgraph_node::get_availability): Check ifunc_resolver instead
4923 of looking up ifunc attribute.
4924 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4925 * varasm.c (do_assemble_alias): Likewise.
4926 (assemble_alias): Likewise.
4927 (default_binds_local_p_3): Likewise.
4928 * cgraph.h (cgraph_node): Add ifunc_resolver.
4929 (cgraph_node::only_called_directly_or_aliased_p): Return false
4931 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4933 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4934 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4935 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4936 instead of looking up ifunc attribute.
4938 2018-05-22 Luis Machado <luis.machado@linaro.org>
4940 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4942 2018-05-22 Martin Sebor <msebor@redhat.com>
4945 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4946 only when expasion succeeds.
4947 (expand_builtin_strcmp): Same.
4948 (expand_builtin_strncmp): Same.
4950 2018-05-22 Martin Sebor <msebor@redhat.com>
4952 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4954 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4955 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4957 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4958 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4960 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4961 Add check for legitimate_address.
4962 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4963 (aarch64_swap_ldrstr_operands): New.
4964 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4967 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4968 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4970 * config/aarch64/aarch64.md: New patterns to generate stp
4972 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4973 single words and double words.
4974 (load_pair_sw, load_pair_dw): Likewise.
4975 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4977 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4979 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4980 for different mode ldpstp and add peephole for merged zero stores.
4982 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4984 (aarch64_gen_store_pair): Rename calls to match new patterns.
4985 (aarch64_gen_load_pair): Rename calls to match new patterns.
4986 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4987 (load_pair<DREG:mode><DREG2:mode>): ... This.
4988 (store_pair<mode>): Rename to...
4989 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4990 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4992 (V_INT_EQUIV): Handle SImode.
4993 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4996 2018-05-22 Martin Sebor <msebor@redhat.com>
4999 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5000 or ajust the presumed bound on an operation to avoid unnecessary
5003 2018-05-22 Martin Sebor <msebor@redhat.com>
5005 PR tree-optimization/85826
5006 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5007 assuming that a DECL necesarily has a constant size.
5009 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5012 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5014 2018-05-22 Richard Biener <rguenther@suse.de>
5016 PR tree-optimization/85834
5017 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5018 non-constant and non-zero memset arguments.
5020 2018-05-22 Martin Liska <mliska@suse.cz>
5023 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5025 2018-05-22 Richard Biener <rguenther@suse.de>
5027 PR tree-optimization/85863
5028 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5029 comparisons when vectype is specified.
5030 (vectorizable_condition): Do not specify vectype for
5031 vect_is_simple_cond when SLP vectorizing.
5033 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5036 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5037 define __ibm128 as long double.
5038 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5039 as a distinct type when IEEE 128-bit support is enabled.
5040 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5041 128-bit types to use the correct functions.
5042 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5043 convert between 128-bit floating point types that have different
5044 modes but the same representation, instead of using gen_lowpart to
5046 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5048 (IFKF_reg): New attributes to give the register constraints for
5050 (extend<mode>tf2_internal): New insns to mark an explicit
5051 conversion between 128-bit floating point types that have a
5052 different mode but share the same representation.
5054 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5056 PR tree-optimization/85814
5057 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5058 a null return from get_strinfo when unsharing the next
5059 strinfo in the chain.
5061 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5064 * varasm.c (weak_finish): Clean up weak_decls.
5066 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5068 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5069 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5070 UNSPEC_UADALP values.
5071 * config/aarch64/iterators.md (ABAL): New int iterator.
5074 (sur): Add mappings for the above.
5075 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5077 (aarch64_<sur>abal<mode>_4): Likewise.
5078 (aarch64_<sur>adalp<mode>_3): Likewise.
5079 (<sur>sadv16qi): New define_expand.
5081 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5083 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5084 (*movdf_internal): Ditto.
5085 (*rcpsf2_sse): Ditto.
5086 (*rsqrtsf2_sse): Ditto.
5087 (*sqrt<mode>2_sse): Ditto.
5089 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5091 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5093 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5094 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5095 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5096 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5098 * config/aarch64/arm_neon.h: Likewise.
5099 * config/aarch64/iterators.md (VQ_I): New.
5101 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5103 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5105 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5108 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5110 * config/nds32/constraints.md (S): New constraint.
5111 * config/nds32/nds32.md (call_internal): Use constraint S.
5112 (call_value_internal): Likewise.
5113 (sibcall_internal): Likewise.
5114 (sibcall_value_internal): Likewise.
5116 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5117 Chung-Ju Wu <jasonwucj@gmail.com>
5119 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5122 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5123 Chung-Ju Wu <jasonwucj@gmail.com>
5125 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5126 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5127 (nds32_rtx_costs_impl): Simplify.
5128 (nds32_address_cost_impl): Simplify.
5129 (nds32_init_rtx_costs): New function.
5130 (nds32_rtx_costs_speed_prefer): Likewise.
5131 (nds32_rtx_costs_size_prefer): Likewise.
5132 (nds32_address_cost_speed_prefer): Likewise.
5133 (nds32_address_cost_speed_fwprop): Likewise.
5134 (nds32_address_cost_size_prefer): Likewise.
5135 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5136 * config/nds32/nds32.c (nds32_option_override): Use
5137 nds32_init_rtx_costs function.
5139 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5141 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5142 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5143 (TARGET_PIPELINE_N8): Likewise.
5144 (TARGET_PIPELINE_N10): Likewise.
5145 (TARGET_PIPELINE_N13): Likewise.
5146 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5148 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5150 * config/nds32/nds32-fpu.md: Update copyright year.
5152 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5154 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5156 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5158 * config/nds32/nds32.c
5159 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5160 * config/nds32/nds32.opt (minline-asm-r15): New option.
5162 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5164 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5166 * config/nds32/nds32.md (abssi2): New pattern.
5168 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5170 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5171 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5172 Merge insn pattern from sse_cvtsi2ss<round_name> and
5173 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5174 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5175 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5176 using SWI48 mode iterator.
5177 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5178 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5179 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5180 pattern from sse_cvttss2si<round_saeonly_name>
5181 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5182 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5183 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5184 using SWI48 mode iterator.
5185 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5186 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5187 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5188 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5189 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5190 using SWI48 mode iterator.
5191 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5192 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5193 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5194 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5195 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5196 SWI48 mode iterator.
5197 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5198 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5199 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5200 pattern from sse_cvttsd2si<round_saeonly_name>
5201 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5203 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32-md-auxiliary.c
5206 (nds32_valid_smw_lwm_base_p): Refine.
5207 (nds32_output_smw_single_word): Refine.
5208 (nds32_output_smw_double_word): New.
5209 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5211 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5213 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5214 (nds32_output_stack_pop): Refine.
5215 (nds32_expand_unaligned_load): Refine.
5216 (nds32_expand_unaligned_store): Refine.
5218 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5219 Chung-Ju Wu <jasonwucj@gmail.com>
5221 * config/nds32/constants.md: Add TP_REGNUM constant.
5222 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5223 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5225 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5226 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5227 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5228 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5229 and PIC code generation.
5230 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5232 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5234 * config/nds32/nds32.md: Support TLS and PIC.
5235 * config/nds32/nds32.c: Support TLS and PIC.
5236 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5237 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5240 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5242 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5243 mode with E_ prefix.
5245 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5246 Chung-Ju Wu <jasonwucj@gmail.com>
5248 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5249 * config/nds32/nds32-md-auxiliary.c
5250 (symbolic_reference_mentioned_p): New.
5251 (nds32_legitimize_ict_address): New.
5252 (nds32_expand_ict_move): New.
5253 (nds32_indirect_call_referenced_p): New.
5254 (nds32_symbol_binds_local_p): Delete.
5255 (nds32_long_call_p): Modify.
5256 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5257 * config/nds32/nds32-protos.h
5258 (symbolic_reference_mentioned_p): Declare.
5259 (nds32_legitimize_ict_address): Declare.
5260 (nds32_expand_ict_move): Declare.
5261 (nds32_indirect_call_referenced_p): Declare.
5262 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5263 (nds32_relax_group): Use nds32_ict_const_p as condition.
5264 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5265 (nds32_asm_file_start): Output ict_model directive in asm code.
5266 (nds32_legitimate_address_p): Consider indirect call.
5267 (nds32_print_operand): Consider indirect call.
5268 (nds32_print_operand_address): Consider indirect call.
5269 (nds32_insert_attributes): Handle "indirect_call" attribute.
5270 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5271 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5272 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5273 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5274 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5275 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5276 (TARGET_ICT_MODEL_SMALL): Define.
5277 (TARGET_ICT_MODEL_LARGE): Define.
5278 * config/nds32/nds32.md (movsi): Consider ict model.
5279 (call, call_value): Consider ict model.
5280 (sibcall, sibcall_value): Consider ict model.
5281 * config/nds32/nds32.opt (mict-model): New option.
5282 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5285 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5286 Monk Chiang <sh.chiang04@gmail.com>
5287 Jim Wilson <jimw@sifive.com>
5289 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5290 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5291 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5292 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5293 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5294 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5295 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5296 compute save_libcall_adjustment properly.
5297 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5298 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5299 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5300 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5301 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5302 (ABI_SPEC): Handle mabi=ilp32e.
5303 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5304 (RVE): Add RVE mask.
5305 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5306 <-march>: Add rv32e as an example.
5308 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5311 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5312 (intra_create_variable_infos): Handle C++ constructors.
5314 2018-05-18 Martin Liska <mliska@suse.cz>
5316 * passes.def: Remove a redundant pass.
5318 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5321 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5323 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5325 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5327 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5329 (armv2, armv2a, armv3, armv3m): Delete architectures.
5330 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5331 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5332 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5334 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5335 (*mulsidi3adddi): Likewise.
5336 (mulsidi3): Likewise.
5337 (*mulsidi3_nov6): Likewise.
5338 (umulsidi3): Likewise.
5339 (umulsidi3_nov6): Likewise.
5340 (umaddsidi4): Likewise.
5341 (*umulsidi3adddi): Likewise.
5342 (smulsi3_highpart): Likewise.
5343 (*smulsi3_highpart_nov6): Likewise.
5344 (umulsi3_highpart): Likewise.
5345 (*umulsi3_highpart_nov6): Likewise.
5346 * config/arm/arm.h (arm_arch3m): Delete.
5347 * config/arm/arm.c (arm_arch3m): Delete.
5348 (arm_option_override_internal): Update armv3-related comment.
5349 (arm_configure_build_target): Delete use of isa_bit_mode32.
5350 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5351 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5352 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5354 (mulusa3): Likewise.
5355 * config/arm/arm-protos.h (arm_arch3m): Delete.
5356 * config/arm/arm-tables.opt: Regenerate.
5357 * config/arm/arm-tune.md: Likewise.
5358 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5359 deleted architectures.
5361 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5363 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5364 (armv5t, armv5te): New features.
5365 (ARMv5, ARMv5e): Delete fgroups.
5366 (ARMv5t, ARMv5te): Adjust for above changes.
5368 (armv5, armv5e): Delete arches.
5369 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5371 (*call_reg_arm): Likewise.
5372 (*call_value_reg_armv5): Likewise.
5373 (*call_value_reg_arm): Likewise.
5374 (*call_symbol): Likewise.
5375 (*call_value_symbol): Likewise.
5376 (*sibcall_insn): Likewise.
5377 (*sibcall_value_insn): Likewise.
5379 (prefetch): Likewise.
5380 (define_split and define_peephole2 dependent on arm_arch5):
5382 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5384 (TARGET_ARM_QBIT): Likewise.
5385 (TARGET_DSP_MULTIPLY): Likewise.
5386 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5387 (arm_arch5, arm_arch5e): Delete.
5388 (arm_arch5t, arm_arch5te): Declare.
5389 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5390 (arm_arch5t): Declare.
5391 (arm_option_reconfigure_globals): Update for the above.
5392 (arm_options_perform_arch_sanity_checks): Update comment, replace
5393 use of arm_arch5 with arm_arch5t.
5394 (use_return_insn): Likewise.
5395 (arm_emit_call_insn): Likewise.
5396 (output_return_instruction): Likewise.
5397 (arm_final_prescan_insn): Likewise.
5398 (arm_coproc_builtin_available): Likewise.
5399 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5400 arm_arch5e with arm_arch5t and arm_arch5te.
5401 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5402 (arm_arch5t, arm_arch5te): Declare.
5403 * config/arm/arm-tables.opt: Regenerate.
5404 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5405 * config/arm/t-multilib: Likewise.
5406 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5407 instead of arm_arch5.
5408 (*call_reg_thumb1): Likewise.
5409 (*call_value_reg_thumb1_v5): Likewise.
5410 (*call_value_reg_thumb1): Likewise.
5411 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5413 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5415 2018-05-18 Martin Liska <mliska@suse.cz>
5417 PR gcov-profile/84846
5418 * doc/gcov.texi: Document -t option of gcov tool.
5420 2018-05-18 Martin Liska <mliska@suse.cz>
5422 PR gcov-profile/84846
5423 * gcov.c (print_usage): Add new -t option.
5424 (process_args): Handle the option.
5425 (generate_results): Use stdout as output when requested by
5428 2018-05-18 Martin Liska <mliska@suse.cz>
5430 PR gcov-profile/84846
5431 * coverage.c (coverage_init): Write PWD to .gcno file.
5432 * doc/gcov.texi: Document how working directory is printed.
5433 * gcov-dump.c (dump_gcov_file): Print PWD.
5434 * gcov.c (output_intermediate_file): Likewise.
5435 (read_graph_file): Read PWD string.
5436 (output_lines): Print PWD.
5438 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5441 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5442 for retval and return false if all args to phi are zero.
5444 2018-05-18 Richard Biener <rguenther@suse.de>
5446 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5448 (evrp_dom_walker::before_dom_children): Call it.
5450 2018-05-18 Richard Biener <rguenther@suse.de>
5452 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5453 results when processing array refs with variable index.
5455 2018-05-18 Toon Moene <toon@moene.org>
5457 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5458 directly after that of -floop-interchange. Indicate that both
5459 options are enabled by default when specifying -O3.
5461 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5463 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5464 iterator. Delete separate integer-mode vec_set<mode> expander.
5465 (aarch64_simd_vec_setv2di): Delete.
5466 (vec_setv2di): Delete.
5467 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5468 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5469 the "w, r" alternative.
5471 2018-05-18 Martin Liska <mliska@suse.cz>
5473 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5474 * tree-pass.h (make_pass_lower_switch_O0): New function.
5475 * tree-switch-conversion.c (node_has_low_bound): Remove.
5476 (node_has_high_bound): Likewise.
5477 (node_is_bounded): Likewise.
5478 (class pass_lower_switch): Make it a template type and create
5480 (pass_lower_switch::execute): Add template argument.
5481 (make_pass_lower_switch): New function.
5482 (make_pass_lower_switch_O0): New function.
5483 (do_jump_if_equal): Remove.
5484 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5485 all the hard work to tree optimization passes.
5487 2018-05-18 Martin Liska <mliska@suse.cz>
5489 * dbgcnt.c (limit_low): Renamed from limit.
5490 (limit_high): New variable.
5491 (dbg_cnt_is_enabled): Check for upper limit.
5492 (dbg_cnt): Adjust dumping.
5493 (dbg_cnt_set_limit_by_index): Add new argument for high
5495 (dbg_cnt_set_limit_by_name): Likewise.
5496 (dbg_cnt_process_single_pair): Parse new format.
5497 (dbg_cnt_process_opt): Use strtok.
5498 (dbg_cnt_list_all_counters): Remove 'value' and add
5500 * doc/invoke.texi: Document changes.
5502 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5504 * doc/sourcebuild.texi (scalar_all_fma): Document.
5505 * tree.def (FMA_EXPR): Delete.
5506 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5507 * internal-fn.c (ternary_direct): New macro.
5508 (expand_ternary_optab_fn): Likewise.
5509 (direct_ternary_optab_supported_p): Likewise.
5510 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5511 * builtins.c (fold_builtin_fma): Delete.
5512 (fold_builtin_3): Don't call it.
5513 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5514 * expr.c (expand_expr_real_2): Likewise.
5515 * fold-const.c (operand_equal_p): Likewise.
5516 (fold_ternary_loc): Likewise.
5517 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5518 * gimple.c (DEFTREECODE): Likewise.
5519 * gimplify.c (gimplify_expr): Likewise.
5520 * optabs-tree.c (optab_for_tree_code): Likewise.
5521 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5522 * tree-eh.c (operation_could_trap_p): Likewise.
5523 (stmt_could_throw_1_p): Likewise.
5524 * tree-inline.c (estimate_operator_cost): Likewise.
5525 * tree-pretty-print.c (dump_generic_node): Likewise.
5526 (op_code_prio): Likewise.
5527 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5528 * tree-ssa-operands.c (get_expr_operands): Likewise.
5529 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5530 * fold-const-call.h (fold_fma): Delete.
5531 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5532 CFN_FNMA and CFN_FNMS.
5534 * genmatch.c (combined_fn): New enum.
5535 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5536 (commutative_op): New function.
5537 (commutate): Use it. Handle more than 2 operands.
5538 (dt_operand::gen_gimple_expr): Use commutative_op.
5539 (parser::parse_expr): Allow :c to be used with non-binary
5540 operators if the commutative operand is known.
5541 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5542 CFN_FMS, CFN_FNMA and CFN_FNMS.
5543 (backprop::process_assign_use): Remove FMA_EXPR handling.
5544 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5545 (gen_hsa_fma): New function.
5546 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5547 IFN_FNMA and IFN_FNMS.
5548 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5549 * gimple-fold.h (follow_all_ssa_edges): Declare.
5550 * gimple-fold.c (follow_all_ssa_edges): New function.
5551 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5552 gimple_build interface and use follow_all_ssa_edges to fold the result.
5553 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5554 instead of checking for optabs directly.
5555 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5556 rather than FMA_EXPRs.
5557 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5558 call to IFN_FMA instead of an FMA_EXPR.
5560 2018-05-17 Jim Wilson <jimw@sifive.com>
5562 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5563 sign extended promoted subreg, and the range does not have the sign bit
5564 set, then do a sign extend.
5566 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5567 test, check for sign extended subreg and/or constant operands, and
5568 do a sign extend in that case.
5570 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5572 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5573 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5575 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5576 Change logics_shift_reg to logics_shift_imm.
5577 (thunderx2t99_fp_loadpair_basic): Delete.
5578 (thunderx2t99_fp_storepair_basic): Delete.
5579 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5580 (thunderx2t99_asimd_polynomial): Delete.
5581 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5582 and neon_fp_mul_d_scalar_q.
5583 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5584 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5585 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5586 (thunderx2t99_asimd_lut): Add missing tbl types.
5587 (thunderx2t99_asimd_ext): Delete.
5588 (thunderx2t99_asimd_load1_1_mult): Delete.
5589 (thunderx2t99_asimd_load1_2_mult): Delete.
5590 (thunderx2t99_asimd_load1_ldp): New.
5591 (thunderx2t99_asimd_load1): New.
5592 (thunderx2t99_asimd_load2): Add missing *load2* types.
5593 (thunderx2t99_asimd_load3): New.
5594 (thunderx2t99_asimd_load4): New.
5595 (thunderx2t99_asimd_store1_1_mult): Delete.
5596 (thunderx2t99_asimd_store1_2_mult): Delete.
5597 (thunderx2t99_asimd_store2_mult): Delete.
5598 (thunderx2t99_asimd_store2_onelane): Delete.
5599 (thunderx2t99_asimd_store_stp): New.
5600 (thunderx2t99_asimd_store1): New.
5601 (thunderx2t99_asimd_store2): New.
5602 (thunderx2t99_asimd_store3): New.
5603 (thunderx2t99_asimd_store4): New.
5605 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5607 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5608 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5610 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5611 Segher Boessenkool <segher@kernel.crashing.org>
5614 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5617 2018-05-17 Richard Biener <rguenther@suse.de>
5619 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5620 for pruning loop and prune defs feeding only already visited PHIs.
5622 2018-05-17 Richard Biener <rguenther@suse.de>
5624 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5626 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5627 Richard Biener <rguenther@suse.de>
5629 PR tree-optimization/85793
5630 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5631 for VMAT_ELEMENTWISE.
5633 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5635 * internal-fn.h (lookup_internal_fn): Declare
5636 * internal-fn.c (lookup_internal_fn): New function.
5637 * gimple.c (gimple_build_call_from_tree): Handle calls to
5639 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5640 internal function names.
5641 * tree-pretty-print.c (dump_generic_node): Likewise.
5642 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5644 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5646 * gimple-fold.h (gimple_build): Make the function forms take
5647 combined_fn rather than built_in_function.
5648 (gimple_simplify): Likewise.
5649 * gimple-match-head.c (gimple_simplify): Likewise.
5650 * gimple-fold.c (gimple_build): Likewise.
5651 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5652 rather than gimple_build_call_internal.
5653 (get_initial_defs_for_reduction): Likewise.
5654 (vect_create_epilog_for_reduction): Likewise.
5655 (vectorizable_live_operation): Likewise.
5657 2018-05-17 Martin Liska <mliska@suse.cz>
5659 * gimple-ssa-sprintf.c (format_directive): Do not use
5660 space in between 'G_' and '('.
5662 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5665 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5666 even if the mask is not all ones.
5669 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5671 (ix86_gimple_fold_builtin): Likewise.
5674 * config/i386/i386.c: Include tree-vector-builder.h.
5675 (ix86_vector_shift_count): New function.
5676 (ix86_fold_builtin): Fold shift builtins by scalar count.
5677 (ix86_gimple_fold_builtin): Likewise.
5679 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5680 _mm512_setzero): New intrinsics.
5682 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5683 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5685 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5686 code generation for cases where splatting a value is not useful.
5687 * simplify-rtx.c (simplify_ternary_operation): Simplify
5688 vec_merge across a vec_duplicate and a paradoxical subreg forming
5689 a vector mode to a vec_concat.
5691 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5693 * config.gcc: Support "goldmont-plus".
5694 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5696 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5697 PROCESSOR_GOLDMONT_PLUS.
5698 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5699 (processor_target_table): Add "goldmont-plus".
5700 (PTA_GOLDMONT_PLUS): Define.
5701 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5702 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5703 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5704 (fold_builtin_cpu): Add "goldmont-plus".
5705 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5706 (ix86_option_override_internal): Add "goldmont-plus".
5707 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5708 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5709 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5710 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5712 2018-05-17 Richard Biener <rguenther@suse.de>
5714 PR tree-optimization/85757
5715 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5716 remove defs that only feed that PHI from further processing.
5718 2018-05-16 Jim Wilson <jimw@sifive.com>
5720 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5722 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5724 2018-05-16 Mark Wielaard <mark@klomp.org>
5726 * dwarf2out.c (count_index_strings): New function.
5727 (output_indirect_strings): Call count_index_strings and generate
5728 header for dwarf_version >= 5.
5730 2018-05-16 Mark Wielaard <mark@klomp.org>
5732 * dwarf2out.c (dwarf_FORM): New function.
5733 (set_indirect_string): Use dwarf_FORM.
5734 (reset_indirect_string): Likewise.
5735 (size_of_die): Likewise.
5736 (value_format): Likewise.
5737 (output_die): Likewise.
5738 (add_skeleton_AT_string): Likewise.
5739 (output_macinfo_op): Likewise.
5740 (index_string): Likewise.
5741 (output_index_string_offset): Likewise.
5742 (output_index_string): Likewise.
5743 (count_index_strings): Likewise.
5745 2018-05-16 Carl Love <cel@us.ibm.com>
5747 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5748 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5750 2018-05-16 Martin Jambor <mjambor@suse.cz>
5752 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5753 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5755 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5757 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5758 (fnma<mode>4): Likewise.
5759 (fms<mode>4): Likewise.
5760 (fnms<mode>4): Likewise.
5761 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5762 (aarch64_fnma<mode>4): Likewise.
5763 (aarch64_fms<mode>4): Likewise.
5764 (aarch64_fnms<mode>4): Likewise.
5765 (aarch64_fnmadd<mode>4): Likewise.
5767 2018-05-16 Jason Merrill <jason@redhat.com>
5769 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5771 2018-05-16 Richard Biener <rguenther@suse.de>
5773 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5774 (dump_stmt_cost): Declare.
5775 (add_stmt_cost): Dump cost we add.
5776 (add_stmt_costs): New function.
5777 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5779 (vect_analyze_stmt): Adjust prototype.
5780 (vectorizable_condition): Likewise.
5781 (vectorizable_live_operation): Likewise.
5782 (vectorizable_reduction): Likewise.
5783 (vectorizable_induction): Likewise.
5784 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5785 cost vector to pass to vectorizable_ and record afterwards.
5786 (vect_model_reduction_cost): Take cost vector argument and adjust.
5787 (vect_model_induction_cost): Likewise.
5788 (vectorizable_reduction): Likewise.
5789 (vectorizable_induction): Likewise.
5790 (vectorizable_live_operation): Likewise.
5791 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5792 SLP_TREE_NUMBER_OF_VEC_STMTS.
5793 (vect_analyze_slp_cost_1): Remove.
5794 (vect_analyze_slp_cost): Likewise.
5795 (vect_slp_analyze_node_operations): Take visited args and
5796 a target cost vector. Avoid processing already visited stmt sets.
5797 (vect_slp_analyze_operations): Use a local cost vector to gather
5798 costs and register those of non-discarded instances.
5799 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5800 (vect_schedule_slp_instance): Remove copying of
5801 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5803 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5804 adding cost. Record cost entry location.
5805 (vect_prologue_cost_for_slp_op): Function to compute cost of
5806 a constant or invariant generated for SLP vect in the prologue,
5807 split out from vect_analyze_slp_cost_1.
5808 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5809 (vect_model_promotion_demotion_cost): Likewise.
5810 (vect_model_store_cost): Likewise, make static.
5811 (vect_model_load_cost): Likewise.
5812 (vectorizable_bswap): Add cost vector arg and adjust.
5813 (vectorizable_call): Likewise.
5814 (vectorizable_simd_clone_call): Likewise.
5815 (vectorizable_conversion): Likewise.
5816 (vectorizable_assignment): Likewise.
5817 (vectorizable_shift): Likewise.
5818 (vectorizable_operation): Likewise.
5819 (vectorizable_store): Likewise.
5820 (vectorizable_load): Likewise.
5821 (vectorizable_condition): Likewise.
5822 (vectorizable_comparison): Likewise.
5823 (can_vectorize_live_stmts): Likewise.
5824 (vect_analyze_stmt): Likewise.
5825 (vect_transform_stmt): Adjust calls to vectorizable_*.
5826 * tree-vectorizer.c: Include gimple-pretty-print.h.
5827 (dump_stmt_cost): New function.
5829 2018-05-16 Richard Biener <rguenther@suse.de>
5831 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5832 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5833 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5834 (check_name): New callback.
5835 (dse_classify_store): Track cycles via a visited bitmap of PHI
5836 defs and simplify handling of in-loop and across loop dead stores
5837 and properly fail for loop-variant refs. Handle byte-tracking with
5838 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5841 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5843 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5844 (vect_get_mask_type_for_stmt): Likewise.
5845 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5847 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5848 to determine the statement's vector type and the vector type that
5849 should be used for calculating nunits. Deal with cases in which
5850 the type has to be deferred.
5851 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5852 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5853 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5854 (vect_determine_vf_for_stmt): New functions, split out from...
5855 (vect_determine_vectorization_factor): ...here.
5856 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5857 (vect_get_mask_type_for_stmt): New functions, split out from
5858 vect_determine_vectorization_factor.
5860 2018-05-16 Richard Biener <rguenther@suse.de>
5862 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5863 verify the [VEC_]COND_EXPR embedded comparison.
5865 2018-05-15 Martin Sebor <msebor@redhat.com>
5867 PR tree-optimization/85753
5868 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5869 RECORD_TYPE in addition to ARRAY_TYPE.
5871 2018-05-15 Martin Sebor <msebor@redhat.com>
5874 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5876 2018-05-15 Richard Biener <rguenther@suse.de>
5878 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5879 add by_clobber_p one. Change algorithm to collect all defs
5880 representing uses we need to walk and try reducing them to
5881 a single one before failing.
5882 (dse_dom_walker::dse_optimize_stmt): Adjust.
5884 2018-05-13 Mark Wielaard <mark@klomp.org>
5886 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5887 (size_of_loc_descr): Likewise.
5888 (output_loc_operands): Likewise.
5889 (output_loc_operands_raw): Likewise.
5890 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5891 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5892 (hash_loc_operands): Likewise.
5893 (compare_loc_operands): Likewise.
5895 2018-05-14 Mark Wielaard <mark@klomp.org>
5897 * dwarf2out.c (count_index_addrs): New function.
5898 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5900 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5902 PR tree-optimization/83648
5903 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5904 return value as malloc candidate.
5906 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5909 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5910 param as true in call to suggest_attribute.
5912 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5914 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5915 -mreadonly-in-sdata.
5917 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5919 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5921 (aarch64_crypto_aesd_fused): Likewise.
5923 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5925 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5926 (movsi_aarch64): Likewise.
5927 (load_pairsi): Likewise.
5928 (load_pairdi): Likewise.
5929 (store_pairsi): Likewise.
5930 (store_pairdi): Likewise.
5931 (load_pairsf): Likewise.
5932 (load_pairdf): Likewise.
5933 (store_pairsf): Likewise.
5934 (store_pairdf): Likewise.
5935 (zero_extend): Likewise.
5936 (trunc): Swap alternatives.
5937 (fcvt_target): Add '?' to prefer w over r.
5939 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5942 * config/i386/i386.md: Disallow non-commutative arithmetics in
5943 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5944 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5945 in the peephole2 before it.
5947 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5949 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5950 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5951 (ix86_handle_option): Handle -mcldemote.
5952 * config.gcc: New header.
5953 * config/i386/cldemoteintrin.h: New file.
5954 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5955 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5957 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5958 OPTION_MASK_ISA_CLDEMOTE.
5959 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5960 (ix86_valid_target_attribute_inner_p): Ditto.
5961 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5962 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5963 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5964 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5965 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5967 * config/i386/i386.opt: Add -mcldemote.
5968 * config/i386/x86intrin.h: New header.
5969 * doc/invoke.texi: Add -mcldemote.
5971 2018-05-14 Richard Biener <rguenther@suse.de>
5973 * doc/match-and-simplify.texi: Adjust :s documentation.
5975 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5977 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5978 intended memcpy size.
5979 (REORDER_45): Likewise.
5981 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5983 * sort.cc: New file.
5984 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5985 * vec.c (qsort_chk): Use gcc_qsort.
5986 * Makefile.in (OBJS-libcommon): Add sort.o.
5987 (build/sort.o): New target. Use it...
5988 (BUILD_RTL): ... here, and...
5989 (build/gencfn-macros): ... here, and...
5990 (build/genmatch): ... here.
5992 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5993 Chung-Ju Wu <jasonwucj@gmail.com>
5995 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5996 * config/nds32/nds32-graywolf.md: New file.
5997 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5998 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6000 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6001 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6002 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6003 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6004 * config/nds32/pipelines.md: Include n15 settings.
6006 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6007 Chung-Ju Wu <jasonwucj@gmail.com>
6009 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6010 * config/nds32/nds32-n13.md: New file.
6011 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6012 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6014 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6015 * config/nds32/nds32.md (pipeline_model): Add n13.
6016 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6017 * config/nds32/pipelines.md: Include n13 settings.
6019 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6020 Chung-Ju Wu <jasonwucj@gmail.com>
6022 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6023 * config/nds32/nds32-n10.md: New file.
6024 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6025 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6027 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6028 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6029 * config/nds32/nds32.md (pipeline_model): Add n10.
6030 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6031 * config/nds32/pipelines.md: Include n10 settings.
6033 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6034 Kito Cheng <kito.cheng@gmail.com>
6035 Chung-Ju Wu <jasonwucj@gmail.com>
6037 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6038 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6039 Add enum values for DSP extension instructions.
6040 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6042 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6043 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6045 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6046 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6047 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6048 * config/nds32/nds32-intrinsic.md: Likewise.
6049 * config/nds32/nds32_intrinsic.h: Likewise.
6050 * config/nds32/nds32-md-auxiliary.c: Likewise.
6051 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6052 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6053 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6054 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6055 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6056 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6057 TYPE_DMAC in switch statement.
6058 * config/nds32/nds32.c: New checking and implementation for DSP
6059 extension instructions.
6060 * config/nds32/nds32.h: Likewise.
6061 * config/nds32/nds32.md: Likewise.
6062 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6063 * config/nds32/predicates.md: Implement new predicates for DSP
6066 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6068 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6069 Reformat alternatives and attributes so it is easier to identify
6070 which constraints/attributes go with which instruction.
6071 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6072 (mov<mode>_softfloat32, FMOVE64): Likewise.
6073 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6074 (mov<mode>_softfloat64, FMOVE64): Likewise.
6076 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6078 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6080 (Basic PowerPC Built-in Functions): The new name of the
6081 subsection previously known as "PowerPC Built-in Functions".
6082 (Basic PowerPC Built-in Functions Available on all Configurations):
6084 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6085 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6086 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6087 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6089 2018-05-11 Martin Jambor <mjambor@suse.cz>
6092 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6095 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6098 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6100 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6102 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6103 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6104 (ix86_handle_option): Handle -mwaitpkg.
6105 * config.gcc: New header.
6106 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6107 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6108 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6110 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6111 OPTION_MASK_ISA_WAITPKG.
6112 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6113 (ix86_option_override_internal): Add PTA_WAITPKG.
6114 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6115 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6116 IX86_BUILTIN_TPAUSE.
6117 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6118 __builtin_ia32_umwait and __builtin_ia32_tpause.
6119 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6120 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6121 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6122 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6123 UNSPECV_TPAUSE): New.
6124 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6125 * config/i386/i386.opt: Add -mwaitpkg.
6126 * config/i386/waitpkgintrin.h: New file.
6127 * config/i386/x86intrin.h: New header.
6128 * doc/invoke.texi: Add -mwaitpkg.
6130 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6133 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6135 (cortex-m0): Use armv6s-m isa.
6136 (cortex-m0plus): Likewise.
6137 (cortex-m1): Likewise.
6138 (cortex-m0.small-multiply): Likewise.
6139 (cortex-m0plus.small-multiply): Likewise.
6140 (cortex-m1.small-multiply): Likewise.
6142 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6143 Jakub Jelinek <jakub@redhat.com>
6145 PR tree-optimization/85692
6146 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6147 source permute as well.
6149 2018-05-11 Martin Liska <mliska@suse.cz>
6152 * doc/extend.texi: Document LLVM style format for no_sanitize
6155 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6157 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6158 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6159 (mode_supports_vsx_dform_quad): Likewise.
6160 (mode_supports_vmx_dform): Move these functions to be next to the
6161 other mode_supports functions.
6162 (mode_supports_dq_form): Likewise.
6163 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6164 mode_supports_dq_form.
6165 (reg_offset_addressing_ok_p): Likewise.
6166 (offsettable_ok_by_alignment): Likewise.
6167 (rs6000_legitimate_offset_address_p): Likewise.
6168 (legitimate_lo_sum_address_p): Likewise.
6169 (rs6000_legitimize_address): Likewise.
6170 (rs6000_legitimize_reload_address): Likewise.
6171 (rs6000_secondary_reload_inner): Likewise.
6172 (rs6000_preferred_reload_class): Likewise.
6173 (rs6000_output_move_128bit): Likewise.
6175 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6177 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6178 Generate SImode target register for null target.
6179 <case IX86_BUILTIN_XGETBV>: Ditto.
6180 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6181 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6183 2018-05-10 Carl Love <cel@us.ibm.com>
6185 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6186 dcbtt and dcbtstt if operands[2] is 0.
6188 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6191 * config/i386/sse.md (usadv64qi): New expander.
6193 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6195 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6196 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6197 -maltivec=be support.
6198 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6199 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6200 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6201 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6202 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6203 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6204 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6205 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6206 altivec_vsumsws): Adjust.
6207 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6208 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6209 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6211 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6212 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6213 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6214 (altivec_lve<VI_char>x): Delete expand.
6215 (*altivec_lve<VI_char>x_internal): Rename to...
6216 (altivec_lve<VI_char>x): ... this.
6217 (altivec_lvxl_<mode>): Delete expand.
6218 (*altivec_lvxl_<mode>_internal): Rename to ...
6219 (altivec_lvxl_<mode>): ... this.
6220 (altivec_stvxl_<mode>): Delete expand.
6221 (*altivec_stvxl_<mode>_internal): Rename to ...
6222 (altivec_stvxl_<mode>): ... this.
6223 (altivec_stve<VI_char>x): Delete expand.
6224 (*altivec_stve<VI_char>x_internal): Rename to ...
6225 (altivec_stve<VI_char>x): ... this.
6226 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6227 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6228 reduc_plus_scal_<mode>): Adjust.
6229 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6231 (rs6000_cpu_cpp_builtins): Adjust.
6232 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6233 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6234 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6235 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6236 -maltivec=be support.
6237 (rs6000_split_vec_extract_var): Adjust.
6238 (rs6000_split_v4si_init): Adjust.
6239 (swap_selector_for_mode): Delete.
6240 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6241 altivec_expand_stvex_be): Delete.
6242 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6243 -maltivec=be support.
6244 (rs6000_gimple_fold_builtin): Ditto.
6245 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6247 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6248 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6249 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6250 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6251 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6252 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6253 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6254 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6255 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6256 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6257 anonymous split): Adjust.
6258 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6259 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6261 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6263 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6264 when --with-gxx-include-dir is also specified.
6265 * configure: Regenerate.
6267 2018-05-09 Jim Wilson <jimw@sifive.com>
6270 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6271 * config/riscv/t-withmultilib: New.
6272 * config/riscv/withmultilib.h: New.
6273 * doc/install.texi: Document RISC-V --with-multilib-list support.
6275 2018-05-09 Richard Biener <rguenther@suse.de>
6277 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6279 (vect_bb_vectorization_profitable_p): Adjust. Compute
6280 actual scalar cost using the cost vector and the add_stmt_cost
6283 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6285 PR rtl-optimization/85645
6286 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6287 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6289 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6291 PR rtl-optimization/85645
6292 * shrink-wrap.c (spread_components): Return a boolean saying if
6293 anything was changed.
6294 (try_shrink_wrapping_separate): Iterate spread_components until
6295 nothing changes anymore.
6297 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6299 PR rtl-optimization/85645
6300 * regrename.c (build_def_use): Also kill the chains that include the
6301 destination of a REG_CFA_REGISTER note.
6303 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6305 PR rtl-optimization/85645
6306 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6307 insn that has a REG_CFA_REGISTER note.
6309 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6311 * cfgexpand.c (expand_clobber): New function.
6312 (expand_gimple_stmt_1): Use it.
6313 * tree-vect-stmts.c (vect_clobber_variable): New function,
6315 (vectorizable_simd_clone_call): ...here.
6316 (vectorizable_store): Emit a clobber either side of an
6317 IFN_STORE_LANES sequence.
6318 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6320 2018-05-09 Tom de Vries <tom@codesourcery.com>
6323 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6324 (define_insn "trap_if_false"): Add exit after trap.
6326 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6328 PR rtl-optimization/85638
6329 * bb-reorder.c: Include common/common-target.h.
6330 (create_forwarder_block): New function extracted from...
6331 (fix_up_crossing_landing_pad): ...here. Rename into...
6332 (dw2_fix_up_crossing_landing_pad): ...this.
6333 (sjlj_fix_up_crossing_landing_pad): New function.
6334 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6335 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6336 from both partitions and exit the loop after one iteration.
6338 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6341 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6343 (Basic PowerPC Built-in Functions): The new name of the
6344 subsection previously known as "PowerPC Built-in Functions".
6345 (Basic PowerPC Built-in Functions Available on all Configurations):
6347 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6349 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6351 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6353 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6356 2018-05-08 Jim Wilson <jimw@sifive.com>
6358 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6359 (LD_EMUL_SUFFIX): New.
6360 (LINK_SPEC): Use it.
6362 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6364 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6366 (Basic PowerPC Built-in Functions): The new name of the
6367 subsection previously known as "PowerPC Built-in Functions".
6368 (Basic PowerPC Built-in Functions Available on all Configurations):
6370 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6372 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6374 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6376 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6379 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6382 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6383 after cmpelim optimization.
6385 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6387 * config.gcc: Support "goldmont".
6388 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6389 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6391 * config/i386/i386.c (m_GOLDMONT): Define.
6392 (processor_target_table): Add "goldmont".
6393 (PTA_GOLDMONT): Define.
6394 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6395 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6396 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6397 (fold_builtin_cpu): Add "goldmont".
6398 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6399 (ix86_option_override_internal): Add "goldmont".
6400 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6401 (processor_type): Add PROCESSOR_GOLDMONT.
6402 * config/i386/i386.md: Add CPU "glm".
6403 * config/i386/glm.md: New file.
6404 * config/i386/x86-tune.def: Add m_GOLDMONT.
6405 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6407 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6410 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6412 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6413 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6414 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6417 * config/i386/i386.c (ix86_fold_builtin): Handle
6418 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6421 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6422 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6424 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6427 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6428 (check_arch): Likewise.
6429 (check_fpu): Return the result rather than printing it.
6430 (end arch): Fix operator precedence.
6431 (end cpu): Likewise.
6432 (END): Print the result from check_fpu.
6434 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6435 Alan Hayward <alan.hayward@arm.com>
6436 David Sherwood <david.sherwood@arm.com>
6438 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6439 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6440 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6441 (*fcmuo<mode>_and): New patterns.
6443 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6445 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6446 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6447 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6448 (cmp_op, sve_imm_con): New code attributes.
6449 (SVE_COND_INT_CMP, imm_con): Delete.
6450 (cmp_op): Remove above unspecs from int attribute.
6451 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6453 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6454 comparison-specific unspecs.
6455 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6456 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6457 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6458 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6459 (*vec_fcm<cmp_op><mode>): Rename to...
6460 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6461 (*vec_fcmuo<mode>): Rename to...
6462 (*fcmuo<mode>): ...this and adjust likewise.
6463 (*pred_fcm<cmp_op><mode>): New pattern.
6464 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6465 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6467 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6469 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6470 (aarch64_emit_sve_predicated_cond): New function.
6471 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6472 (aarch64_emit_unspec_cond_or): Replace with...
6473 (aarch64_emit_sve_or_conds): ...this new function. Use
6474 aarch64_emit_sve_ptrue_op for the individual comparisons and
6475 aarch64_emit_binop to OR them together.
6476 (aarch64_emit_inverted_unspec_cond): Replace with...
6477 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6478 aarch64_emit_sve_ptrue_op for the comparison and
6479 aarch64_emit_unop to invert the result.
6480 (aarch64_expand_sve_vec_cmp_float): Update after the above
6481 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6483 2018-05-07 Nathan Sidwell <nathan@acm.org>
6485 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6486 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6487 (Backwards Compatibility): Likewise.
6489 2018-05-07 Luis Machado <luis.machado@linaro.org>
6493 2018-05-07 Luis Machado <luis.machado@linaro.org>
6495 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6496 <prefetch_dynamic_strides>: New const bool field.
6497 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6498 prefetch_dynamic_strides.
6499 (exynosm1_prefetch_tune): Likewise.
6500 (thunderxt88_prefetch_tune): Likewise.
6501 (thunderx_prefetch_tune): Likewise.
6502 (thunderx2t99_prefetch_tune): Likewise.
6503 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6505 (aarch64_override_options_internal): Update to set
6506 PARAM_PREFETCH_DYNAMIC_STRIDES.
6507 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6508 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6509 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6510 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6511 prefetch-dynamic-strides setting.
6513 2018-05-07 Luis Machado <luis.machado@linaro.org>
6515 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6516 <minimum_stride>: New const int field.
6517 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6518 minimum_stride field.
6519 (exynosm1_prefetch_tune): Likewise.
6520 (thunderxt88_prefetch_tune): Likewise.
6521 (thunderx_prefetch_tune): Likewise.
6522 (thunderx2t99_prefetch_tune): Likewise.
6523 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6524 (aarch64_override_options_internal): Update to set
6525 PARAM_PREFETCH_MINIMUM_STRIDE.
6526 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6527 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6528 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6529 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6530 stride is constant and is below the minimum stride threshold.
6532 2018-05-07 Luis Machado <luis.machado@linaro.org>
6534 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6537 2018-05-07 Luis Machado <luis.machado@linaro.org>
6539 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6540 <prefetch_dynamic_strides>: New const bool field.
6541 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6542 prefetch_dynamic_strides.
6543 (exynosm1_prefetch_tune): Likewise.
6544 (thunderxt88_prefetch_tune): Likewise.
6545 (thunderx_prefetch_tune): Likewise.
6546 (thunderx2t99_prefetch_tune): Likewise.
6547 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6549 (aarch64_override_options_internal): Update to set
6550 PARAM_PREFETCH_DYNAMIC_STRIDES.
6551 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6552 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6553 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6554 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6555 prefetch-dynamic-strides setting.
6557 2018-05-07 Luis Machado <luis.machado@linaro.org>
6559 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6560 <minimum_stride>: New const int field.
6561 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6562 minimum_stride field.
6563 (exynosm1_prefetch_tune): Likewise.
6564 (thunderxt88_prefetch_tune): Likewise.
6565 (thunderx_prefetch_tune): Likewise.
6566 (thunderx2t99_prefetch_tune): Likewise.
6567 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6568 (aarch64_override_options_internal): Update to set
6569 PARAM_PREFETCH_MINIMUM_STRIDE.
6570 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6571 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6572 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6573 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6574 stride is constant and is below the minimum stride threshold.
6576 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6579 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6580 the type is addressable. Don't force op into register if it has
6583 2018-05-05 Roland McGrath <mcgrathr@google.com>
6586 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6587 any section for which we don't know a specific type it should have,
6588 regardless of name. Previously this was done only for the exact
6589 names ".init_array", ".fini_array", and ".preinit_array".
6590 (default_elf_asm_named_section): Add comment about
6591 relationship with default_section_type_flags and SECTION_NOTYPE.
6592 (get_section): Don't consider it a type conflict if one side has
6593 SECTION_NOTYPE and the other doesn't, as long as neither has the
6594 SECTION_BSS et al used in the default_section_type_flags logic.
6596 2018-05-05 Tom de Vries <tom@codesourcery.com>
6599 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6600 (workaround_barsyncs): New function.
6601 (nvptx_reorg): Use workaround_barsyncs.
6602 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6603 (define_expand "nvptx_membar_cta"): New define_expand.
6604 (define_insn "*nvptx_membar_cta"): New insn.
6606 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6608 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6609 To improve optimization opportunities.
6610 * builtin-types.def: The new needed builtin types for the above.
6612 2018-05-04 Richard Biener <rguenther@suse.de>
6614 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6615 * gimple-ssa-store-merging.c
6616 (imm_store_chain_info::output_merged_store): Remove redundant create,
6617 release split_store vector contents on failure.
6618 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6619 scalar stmt vector on cache hit.
6621 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6623 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6625 * config.gcc (powerpc-xilinx-eabi*): Remove.
6626 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6628 (fusion_addis_mem_combo_load): Ditto.
6629 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6631 (rs6000_cpu_cpp_builtins): Ditto.
6632 * config/rs6000/rs6000-linux.c
6633 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6634 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6635 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6637 (rs6000_setup_reg_addr_masks): Ditto.
6638 (rs6000_init_hard_regno_mode_ok): Ditto.
6639 (rs6000_option_override_internal): Ditto.
6640 (legitimate_lo_sum_address_p): Ditto.
6641 (rs6000_legitimize_address): Ditto.
6642 (rs6000_legitimize_reload_address): Ditto.
6643 (rs6000_legitimate_address_p): Ditto.
6644 (abi_v4_pass_in_fpr): Ditto.
6645 (setup_incoming_varargs): Ditto.
6646 (rs6000_gimplify_va_arg): Ditto.
6647 (rs6000_split_multireg_move): Ditto.
6648 (rs6000_savres_strategy): Ditto.
6649 (rs6000_emit_prologue_components): Ditto.
6650 (rs6000_emit_epilogue_components): Ditto.
6651 (rs6000_emit_prologue): Ditto.
6652 (rs6000_emit_epilogue): Ditto.
6653 (rs6000_elf_file_end): Ditto.
6654 (rs6000_function_value): Ditto.
6655 (rs6000_libcall_value): Ditto.
6656 * config/rs6000/rs6000.h: Ditto.
6657 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6658 (TARGET_MINMAX): ... this. New.
6659 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6660 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6661 (*movsi_internal1_single): Delete.
6662 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6663 mfpu=, mxilinx-fpu): Delete.
6664 * config/rs6000/singlefp.h: Delete.
6665 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6666 * config/rs6000/t-rs6000: Ditto.
6667 * config/rs6000/t-xilinx: Delete.
6668 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6669 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6670 (VStype_simple): Delete.
6671 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6672 * config/rs6000/xfpu.h: Delete.
6673 * config/rs6000/xfpu.md: Delete.
6674 * config/rs6000/xilinx.h: Delete.
6675 * config/rs6000/xilinx.opt: Delete.
6676 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6677 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6679 2018-05-04 Tom de Vries <tom@codesourcery.com>
6682 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6685 2018-05-04 Richard Biener <rguenther@suse.de>
6688 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6689 (expand_complex_div_wide): Likewise.
6690 (expand_complex_operations_1): Likewise.
6691 (expand_complex_libcall): Preserve EH info of the original stmt.
6692 (tree_lower_complex): Handle removed blocks.
6693 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6694 on complex multiplication and division libcall builtins.
6696 2018-05-04 Richard Biener <rguenther@suse.de>
6699 * fold-const.c (negate_expr_p): Restrict negation of operand
6700 zero of a division to when we know that can happen without
6702 (fold_negate_expr_1): Likewise.
6704 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6707 * real.h (real_nextafter): Declare.
6708 * real.c (real_nextafter): New function.
6709 * fold-const-call.c (fold_const_nextafter): New function.
6710 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6711 CASE_CFN_NEXTTOWARD.
6712 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6713 even when arg1_mode is different from arg0_mode.
6715 2018-05-03 Nathan Sidwell <nathan@acm.org>
6717 * doc/extend.texi (Deprecated Features): Remove
6719 (Backwards Compatibility): Likewise.
6720 * doc/invoke.texi (C++ Language Options): Likewise.
6721 (C++ Dialect Options): Likewise.
6723 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6726 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6727 _mm512_mask_mullox_epi64): New intrinsics.
6729 2018-05-03 Tom de Vries <tom@codesourcery.com>
6732 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6733 dump files): Add offload-tree.
6735 2018-05-03 Richard Biener <rguenther@suse.de>
6737 PR tree-optimization/85615
6738 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6739 to loops not nested in BBs loop father to avoid creating multi-entry
6742 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6744 PR tree-optimization/70291
6745 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6746 arguments. Change return type to tree. Emit libcall as a new
6747 statement rather than replacing existing one when inplace_p is true.
6748 (expand_complex_multiplication_components): New function.
6749 (expand_complex_multiplication): Expand floating-point complex
6750 multiplication using the above.
6751 (expand_complex_division): Rename inner_type parameter to type.
6752 Update expand_complex_libcall call-site.
6753 (expand_complex_operations_1): Update expand_complex_multiplication
6754 and expand_complex_division call-sites.
6756 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6759 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6760 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6761 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6762 the highest significant bit of the shift count mask is clear. In
6763 check whether and[sq]i3 is needed verify that all significant bits
6764 of the shift count other than the highest are set.
6766 2018-05-02 Tom de Vries <tom@codesourcery.com>
6769 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6770 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6771 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6772 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6773 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6774 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6775 __builtin_goacc_parlevel_size.
6777 2018-05-02 Richard Biener <rguenther@suse.de>
6779 PR tree-optimization/85597
6780 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6781 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6784 2018-05-02 Tom de Vries <tom@codesourcery.com>
6787 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6788 dump files): Add ltrans-tree.
6790 2018-05-02 Tom de Vries <tom@codesourcery.com>
6793 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6794 dump files): Add wpa-ipa.
6796 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6798 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6799 powerpc*-*-linux*paired* target.
6800 * config/rs6000/750cl.h: Delete.
6801 * config/rs6000/paired.h: Delete.
6802 * config/rs6000/paired.md: Delete.
6803 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6805 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6806 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6807 comment. Remove paired float support.
6808 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6809 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6811 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6812 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6814 * config/rs6000/rs6000.c: Remove paired float support.
6815 (paired_expand_vector_init, paired_expand_vector_move,
6816 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6817 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6818 paired_expand_builtin, paired_expand_predicate_builtin,
6819 paired_init_builtins): Delete.
6820 * config/rs6000/rs6000.h: Remove paired float support.
6821 * config/rs6000/rs6000.md: Remove paired float support.
6822 (move_from_CR_ov_bit): Delete.
6823 * config/rs6000/rs6000.opt (mpaired): Delete.
6824 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6825 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6827 2018-05-02 Richard Biener <rguenther@suse.de>
6830 * gimplify.c (gimplify_save_expr): When in SSA form allow
6831 SAVE_EXPRs to compute to SSA vars.
6833 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6836 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6837 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6838 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6839 clobber operands[2], instead use a new pseudo. Formatting fixes.
6841 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6843 PR tree-optimization/85586
6844 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6845 exit early for statements in the same group if the accesses are
6848 2018-05-02 Tom de Vries <tom@codesourcery.com>
6851 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6854 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6856 PR tree-optimization/85143
6857 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6859 2018-05-01 Tom de Vries <tom@codesourcery.com>
6862 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6863 not found" error message.
6865 2018-05-01 Tom de Vries <tom@codesourcery.com>
6868 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6869 * vec.c (test_ordered_remove_if): New function.
6870 (vec_c_tests): Call test_ordered_remove_if.
6871 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6872 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6873 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6874 VEC_ORDERED_REMOVE_IF.
6876 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6878 PR tree-optimization/82665
6879 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6880 pointer subtraction where arguments come from a memchr call.
6882 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6884 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6885 --push-state --as-needed and --pop-state instead of --as-needed and
6886 --no-as-needed if ld supports it.
6887 * configure: Regenerated.
6890 * doc/install.texi2html: Replace _002d with - and _002a with * in
6891 generated html files using sed.
6893 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6896 * gcc-rich-location.c (blank_line_before_p): New function.
6897 (use_new_line): New function.
6898 (gcc_rich_location::add_fixit_insert_formatted): New function.
6899 * gcc-rich-location.h
6900 (gcc_rich_location::add_fixit_insert_formatted): New function.
6902 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6904 * selftest.c (assert_streq): Rename "expected" and "actual" to
6905 "val1" and "val2". Extend NULL-handling to cover both inputs
6906 symmetrically, while still requiring both to be non-NULL for a pass.
6907 * selftest.h (assert_streq): Rename "expected" and "actual" to
6909 (ASSERT_EQ): Likewise.
6910 (ASSERT_EQ_AT): Likewise.
6911 (ASSERT_KNOWN_EQ): Likewise.
6912 (ASSERT_KNOWN_EQ_AT): Likewise.
6913 (ASSERT_NE): Likewise.
6914 (ASSERT_MAYBE_NE): Likewise.
6915 (ASSERT_MAYBE_NE_AT): Likewise.
6916 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6917 the assertion to pass.
6918 (ASSERT_STREQ_AT): Likewise.
6920 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6922 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6923 interaction with -pie.
6925 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6927 * selftest.h: Fix alphabetization of per-source-file selftest
6930 2018-04-30 Jason Merrill <jason@redhat.com>
6932 PR c++/61982 - dead stores to destroyed objects.
6933 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6936 2018-04-30 Jason Merrill <jason@redhat.com>
6938 * tree.c (build_clobber): New.
6939 * tree.h: Declare it.
6940 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6942 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6944 * diagnostic-show-locus.c (layout::layout): Update for
6945 location_get_source_line returning a char_span.
6946 (struct char_span): Move to input.h.
6947 (struct correction): Update for fields in char_span becoming
6949 (struct source_line): Update for location_get_source_line
6950 returning a char_span.
6951 (layout::print_line): Likewise.
6952 * edit-context.c (edited_file::print_content): Likewise.
6953 (edited_file::print_diff_hunk): Likewise.
6954 (edited_file::print_run_of_changed_lines): Likewise.
6955 (edited_file::get_num_lines): Likewise.
6956 (edited_line::edited_line): Likewise.
6957 * final.c (asm_show_source): Likewise.
6958 * input.c (location_get_source_line): Convert return type
6959 from const char * to char_span, losing the final "line_len"
6961 (dump_location_info): Update for the above.
6962 (get_substring_ranges_for_loc): Likewise. Use a char_span
6963 when handling the literal within the line.
6964 (test_reading_source_line): Update for location_get_source_line
6965 returning a char_span.
6966 * input.h (class char_span): Move here from
6967 diagnostic-show-locus.c, converting from a struct to a class.
6968 Make data members private.
6969 (char_span::operator bool): New.
6970 (char_span::length): New.
6971 (char_span::get_buffer): New.
6972 (char_span::operator[]): New.
6973 (char_span::subspan): Make const.
6974 (char_span::xstrdup): New.
6975 (location_get_source_line): Convert return type from const char *
6976 to char_span, losing the final "line_size" param.
6978 2018-04-30 Jan Hubicka <jh@suse.cz>
6980 * lto-wrapper.c (ltrans_priorities): New static var.
6981 (cmp_priority): New.
6982 (run_gcc): Read priorities and if doing parallel build order
6983 the Makefile by them.
6985 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6987 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6989 2018-04-30 Richard Biener <rguenther@suse.de>
6991 * tree-cfg.c (verify_address): Remove base argument, add
6992 flag whether to check TREE_ADDRESSABLE and do that.
6993 (verify_expr): Remove.
6994 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6995 (verify_gimple_assign_single): Likewise.
6996 (verify_gimple_switch): Likewise.
6997 (verify_expr_location_1): Dereference tp once. Add (disabled)
6998 piece from verify_expr.
6999 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7001 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7003 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7005 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7007 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7008 (small_data_pattern): Likewise.
7009 (arc_rewrite_small_data): Likewise.
7010 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7011 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7012 (get_symbol_alignment): New function.
7013 (legitimate_small_data_address_p): Likewise.
7014 (legitimate_scaled_address): Update, call
7015 legitimate_small_data_address_p.
7016 (output_sdata): New static variable.
7017 (arc_print_operand): Update how we handle small data operands.
7018 (arc_print_operand_address): Likewise.
7019 (arc_legitimate_address_p): Update, use
7020 legitimate_small_data_address_p.
7021 (arc_rewrite_small_data_p): Remove.
7022 (arc_rewrite_small_data_1): Likewise.
7023 (arc_rewrite_small_data): Likewise.
7024 (small_data_pattern): Likewise.
7025 (compact_sda_memory_operand): Update to use
7026 legitimate_small_data_address_p and get_symbol_alignment.
7027 (prepare_move_operands): Don't rewite sdata pattern.
7028 (prepare_extend_operands): Remove.
7029 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7031 (zero_extendqisi2): Likewise.
7032 (zero_extendhisi2): Likewise.
7033 (extendqihi2): Likewise.
7034 (extendqisi2): Likewise.
7035 (extendhisi2): Likewise.
7039 * config/arc/constraints.md (Usd): Change it to memory constraint.
7041 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7043 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7044 as source of std instructions.
7045 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7046 allow 6-bit constants as source for store instructions.
7047 (movdi_insn): Update instruction pattern to allow 6-bit constants
7048 as source for store instructions.
7050 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7052 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7054 2018-04-30 Nathan Sidwell <nathan@acm.org>
7055 Sandra Loosemore <sandra@codesourcery.com>
7057 * dumpfile.c (dump_open): Allow '-' for stdout.
7058 * doc/invoke.texi (Developer Options): Document dump filename
7059 determination early. Document stdin/stdout selection.
7061 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7063 Microblaze Target: PIC data text relative
7065 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7066 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7068 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7069 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7070 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7071 New addressing mode for data-text relative position indepenedent code.
7072 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7073 'ADDRESS_SYMBOLIC_TXT_REL'.
7074 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7075 (microblaze_legitimate_pic_operand): Exclude function calls from
7076 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7077 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7079 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7080 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7081 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7082 for 'address + offset'.
7083 (microblaze_expand_prologue): Add new function prologue call for
7085 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7086 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7087 table in case of TARGET_PIC_DATA_TEXT_REL.
7088 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7089 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7090 Add new macros 'UNSPEC_TEXT',
7091 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7092 + exclude function calls from 'UNSPEC_PLT' in case of data text
7094 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7095 new target hook for generating address diff vector tables in case of
7097 * doc/tm.texi : Regenerate.
7098 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7099 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7100 of addr diff vector generation.
7101 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7102 target hook definition.
7103 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7104 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7105 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7106 option for data text relative.
7108 2018-04-30 Richard Biener <rguenther@suse.de>
7110 * tree-chrec.h (evolution_function_is_constant_p): Remove
7112 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7114 2018-04-30 Richard Biener <rguenther@suse.de>
7117 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7119 2018-04-30 Richard Biener <rguenther@suse.de>
7121 PR tree-optimization/28364
7122 PR tree-optimization/85275
7123 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7124 copying first exit test.
7126 2018-04-28 Mark Wielaard <mark@klomp.org>
7128 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7130 (dwarf_AT): Handle DW_AT_addr_base.
7131 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7133 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7136 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7137 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7138 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7139 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7141 2018-04-28 Richard Biener <rguenther@suse.de>
7143 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7144 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7145 to reflect use. Only add interesting stmts.
7147 2018-04-27 Martin Jambor <mjambor@suse.cz>
7150 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7151 the jump function allows for passing through aggregate values.
7153 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7155 * input.h (in_system_header_at): Convert from macro to inline
7157 (from_macro_expansion_at): Likewise.
7158 (from_macro_definition_at): Likewise.
7160 2018-04-27 Jeff Law <law@redhat.com>
7162 * config.gcc: Mark tile* targets as deprecated/obsolete.
7164 2018-04-27 Richard Biener <rguenther@suse.de>
7166 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7169 2018-04-27 Richard Biener <rguenther@suse.de>
7171 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7173 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7175 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7176 with Yd constraint. Set "preferred_for_speed" attribute from
7177 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7179 (*movdi_internal): Ditto.
7180 (movti_interunit splitters): Remove
7181 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7182 (movdi_interunit splitters): Ditto.
7183 * config/i386/constraints.md (Ye): Remove.
7184 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7186 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7189 * config/aarch64/constraints.md (Usg): Limit to 31.
7192 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7194 PR tree-optimization/85529
7195 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7196 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7197 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7198 zero extension or masking of the MSB bit.
7199 (optimize_range_tests): Add FIRST_BB argument, pass it through
7200 to optimize_range_tests_var_bound.
7201 (maybe_optimize_range_tests, reassociate_bb): Adjust
7202 optimize_range_tests callers.
7204 2018-04-26 Richard Biener <rguenther@suse.de>
7205 Jakub Jelinek <jakub@redhat.com>
7207 * cgraph.h (symbol_table): Just declare debug method here.
7208 * symtab.c (symbol_table::debug): Define.
7210 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7212 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7214 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7216 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7217 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7218 (*movdi_internal): Substitute Yi and Yj constraint with x
7219 and Ym and Yn constraint with y constraint. Update "isa"
7220 attribute and set "preferred_for_speed" attribute from
7221 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7222 (*movsi_internal): Ditto.
7223 (*movdf_internal): Ditto.
7224 (*movsf_internal): Ditto.
7225 (*zero_extendsidi2): Ditto.
7226 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7227 (sse2_loadld): Ditto.
7228 (*vec_extract<ssevecmodelower>_0): Ditto.
7229 (*vec_extractv4si_0_zext_sse4): Ditto.
7230 (vec_concatv2di): Ditto.
7231 (*vec_dup<mode>): Ditto.
7232 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7233 * config/i386/constraints.md (Yi): Remove.
7238 2018-04-26 Nathan Sidwell <nathan@acm.org>
7240 * dumpfile.c (dump_open): New.
7241 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7242 (dump_finish): Detect stdio/stderr by value not name.
7244 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7246 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7248 2018-04-26 Tom de Vries <tom@codesourcery.com>
7251 * config/nvptx/nvptx.c (verify_neutering_jumps)
7252 (verify_neutering_labels): New function
7253 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7255 2018-04-26 Tom de Vries <tom@codesourcery.com>
7258 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7259 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7262 2018-04-26 Richard Biener <rguenther@suse.de>
7263 Tom de Vries <tom@codesourcery.com>
7266 * lto-streamer-out.c (output_function): Fixup loops if required to match
7267 discovery done in the reader.
7269 2018-04-26 Richard Biener <rguenther@suse.de>
7271 PR tree-optimization/85116
7272 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7273 have a loop exit from the single latch predecessor. Remove
7274 case of header with just condition.
7275 (ch_base::copy_headers): Exclude infinite loops from any
7277 (pass_ch::execute): Record exits.
7279 2018-04-26 Richard Biener <rguenther@suse.de>
7281 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7282 prologue cost vector and pass it to vect_get_load_cost.
7283 (vect_get_peeling_costs_all_drs): Likewise.
7284 (vect_peeling_hash_get_lowest_cost): Likewise.
7285 (vect_enhance_data_refs_alignment): Likewise.
7287 2018-04-26 Richard Biener <rguenther@suse.de>
7290 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7291 checking of integer<->pointer conversions.
7292 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7293 sign-/zero-extending pointer types.
7294 (expand_omp_for_static_chunk): Likewise.
7296 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7297 Jean Lee <xiaoyur347@gmail.com>
7299 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7300 (TARGET_ASAN_SHADOW_OFFSET): Define.
7301 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7302 true for -fsanitize=address.
7304 2018-04-25 Mark Wielaard <mark@klomp.org>
7306 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7309 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7311 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7312 than "alu", remove explicit "memory" and "imm_disp" attributes.
7313 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7316 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7317 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7318 gen_lowpart_no_emit.
7320 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7323 * config/i386/i386.c (ix86_expand_builtin): Change memory
7324 operand to XI, extend p0 to Pmode.
7325 * config/i386/i386.md: Change unspec volatile and operand
7326 1 mode to XI, change operand 0 mode to P.
7328 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7330 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7331 GET_MODE_MASK before any checking.
7332 (nds32_can_use_bset_p): Likewise.
7333 (nds32_can_use_btgl_p): Likewise.
7335 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7337 * config/nds32/nds32-doubleword.md: New define_split pattern for
7338 illegal register number.
7340 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7342 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7344 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7346 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7348 2018-04-25 Richard Biener <rguenther@suse.de>
7350 * lto-streamer.h (LTO_major_version): Bump to 8.
7352 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7354 * BASE-VER: Set to 9.0.0.
7356 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7358 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7359 in __abskf2 and __powikf2.
7361 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7364 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7365 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7366 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7367 Use the above on operand 2. Reindent.
7368 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7370 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7373 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7375 * config/i386/i386.opt (mcet): Removed.
7376 * doc/install.texi: Remove -mcet documentation.
7377 * doc/invoke.texi: Likewise.
7379 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7382 * doc/install.texi: Remove -mcet from bootstrap-cet.
7384 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7387 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7388 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7392 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7393 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7394 containing a CONST_VECTOR.
7396 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7398 * doc/install.texi: Update newlib dependency for nvptx.
7400 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7403 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7404 instead of INTVAL when shifting x left.
7406 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7408 PR tree-optimization/85478
7409 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7410 vect_grouped_store_supported for single element vectors.
7412 2018-04-24 Richard Biener <rguenther@suse.de>
7415 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7416 load cost increase to the case of non-constant step.
7418 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7421 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7422 destination if any_malformed_asm.
7424 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7427 * expr.c (store_field): In the bitfield case, if the value comes from
7428 a function call and is returned in registers by means of a PARALLEL,
7429 do not change the mode of the temporary unless BLKmode and VOIDmode.
7431 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7433 PR rtl-optimization/85423
7434 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7435 dependencies to debug insns when the previous insn is non-debug.
7437 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7439 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7440 enums into a single definition.
7441 (fls): Fix predicates and printing.
7444 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7446 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7447 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7448 and short u6 immediate.
7449 (check_if_valid_sleep_operand): Remove.
7450 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7452 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7454 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7455 flag_always_save_lp condition.
7456 * config/nds32/nds32.opt (malways-save-lp): New option.
7458 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7460 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7461 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7462 * config/nds32/nds32.h
7463 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7464 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7466 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7468 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7469 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7471 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7472 Chung-Ju Wu <jasonwucj@gmail.com>
7474 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7476 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7477 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7479 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7481 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7483 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7485 * config/nds32/nds32-protos.h (nds32_data_alignment,
7486 nds32_local_alignment): Declare.
7487 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7488 nds32_local_alignment): New functions.
7489 (TARGET_CONSTANT_ALIGNMENT): Define.
7490 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7492 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7494 * config/nds32/nds32.c
7495 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7496 (TARGET_MODES_TIEABLE_P): Likewise.
7498 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7500 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7503 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7504 Chung-Ju Wu <jasonwucj@gmail.com>
7506 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7507 for unaligned access.
7508 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7510 * config/nds32/nds32-intrinsic.md: Likewise.
7511 * config/nds32/nds32_intrinsic.h: Likewise.
7512 * config/nds32/nds32.h (nds32_builtins): Likewise.
7513 * config/nds32/nds32.opt (munaligned-access): New option.
7514 * config/nds32/nds32.c (nds32_asm_file_start): Display
7515 flag_unaligned_access status.
7517 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7519 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7520 -mno-relax is present.
7521 * config/riscv/linux.h (LINK_SPEC): Ditto.
7523 2018-04-20 Martin Sebor <msebor@redhat.com>
7526 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7528 (gimple_fold_builtin_stxcpy_chk): Same.
7529 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7531 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7534 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7535 __powikf2 when long double is IEEE 128-bit.
7537 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7539 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7540 step to make sure stack always aligned.
7542 2018-04-20 Carl Love <cel@us.ibm.com>
7545 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7546 size check for arg0.
7548 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7549 Tom de Vries <tom@codesourcery.com>
7552 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7553 Emit insns for calls too.
7554 (nvptx_find_par): Always look for worker-level predecessor insn.
7555 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7557 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7558 (nvptx_process_pars): Propagate frames for calls.
7560 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7563 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7565 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7566 (ix86_handle_option): Don't handle OPT_mibt.
7567 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7569 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7571 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7572 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7573 (ix86_target_macros): Define __CET__ with flag_cf_protection
7574 for -fcf-protection.
7575 * config/i386/i386.c (isa2_opts): Remove -mibt.
7576 * config/i386/i386.h (TARGET_IBT): Removed.
7577 (TARGET_IBT_P): Likewise.
7578 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7579 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7580 * config/i386/i386.opt (mcet): Update help message.
7583 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7584 -mcet as an alias for -mshstk.
7586 2018-04-20 Richard Biener <rguenther@suse.de>
7589 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7590 complexity by forcing a single use of the multiply operand.
7592 2018-04-20 Martin Jambor <mjambor@suse.cz>
7595 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7596 recursion dependency to only apply to non-clones.
7598 2018-04-20 Martin Jambor <mjambor@suse.cz>
7601 * ipa-cp.c (create_specialized_node): Check that clones of
7602 self-recursive edges exist during IPA-CP.
7604 2018-04-19 Toon Moene <toon@moene.org>
7606 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7609 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7611 PR tree-optimization/85467
7612 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7613 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7614 VECTOR_CST element to type.
7616 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7619 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7620 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7621 (builtin_longjmp): Likewise.
7622 (save_stack_nonlocal): New pattern.
7623 (restore_stack_nonlocal): Likewise.
7625 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7628 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7629 Replace ASM_OUTPUT_LABEL with fprintf.
7631 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7634 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7635 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7636 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7637 define __IBT__ and __SHSTK__ for -fcf-protection.
7638 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7640 (ix86_trampoline_init): Likewise.
7641 (x86_output_mi_thunk): Likewise.
7642 (ix86_notrack_prefixed_insn_p): Likewise.
7643 (ix86_option_override_internal): Don't disallow -fcf-protection.
7644 * config/i386/i386.md (rdssp<mode>): Also enable for
7646 (incssp<mode>): Likewise.
7647 (nop_endbr): Likewise.
7648 * config/i386/i386.opt (mcet): Change help message to built-in
7652 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7653 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7654 enable CET built-in functions.
7656 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7658 * common/config/i386/i386-common.c
7659 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7660 OPTION_MASK_ISA_MOVDIRI_UNSET,
7661 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7662 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7663 * config.gcc (movdirintrin.h): New header.
7664 * config/i386/cpuid.h (bit_MOVDIRI,
7665 bit_MOVDIR64B): New bits.
7666 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7668 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7669 (VOID, PVOID, PCVOID)): New function types.
7670 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7671 __builtin_ia32_directstoreu_u64,
7672 __builtin_ia32_movdir64b): New builtins.
7673 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7674 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7676 (ix86_valid_target_attribute_inner_p): Ditto.
7677 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7678 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7679 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7680 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7681 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7682 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7683 (movdiri<mode>, movdir64b_<mode>): New.
7684 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7685 * config/i386/immintrin.h: Include movdirintrin.h.
7686 * config/i386/movdirintrin.h: New file.
7687 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7689 2018-04-19 Richard Biener <rguenther@suse.de>
7692 * cfg.c (clear_bb_flags): When loop state says we have
7693 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7695 2018-04-19 Richard Biener <rguenther@suse.de>
7697 PR tree-optimization/84737
7698 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7699 copying restrict info.
7700 (vect_setup_realignment): Use it.
7701 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7702 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7703 the first DR to all generated stores.
7704 (vectorizable_load): Likewise for loads.
7706 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7708 PR tree-optimization/85446
7709 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7710 the integral and pointer types to have the same precision.
7712 * doc/install.texi: Document --disable-cet being the default and
7715 2018-04-18 Martin Liska <mliska@suse.cz>
7717 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7720 2018-04-18 Martin Liska <mliska@suse.cz>
7723 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7726 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7727 arguments if they are comparable.
7729 2018-04-18 Martin Liska <mliska@suse.cz>
7732 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7735 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7738 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7741 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7742 ENDBR after calling __morestack.
7744 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7747 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7748 by using gcc_base_ver to generate a gcc_driver_version, and use
7749 it when generating GCC_DRIVER_NAME.
7750 * configure: Regenerate.
7752 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7755 * config.gcc: Obsolete powerpc*-*-*spe*.
7757 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7760 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7761 (stabstr_D): Change type of unum from unsigned int to
7762 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7765 2018-04-17 Jim Wilson <jimw@sifive.com>
7768 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7769 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7770 Set arg_pointer_offset after using pretend_args_size.
7772 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7774 PR rtl-optimization/85431
7775 * dse.c (record_store): Ignore zero width stores.
7778 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7779 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7780 __builtin_stack_restore rather than after it.
7781 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7782 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7783 argument instead of virtual_dynamic_stack_rtx.
7785 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7787 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7789 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7790 Add note to error message to explain internal mapping of overloaded
7791 built-in function name to non-overloaded built-in function name.
7792 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7795 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7798 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7799 where the inputs overlap with the output.
7801 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7803 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7804 (=v, v) alternative and explicit "memory" attribute.
7805 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7806 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7808 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7809 "sselog1" type instead of "sselog".
7810 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7811 "sselog". Remove explicit "memory" attribute.
7812 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7813 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7815 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7816 "sselog1" type instead of "sselog". Remove explicit "memory"
7818 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7819 use "sselog1" type instead of "sselog". Remove explicit "memory"
7821 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7822 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7824 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7825 "sselog1" type instead of "sselog". Remove explicit "memory"
7827 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7828 use "sselog1" type instead of "sselog". Remove explicit "memory"
7832 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7835 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7838 2018-04-17 Martin Jambor <mjambor@suse.cz>
7841 * ipa-cp.c (create_specialized_node): Call
7842 expand_all_artificial_thunks if necessary.
7844 2018-04-17 Martin Liska <mliska@suse.cz>
7847 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7848 in message, remote space in between '_G' and '('.
7850 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7853 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7854 avx512f_vmcmp<mode>3<round_saeonly_name>,
7855 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7856 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7857 avx512f_rndscale<mode><round_saeonly_name>,
7858 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7859 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7860 Use %<iptr>2 instead of %2 for -masm=intel.
7861 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7862 avx512f_vcvttss2usi<round_saeonly_name>,
7863 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7865 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7866 avx512f_vcvttsd2usi<round_saeonly_name>,
7867 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7868 Use %q1 instead of %1 for -masm=intel.
7869 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7870 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7871 of %3 for -masm=intel.
7872 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7874 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7876 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7878 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7879 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7881 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7883 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7884 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7885 %g1 and one with %0 and %1.
7886 (avx512er_vmrcp28<mode><round_saeonly_name>,
7887 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7889 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7890 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7891 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7892 of %0 and %{%4%} for -masm=intel.
7893 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7894 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7895 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7896 order of %0 and %{%5%}%{z%} for -masm=intel.
7898 2018-04-17 Jan Hubicka <jh@suse.cz>
7901 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7903 2018-04-17 Martin Liska <mliska@suse.cz>
7906 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7907 for target_clone error message. Make default implementation
7908 clone to be a local declaration.
7909 (separate_attrs): Add new argument and check for an empty
7911 (expand_target_clones): Handle it.
7912 (ipa_target_clone): Make redirection just for target_clones
7915 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7916 Tom de Vries <tom@codesourcery.com>
7919 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7920 tiled basic blocks without omp continue statements.
7922 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7925 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7926 vec_extract expression as having side effects to make sure it gets
7929 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7932 * config/i386/i386.c (get_builtin_code_for_version): Check
7935 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7938 * gcc/config.gcc: Support "skylake".
7939 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7941 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7942 (processor_target_table): Add "skylake".
7943 (ix86_option_override_internal): Add "skylake".
7944 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7945 PROCESSOR_CANNONLAKE.
7946 (get_builtin_code_for_version): Fix priority for
7947 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7948 PROCESSOR_SKYLAKE-AVX512.
7949 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7950 (processor_type): Add PROCESSOR_SKYLAKE.
7952 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7953 Jason Merrill <jason@redhat.com>
7956 * convert.c (convert_to_integer_1): Use direct recursion for
7957 enumeral types and types with a precision less than the number
7958 of bits in their mode.
7960 2018-04-16 Julia Koval <julia.koval@intel.com>
7963 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7964 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7966 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7969 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7970 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7971 and -mno-direct-move.
7973 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7976 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7977 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7978 Ensure negative shifts result in {0}.
7980 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7982 PR rtl-optimization/79916
7983 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7984 regs (if any) to define how to gnerate SD moves when LRA is in
7987 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7989 PR rtl-optimization/85393
7990 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7991 * except.c (expand_dw2_landing_pad_for_region): Make static.
7992 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7993 a label and unconditional jump to old_bb, rather than
7994 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7997 PR rtl-optimization/85376
7998 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7999 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8000 instead of a specific value.
8002 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8003 Bin Cheng <bin.cheng@arm.com>
8005 PR tree-optimization/82965
8006 PR tree-optimization/83991
8007 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8008 by_profile_only parameter.
8009 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8010 information if the loop was predicted to iterate too many times.
8011 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8013 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8016 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8019 2018-04-13 Martin Liska <mliska@suse.cz>
8020 Jakub Jelinek <jakub@redhat.com>
8023 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8024 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8025 * builtins.c (expand_builtin_memory_copy_args): Use
8026 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8027 handle dest_addr == pc_rtx.
8029 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8032 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8033 asked to not generate direct moves.
8034 (fix_trunc<mode>si2_stfiwx): Similar.
8035 (fix_trunc<mode>si2_internal): Similar.
8037 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8040 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8041 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8042 lookup if dest in some wider mode is known to be const0_rtx and
8043 if so, record permanent equivalence for it to be ZERO_EXTEND of
8044 the narrower mode destination.
8046 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8048 * lto-streamer-out.c (output_function): Revert 259346.
8049 * omp-expand.c (expand_oacc_for): Likewise.
8051 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8053 PR rtl-optimization/85354
8054 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8055 * sel-sched.c (sel_global_init): ... here.
8057 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8060 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8061 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8062 mode for PE-COFF targets.
8063 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8064 (i386_pe_asm_lto_end): Likewise.
8065 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8066 (TARGET_ASM_LTO_END): Likewise.
8067 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8068 (i386_pe_asm_lto_start): New function.
8069 (i386_pe_asm_lto_end): Likewise.
8071 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8072 Richard Biener <rguenther@suse.de>
8075 * lto-streamer-out.c (output_function): Fix CFG loop state before
8077 * omp-expand.c (expand_oacc_for): Handle calls to internal
8078 functions like regular functions.
8080 2018-04-12 Richard Biener <rguenther@suse.de>
8083 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8084 for the early LTO debug to properly generate references to it
8085 during DIE emission. Do not re-use that for the skeleton for
8087 (dwarf2out_early_finish): Likewise.
8089 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8092 * config/i386/sse.md
8093 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8094 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8095 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8096 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8097 and output is a reg, avoid creating invalid lowpart subreg, but
8098 instead split into a 512-bit move. Don't split if not AVX512VL,
8099 input is xmm16+ reg and output is a mem.
8100 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8101 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8102 xmm16+ reg and output is a mem.
8104 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8106 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8107 also for flag_dwarf2_cfi_asm.
8109 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8111 PR rtl-optimization/85342
8112 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8113 a bool scalar var inside of the loop instead. Don't try to update
8114 recog_data.operand after failed apply_change_group.
8116 2018-04-12 Tom de Vries <tom@codesourcery.com>
8119 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8120 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8121 array with flexible array member as array without given dimension.
8122 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8123 argument for undefined param to true.
8125 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8128 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8129 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8130 from PowerPC section.
8131 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8132 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8133 help text that is too long.
8134 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8135 help text that is too long.
8136 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8137 help text that is too long.
8139 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8141 * config/alpha/alpha.md (stack_probe_internal): Rename
8142 from "probe_stack". Update all callers.
8144 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8146 PR rtl-optimization/84566
8147 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8148 sched_macro_fuse_insns.
8150 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8153 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8154 (compute_block_dependences): ... from here.
8156 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8158 PR tree-optimization/85331
8159 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8160 from int to HOST_WIDE_INT.
8162 2018-04-11 Martin Jambor <mjambor@suse.cz>
8165 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8166 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8167 not the same as the source val.
8168 (cgraph_edge_brings_value_p): New parameter.
8169 (gather_edges_for_value): Pass destination value to
8170 cgraph_edge_brings_value_p.
8171 (perhaps_add_new_callers): Likewise.
8172 (get_info_about_necessary_edges): Likewise and exclude values brought
8173 only by self-recursive edges.
8174 (create_specialized_node): Redirect only clones of self-calling edges.
8175 (+self_recursive_pass_through_p): New function.
8176 (find_more_scalar_values_for_callers_subset): Use it.
8177 (find_aggregate_values_for_callers_subset): Likewise.
8178 (known_aggs_to_agg_replacement_list): Removed.
8179 (decide_whether_version_node): Re-calculate known constants for all
8180 remaining context clones.
8182 2018-04-11 Richard Biener <rguenther@suse.de>
8185 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8186 from early DWARF output.
8187 (dwarf2out_early_finish): Output line info unconditionally into
8188 early DWARF and add reference to it.
8190 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8193 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8194 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8195 other than V2DFmode using iptr mode attribute.
8196 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8198 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8200 PR rtl-optimization/84659
8201 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8203 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8206 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8208 (output_loc_list): Pass address of a dummy size variable even in the
8209 locview handling loop.
8210 (index_location_lists): Add comment on why skip_loc_list_entry can't
8213 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8216 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8219 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8222 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8223 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8224 and -mstring-compare-inline-limit.
8226 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8229 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8230 for stack clash protection in a register whenever we need it to be in
8233 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8235 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8236 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8238 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8241 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8243 (mlong-double-): Ditto.
8244 * config/rs6000/sysv4.opt (msdata=): Ditto.
8245 (mtls-size=): Ditto.
8247 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8249 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8250 erroneous entries for
8251 "vector int vec_ldl (int, long int *)", and
8252 "vector unsigned int vec_ldl (int, unsigned long int *)".
8253 Add comments and entries for
8254 "vector bool char vec_ldl (int, bool char *)",
8255 "vector bool short vec_ldl (int, bool short *)",
8256 "vector bool int vec_ldl (int, bool int *)",
8257 "vector bool long long vec_ldl (int, bool long long *)",
8258 "vector pixel vec_ldl (int, pixel *)",
8259 "vector long long vec_ldl (int, long long *)",
8260 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8261 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8262 type tree bool_long_long_type_node and correct definition of
8263 bool_V2DI_type_node to make reference to this new type tree.
8264 (rs6000_mangle_type): Replace erroneous reference to
8265 bool_long_type_node with bool_long_long_type_node.
8266 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8267 comments to emphasize sign distinctions for char and int types and
8268 replace RS6000_BTI_bool_long constant with
8269 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8270 use of RS6000_BTI_pixel.
8271 (bool_long_type_node): Remove this macro definition.
8272 (bool_long_long_type_node): New macro definition
8274 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8276 PR rtl-optimization/85300
8277 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8278 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8279 simplify_unary_operation fails.
8281 2018-04-10 Martin Liska <mliska@suse.cz>
8283 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8284 cgraph_edge and ipa_ref.
8286 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8290 * config/i386/sse.md
8291 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8292 computation of the VEC_MERGE selector from mask.
8293 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8294 Fix decoding of the VEC_MERGE selector into mask.
8296 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8298 PR tree-optimization/85286
8299 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8301 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8303 * final.c (final_1): Set insn_last_address as well as
8304 insn_current_address.
8306 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8309 * explow.c (emit_stack_probe): Call validize_mem on memory location
8310 before passing it to gen_probe_stack. Create address operand and
8311 legitimize it for the probe_stack_address case.
8313 2018-04-09 Jan Hubicka <jh@suse.cz>
8316 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8317 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8318 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8320 (free_lang_data): Rebuild type inheritance graph.
8322 2018-04-09 Martin Sebor <msebor@redhat.com>
8324 * invoke.texi (-finline-small-functions): Mention other optimization
8326 (-findirect-inlining, -fpartial-inlining): Same.
8327 (-finline-functions-called-once): Same.
8328 (-freorder-blocks-and-partition): Same.
8330 2018-04-09 Jan Hubicka <jh@suse.cz>
8333 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8334 jumps; choose last target that matches the criteria (i.e.
8335 no partition changes for non-crossing jumps).
8336 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8337 support for redirecting crossing jumps to non-crossing.
8339 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8341 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8342 also for naked functions.
8344 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8346 * config/arc/arc.md (add_shift): New pattern.
8347 (add_shift2): Likewise.
8348 (sub_shift): Likewise.
8349 (sub_shift_cmp0_noout): Likewise.
8350 (compare_si_ashiftsi): Likewise.
8351 (xbfu_cmp0_noout): New combine pattern.
8352 (xbfu_cmp0"): Likewise.
8353 (movsi_set_cc_insn): Place the predicable variant first.
8354 (commutative_binary_cmp0_noout): Remove clobber.
8355 (commutative_binary_cmp0): New pattern.
8356 (noncommutative_binary_cmp0): Likewise.
8357 (noncommutative_binary_cmp0_noout): Likewise.
8358 (noncommutative_binary_comparison_result_used): Removed.
8359 (rsub_cmp0): New pattern.
8360 (rsub_cmp0_noout): Likewise.
8361 (extzvsi): Changed, keep only meaningful variants.
8362 (SQH, SEZ): New iterators.
8363 (SQH_postfix): New mode attribute.
8364 (SEZ_prefix): New code attribute.
8365 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8366 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8367 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8369 (noncommutative_operator): Check the availability of barrel
8372 2018-04-09 Richard Biener <rguenther@suse.de>
8374 PR tree-optimization/85284
8375 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8376 Only use the niter constraining form of simple_iv when the exit
8379 2018-04-09 Tom de Vries <tom@codesourcery.com>
8382 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8383 (define_expand "*memory_barrier"): New define_expand.
8384 (define_insn "memory_barrier"): New insn.
8386 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8388 PR rtl-optimization/80463
8389 PR rtl-optimization/83972
8390 PR rtl-optimization/83480
8392 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8393 correct producer for the insn.
8394 (tidy_control_flow): Fixup seqnos in case of debug insns.
8396 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8398 PR rtl-optimization/83913
8400 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8401 different sched-times when merging exprs.
8403 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8405 PR rtl-optimization/83962
8407 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8408 tidy_fallthru_edge and tidy_control_flow.
8410 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8412 PR rtl-optimization/83530
8414 * sel-sched.c (force_next_insn): New global variable.
8415 (remove_insn_for_debug): When force_next_insn is true, also leave only
8416 next insn in the ready list.
8417 (sel_sched_region): When the region wasn't scheduled, make another pass
8418 over it with force_next_insn set to 1.
8420 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8422 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8424 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8425 for interrupt control.
8426 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8427 functions for interrupt control.
8428 * config/nds32/nds32-intrinsic.md: Likewise.
8429 * config/nds32/nds32_intrinsic.h: Likewise.
8430 * config/nds32/nds32.h (nds32_builtins): Likewise.
8432 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8434 * config/nds32/nds32.c (nds32_init_machine_status,
8435 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8436 strict_aligned_p field.
8437 (nds32_expand_to_rtl_hook): New function.
8438 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8439 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8441 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8442 Chung-Ju Wu <jasonwucj@gmail.com>
8444 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8445 * config/nds32/nds32-n7.md: New file.
8446 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8447 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8449 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8450 * config/nds32/nds32.md (pipeline_model): Add n7.
8451 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8452 * config/nds32/pipelines.md: Include n7 settings.
8454 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8455 Chung-Ju Wu <jasonwucj@gmail.com>
8457 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8458 * config/nds32/nds32-e8.md: New file.
8459 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8460 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8462 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8463 * config/nds32/nds32.md (pipeline_model): Add e8.
8464 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8465 * config/nds32/pipelines.md: Include e8 settings.
8467 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8468 Chung-Ju Wu <jasonwucj@gmail.com>
8470 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8471 * config/nds32/nds32-n8.md: New file.
8472 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8473 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8475 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8476 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8477 * config/nds32/nds32.md (pipeline_model): Add n8.
8478 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8479 * config/nds32/pipelines.md: Include n8 settings.
8481 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8482 Chung-Ju Wu <jasonwucj@gmail.com>
8484 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8485 * config/nds32/nds32-n9-2r1w.md: New file.
8486 * config/nds32/nds32-n9-3r2w.md: New file.
8487 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8488 nds32_register_ports): New or modify for cpu n9.
8489 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8491 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8492 * config/nds32/nds32-utils.c: New file.
8493 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8494 TARGET_MUL_SLOW): Define.
8495 * config/nds32/nds32.md (pipeline_model): New attribute.
8496 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8497 New options that support cpu n9.
8498 * config/nds32/pipelines.md: Include n9 settings.
8499 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8501 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8503 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8504 information if necessary.
8505 (output_cond_branch_compare_zero): Likewise.
8506 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8507 (nds32_target_alignment): Refine for alignment.
8508 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8509 (FUNCTION_BOUNDARY): Modify.
8510 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8512 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8514 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8516 * config/nds32/constants.md (unspec_volatile_element): Add values for
8517 TLB operation and data prefetch.
8518 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8519 functions for TLB operation and data prefetch.
8520 * config/nds32/nds32-intrinsic.md: Likewise.
8521 * config/nds32/nds32_intrinsic.h: Likewise.
8522 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8523 (nds32_print_operand): Likewise.
8524 * config/nds32/nds32.h (nds32_builtins): Likewise.
8526 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8527 Andrew Pinski <pinsika@gcc.gnu.org>
8530 * match.pd: Use constant_boolean_node of correct type instead of
8531 boolean_true_node or boolean_false_node for simplifying
8532 pointer comparisons to zero.
8534 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8536 PR tree-optimization/80021
8537 * tree.c (verify_type_variant): Make error call in verify_variant_match
8538 translatable and remove final full stop.
8540 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8542 * config/nds32/constants.md (unspec_volatile_element): Add
8543 UNSPEC_VOLATILE_EH_RETURN.
8544 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8545 nds32_output_stack_pop): Support dwarf exception handling process.
8546 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8547 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8548 exception handling process.
8549 (nds32_compute_stack_frame): Likewise.
8550 (nds32_return_addr_rtx): Likewise.
8551 (nds32_initial_elimination_offset): Likewise.
8552 (nds32_expand_prologue): Likewise.
8553 (nds32_expand_epilogue): Likewise.
8554 (nds32_dynamic_chain_address): New function.
8555 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8557 (DYNAMIC_CHAIN_ADDRESS): Define.
8558 (EH_RETURN_DATA_REGNO): Define.
8559 (EH_RETURN_STACKADJ_RTX): Define.
8560 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8561 patterns for dwarf exception handling.
8563 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8565 * config/nds32/nds32.h: Clean up obsolete macros.
8567 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8569 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8570 Add enum values for particular instructions.
8571 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8572 particular intrinsic functions.
8573 * config/nds32/nds32-intrinsic.md: Likewise.
8574 * config/nds32/nds32_intrinsic.h: Likewise.
8575 * config/nds32/nds32.h (nds32_builtins): Likewise.
8576 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8577 (btst, ave): New patterns for particular instructions.
8579 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8581 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8582 Add enum values for atomic load/store and memory sync.
8583 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8585 * config/nds32/nds32-intrinsic.md: Likewise.
8586 * config/nds32/nds32_intrinsic.h: Likewise.
8587 * config/nds32/nds32.h (nds32_builtins): Likewise.
8589 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8591 PR tree-optimization/85257
8592 * fold-const.c (native_encode_vector): If not all elts could fit
8593 and off is -1, return 0 rather than offset.
8594 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8595 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8596 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8597 adjust buffer in native_interpret_expr call.
8599 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8601 * config/nds32/constants.md (unspec_volatile_element): Add cache
8602 control enum values.
8603 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8604 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8605 * config/nds32/nds32.c (nds32_cctl_names): New.
8606 (nds32_print_operand): Handle cache control register names.
8607 * config/nds32/nds32.h (nds32_builtins): New enum values.
8608 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8610 * config/nds32/nds32.md (type): Add mmu.
8611 * config/nds32/pipelines.md (simple_insn): Add mmu.
8613 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8615 * config/nds32/nds32.md (type): Remove call.
8616 * config/nds32/pipelines.md (simple_insn): Likewise.
8618 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8620 * config/nds32/constants.md (unspec_volatile_element): Add
8621 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8622 UNSPEC_VOLATILE_FMFCFG.
8623 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8624 description for fmfcfg and fmfcsr.
8625 (bdesc_1arg): Add fmtcsr.
8626 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8627 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8628 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8629 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8630 unspec_fmfcfg): New patterns.
8631 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8632 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8633 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8634 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8635 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8636 __nds32__fmfcfg): Define.
8638 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8640 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8641 intrinsic register names.
8642 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8643 intrinsic register enum values and macros.
8645 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8647 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8648 for load/store addressing form.
8649 (nds32_print_operand_address): Likewise.
8651 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8654 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8655 based on LABEL_REF. Remove useless assertion.
8656 (pic_address_needs_scratch): Fix formatting.
8657 (sparc_legitimize_pic_address): Minor tweaks.
8658 (sparc_delegitimize_address): Adjust assertion accordingly.
8659 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8660 into symbolic_operand.
8661 (movsi_high_pic_label_ref): Likewise.
8662 (movsi_lo_sum_pic_label_ref): Likewise.
8663 (movdi_pic_label_ref): Likewise.
8664 (movdi_high_pic_label_ref): Likewise.
8665 (movdi_lo_sum_pic_label_ref): Likewise.
8667 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8669 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8670 custom LIB_SPEC setup.
8672 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8673 Kito Cheng <kito.cheng@gmail.com>
8675 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8676 * config/riscv/freebsd.h: New.
8678 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8680 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8681 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8684 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8685 Kito Cheng <kito.cheng@gmail.com>
8687 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8688 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8689 * config/nds32/nds32-protos.h (nds32_output_call,
8690 nds32_output_return): Declare.
8691 * config/nds32/nds32.md: Refine all the call and return patterns.
8693 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8696 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8697 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8699 PR rtl-optimization/84872
8700 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8701 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8704 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8706 * expr.c (copy_blkmode_to_reg): Revert 254862.
8707 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8709 2018-04-06 Richard Biener <rguenther@suse.de>
8712 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8713 after seeing a component reference with an adjacent field. Treat
8714 refs to arrays at struct end of external decls similar to
8715 refs to unconstrained commons.
8717 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8720 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8721 look through SAVE_EXPRs with non-side-effects argument. Adjust
8723 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8726 2018-04-06 Richard Biener <rguenther@suse.de>
8729 * alias.c (find_base_term): New wrapper around find_base_term
8730 unwinding CSELIB_VAL_PTR changes.
8731 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8734 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8736 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8738 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8739 constant definitions.
8740 ("nop"): lr 0,0 -> nopr r0
8741 ("nop_lr0", "nop_lr1"): New insn definitions.
8743 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8745 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8746 NDS32_V3PUSH_AVAILABLE_P macro.
8748 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8749 Chung-Ju Wu <jasonwucj@gmail.com>
8751 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8752 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8753 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8754 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8755 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8756 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8757 * config/nds32/constraints.md: New constraints and checking for hard
8758 float configuration.
8759 * config/nds32/iterators.md: New mode iterator and attribute for hard
8760 float configuration.
8761 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8763 * config/nds32/nds32-fpu.md: New file.
8764 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8765 deal with hard float code generation.
8766 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8768 (abi_type, float_reg_number): New enum type.
8769 * config/nds32/nds32-predicates.c: New predicates for hard float.
8770 * config/nds32/nds32-protos.h: Declare functions for hard float.
8771 * config/nds32/nds32.c: Implementation for hard float configuration.
8772 * config/nds32/nds32.h: Definitions for hard float configuration.
8773 * config/nds32/nds32.md: Include hard float machine description and
8774 modify patterns for hard float configuration.
8775 * config/nds32/nds32.opt: New options for hard float configuration.
8776 * config/nds32/predicates.md: New predicates for hard float
8779 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8781 * common/config/nds32/nds32-common.c
8782 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8784 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8787 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8788 CONSTRUCTOR_ELT (ctor, ...)->value.
8790 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8793 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8795 2018-04-05 Tom de Vries <tom@codesourcery.com>
8798 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8801 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8802 Kito Cheng <kito.cheng@gmail.com>
8804 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8805 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8806 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8807 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8809 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8810 Kito Cheng <kito.cheng@gmail.com>
8812 * config/nds32/constraints.md (Ufe): New memory constraint.
8813 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8814 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8815 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8817 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8818 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8820 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8822 * config/nds32/nds32.md: Use optimize_size in the condition for
8823 alu-shift instructions.
8825 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8827 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8829 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8831 * config/nds32/nds32.md (negsi2): Refine pattern.
8833 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8834 Chung-Ju Wu <jasonwucj@gmail.com>
8836 * config/nds32/iterators.md (shift_rotate): New code iterator.
8837 (shift): New code attribute.
8838 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8839 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8840 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8841 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8842 bit-wise operations.
8843 (andsi3, *andsi3): Ditto.
8844 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8845 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8846 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8847 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8848 nds32_ior_operand, nds32_xor_operand): New predicates.
8850 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8852 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8853 (addsi3, subsi3): ... this.
8855 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8857 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8859 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8861 * config/nds32/nds32.md: Adjust indention.
8863 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8865 * config/nds32/nds32.md (feature): New attribute.
8867 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8869 * config/nds32/nds32.md (subtype): New attribute.
8871 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8874 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8875 expansion to perform a bitwise AND of the argument followed by a
8876 boolean negation of the result.
8878 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8880 PR rtl-optimization/84878
8881 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8882 the basic block. Assert the use reference is not artificial and that
8883 it has an associated insn.
8885 2018-04-04 Michael Matz <matz@suse.de>
8887 * builtins.c (compute_objsize): Pass correct operand
8888 to array_at_struct_end_p.
8890 2018-04-04 Richard Biener <rguenther@suse.de>
8893 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8894 from contexts for DINFO_LEVEL_TERSE and below.
8896 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8898 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8899 resiter_operand condition.
8900 * config/nds32/nds32.md (*move<mode>): Ditto.
8902 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8903 Monk Chiang <sh.chiang04@gmail.com>
8905 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8907 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8909 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8911 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8912 Kito Cheng <kito.cheng@gmail.com>
8914 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8915 nds32_cond_code_str, output_cond_branch,
8916 output_cond_branch_compare_zero, nds32_expand_cbranch,
8917 nds32_expand_cstore, nds32_expand_movcc,
8918 nds32_output_cbranchsi4_equality_zero,
8919 nds32_output_cbranchsi4_equality_reg,
8920 nds32_output_cbranchsi4_equality_reg_or_const_int,
8921 nds32_output_cbranchsi4_greater_less_zero: New functions.
8922 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8923 nds32_expand_cstore, nds32_expand_movcc,
8924 nds32_output_cbranchsi4_equality_zero,
8925 nds32_output_cbranchsi4_equality_reg,
8926 nds32_output_cbranchsi4_equality_reg_or_const_int,
8927 nds32_output_cbranchsi4_greater_less_zero): Declare.
8928 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8929 nds32_rimm11s_operand): New predicates.
8930 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8931 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8934 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8936 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8937 * config/nds32/nds32.md: Ditto.
8938 * config/nds32/pipelines.md: Ditto.
8940 2018-04-04 Richard Biener <rguenther@suse.de>
8942 PR tree-optimization/85168
8943 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8944 propagating abnormals.
8946 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8948 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8950 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8951 Kito Cheng <kito.cheng@gmail.com>
8953 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8954 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8955 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8956 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8957 * config/nds32/nds32.md (sibcall_internal): New.
8958 (sibcall_register): Remove.
8959 (sibcall_immediate): Remove.
8960 (sibcall_value_internal): New.
8961 (sibcall_value_register): Remove.
8962 (sibcall_value_immediate): Remove.
8963 * config/nds32/predicates.md (nds32_general_register_operand): New.
8964 (nds32_call_address_operand): New.
8966 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8968 PR rtl-optimization/85167
8969 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8970 bb_defs if *split_p, instead preinitialize it to NULL.
8972 PR tree-optimization/85156
8973 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8974 evaluating the argument multiple times.
8976 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8978 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8980 (_mm_cvtpd_ps): Likewise.
8981 (_mm_cvttpd_epi32): Likewise.
8982 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8983 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8984 vector, pixel, and bool following altivec.h include.
8986 2018-04-03 Martin Sebor <msebor@redhat.com>
8988 * doc/extend.texi (Common Function Attributes): Clarify.
8989 (const attribute): Likewise.
8990 (pure attribute): Likewise.
8992 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8995 * config/i386/i386.c (ix86_expand_vector_set): Use
8996 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8998 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9000 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9001 instructions when changing rounding bits to preserve precision bits
9002 in the x87 control word.
9004 2018-04-03 Martin Liska <mliska@suse.cz>
9006 PR tree-optimization/82491
9007 * rtl.h (strip_offset_and_add): Replace += suboffset with
9008 poly_uint64 () + suboffset.
9010 2018-03-29 Martin Liska <mliska@suse.cz>
9011 Martin Jambor <mjambor@suse.cz>
9014 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9015 param_type is not an integral or pointer type.
9017 2018-04-03 Richard Biener <rguenther@suse.de>
9019 * sese.h (recompute_all_dominators): Remove.
9021 2018-04-02 Martin Sebor <msebor@redhat.com>
9023 * doc/invoke.texi (-Wrestrict): Fix typos.
9025 2018-04-02 Jim Wilson <jimw@sifive.com>
9027 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9028 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9029 (<optab>di3, <optab>si3_extend): Likewise.
9030 (<optab>si3_mask, <optab>si3_mask_1): New.
9031 (<optab>di3_mask, <optab>di3_mask_1): New.
9032 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9033 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9034 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9036 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9038 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9041 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9043 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9044 (nds32_canonicalize_comparison): New function.
9046 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9047 Kito Cheng <kito.cheng@gmail.com>
9048 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9050 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9051 * config/nds32/constants.md (unspec_volatile_element): Add
9052 UNSPEC_VOLATILE_RELAX_GROUP.
9053 * config/nds32/nds32-relax-opt.c: New file.
9054 * config/nds32/nds32-predicates.c
9055 (nds32_symbol_load_store_p): New function.
9056 * config/nds32/nds32-protos.h
9057 (nds32_symbol_load_store_p): Declare function.
9058 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9059 * config/nds32/nds32.c
9060 (nds32_register_pass): New function to register pass.
9061 (nds32_register_passes): New function to register passes.
9062 * config/nds32/nds32.md (relax_group): New pattern.
9063 * config/nds32/nds32.opt (mrelax-hint): New option.
9064 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9066 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9068 * config/nds32/t-nds32: Modify files dependency.
9070 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9072 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9073 (PROFILE_HOOK): Define its implementation.
9075 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9077 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9078 type and 32-bit size.
9080 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9083 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9084 (V_128_256): New mode iterator.
9085 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9086 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9087 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9089 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9090 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9092 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9095 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9096 NaN inputs correctly.
9098 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9101 * config/rs6000/vsx.md (??r): New mode attribute.
9102 (*vsx_mov<mode>_64bit): Use it.
9103 (*vsx_mov<mode>_32bit): Likewise.
9105 2018-03-30 Martin Sebor <msebor@redhat.com>
9107 PR tree-optimization/84818
9108 * builtins.c (check_access): Use warning_n.
9110 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9113 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9115 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9118 2018-03-30 Julia Koval <julia.koval@intel.com>
9121 * x86-tune.def (movx, partial_reg_dependency): Enable for
9124 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9127 * lra-constraints.c (process_alt_operands): Move setting
9128 this_alternative_matches below.
9130 2018-03-29 Martin Liska <mliska@suse.cz>
9133 * doc/invoke.texi: Document how LTO works with debug info.
9134 Describe auto-load support of binutils. Mention 'x86-64'
9135 as valid option value of -march option.
9137 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9139 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9142 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9143 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9144 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9147 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9150 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9151 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9152 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9153 for RS6000_BTM_POWERPC64.
9154 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9155 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9156 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9161 2018-03-28 Carl Love <cel@us.ibm.com>
9164 2017-09-27 Carl Love <cel@us.ibm.com>
9166 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9167 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9168 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9171 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9173 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9174 instead of __vector bool.
9175 (_mm_max_pu8): Likewise.
9176 (_mm_min_pi16): Likewise.
9178 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9181 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9182 (DIVWEUO): Likewise.
9184 (DIVDEUO): Likewise.
9185 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9186 DIVWEUO and DIVDEUO.
9187 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9188 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9189 (div_extend): Likewise.
9190 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9192 (__builtin_divweuo): Likewise.
9193 (__builtin_divdeo): Likewise.
9194 (__builtin_divdeuo): Likewise.
9196 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9199 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9200 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9202 PR tree-optimization/82004
9203 * gimple-match-head.c (optimize_pow_to_exp): New function.
9204 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9205 Don't fold to exp if optimize_pow_to_exp is false.
9207 2018-03-28 Martin Liska <mliska@suse.cz>
9210 * calls.c (initialize_argument_information): Fix trailing space.
9211 * common.opt: Fix typo and provide better explanation for
9212 -fsanitize-coverage option.
9213 * config/i386/i386.opt: Fix typo.
9215 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9216 Martin Liska <mliska@suse.cz>
9219 * gimplify.c (asan_poison_variable): Don't do the check for
9220 gimplify_omp_ctxp here.
9221 (gimplify_decl_expr): Do it here.
9222 (gimplify_target_expr): Likewise.
9224 2018-03-28 Martin Liska <mliska@suse.cz>
9227 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9228 chkp_type_bounds_count if MPX is not enabled.
9230 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9232 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9234 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9237 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9238 function to create the function decl for complex long double
9239 multiply and divide for -mabi=ieeelongdouble.
9240 (init_float128_ieee): Call it.
9242 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9245 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9246 -fcf-protection=branch -mibt.
9247 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9249 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9252 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9253 arm_word_relocations.
9255 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9258 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9259 extern array declarations.
9261 2018-03-27 Richard Biener <rguenther@suse.de>
9264 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9265 explicit single_use checks.
9267 2018-03-27 Richard Biener <rguenther@suse.de>
9269 PR tree-optimization/85082
9270 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9273 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9275 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9276 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9277 Turn on fasynchronous-unwind-tables and funwind-tables.
9279 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9282 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9283 (*bmi_blsr_<mode>_ccz): Ditto.
9285 2018-03-26 Tom de Vries <tom@codesourcery.com>
9287 PR tree-optimization/85063
9288 * omp-general.c (offloading_function_p): New function. Factor out
9290 * omp-offload.c (pass_omp_target_link::gate): ... here.
9291 * omp-general.h (offloading_function_p): Declare.
9292 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9293 with attribute omp declare target for offloading functions.
9295 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9297 PR tree-optimization/84005
9298 * tree-data-ref.h (get_base_for_alignment): Declare.
9299 * tree-data-ref.c (get_base_for_alignment_1): New function.
9300 (get_base_for_alignment): Likewise.
9301 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9302 get_base_for_alignment to find a suitable base object, instead
9303 of always using drb->base_address.
9305 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9308 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9309 known size by default.
9311 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9314 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9315 and non BLKmode operands.
9317 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9320 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9321 Clean up attributes.
9323 2018-03-23 Richard Biener <rguenther@suse.de>
9326 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9327 we are going to emit early debug for LTO.
9329 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9332 * function.c (match_asm_constraints_1): Don't optimize if input
9333 doesn't satisfy general_operand predicate for output's mode.
9336 * alias.c (write_dependence_p): Don't require for x_canonicalized
9337 non-VOIDmode if x has VOIDmode.
9340 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9341 just don't try to optimize it rather than assert it never happens.
9343 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9345 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9346 macro expansions for definition of ST_INTERNAL_<mode> and
9347 LD_INTERNAL_<mode> builtins.
9348 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9350 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9352 (altivec_expand_st_builtin): Likewise.
9353 (altivec_expand_builtin): Remove calls to deleted functions.
9354 (rs6000_address_for_altivec): Delete this function.
9355 * config/rs6000/vector.md: Remove expands for
9356 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9358 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9361 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9362 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9363 re-computing once computed.
9364 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9365 (arm_init_machine_status): Initialize
9366 machine->static_chain_stack_bytes.
9368 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9371 * doc/extend.texi: Add four new prototypes for vec_ld.
9372 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9373 definitions for more logical presentation.
9374 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9375 entries for V1TI variants of __builtin_altivec_ld builtin.
9376 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9377 handling of V1TI variant of LVX icode pattern.
9378 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9379 (rs6000_gimple_fold_builtin): Likewise.
9380 (altivec_init_builtins): Add code to define
9381 __builtin_altivec_lvx_v1ti function.
9383 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9386 * function.c (match_asm_constraints_1): Don't do the optimization
9387 if input isn't a REG, SUBREG, MEM or constant.
9389 2018-03-22 Tom de Vries <tom@codesourcery.com>
9391 PR tree-optimization/84956
9392 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9393 bb_has_abnormal_pred.
9395 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9398 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9399 DECL_INITIAL (decl) to decl at the end.
9400 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9403 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9405 * doc/extend.texi (__builtin_tgmath): Document when complex
9406 integer types are treated as _Complex _Float64.
9408 2018-03-21 Tom de Vries <tom@codesourcery.com>
9410 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9412 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9414 PR tree-optimization/84960
9415 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9416 if it is ENTRY block, move them into single succ of ENTRY in that case.
9418 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9420 PR tree-optimization/84811
9421 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9422 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9423 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9424 (known_subrange_p): Update accordingly. Cast each value involved
9425 in the size comparison, rather than casting the result of the
9428 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9430 PR tree-optimization/84982
9431 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9432 by flipping the least significant bit rather than all bits from
9433 bitpos to bitpos + bitsize - 1.
9435 2018-03-21 Nathan Sidwell <nathan@acm.org>
9437 * doc/extend.texi (Deprecated Features): Remove mention of
9438 long-deleted deprecations.
9440 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9443 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9444 * configure: Regenerate.
9446 2018-03-21 Tom de Vries <tom@codesourcery.com>
9448 PR tree-optimization/83126
9449 * tree-parloops.c (num_phis): New function.
9450 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9452 2018-03-21 Nathan Sidwell <nathan@acm.org>
9454 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9455 mention anon-struct/union members and trailing attributes.
9457 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9459 PR tree-optimization/84969
9460 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9461 builtin memset partitions if they set different rhs values.
9463 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9465 PR rtl-optimization/84989
9466 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9467 VEC_DUPLICATE with scalar result mode.
9469 2018-03-21 Martin Liska <mliska@suse.cz>
9472 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9473 not intended return statement.
9475 2018-03-21 Martin Liska <mliska@suse.cz>
9478 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9479 (chkp_find_bound_slots_1): Limit number of iterations.
9481 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9484 * Minor grammar fixes for x86 options.
9486 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9489 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9490 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9492 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9495 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9496 (altivec_lvx_<mode>_1op): Likewise.
9497 (altivec_stvx_<mode>_2op): Likewise.
9498 (altivec_stvx_<mode>_1op): Likewise.
9499 (altivec_lvx_<VM2:mode>): New define_expand.
9500 (altivec_stvx_<VM2:mode>): Likewise.
9501 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9502 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9503 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9504 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9505 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9506 (rs6000_gen_lvx): Likewise.
9507 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9508 (altivec_expand_stv_builtin): Likewise.
9509 (altivec_expand_builtin): Likewise.
9510 * config/rs6000/vector.md: Likewise.
9512 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9515 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9518 2018-03-20 Richard Biener <rguenther@suse.de>
9521 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9522 sign-conversions as zero, fall back to standard scalar_stmt
9525 2018-03-20 Martin Liska <mliska@suse.cz>
9528 * predict.c (rebuild_frequencies): Handle case when we have
9529 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9531 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9534 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9535 flag_section_anchors.
9536 * varasm.c (use_blocks_for_decl_p): Remove hack for
9537 dw2_force_const_mem.
9540 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9542 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9543 be created, use lowpart_subreg of operands[0] rather than operands[0]
9545 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9546 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9547 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9548 and n constraint instead of aarch64_shift_imm_di and Usd.
9549 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9550 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9552 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9555 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9556 to favor GPR over NEON registers.
9557 (<shift>di3_neon): Likewise.
9559 2018-03-20 Tom de Vries <tom@codesourcery.com>
9562 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9563 (nvptx_process_pars): Emit bar.sync asap and alap.
9565 2018-03-20 Tom de Vries <tom@codesourcery.com>
9568 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9569 seen_label if seen_label is already set.
9571 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9574 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9575 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9576 Use 1U instead of 1. Formatting fixes.
9579 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9580 instead of TREE_TYPE (s1) for the return value.
9582 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9584 PR tree-optimization/84946
9585 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9586 bitsize + bitsize in poly_uint64 rather than poly_int64.
9589 * dwarf2asm.c: Include fold-const.c.
9590 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9591 of decl rather than decl itself.
9593 PR rtl-optimization/84643
9594 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9596 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9599 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9600 calling assemble_variable.
9602 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9605 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9606 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9608 2018-03-19 Jim Wilson <jimw@sifive.com>
9611 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9612 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9613 (riscv_first_stack_step): Likewise.
9614 (riscv_option_override): Use STACK_BOUNDARY instead of
9616 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9618 (BIGGEST_ALIGNMENT): Set to 128.
9619 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9620 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9623 2018-03-19 Richard Biener <rguenther@suse.de>
9625 PR tree-optimization/84933
9626 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9627 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9629 2018-03-19 Richard Biener <rguenther@suse.de>
9631 PR tree-optimization/84859
9632 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9633 (cond_if_else_store_replacement): Perform sinking operation on
9634 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9635 Generalize what a BB with a single eligible store is.
9637 2018-03-19 Richard Biener <rguenther@suse.de>
9639 PR tree-optimization/84929
9640 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9641 chrec_is_positive against non-chrec arg.
9643 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9646 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9648 2018-03-18 Martin Liska <mliska@suse.cz>
9650 PR rtl-optimization/84635
9651 * regrename.c (build_def_use): Use matches_mode only when
9654 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9656 PR tree-optimization/84913
9657 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9658 vectorize chains of COND_EXPRs.
9660 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9662 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9664 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9666 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9668 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9670 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9672 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9673 Kito Cheng <kito.cheng@gmail.com>
9675 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9676 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9677 (nds32_adjust_reg_alloc_order): New function.
9678 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9680 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9682 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9683 nds32_print_operand, nds32_print_operand_address): Use
9684 HOST_WIDE_INT_PRINT_DEC instead.
9686 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9688 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9690 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9693 * config/i386/i386.c (initial_ix86_tune_features,
9694 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9696 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9697 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9698 rather than 1u << ix86_tune. Formatting fix.
9699 (ix86_option_override_internal): Change ix86_arch_mask from
9700 unsigned int to unsigned HOST_WIDE_INT, initialize to
9701 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9702 (ix86_function_specific_restore): Likewise.
9704 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9707 * postreload.c (reload_combine_recognize_pattern): Perform
9708 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9709 truncate_int_for_mode the result for the destination's mode.
9712 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9713 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9715 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9718 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9721 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9722 regno_allocno_class_array and sorted_pseudos.
9723 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9724 insns where regno is used.
9726 2018-03-16 Martin Liska <mliska@suse.cz>
9729 * multiple_target.c (create_dispatcher_calls): Redirect
9730 reference in the symbol table.
9732 2018-03-16 Martin Liska <mliska@suse.cz>
9735 * multiple_target.c (create_dispatcher_calls): Redirect also
9738 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9742 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9743 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9744 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9746 2018-03-16 Julia Koval <julia.koval@intel.com>
9748 * doc/invoke.texi (Skylake Server): Add CLWB.
9749 Cannonlake): Remove CLWB.
9751 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9753 PR tree-optimization/84841
9754 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9756 (FLOAT_ONE_CONST_TYPE): Define.
9757 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9758 (sort_by_operand_rank): Put entries with higher constant_type last
9759 rather than first to match comments.
9761 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9763 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9766 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9769 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9770 check and use address of target always.
9772 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9775 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9776 (indirect_thunk_bnd_needed): Likewise.
9777 (indirect_thunks_used): Likewise.
9778 (indirect_thunks_bnd_used): Likewise.
9779 (indirect_return_needed): New.
9780 (indirect_return_bnd_needed): Likewise.
9781 (output_indirect_thunk_function): Add a bool argument for
9783 (output_indirect_thunk_function): Don't generate alias for
9784 function return thunk.
9785 (ix86_code_end): Call output_indirect_thunk_function to generate
9786 function return thunks.
9787 (ix86_output_function_return): Set indirect_return_bnd_needed
9788 and indirect_return_needed instead of indirect_thunk_bnd_needed
9789 and indirect_thunk_needed.
9791 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9793 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9794 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9795 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9797 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9798 Paul Hua <paul.hua.gm@gmail.com>
9801 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9803 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9805 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9806 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9809 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9812 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9813 instead of GET_MODE_SIZE when comparing Units.
9815 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9818 * varasm.c (hash_section): Return an unchangeble hash value
9819 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9820 Return !aarch64_can_use_per_function_literal_pools_p ().
9822 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9825 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9826 rather than address of cmode as last argument to prepare_cmp_insn.
9828 2018-03-15 Julia Koval <julia.koval@intel.com>
9830 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9831 F_AVX512VNNI, F_AVX512BITALG): New.
9833 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9836 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9837 insn for floating-point loads and stores.
9839 2018-03-14 Carl Love <cel@us.ibm.com>
9841 * config/rs6000/rs6000-c.c: Add macro definitions for
9842 ALTIVEC_BUILTIN_VEC_PERMXOR.
9843 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9844 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9845 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9846 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9848 * config/doc/extend.texi: Add prototypes for vec_permxor.
9850 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9853 * diagnostic-show-locus.c (class layout_point): Convert m_line
9854 from int to linenum_type.
9855 (line_span::comparator): Use linenum "compare" function when
9856 comparing line numbers.
9857 (test_line_span): New function.
9858 (layout_range::contains_point): Convert param "row" from int to
9860 (layout_range::intersects_line_p): Likewise.
9861 (layout::will_show_line_p): Likewise.
9862 (layout::print_source_line): Likewise.
9863 (layout::should_print_annotation_line_p): Likewise.
9864 (layout::print_annotation_line): Likewise.
9865 (layout::print_leading_fixits): Likewise.
9866 (layout::annotation_line_showed_range_p): Likewise.
9867 (struct line_corrections): Likewise for field m_row.
9868 (line_corrections::line_corrections): Likewise for param "row".
9869 (layout::print_trailing_fixits): Likewise.
9870 (layout::get_state_at_point): Likewise.
9871 (layout::get_x_bound_for_row): Likewise.
9872 (layout::print_line): Likewise.
9873 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9874 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9875 * input.c (selftest::test_linenum_comparisons): New function.
9876 (selftest::input_c_tests): Call it.
9877 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9878 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9879 * selftest.h (ASSERT_GT): New macro.
9880 (ASSERT_GT_AT): New macro.
9881 (ASSERT_LT): New macro.
9882 (ASSERT_LT_AT): New macro.
9884 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9886 PR rtl-optimization/84780
9887 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9889 2018-03-14 Martin Liska <mliska@suse.cz>
9891 * tree.c (record_node_allocation_statistics): Use
9892 get_stats_node_kind.
9893 (get_stats_node_kind): New function extracted from
9894 record_node_allocation_statistics.
9895 (free_node): Use get_stats_node_kind.
9897 2018-03-14 Richard Biener <rguenther@suse.de>
9899 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9900 that the value-set of ANTIC_IN doesn't grow.
9903 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9905 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9906 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9908 2018-03-14 Julia Koval <julia.koval@intel.com>
9910 * config.gcc (icelake-client, icelake-server): New.
9912 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9913 (initial_ix86_arch_features): Ditto.
9914 (PTA_SKYLAKE): Add SGX.
9915 (PTA_ICELAKE): Remove.
9916 (PTA_ICELAKE_CLIENT): New.
9917 (PTA_ICELAKE_SERVER): New.
9918 (ix86_option_override_internal): Split up icelake on icelake client and
9920 (get_builtin_code_for_version): Ditto.
9921 (fold_builtin_cpu): Ditto.
9922 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9923 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9924 * config/i386/i386.h (processor_type): Ditto.
9925 * doc/invoke.texi: Ditto.
9927 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9930 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9931 INTEGER_CST offset, add it together with bitpos / 8 and
9932 sign extend based on POINTER_SIZE.
9936 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9939 * config/i386/constraints.md (Yc): New register constraint.
9940 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9941 Use Yc constraint for alternative 2 of operand 0. Remove
9942 preferred_for_speed attribute.
9944 2018-03-14 Richard Biener <rguenther@suse.de>
9946 PR tree-optimization/84830
9947 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9948 with the old one to avoid oscillations.
9950 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9953 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9955 (assign_by_spills): Return a flag of reload assignment failure.
9956 Do not process the reload assignment failures. Do not spill other
9957 reload pseudos if they has the same reg class. Update n if
9959 (lra_assign): Add a return arg. Set up from the result of
9960 assign_by_spills call.
9961 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9962 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9963 usage_insns if it is not NULL.
9964 (spill_hard_reg_in_range): New function.
9965 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9966 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9967 function prototypes.
9968 (lra_assign): Change prototype.
9969 * lra.c (lra): Add code to deal with fails by splitting hard reg
9972 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9974 * config/riscv/riscv.opt (mrelax): New option.
9975 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9976 "norelax" when riscv_mrelax is disabled.
9977 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9979 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9982 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9983 reassociation for int modes.
9985 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9987 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9988 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9990 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9991 * config/aarch64/aarch64-sve.md
9992 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9993 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9994 (*extend<mode><Vwide>2): Rename to...
9995 (aarch64_sve_extend<mode><Vwide>2): ...this.
9996 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9997 renaming the old pattern to...
9998 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10000 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10001 define_expand, renaming the old pattern to...
10002 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10003 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10004 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10005 account when deciding which SVE instruction the optab should use.
10006 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10008 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10010 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10011 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10012 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10013 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10014 (tlsdesc_small_<mode>): Turn a define_expand and use
10015 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10016 (tlsdesc_small_advsimd_<mode>): ...this.
10017 (tlsdesc_small_sve_<mode>): New pattern.
10019 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10021 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10022 (UNSPEC_UMUL_HIGHPART): New constants.
10023 (MUL_HIGHPART): New int iteraor.
10024 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10025 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10027 (*<su>mul<mode>3_highpart): New define_insn.
10029 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10032 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10035 2018-03-13 Martin Liska <mliska@suse.cz>
10038 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10040 (sem_item_optimizer::~sem_item_optimizer): Release it.
10041 (sem_item_optimizer::merge_classes): Register variable aliases.
10042 (sem_item_optimizer::fixup_pt_set): New function.
10043 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10044 * ipa-icf.h: Declare new variables and functions.
10046 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10048 PR middle-end/84834
10049 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10050 integer_pow2p@2 and test integer_pow2p in condition.
10051 (A < 0 ? C : 0): Similarly for @1.
10053 PR middle-end/84831
10054 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10055 characters starting at p contain '\0' character, don't look beyond
10059 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10060 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10063 * reg-stack.c (change_stack): Change update_end var from int to
10064 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10065 also call set_block_for_insn on the newly added insns and rescan.
10068 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10069 on the last operand.
10072 * tree.c (stabilize_reference_1): Return save_expr (e) for
10073 STATEMENT_LIST even if it doesn't have side-effects.
10075 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10077 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10079 2018-03-12 Renlin Li <renlin.li@arm.com>
10081 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10082 aarch64_output_scalar_simd_mov_immediate.
10084 2018-03-12 Martin Sebor <msebor@redhat.com>
10086 PR tree-optimization/83456
10087 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10088 for perfectly overlapping calls to memcpy.
10089 (gimple_fold_builtin_memory_chk): Same.
10090 (gimple_fold_builtin_strcpy): Handle no-warning.
10091 (gimple_fold_builtin_stxcpy_chk): Same.
10092 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10094 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10096 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10097 parameter. Use it for SFmode.
10098 (rs6000_function_arg_advance_1): Adjust.
10099 (rs6000_function_arg): Adjust.
10100 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10102 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10104 PR rtl-optimization/84169
10105 PR rtl-optimization/84780
10106 * combine.c (can_combine_p): Check for a 2-insn combination whether
10107 the destination register is used between the two insns, too.
10109 2018-03-12 Richard Biener <rguenther@suse.de>
10111 PR tree-optimization/84803
10112 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10113 for refs DR analysis didn't process.
10115 2018-03-12 Richard Biener <rguenther@suse.de>
10117 PR tree-optimization/84777
10118 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10119 force-vectorize loops ignore whether we are optimizing for size.
10121 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10123 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10124 (TARGET_MD_ASM_ADJUST): Define.
10126 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10127 Kito Cheng <kito.cheng@gmail.com>
10128 Chung-Ju Wu <jasonwucj@gmail.com>
10130 * config/nds32/nds32.c (nds32_compute_stack_frame,
10131 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10132 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10133 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10134 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10135 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10136 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10137 * config/nds32/nds32.md (prologue, epilogue): Use macro
10138 NDS32_V3PUSH_AVAILABLE_P to do checking.
10140 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10143 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10144 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10145 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10146 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10147 addition of most attributes on !orig_type_die or the attribute not
10148 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10150 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10151 Chung-Ju Wu <jasonwucj@gmail.com>
10153 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10154 __NDS32_VH__ macro.
10155 * config/nds32/nds32.opt (mvh): New option.
10157 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10158 Chung-Ju Wu <jasonwucj@gmail.com>
10160 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10162 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10163 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10166 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10167 Chung-Ju Wu <jasonwucj@gmail.com>
10169 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10171 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10172 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10174 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10175 Kito Cheng <kito.cheng@gmail.com>
10176 Chung-Ju Wu <jasonwucj@gmail.com>
10178 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10179 UNSPEC_FFMISM and UNSPEC_FLMISM.
10180 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10181 for ffb, ffmism and flmism.
10182 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10183 (unspec_ffmism): Ditto.
10184 (unspec_flmism): Ditto.
10185 (nds32_expand_builtin_impl): Check if string extension is available.
10186 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10187 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10189 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10192 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10195 * lra-assigns.c (assign_by_spills): Return a flag of reload
10196 assignment failure. Do not process the reload assignment
10197 failures. Do not spill other reload pseudos if they has the same
10199 (lra_assign): Add a return arg. Set up from the result of
10200 assign_by_spills call.
10201 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10202 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10203 usage_insns if it is not NULL.
10204 (spill_hard_reg_in_range): New function.
10205 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10206 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10207 function prototypes.
10208 (lra_assign): Change prototype.
10209 * lra.c (lra): Add code to deal with fails by splitting hard reg
10212 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10215 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10217 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10220 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10221 (dw_val_node): Add val_symbolic_view.
10222 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10223 (symview_upper_bound): New.
10224 (new_line_info_table): Initialize symviews_since_reset.
10225 (dwarf2out_source_line): Count symviews_since_reset and set
10226 symview_upper_bound.
10227 (dw_val_equal_p): Handle symview.
10228 (add_AT_symview): New.
10229 (print_dw_val): Handle symview.
10230 (attr_checksum, attr_checksum_ordered): Likewise.
10231 (same_dw_val_p, size_of_die): Likewise.
10232 (value_format, output_die): Likewise.
10233 (add_high_low_attributes): Use add_AT_symview for entry_view.
10234 (dwarf2out_finish): Reset symview_upper_bound, clear
10237 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10240 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10241 Add strict argument and use it.
10242 (rs6000_split_multireg_move): Update for new strict argument.
10243 (mem_operand_gpr): Disallow all non-offsettable addresses.
10244 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10246 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10249 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10250 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10251 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10254 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10255 decl, use remap_type if we want to use the type.
10257 2018-03-09 Martin Sebor <msebor@redhat.com>
10259 PR tree-optimization/84526
10260 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10262 (builtin_access::generic_overlap): Be prepared to handle non-array
10265 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10267 PR rtl-optimization/84682
10268 * lra-constraints.c (process_address_1): Check is_address flag
10269 for address constraints.
10270 (process_alt_operands): Likewise.
10271 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10272 preprocess_constraints.
10273 * recog.h (preprocess_constraints): Add oploc parameter.
10275 * recog.c (preprocess_constraints): Test address_operand for
10276 CT_ADDRESS constraints.
10278 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10281 * lra-assigns.c (assign_by_spills): Return a flag of reload
10282 assignment failure. Do not process the reload assignment
10283 failures. Do not spill other reload pseudos if they has the same
10285 (lra_assign): Add a return arg. Set up from the result of
10286 assign_by_spills call.
10287 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10288 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10289 usage_insns if it is not NULL.
10290 (spill_hard_reg_in_range): New function.
10291 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10292 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10293 function prototypes.
10294 (lra_assign): Change prototype.
10295 * lra.c (lra): Add code to deal with fails by splitting hard reg
10298 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10301 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10302 Accept complain bool parameter. Only emit errors if it is true.
10303 (arm_parse_cpu_option_name): Likewise.
10304 (arm_target_thumb_only): Adjust callers of the above.
10305 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10306 prototype to take a default true bool parameter.
10307 (arm_parse_arch_option_name): Likewise.
10309 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10310 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10314 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10315 * configure: Regenerate.
10316 * configure.ac ("linker --version-script option"): New.
10317 ("linker soname option"): New.
10319 2018-03-09 Richard Biener <rguenther@suse.de>
10321 PR tree-optimization/84775
10322 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10323 immediate uses of predicate stmts and mark them modified.
10326 PR tree-optimization/84178
10327 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10329 (version_loop_for_if_conversion): Delay update_ssa call.
10330 (tree_if_conversion): Delay update_ssa until after predicate
10333 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10336 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10337 when the function accesses prior frames.
10339 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10342 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10343 gen_llsym, otherwise call maybe_gen_llsym.
10345 PR inline-asm/84742
10346 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10347 has ',' character inside of it.
10349 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10352 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10353 as clobbering CC_REGNUM.
10355 2018-03-08 Richard Biener <rguenther@suse.de>
10357 PR middle-end/84552
10358 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10359 (follow_copies_to_constant): Do not follow SSA names registered
10362 2018-03-08 Richard Biener <rguenther@suse.de>
10364 PR tree-optimization/84178
10365 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10367 (version_loop_for_if_conversion): Delay update_ssa call.
10368 (tree_if_conversion): Delay update_ssa until after predicate
10371 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10373 PR tree-optimization/84178
10374 * tree-if-conv.c (release_bb_predicate): Remove the
10375 the assertion that the stmts have NULL use_ops.
10376 Discard the statements, asserting that they haven't
10377 yet been added to a BB.
10379 2018-03-08 Richard Biener <rguenther@suse.de>
10381 PR tree-optimization/84746
10382 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10383 (phi_translate): Pass in destination ANTIC_OUT set.
10384 (phi_translate_1): Likewise. For a simplified result lookup
10385 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10386 (phi_translate_set): Adjust.
10387 (do_pre_regular_insertion): Likewise.
10388 (do_pre_partial_partial_insertion): Likewise.
10390 2018-03-08 Martin Liska <mliska@suse.cz>
10392 PR gcov-profile/84735
10393 * doc/gcov.texi: Document usage of profile files.
10394 * gcov-io.h: Document changes in the format.
10396 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10400 * dwarf2out.c (struct dw_line_info_table): Update comments for
10402 (FORCE_RESET_NEXT_VIEW): New.
10403 (FORCE_RESETTING_VIEW_P): New.
10404 (RESETTING_VIEW_P): Check for -1 too.
10405 (ZERO_VIEW_P): Likewise.
10406 (new_line_info_table): Force-reset next view.
10407 (dwarf2out_begin_function): Likewise.
10408 (dwarf2out_source_line): Simplify zero_view_p initialization.
10409 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10410 view directly. Omit view when omitting .loc at line 0.
10412 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10414 PR tree-optimization/84740
10415 * tree-switch-conversion.c (process_switch): Call build_constructors
10416 only if info.phi_count is non-zero.
10418 PR tree-optimization/84739
10419 * tree-tailcall.c (find_tail_calls): Check call arguments against
10420 DECL_ARGUMENTS (current_function_decl) rather than
10421 DECL_ARGUMENTS (func) when checking for tail recursion.
10423 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10425 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10426 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10427 Volker Reichelt's entry and add entries for people that perform
10428 GCC fuzzy testing and report numerous bugs.
10430 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10433 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10434 readonly data in sdata, if that is disabled.
10435 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10436 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10437 -mreadonly-in-sdata option.
10439 2018-03-07 Martin Sebor <msebor@redhat.com>
10441 PR tree-optimization/84468
10442 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10443 basic block when looking for nul assignment.
10445 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10448 * except.h (output_function_exception_table): Adjust prototype.
10449 * except.c (output_function_exception_table): Remove FNNAME parameter
10450 and add SECTION parameter. Ouput one part of the table at a time.
10451 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10452 the first part of the exception table and emit unwind directives.
10453 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10454 (i386_pe_seh_cold_init): Likewise.
10455 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10456 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10457 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10458 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10459 * config/i386/winnt.c: Include except.h.
10460 (struct seh_frame_state): Add reg_offset, after_prologue and
10461 in_cold_section fields.
10462 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10463 (i386_pe_seh_cold_init): New function.
10464 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10465 to seh->in_cold_section.
10466 (seh_emit_push): Record the offset of the push.
10467 (seh_emit_save): Record the offet of the save.
10468 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10469 Test seh->after_prologue to disregard the epilogue.
10470 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10471 (i386_pe_end_cold_function): New function.
10473 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10476 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10477 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10480 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10482 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10483 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10485 PR middle-end/84723
10486 * multiple_target.c: Include tree-inline.h and intl.h.
10487 (expand_target_clones): Diagnose and fail if node->definition and
10488 !tree_versionable_function_p (node->decl).
10490 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10492 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10494 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10495 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10496 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10498 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10501 * combine.c (try_combine): Use reg_or_subregno instead of handling
10502 just paradoxical SUBREGs and REGs.
10504 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10506 * config/arc/arc.c (arc_finalize_pic): Remove function.
10507 (arc_must_save_register): We use single base PIC register, remove
10508 checks to save/restore the PIC register.
10509 (arc_expand_prologue): Likewise.
10510 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10512 (arc_verify_short): Likewise.
10513 (arc_attr_type): Likewise.
10514 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10515 (walk_stores): Likewise.
10516 (arc_address_cost): Make it static.
10517 (arc_verify_short): Likewise.
10518 (branch_dest): Likewise.
10519 (arc_attr_type): Likewise.
10520 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10521 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10522 (arc_final_prescan_insn): Remove inserting the nops due to
10523 hardware hazards. It is done in reorg step.
10524 (insn_length_variant_t): Remove.
10525 (insn_length_parameters_t): Likewise.
10526 (arc_insn_length_parameters): Likewise.
10527 (arc_get_insn_variants): Likewise.
10528 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10530 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10532 PR inline-asm/84683
10533 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10536 PR tree-optimization/84687
10537 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10539 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10541 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10543 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10544 Rename to ppc_speculation_barrier.
10545 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10546 __builtin_ppc_speculation_barrier.
10548 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10551 * combine.c (combine_simplify_rtx): Don't try to simplify if
10552 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10555 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10557 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10558 to 32 bytes when compiling for POWER9.
10560 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10563 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10564 regparm >= 3 with no arg reg available also for calls with
10565 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10568 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10570 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10572 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10575 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10577 2018-03-05 Richard Biener <rguenther@suse.de>
10579 PR tree-optimization/84486
10580 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10581 When inserting a __builtin_assume_aligned call set the LHS
10582 SSA name alignment info accordingly.
10584 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10586 PR tree-optimization/84114
10587 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10588 Avoid reassociation of FLOAT_MODE addition.
10590 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10592 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10593 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10594 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10595 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10596 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10597 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10598 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10600 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10601 __builtin_ia32_wbinvd): New builtins.
10602 (SPECIAL_ARGS2): New.
10603 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10604 (SPECIAL_ARGS2): New.
10605 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10606 (ix86_valid_target_attribute_inner_p): Ditto.
10607 (ix86_init_mmx_sse_builtins): Add special_args2.
10608 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10609 TARGET_WBNOINVD_P): New.
10610 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10611 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10612 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10613 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10614 * config/i386/pconfigintrin.h: New file.
10615 * config/i386/wbnoinvdintrin.h: Ditto.
10616 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10618 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10620 2018-03-05 Richard Biener <rguenther@suse.de>
10622 PR tree-optimization/84670
10623 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10625 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10626 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10627 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10628 doesn't grow if all successors (recursively) were visited at least
10631 2018-03-05 Richard Biener <rguenther@suse.de>
10633 PR tree-optimization/84650
10634 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10635 if executed in the loop pipeline.
10637 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10639 * doc/configfiles.texi (Configuration Files): Move info about
10640 conditionalizing $target-protos.h to...
10641 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10642 differs from $target-protos.h.
10644 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10645 Chung-Ju Wu <jasonwucj@gmail.com>
10647 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10648 * config/nds32/nds32-multiple.md (setmemsi): Define.
10649 * config/nds32/nds32-memory-manipulation.c
10650 (nds32_gen_dup_4_byte_to_word_value): New.
10651 (emit_setmem_word_loop): New.
10652 (emit_setmem_byte_loop): New.
10653 (nds32_expand_setmem_loop): New.
10654 (nds32_expand_setmem_loop_v3m): New.
10655 (nds32_expand_setmem_unroll): New.
10656 (nds32_expand_setmem): New.
10658 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10659 Chung-Ju Wu <jasonwucj@gmail.com>
10661 * config/nds32/nds32-memory-manipulation.c
10662 (nds32_emit_load_store): New.
10663 (nds32_emit_post_inc_load_store): New.
10664 (nds32_emit_mem_move): New.
10665 (nds32_emit_mem_move_block): New.
10666 (nds32_expand_movmemsi_loop_unknown_size): New.
10667 (nds32_expand_movmemsi_loop_known_size): New.
10668 (nds32_expand_movmemsi_loop): New.
10669 (nds32_expand_movmemsi_unroll): New.
10670 (nds32_expand_movmemqi): Rename ...
10671 (nds32_expand_movmemsi): ... to this.
10672 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10673 (movmemsi): ... to this.
10674 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10675 (nds32_expand_movmemsi): ... to this.
10677 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10678 Monk Chiang <sh.chiang04@gmail.com>
10679 Chung-Ju Wu <jasonwucj@gmail.com>
10681 * config/nds32/nds32-protos.h
10682 (nds32_expand_load_multiple): New arguments.
10683 (nds32_expand_store_multiple): Ditto.
10684 (nds32_valid_multiple_load_store): Rename ...
10685 (nds32_valid_multiple_load_store_p): ... to this.
10686 * config/nds32/nds32-memory-manipulation.c
10687 (nds32_expand_load_multiple): Refine implementation.
10688 (nds32_expand_store_multiple): Ditto.
10689 * config/nds32/nds32-multiple.md
10690 (load_multiple): Update nds32_expand_load_multiple interface.
10691 (store_multiple): Update nds32_expand_store_multiple interface.
10692 * config/nds32/nds32-predicates.c
10693 (nds32_valid_multiple_load_store): Rename ...
10694 (nds32_valid_multiple_load_store_p): ... to this and refine
10696 * config/nds32/predicates.md
10697 (nds32_load_multiple_and_update_address_operation): New predicate.
10698 (nds32_store_multiple_and_update_address_operation): New predicate.
10700 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10701 Chung-Ju Wu <jasonwucj@gmail.com>
10703 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10704 (combo): New attribute.
10705 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10707 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10709 * config/nds32/nds32.opt: Change -mcmodel= default value.
10711 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10712 Monk Chiang <sh.chiang04@gmail.com>
10713 Chung-Ju Wu <jasonwucj@gmail.com>
10715 * config/nds32/constants.md (unspec_element): New enum.
10716 * config/nds32/constraints.md (Umw): New constraint.
10717 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10718 * config/nds32/nds32-intrinsic.md: Likewise.
10719 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10720 (nds32_valid_smw_lwm_base_p): New.
10721 (nds32_output_smw_single_word): New.
10722 (nds32_output_lmw_single_word): New.
10723 (nds32_expand_unaligned_load): New.
10724 (nds32_expand_unaligned_store): New.
10725 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10726 (nds32_output_smw_single_word): Declare.
10727 (nds32_output_lmw_single_word): Declare.
10728 (nds32_expand_unaligned_load): Declare.
10729 (nds32_expand_unaligned_store): Declare.
10730 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10731 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10732 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10733 NDS32_BUILTIN_UASTORE_DW.
10734 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10737 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10738 Kito Cheng <kito.cheng@gmail.com>
10739 Chung-Ju Wu <jasonwucj@gmail.com>
10741 * config/nds32/nds32-intrinsic.c
10742 (nds32_expand_builtin_null_ftype_reg): Delete.
10743 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10744 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10745 (nds32_read_argument): New.
10746 (nds32_legitimize_target): Ditto.
10747 (nds32_legitimize_argument): Ditto.
10748 (nds32_check_constant_argument): Ditto.
10749 (nds32_expand_unop_builtin): Ditto.
10750 (nds32_expand_unopimm_builtin): Ditto.
10751 (nds32_expand_binop_builtin): Ditto.
10752 (nds32_builtin_decl_impl): Ditto.
10753 (builtin_description): Ditto.
10754 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10755 (nds32_init_builtins_impl): Ditto.
10756 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10757 (nds32_builtin_decl): New.
10758 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10759 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10761 2018-03-02 Jeff Law <law@redhat.com>
10763 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10764 (redundant_insn, fill_simple_delay_slots): Likewise.
10765 (fill_slots_from_thread): Likewise.
10766 * resource.c (mark_referenced_resources): Likewise.
10767 (mark_set_resources, find_dead_or_set_registers): Likewise.
10769 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10771 * substring-locations.h (format_warning_va): Formatting fix for
10772 ATTRIBUTE_GCC_DIAG.
10773 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10775 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10776 * substring-locations.c: Include intl.h.
10777 (format_warning_va): Turned into small wrapper around
10778 format_warning_n_va, renamed to ...
10779 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10780 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10782 (format_warning_at_substring_n): New function.
10783 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10784 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10785 format_warning_at_substring with just a shorter name instead of
10786 const function pointer.
10787 (fmtwarn_n): New function.
10788 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10789 appropriate, get rid of all the fmtstr temporaries, move conditionals
10790 with G_() wrapped string literals directly into fmtwarn arguments,
10791 cast dir.len to (int), formatting fixes.
10793 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10795 * doc/invoke.texi: Remove "Cilk Plus" references.
10797 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10798 Richard Biener <rguenther@suse.de>
10801 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10802 for error or warning attributes if CALL_FROM_THUNK_P is set.
10805 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10808 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10809 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10812 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10813 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10815 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10816 instead of -1U in last predictors element's probability member.
10818 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10821 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10822 arguments if they are comparable.
10824 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10826 PR tree-optimization/84634
10827 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10828 masks and masked_loop_p with a single loop_masks, making sure it's
10829 null for bb vectorization.
10831 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10833 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10834 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10835 loop->force_vectorize to check whether there is no alias.
10837 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10840 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10842 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10844 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10845 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10846 instead of a loop around prev_real_insn.
10847 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10850 PR inline-asm/84625
10851 * config/i386/i386.c (ix86_print_operand): Use conditional
10852 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10855 2018-03-02 Richard Biener <rguenther@suse.de>
10857 PR tree-optimization/84427
10858 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10859 (bitmap_set_subtract_values): Rewrite to handle multiple
10862 (prune_clobbered_mems): Likewise.
10863 (phi_translate): Take edge instead of pred/phiblock.
10864 (phi_translate_1): Likewise.
10865 (phi_translate_set): Likewise. Insert all translated
10866 exprs for a value into the set, keeping possibly multiple
10867 expressions per value.
10868 (compute_antic_aux): Adjust for phi_translate changes.
10869 When intersecting union the expressions and prune those
10870 not in the final value set, keeping possibly multiple
10871 expressions per value. Do not use value-insertion
10872 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10873 all expressions. Add verification that the value-sets
10874 only shrink during iteration.
10875 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10876 (do_pre_regular_insertion): Likewise.
10877 (do_pre_partial_partial_insertion): Likewise.
10879 2018-03-02 Richard Biener <rguenther@suse.de>
10882 * config/darwin.c (saved_debug_info_level): New static global.
10883 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10884 (darwin_asm_lto_end): Restore debug info generation settings.
10886 2018-03-01 Martin Liska <mliska@suse.cz>
10889 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10890 volatile arguments.
10892 2018-03-01 Richard Biener <rguenther@suse.de>
10895 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10896 type annotation with LTO when debug was disabled at compile-time.
10898 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10900 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10902 (mips_final_postscan_insn): Likewise.
10904 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10906 PR rtl-optimization/84528
10907 * alias.c (init_alias_target): Add commentary.
10908 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10909 a unique base value if the frame pointer is not eliminated
10910 to the stack pointer.
10912 2018-03-01 Tom de Vries <tom@codesourcery.com>
10914 PR rtl-optimization/83327
10915 * lra-int.h (hard_regs_spilled_into): Declare.
10916 * lra.c (hard_regs_spilled_into): Define.
10917 (init_reg_info): Init hard_regs_spilled_into.
10918 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10919 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10920 (process_bb_lives): Handle hard_regs_spilled_into.
10921 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10922 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10924 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10926 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10927 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10928 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10929 * config/rs6000/aix72.h: New file.
10931 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10933 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10934 instead of warning_at with conditional singular and plural messages
10938 * stor-layout.c (update_alignment_for_field): For
10939 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10940 && !DECL_PACKED (field), do the alignment update, just use
10941 only desired_align instead of MAX (type_align, desired_align)
10943 (place_field): Don't do known_align < desired_align handling
10944 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10945 is non-NULL, instead do it after rli->prev_field handling and
10946 only if not within a bitfield word. For DECL_PACKED (field)
10947 use type_align of BITS_PER_UNIT.
10949 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10951 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10952 superfluous parentheses and trailing spaces.
10954 2018-02-28 Richard Biener <rguenther@suse.de>
10956 PR tree-optimization/84584
10957 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10958 SCoPs with fake exit edge.
10960 2018-02-28 Martin Liska <mliska@suse.cz>
10963 * timevar.c (timer::print): Fix format to properly print 100%
10966 2018-02-28 Richard Biener <rguenther@suse.de>
10968 PR middle-end/84607
10969 * genmatch.c (capture_info::walk_match): Do not mark
10970 captured expressions without operands as expr_p given
10971 they act more like predicates and should be subject to
10972 "lost tail" side-effect preserving.
10974 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10976 PR rtl-optimization/81611
10977 * auto-inc-dec.c (attempt_change): Move dead note from
10978 mem_insn if it's the next use of regno
10979 (find_address): Take address use of reg holding
10980 non-incremented value. Add parm to limit search to the named
10982 (merge_in_block): Attempt to use a mem insn that is the next
10983 use of the original regno.
10985 2018-02-27 Martin Sebor <msebor@redhat.com>
10988 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10989 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10991 2018-02-27 Martin Sebor <msebor@redhat.com>
10993 PR translation/84207
10994 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10995 n argument to unsigned HOST_WIDE_INT.
10996 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10997 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10998 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10999 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11001 2018-02-27 Richard Biener <rguenther@suse.de>
11003 PR tree-optimization/84512
11004 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11005 Do not use the estimate returned from record_stmt_cost for
11006 the scalar iteration cost but sum properly using add_stmt_cost.
11008 2018-02-27 Richard Biener <rguenther@suse.de>
11010 PR tree-optimization/84466
11011 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11012 Adjust last change to less strictly validate use operands.
11014 2018-02-27 Martin Liska <mliska@suse.cz>
11016 PR gcov-profile/84548
11017 * gcov.c (process_file): Allow partial overlap and consider it
11018 also as group functions.
11019 (output_lines): Properly calculate range of lines for a group.
11021 2018-02-27 Martin Liska <mliska@suse.cz>
11023 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11024 'ggc' suffixes. Change first column width.
11025 (timer::print): Fix formatting of the column.
11027 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11029 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11030 preserve inline entry blocks for the sake of debug inline
11031 entry point markers alone.
11032 (remove_unused_locals): Suggest in comments a better place to
11033 force the preservation of inline entry blocks that are
11034 otherwise unused, but do not preserve them.
11036 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11038 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11040 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11043 * config/i386/constraints.md (Bs): Replace
11044 ix86_indirect_branch_register with
11045 TARGET_INDIRECT_BRANCH_REGISTER.
11047 * config/i386/i386.md (indirect_jump): Likewise.
11048 (tablejump): Likewise.
11049 (*sibcall_memory): Likewise.
11050 (*sibcall_value_memory): Likewise.
11051 Peepholes of indirect call and jump via memory: Likewise.
11052 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11053 (*sibcall_value_GOT_32): Likewise.
11054 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11055 (GOT_memory_operand): Likewise.
11056 (call_insn_operand): Likewise.
11057 (sibcall_insn_operand): Likewise.
11058 (GOT32_symbol_operand): Likewise.
11059 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11061 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11063 PR rtl-optimization/83496
11064 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11065 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11066 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11067 redundant insn, if any.
11068 (relax_delay_slots): Likewise.
11069 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11071 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11073 PR tree-optimization/83965
11074 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11075 that grouped statements are part of a reduction chain. Return
11076 true if the statement is not marked as a reduction itself but
11077 is part of a group.
11078 (vect_recog_dot_prod_pattern): Don't check whether the statement
11079 is part of a group here.
11080 (vect_recog_sad_pattern): Likewise.
11081 (vect_recog_widen_sum_pattern): Likewise.
11083 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11086 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11088 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11091 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11093 (ix86_output_indirect_function_return): New prototype.
11094 (ix86_split_simple_return_pop_internal): Likewise.
11095 * config/i386/i386.c (indirect_return_via_cx): New.
11096 (indirect_return_via_cx_bnd): Likewise.
11097 (indirect_thunk_name): Handle return va CX_REG.
11098 (output_indirect_thunk_function): Create alias for
11099 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11100 (ix86_output_indirect_jmp): Remove the bool argument.
11101 (ix86_output_indirect_function_return): New function.
11102 (ix86_split_simple_return_pop_internal): Likewise.
11103 * config/i386/i386.md (*indirect_jump): Don't pass false
11104 to ix86_output_indirect_jmp.
11105 (*tablejump_1): Likewise.
11106 (simple_return_pop_internal): Change it to define_insn_and_split.
11107 Call ix86_split_simple_return_pop_internal to split it for
11108 -mfunction-return=.
11109 (simple_return_indirect_internal): Call
11110 ix86_output_indirect_function_return instead of
11111 ix86_output_indirect_jmp.
11113 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11116 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11117 memset and value initialization afterwards.
11119 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11121 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11123 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11126 * common/config/aarch64/aarch64-common.c
11127 (aarch_option_optimization_table[]): Switch
11128 off fomit-frame-pointer
11130 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11131 Chung-Ju Wu <jasonwucj@gmail.com>
11133 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11135 (store_multiple): Ditto.
11137 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11139 * config.gcc: Add --with-cpu support for nds32 target.
11140 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11141 * config/nds32/nds32.opt: Add -mcpu= option.
11143 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11145 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11146 isel=yes): Warn for these deprecated options.
11148 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11150 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11151 ISA_2_5_MASKS_EMBEDDED.
11153 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11155 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11156 p->max as pointers rather than using iterative_hash_expr.
11158 2018-02-23 Carl Love <cel@us.ibm.com>
11160 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11161 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11163 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11164 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11165 P8V_BUILTIN_VEC_VUNSIGNED2.
11167 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11170 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11171 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11173 * lra-constraints.c (curr_insn_transform): Set up
11174 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11176 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11177 * lra-eliminations.c (spill_pseudos): Ditto.
11178 (process_insn_for_elimination): Ditto.
11179 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11180 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11181 LRA_NON_CLOBBERED_ALT.
11183 2018-02-22 Martin Sebor <msebor@redhat.com>
11185 PR tree-optimization/84480
11186 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11187 to maybe_diag_stxncpy_trunc. Call it.
11188 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11189 from gimple_fold_builtin_strcpy. Print inlining stack.
11190 (handle_builtin_stxncpy): Print inlining stack.
11191 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11193 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11196 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11197 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11198 and -fcheck-pointer-bounds are used together.
11199 (indirect_thunk_prefix): New enum.
11200 (indirect_thunk_need_prefix): New function.
11201 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11202 "_nt" instead of "_bnd" for NOTRACK prefix.
11203 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11204 (output_indirect_thunk_function): Likewise.
11206 (ix86_code_end): Update output_indirect_thunk_function calls.
11207 (ix86_output_indirect_branch_via_reg): Replace
11208 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11209 (ix86_output_indirect_branch_via_push): Likewise.
11210 (ix86_output_function_return): Likewise.
11211 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11212 incompatible with -fcf-protection=branch and
11213 -fcheck-pointer-bounds.
11215 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11218 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11219 Change gcc_assert call to output_operand_lossage.
11221 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11223 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11225 2018-02-22 DJ Delorie <dj@redhat.com>
11226 Sebastian Perta <sebastian.perta@renesas.com>
11227 Oleg Endo <olegendo@gcc.gnu.org>
11229 * config/rx/rx.c (rx_rtx_costs): New function.
11230 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11232 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11234 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11236 2018-02-22 Martin Liska <mliska@suse.cz>
11239 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11240 Add "native" as a possible value.
11242 2018-02-22 Martin Liska <mliska@suse.cz>
11245 * config/i386/i386.c (ix86_option_override_internal):
11246 Add "native" as a possible value for -march and -mtune.
11248 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11251 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11252 to all type variants.
11254 PR tree-optimization/84503
11255 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11256 width as info->bitpos + info->bitsize - start.
11257 (merged_store_group::merge_overlapping): Simplify width computation.
11258 (check_no_overlap): New function.
11259 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11260 start + width and last_order of the group, fail if check_no_overlap
11262 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11263 to group if check_no_overlap fails.
11265 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11267 * config/rs6000/altivec.md: Delete contraint arguments to
11268 define_expand, define_split, and define_peephole2, and in
11269 define_insn_and_split if always unused.
11270 * config/rs6000/darwin.md: Ditto.
11271 * config/rs6000/dfp.md: Ditto.
11272 * config/rs6000/rs6000.md: Ditto.
11273 * config/rs6000/sync.md: Ditto.
11274 * config/rs6000/vector.md: Ditto.
11275 * config/rs6000/vsx.md: Ditto.
11277 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11279 * config/rs6000/altivec.md: Write output control strings as braced
11280 blocks instead of double-quoted strings.
11281 * config/rs6000/darwin.md: Ditto.
11282 * config/rs6000/rs6000.md: Ditto.
11283 * config/rs6000/vector.md: Ditto.
11284 * config/rs6000/vsx.md: Ditto.
11286 2018-02-21 Jason Merrill <jason@redhat.com>
11288 PR c++/84314 - ICE with templates and fastcall attribute.
11289 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11291 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11293 * ipa-cp.c (determine_versionability): Fix comment typos.
11295 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11298 * ipa-cp.c (determine_versionability): Do not version functions caling
11301 2018-02-21 Martin Liska <mliska@suse.cz>
11304 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11305 Add "native" as a possible value.
11306 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11307 the macro when native cpu detection is available.
11309 2018-02-21 Martin Liska <mliska@suse.cz>
11312 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11313 Add "native" as a possible value.
11314 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11315 when native cpu detection is available.
11317 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11318 Martin Sebor <msebor@redhat.com>
11320 PR tree-optimization/84478
11321 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11323 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11324 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11325 support which is conservatively correct, for 2 only stay conservative
11326 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11327 argument to the 2 argument get_range_strlen, adjust 6 arg
11328 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11330 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11331 (gimple_fold_builtin_strlen): Pass true as last argument to
11334 2018-02-20 Martin Sebor <msebor@redhat.com>
11336 PR middle-end/84095
11337 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11338 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11339 (builtin_memref::builtin_memref): Factor out parts into
11340 set_base_and_offset and call it.
11342 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11344 PR middle-end/84406
11345 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11346 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11347 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11348 search at the associated MODE_INT.
11350 2018-02-20 Jeff Law <law@redhat.com>
11352 PR middle-end/82123
11353 PR tree-optimization/81592
11354 PR middle-end/79257
11355 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11356 for range data rather than using global data.
11357 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11358 range data rather than using global data.
11359 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11360 pass it to children as needed.
11361 (struct directive::fmtresult): Similarly.
11362 (struct directive::set_width): Similarly.
11363 (struct directive::set_precision): Similarly.
11364 (format_integer, format_directive, parse_directive): Similarly.
11365 (format_none): Accept unnamed vr_values parameter.
11366 (format_percent, format_floating, format_character): Similarly.
11367 (format_string, format_plain): Similarly.
11368 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11369 the EVRP range analyzer for range data rather than using global data.
11370 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11371 gimple-ssa-evrp-analyze.h
11372 (class sprintf_dom_walker): Add after_dom_children member function.
11373 Add evrp_range_analyzer member.
11374 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11375 range analyzer as needed.
11376 (sprintf_dom_walker::after_dom_children): New member function.
11377 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11379 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11380 (evrp_range_analyzer::pop_to_marker): Likewise.
11382 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11384 PR tree-optimization/84419
11385 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11386 with the required type if its current type is compatible but
11389 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11391 PR middle-end/82004
11392 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11393 after vectorization.
11395 2018-02-20 Martin Liska <mliska@suse.cz>
11398 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11399 possible values if we don't have a hint.
11401 2018-02-20 Martin Liska <mliska@suse.cz>
11405 * final.c (shorten_branches): Build align_tab array with one
11407 * opts.c (finish_options): Add alignment option limit check.
11408 (MAX_CODE_ALIGN): Likewise.
11409 (MAX_CODE_ALIGN_VALUE): Likewise.
11410 * doc/invoke.texi: Document maximum allowed option value for
11411 all -falign-* options.
11413 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11416 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11417 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11418 * var-tracking.c (emit_note_insn_var_location): Remove all references
11419 to NOTE_INSN_CALL_ARG_LOCATION.
11420 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11421 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11422 Use copy_rtx_if_shared.
11423 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11424 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11425 (dwarf2out_var_location): Remove handling of
11426 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11428 * final.c (final_scan_insn): Remove all references to
11429 NOTE_INSN_CALL_ARG_LOCATION.
11430 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11431 before dumping final insns.
11432 * except.c (emit_note_eh_region_end): Remove all references to
11433 NOTE_INSN_CALL_ARG_LOCATION.
11434 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11435 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11436 * config/arc/arc.c (hwloop_optimize): Likewise.
11437 * config/arm/arm.c (create_fix_barrier): Likewise.
11438 * config/s390/s390.c (s390_chunkify_start): Likewise.
11439 * config/sh/sh.c (find_barrier): Likewise.
11440 * config/i386/i386.c (rest_of_insert_endbranch,
11441 ix86_seh_fixup_eh_fallthru): Likewise.
11442 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11443 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11444 * config/frv/frv.c (frv_function_prologue): Likewise.
11445 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11447 (note_outside_basic_block_p): Remove all references to
11448 NOTE_INSN_CALL_ARG_LOCATION.
11449 * gengtype.c (adjust_field_rtx_def): Likewise.
11450 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11452 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11453 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11456 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11459 PR tree-optimization/84452
11460 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11461 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11464 2018-02-19 Martin Liska <mliska@suse.cz>
11467 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11469 2018-02-19 Martin Liska <mliska@suse.cz>
11470 Richard Sandiford <richard.sandiford@linaro.org>
11472 PR tree-optimization/82491
11473 * gimple-fold.c (get_base_constructor): Make earlier bail out
11476 2018-02-19 Carl Love <cel@us.ibm.com>
11478 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11479 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11481 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11482 P8V_BUILTIN_VEC_NEG.
11484 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11486 * config/rl78/rl78.md (movdf): New define expand.
11488 2018-02-19 Martin Liska <mliska@suse.cz>
11491 * doc/invoke.texi: Fix typo.
11492 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11494 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11496 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11497 handle rs6000_single_float and rs6000_double_float specially for
11500 2018-02-16 Jeff Law <law@redhat.com>
11502 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11503 (pop_regs): Use it.
11505 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11508 * ipa-inline.c (inline_small_functions): Fix a typo.
11510 2018-02-16 Nathan Sidwell <nathan@acm.org>
11512 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11514 2018-02-16 Carl Love <cel@us.ibm.com>
11516 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11517 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11518 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11519 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11520 expansion to P8V_BUILTIN_VEC_FLOAT2.
11522 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11524 PR rtl-optimization/70023
11525 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11526 src_regno into account.
11528 2018-02-16 Carl Love <cel@us.ibm.com>
11530 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11531 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11532 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11533 * config/rs6000/rs6000.c: Remove case statements for
11534 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11535 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11536 and P9V_BUILTIN_VEC_VINSERT4B.
11537 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11538 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11539 * config/rs6000/vsx.md:
11540 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11543 2018-02-16 Carl Love <cel@us.ibm.com>
11545 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11547 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11549 * config/rs6000/rs6000-c.c: Add the definitions for
11550 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11551 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11552 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11553 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11554 definition for insert4b and define insn *insert3b_internal.
11555 * doc/extend.texi: Add documentation for vec_extract4b.
11557 2018-02-16 Nathan Sidwell <nathan@acm.org>
11559 * doc/extend.texi (Backwards Compatibility): Mention friend
11560 injection. Note for-scope is deprecated.
11561 * doc/invoke.texi (-ffriend-injection): Deprecate.
11563 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11565 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11566 that moved to I2, also allow destinations that are a paradoxical
11567 subreg (instead of a normal reg).
11569 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11572 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11575 2018-02-16 Richard Biener <rguenther@suse.de>
11577 PR tree-optimization/84037
11578 PR tree-optimization/84016
11580 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11581 Adjust vec_construct for the fact we need additional higher latency
11582 128bit inserts for AVX256 and AVX512 vector builds.
11583 (ix86_add_stmt_cost): Scale vector construction cost for
11586 2018-02-16 Richard Biener <rguenther@suse.de>
11588 PR tree-optimization/84417
11589 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11590 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11591 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11593 2018-02-16 Martin Liska <mliska@suse.cz>
11596 * internal-fn.def (ASAN_CHECK): Set proper flags.
11597 (ASAN_MARK): Likewise.
11599 2018-02-16 Julia Koval <julia.koval@intel.com>
11601 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11602 from PTA_CANNONLAKE.
11604 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11607 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11608 Use ++iter rather than iter++ for std::list iterators.
11609 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11610 defer deleting them until all nodes in the forest are processed. Do
11611 free even leaf nodes. Change to_process into auto_vec.
11614 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11615 * vec.h (vec_default_construct): Use memset instead of placement new
11616 if BROKEN_VALUE_INITIALIZATION is defined.
11617 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11618 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11621 PR rtl-optimization/83723
11622 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11623 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11624 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11626 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11628 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11630 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11632 PR rtl-optimization/81443
11633 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11634 from inner REGs to paradoxical SUBREGs.
11636 2018-02-16 Richard Biener <rguenther@suse.de>
11638 PR tree-optimization/84399
11639 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11640 For operands we can analyze at their definition make sure we can
11641 analyze them at each use as well.
11643 2018-02-16 Richard Biener <rguenther@suse.de>
11645 PR tree-optimization/84190
11646 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11647 volatile accesses if the decl isn't volatile.
11649 2018-02-15 Jason Merrill <jason@redhat.com>
11651 PR c++/84314 - ICE with templates and fastcall attribute.
11652 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11653 TYPE_CANONICAL on an existing type.
11655 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11657 PR tree-optimization/84383
11658 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11659 dstoff nor call operand_equal_p if dstbase is NULL.
11661 PR tree-optimization/84334
11662 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11663 also a CONSTANT_CLASS_P, punt.
11665 2018-02-14 Jim Wilson <jimw@sifive.com>
11667 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11668 first SMALL_OPERAND check. New local min_second_step. Move assert
11669 to where locals are set. Add TARGET_RVC support.
11670 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11672 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11674 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11676 2018-02-14 Martin Sebor <msebor@redhat.com>
11678 PR tree-optimization/83698
11679 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11680 arrays constrain the offset range to their bounds.
11681 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11682 (builtin_access::overlap): Avoid setting the size of overlap if it's
11684 (maybe_diag_overlap): Also consider arrays when deciding what values
11685 of offsets to include in diagnostics.
11687 2018-02-14 Martin Sebor <msebor@redhat.com>
11690 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11691 that correspond to the kind of a declaration.
11693 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11696 * config/pa/pa.md: Load address of PIC label using the linkage table
11697 if the label is nonlocal.
11699 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11701 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11702 warning message if user requests -maltivec=be.
11703 * doc/invoke.texi: Document deprecation of -maltivec=be.
11705 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11708 * config/rs6000/rs6000-c.c: Update definitions for
11709 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11710 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11712 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11715 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11716 add _get_ssp intrinsics. Remove argument from
11717 __builtin_ia32_rdssp[d|q].
11718 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11719 * config/i386/i386-builtin.def: Remove argument from
11720 __builtin_ia32_rdssp[d|q].
11721 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11722 ix86_expand_special_args_builtin for _rdssp[d|q].
11723 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11724 Clear register before usage.
11725 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11726 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11728 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11730 PR tree-optimization/84357
11731 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11732 operand 1 of an ARRAY_REF too.
11734 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11737 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11738 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11740 (set_of_reg): New struct.
11741 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11742 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11743 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11745 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11746 Split into bitclr, bitset, bitinvert patterns if appropriate.
11747 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11748 use rx_fuse_in_memory_bitop.
11749 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11750 to named insn, correct maximum insn length.
11752 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11755 * machmode.def: Define a complex mode for PARTIAL_INT.
11756 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11758 * doc/rtl.texi: Document CSPImode.
11759 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11761 (msp430_hard_regno_nregs_with_padding): Likewise.
11763 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11766 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11768 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11770 PR rtl-optimization/84169
11771 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11772 we generated a parallel as new i3 and we split that to new i2 and i3
11773 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11774 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11775 those to i2, not i1. Partially rewrite this scan code.
11777 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11780 * stor-layout.c (place_field): For variable length fields, adjust
11781 offset_align afterwards not just based on the field's alignment,
11782 but also on the size.
11784 PR middle-end/84309
11785 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11786 of exps and logs in the use_exp2 case.
11788 2018-02-13 Jeff Law <law@redhat.com>
11790 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11791 entry for "vector".
11793 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11796 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11800 * common.opt (gas-loc-support, gas-locview-support): New.
11801 (ginline-points, ginternal-reset-location-views): New.
11802 * doc/invoke.texi: Document them. Use @itemx where intended.
11803 (gvariable-location-views): Adjust.
11804 * target.def (reset_location_view): New.
11805 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11806 (TARGET_RESET_LOCATION_VIEW): New.
11807 * doc/tm.texi: Rebuilt.
11808 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11809 (dwarf2out_default_as_locview_support): New.
11810 (output_asm_line_debug_info): Use option variables.
11811 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11812 (output_loc_list): Likewise.
11813 (add_high_low_attributes): Check option variables.
11814 Don't output entry view attribute in strict mode.
11815 (gen_inlined_subroutine_die): Check option variables.
11816 (dwarf2out_inline_entry): Likewise.
11817 (init_sections_and_labels): Likewise.
11818 (dwarf2out_early_finish): Likewise.
11819 (maybe_reset_location_view): New, from...
11820 (dwarf2out_var_location): ... here. Call it.
11821 * debug.h (dwarf2out_default_as_loc_support): Declare.
11822 (dwarf2out_default_as_locview_support): Declare.
11823 * hooks.c (hook_int_rtx_insn_0): New.
11824 * hooks.h (hook_int_rtx_insn_0): Declare.
11825 * toplev.c (process_options): Take -gas-loc-support and
11826 -gas-locview-support from dwarf2out. Enable
11827 -gvariable-location-views by default only with locview
11828 assembler support. Enable -ginternal-reset-location-views by
11829 default only if the target defines the corresponding hook.
11830 Enable -ginline-points by default if location views are
11831 enabled; force it disabled if statement frontiers are
11833 * tree-inline.c (expand_call_inline): Check option variables.
11834 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11836 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11838 PR tree-optimization/84321
11839 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11840 handling. Also check whether the anti-range contains any values
11841 that satisfy the mask; switch to a VR_RANGE if not.
11843 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11846 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11848 2018-02-13 Martin Jambor <mjambor@suse.cz>
11851 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11852 of call statements, also set location of a load to a temporary.
11854 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11856 * config/rl78/rl78.c (add_vector_labels): New function.
11857 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11858 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11859 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11860 which checks that no arguments are passed.
11861 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11862 * doc/extend.texi: Documentation for the new attribute.
11864 2018-02-13 Andreas Schwab <schwab@suse.de>
11866 * config/riscv/linux.h (CPP_SPEC): Define.
11868 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11871 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11872 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11873 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11874 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11875 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11876 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11877 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11878 temporarily for AES and PCLMUL builtins.
11880 PR tree-optimization/84339
11881 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11882 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11885 PR middle-end/84309
11886 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11887 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11888 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11890 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11892 * omp-simd-clone.h: New file.
11893 * omp-simd-clone.c: Include omp-simd-clone.h.
11894 (expand_simd_clones): No longer static.
11895 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11896 cgraph.h and omp-simd-clone.h.
11897 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11898 (vect_recog_widen_shift_pattern): Formatting fix.
11899 (vect_pattern_recog_1): Don't check optab for calls.
11902 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11903 operands[2] into a REG before using gen_lowpart on it.
11905 2018-02-12 Jeff Law <law@redhat.com>
11908 * config/sh/sh.c (find_barrier): Consider a sibling call
11911 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11912 successfully back substituting a reg.
11914 2018-02-12 Richard Biener <rguenther@suse.de>
11916 PR tree-optimization/84037
11917 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11918 parameter, move visited init to caller.
11919 (vect_slp_analyze_operations): Separate cost from validity
11920 check, initialize visited once for all instances.
11921 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11923 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11925 (vect_model_promotion_demotion_cost): Likewise.
11926 (vectorizable_bswap): Guard cost modeling with !slp_node
11927 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11929 (vectorizable_call): Likewise.
11930 (vectorizable_conversion): Likewise.
11931 (vectorizable_assignment): Likewise.
11932 (vectorizable_shift): Likewise.
11933 (vectorizable_operation): Likewise.
11934 (vectorizable_store): Likewise.
11935 (vectorizable_load): Likewise.
11936 (vectorizable_condition): Likewise.
11937 (vectorizable_comparison): Likewise.
11939 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11942 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11943 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11944 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11946 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11948 PR middle-end/83665
11949 * params.def (inline-min-speedup): Increase from 8 to 15.
11950 (max-inline-insns-auto): Decrease from 40 to 30.
11951 * ipa-split.c (consider_split): Add some buffer for function to
11952 be considered inlining candidate.
11953 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11956 2018-02-12 Richard Biener <rguenther@suse.de>
11958 PR tree-optimization/84037
11959 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11960 matched stmts if we cannot swap the non-matched ones.
11962 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11964 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11965 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11966 _mm_maskz_scalef_round_ss): New intrinsics.
11967 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11968 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11969 __builtin_ia32_scalefss_round): Remove.
11970 (__builtin_ia32_scalefsd_mask_round,
11971 __builtin_ia32_scalefss_mask_round): New intrinsics.
11972 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11973 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11974 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11975 "<round_constraint>")): Changed to ...
11976 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11977 "<round_scalar_constraint>")): ... this.
11978 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11979 %0, %1, %2<round_op3>}"): Changed to ...
11980 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11981 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11982 %2<round_scalar_mask_op3>}"): ... this.
11983 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11985 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11987 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11988 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11989 (_mm_maskz_sqrt_round_ss): New intrinsics.
11990 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11991 (__builtin_ia32_sqrtsd_mask_round)
11992 (__builtin_ia32_sqrtss_mask_round): New builtins.
11993 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11994 (__builtin_ia32_sqrtss_round): Remove.
11995 (__builtin_ia32_sqrtsd_mask_round)
11996 (__builtin_ia32_sqrtss_mask_round): New builtins.
11997 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11998 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11999 ((match_operand:VF_128 1 "vector_operand"
12000 "xBm,<round_constraint>")): Changed to ...
12001 ((match_operand:VF_128 1 "vector_operand"
12002 "xBm,<round_scalar_constraint>")): ... this.
12003 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12004 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12005 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12006 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12007 %<iptr>1<round_scalar_mask_op3>}): ... this.
12008 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12009 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12011 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12014 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12015 Cast vec_cmpeq result to correct type.
12016 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12017 Cast vec_cmpgt result to correct type.
12019 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12021 * final.c (final_scan_insn_1): Renamed from...
12022 (final_scan_insn): ... this. New wrapper, to recover
12023 seen from the outermost call in recursive ones.
12024 * config/sparc/sparc.c (output_return): Drop seen from call.
12025 (output_sibcall): Likewise.
12026 * config/visium/visium.c (output_branch): Likewise.
12028 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12030 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12033 2018-02-10 Alan Modra <amodra@gmail.com>
12036 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12037 Specify LR as an input.
12039 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12042 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12043 remove_member_access_dummy_vars): New functions.
12044 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12045 lower_omp_1, execute_lower_omp): Use them.
12047 PR rtl-optimization/84308
12048 * shrink-wrap.c (spread_components): Release todo vector.
12050 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12052 PR rtl-optimization/57193
12053 * ira-color.c (struct allocno_color_data): Add member
12054 conflict_allocno_hard_prefs.
12055 (update_conflict_allocno_hard_prefs): New.
12056 (bucket_allocno_compare_func): Add a preference based on
12057 conflict_allocno_hard_prefs.
12058 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12059 (color_allocnos): Remove a dead code. Initiate
12060 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12062 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12065 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12066 constraint from =wa to wa. Avoid a subreg on the output operand,
12067 instead use a pseudo and subreg it in a move.
12068 (p9_xxbrd_<mode>): Changed to ...
12069 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12070 (p9_xxbrd_v2df): New expander.
12071 (p9_xxbrw_<mode>): Changed to ...
12072 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12073 (p9_xxbrw_v4sf): New expander.
12075 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12077 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12079 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12082 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12083 multiply in 32-bit mode.
12084 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12085 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12088 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12090 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12091 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12092 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12093 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12095 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12098 * dwarf2out.c (is_trivial_indirect_ref): New function.
12099 (dwarf2out_late_global_decl): Do not generate a location
12100 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12101 and that are not defined in the current unit.
12103 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12105 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12106 instead of a libcall for UNORDERED.
12108 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12111 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12112 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12114 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12117 * config/s390/s390.c (s390_set_current_function): Invoke
12118 s390_indirect_branch_settings also if fndecl didn't change.
12120 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12122 * config/rs6000/rs6000.md (blockage): Set length to zero.
12124 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12126 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12128 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12131 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12132 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12136 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12137 PARALLEL incoming that failed vt_get_decl_and_offset check.
12139 PR middle-end/84237
12140 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12141 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12143 (get_variable_section): For decls in named .bss* sections pass true as
12144 second argument to bss_initializer_p.
12146 2018-02-09 Marek Polacek <polacek@redhat.com>
12147 Jakub Jelinek <jakub@redhat.com>
12150 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12151 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12152 Sync some changes from cxx_fold_indirect_ref.
12154 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12156 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12158 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12159 (BLOCK_INLINE_ENTRY_LABEL): New.
12160 (dwarf2out_var_location): Disregard inline entry markers.
12161 (inline_entry_data): New struct.
12162 (inline_entry_data_hasher): New hashtable type.
12163 (inline_entry_data_hasher::hash): New.
12164 (inline_entry_data_hasher::equal): New.
12165 (inline_entry_data_table): New variable.
12166 (add_high_low_attributes): Add DW_AT_entry_pc and
12167 DW_AT_GNU_entry_view attributes if a pending entry is found
12168 in inline_entry_data_table. Add old entry_pc attribute only
12169 if debug nonbinding markers are disabled.
12170 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12171 markers are enabled.
12172 (block_within_block_p, dwarf2out_inline_entry): New.
12173 (dwarf2out_finish): Check that no entries remained in
12174 inline_entry_data_table.
12175 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12176 (final_scan_insn, notice_source_line): Likewise.
12177 (rest_of_clean_state): Skip inline entry markers.
12178 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12180 * gimple.c (gimple_build_debug_inline_entry): New.
12181 * gimple.h (enum gimple_debug_subcode): Add
12182 GIMPLE_DEBUG_INLINE_ENTRY.
12183 (gimple_build_debug_inline_entry): Declare.
12184 (gimple_debug_inline_entry_p): New.
12185 (gimple_debug_nonbind_marker_p): Adjust.
12186 * insn-notes.def (INLINE_ENTRY): New.
12187 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12188 inline entry marker notes.
12189 (print_insn): Likewise.
12190 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12191 (INSN_DEBUG_MARKER_KIND): Likewise.
12192 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12193 * tree-inline.c (expand_call_inline): Build and insert
12194 debug_inline_entry stmt.
12195 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12196 inline entry blocks early, if nonbind markers are enabled.
12197 (dump_scope_block): Dump fragment info.
12198 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12199 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12200 (gimple_build_debug_inline_entry): New.
12201 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12202 Enable/disable inline entry points too.
12203 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12204 (DEBUG_INSN): Describe inline entry markers.
12206 * common.opt (gvariable-location-views): New.
12207 (gvariable-location-views=incompat5): New.
12208 * config.in: Rebuilt.
12209 * configure: Rebuilt.
12210 * configure.ac: Test assembler for view support.
12211 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12212 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12213 * dwarf2out.c (var_loc_view): New typedef.
12214 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12215 (dwarf2out_locviews_in_attribute): New.
12216 (dwarf2out_locviews_in_loclist): New.
12217 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12218 (enum dw_line_info_opcode): Add LI_adv_address.
12219 (struct dw_line_info_table): Add view.
12220 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12221 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12222 (zero_view_p): New variable.
12223 (ZERO_VIEW_P): New macro.
12224 (output_asm_line_debug_info): New.
12225 (struct var_loc_node): Add view.
12226 (add_AT_view_list, AT_loc_list): New.
12227 (add_var_loc_to_decl): Add view param. Test it against last.
12228 (new_loc_list): Add view params. Record them.
12229 (AT_loc_list_ptr): Handle loc and view lists.
12230 (view_list_to_loc_list_val_node): New.
12231 (print_dw_val): Handle dw_val_class_view_list.
12232 (size_of_die): Likewise.
12233 (value_format): Likewise.
12234 (loc_list_has_views): New.
12235 (gen_llsym): Set vl_symbol too.
12236 (maybe_gen_llsym, skip_loc_list_entry): New.
12237 (dwarf2out_maybe_output_loclist_view_pair): New.
12238 (output_loc_list): Output view list or entries too.
12239 (output_view_list_offset): New.
12240 (output_die): Handle dw_val_class_view_list.
12241 (output_dwarf_version): New.
12242 (output_compilation_unit_header): Use it.
12243 (output_skeleton_debug_sections): Likewise.
12244 (output_rnglists, output_line_info): Likewise.
12245 (output_pubnames, output_aranges): Update version comments.
12246 (output_one_line_info_table): Output view numbers in asm comments.
12247 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12248 Call maybe_gen_llsym.
12249 (loc_list_from_tree_1): Adjust.
12250 (add_AT_location_description): Create view list attribute if
12251 needed, check it's absent otherwise.
12252 (convert_cfa_to_fb_loc_list): Adjust.
12253 (maybe_emit_file): Call output_asm_line_debug_info for test.
12254 (dwarf2out_var_location): Reset views as needed. Precompute
12255 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12256 attribute. Set view.
12257 (new_line_info_table): Reset next view.
12258 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12259 (dwarf2out_source_line): Likewise. Output view resets and labels to
12260 the assembler, or select appropriate line info opcodes.
12261 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12262 (optimize_string_length): Catch it. Adjust.
12263 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12264 dw_val_class_view_list, and remove it if no longer needed.
12265 (hash_loc_list): Hash view numbers.
12266 (loc_list_hasher::equal): Compare them.
12267 (optimize_location_lists): Check whether a view list symbol is
12268 needed, and whether the locview attribute is present, and
12269 whether they match. Remove the locview attribute if no longer
12271 (index_location_lists): Call skip_loc_list_entry for test.
12272 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12273 Use output_dwarf_version.
12274 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12275 (struct dw_val_node): Add val_view_list.
12276 * final.c (SEEN_NEXT_VIEW): New.
12277 (set_next_view_needed): New.
12278 (clear_next_view_needed): New.
12279 (maybe_output_next_view): New.
12280 (final_start_function): Rename to...
12281 (final_start_function_1): ... this. Take pointer to FIRST,
12282 add SEEN parameter. Emit param bindings in the initial view.
12283 (final_start_function): Reintroduce SEEN-less interface.
12284 (final): Rename to...
12285 (final_1): ... this. Take SEEN parameter. Output final pending
12286 next view at the end.
12287 (final): Reintroduce seen-less interface.
12288 (final_scan_insn): Output pending next view before switching
12289 sections or ending a block. Mark the next view as needed when
12290 outputting variable locations. Notify debug backend of section
12291 changes, and of location view changes.
12292 (rest_of_handle_final): Adjust.
12293 * toplev.c (process_options): Autodetect value for debug variable
12294 location views option. Warn on incompat5 without -gdwarf-5.
12295 * doc/invoke.texi (gvariable-location-views): New.
12296 (gvariable-location-views=incompat5): New.
12297 (gno-variable-location-views): New.
12299 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12301 PR tree-optimization/84136
12302 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12303 that the result of find_edge is non-NULL.
12305 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12308 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12309 storing integer register in SImode. Fix cost of 256 and 512
12310 byte aligned SSE register store.
12312 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12314 * config/i386/i386.c (ix86_multiplication_cost): Fix
12315 multiplication cost for TARGET_AVX512DQ.
12317 2018-02-08 Marek Polacek <polacek@redhat.com>
12319 PR tree-optimization/84238
12320 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12323 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12325 PR tree-optimization/84265
12326 * tree-vect-stmts.c (vectorizable_store): Don't treat
12327 VMAT_CONTIGUOUS accesses as grouped.
12328 (vectorizable_load): Likewise.
12330 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12332 PR tree-optimization/81635
12333 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12334 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12335 (test_round_for_mask): New functions.
12336 (wide_int_cc_tests): Call test_round_for_mask.
12337 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12338 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12339 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12340 range returned by get_range_info.
12342 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12345 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12346 * symtab.c: Include builtins.h
12347 (symtab_node::output_to_lto_symbol_table_p): Move here
12348 from lto-streamer-out.c:output_symbol_p.
12349 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12350 (output_symbol_p): Move all logic to symtab.c
12351 (produce_symtab): Update.
12353 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12355 * config/s390/s390-opts.h (enum indirect_branch): Define.
12356 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12357 (s390_indirect_branch_via_thunk)
12358 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12359 (enum s390_indirect_branch_type): Define.
12360 * config/s390/s390.c (struct s390_frame_layout, struct
12361 machine_function): Remove.
12362 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12363 (indirect_branch_table_label_no, indirect_branch_table_name):
12365 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12366 (enum s390_indirect_branch_option): Define.
12367 (s390_return_addr_from_memory): New function.
12368 (s390_handle_string_attribute): New function.
12369 (s390_attribute_table): Add new attribute handler.
12370 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12371 (s390_indirect_branch_via_thunk): New function.
12372 (s390_indirect_branch_via_inline_thunk): New function.
12373 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12374 sibling call optimization for non z10 compiles.
12375 (s390_emit_call): Force indirect branch target to be a single
12376 register. Add r1 clobber for non-z10 compiles.
12377 (s390_emit_epilogue): Emit return jump via return_use expander.
12378 (s390_reorg): Handle JUMP_INSNs as execute targets.
12379 (s390_option_override_internal): Perform validity checks for the
12380 new command line options.
12381 (s390_indirect_branch_attrvalue): New function.
12382 (s390_indirect_branch_settings): New function.
12383 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12384 (s390_output_indirect_thunk_function): New function.
12385 (s390_code_end): Implement target hook.
12386 (s390_case_values_threshold): Implement target hook.
12387 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12389 * config/s390/s390.h (struct s390_frame_layout)
12390 (struct machine_function): Move here from s390.c.
12391 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12392 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12393 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12394 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12395 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12396 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12397 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12398 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12399 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12400 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12401 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12402 (mnemonic attribute): Add values which aren't recognized
12404 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12405 pattern for branch conversion. Fix mnemonic attribute.
12406 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12407 indirect branch via thunk if requested.
12408 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12409 ("*indirect_jump"): Disable for branch conversion using out of
12411 ("indirect_jump_via_thunk<mode>_z10")
12412 ("indirect_jump_via_thunk<mode>")
12413 ("indirect_jump_via_inlinethunk<mode>_z10")
12414 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12415 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12416 ("casesi_jump_via_inlinethunk<mode>_z10")
12417 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12418 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12419 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12420 ("*indirect2_jump"): Disable for branch conversion.
12421 ("casesi_jump"): Turn into expander and expand patterns for branch
12423 ("return_use"): New expander.
12424 ("*return"): Emit return via thunk and rename it to ...
12425 ("*return<mode>"): ... this one.
12426 * config/s390/s390.opt: Add new options and and enum for the
12429 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12431 * lra-constraints.c (match_reload): Unconditionally use
12432 gen_lowpart_SUBREG, rather than selecting between that
12433 and equivalent gen_rtx_SUBREG code.
12435 2018-02-08 Richard Biener <rguenther@suse.de>
12437 PR tree-optimization/84233
12438 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12439 changed flag instead of boguously re-using phi_inserted.
12441 2018-02-08 Martin Jambor <mjambor@suse.cz>
12443 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12444 static local variables.
12446 2018-02-08 Richard Biener <rguenther@suse.de>
12448 PR tree-optimization/84278
12449 * tree-vect-stmts.c (vectorizable_store): When looking for
12450 smaller vector types to perform grouped strided loads/stores
12451 make sure the mode is supported by the target.
12452 (vectorizable_load): Likewise.
12454 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12456 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12457 Increase LDP/STP opportunities by adding adjacent callee-saves.
12459 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12461 PR rtl-optimization/84068
12462 PR rtl-optimization/83459
12463 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12465 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12467 PR tree-optimization/84224
12468 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12469 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12470 non-zero arguments.
12472 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12475 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12476 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12477 count, remove one USE.
12479 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12481 * doc/install.texi (Configuration): Document the
12482 --with-long-double-format={ibm,ieee} PowerPC configuration
12486 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12487 Convert from define_expand to be define_insn_and_split. Rework
12488 float/double/_Float128 conversions to QI/HI/SImode to work with
12489 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12490 conversions to QI/HImode types did a store and then a load to
12491 truncate the value. For conversions to VSX registers, don't split
12492 the insn, instead emit the code directly. Use the code iterator
12493 any_fix to combine signed and unsigned conversions.
12494 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12495 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12496 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12497 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12498 (fix_<mode>di2_hw): Likewise.
12499 (fixuns_<mode>di2_hw): Likewise.
12500 (fix_<mode>si2_hw): Likewise.
12501 (fixuns_<mode>si2_hw): Likewise.
12502 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12503 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12504 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12505 fix<uns>_trunc<SFDF:mode>si2_p8.
12506 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12507 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12508 (fix<uns>_<mode>_mem): Likewise.
12509 (fctiw<u>z_<mode>_mem): Likewise.
12510 (fix<uns>_<mode>_mem): Likewise.
12511 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12512 the register allocator from doing a direct move to the GPRs to do
12513 a store, and instead use the ISA 3.0 store byte/half-word from
12514 vector register instruction. For IEEE 128-bit floating point,
12515 also optimize stores of 32-bit ints.
12516 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12518 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12520 * genextract.c (push_pathstr_operand): New function to support
12522 (walk_rtx): Call push_pathstr_operand.
12523 (print_path): Support [a-zA-Z].
12525 2018-02-07 Richard Biener <rguenther@suse.de>
12527 PR tree-optimization/84037
12528 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12529 (cse_and_gimplify_to_preheader): Declare.
12530 (vect_get_place_in_interleaving_chain): Likewise.
12531 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12533 (_loop_vec_info::~_loop_vec_info): Delete it.
12534 (cse_and_gimplify_to_preheader): New function.
12535 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12536 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12537 (vectorizable_load): Likewise. For grouped stores always base
12538 the IV on the first element.
12539 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12540 condition before gimplifying.
12542 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12544 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12545 *DIV_EXPR and *MOD_EXPR.
12547 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12550 * config/i386/i386.c (ix86_option_override_internal): Mask out
12551 the CF_SET bit when checking -fcf-protection.
12553 2018-02-07 Tom de Vries <tom@codesourcery.com>
12556 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12559 2018-02-07 Richard Biener <rguenther@suse.de>
12561 PR tree-optimization/84204
12562 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12565 PR tree-optimization/84205
12566 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12567 special-case isl_ast_op_zdiv_r.
12569 PR tree-optimization/84223
12570 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12571 Only add conditions from within the region.
12572 (gather_bbs::after_dom_children): Adjust.
12574 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12577 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12578 * config/avr/avr.md: Only post-reload split REG-REG moves if
12579 either register is GENERAL_REG_P.
12581 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12583 PR tree-optimization/84235
12584 * tree-ssa-scopedtables.c
12585 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12586 if the subtraction is performed in floating point type where NaNs are
12587 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12588 build 1. Formatting fix.
12590 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12593 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12594 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12595 and skip it regardless of bb boundaries. Use CALL_P macro,
12596 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12597 unnecessarily, formatting fix.
12599 2018-02-06 Michael Collison <michael.collison@arm.com>
12601 * config/arm/thumb2.md:
12602 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12603 (*thumb_mov_notscc): Ditto.
12605 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12608 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12609 unsigned_fix, not "s".
12611 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12613 * configure.ac (gcc_fn_eh_frame_ro): New function.
12614 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12615 correct .eh_frame permissions.
12616 * configure: Regenerate.
12618 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12620 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12621 irrelevant options.
12623 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12625 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12626 Display warning message for -mno-speculate-indirect-jumps.
12628 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12630 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12632 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12634 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12636 PR tree-optimization/84225
12637 * tree-eh.c (find_trapping_overflow): Only call
12638 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12640 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12643 * config/i386/i386.c: Reimplement the check of possible options
12644 -mibt/-mshstk conbination. Change error messages.
12645 * doc/invoke.texi: Fix a typo: remove extra '='.
12647 2018-02-06 Marek Polacek <polacek@redhat.com>
12649 PR tree-optimization/84228
12650 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12652 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12655 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12656 emitted arch directives.
12657 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12658 __ARM_FEATURE_COPROC before changing architectures.
12660 2018-02-06 Richard Biener <rguenther@suse.de>
12662 * config/i386/i386.c (print_reg): Fix typo.
12663 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12665 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12667 * configure: Regenerate.
12669 2018-02-05 Martin Sebor <msebor@redhat.com>
12671 PR tree-optimization/83369
12672 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12675 2018-02-05 Martin Liska <mliska@suse.cz>
12677 * doc/invoke.texi: Cherry-pick upstream r323995.
12679 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12681 * ira.c (ira_init_register_move_cost): Adjust comment.
12683 2018-02-05 Martin Liska <mliska@suse.cz>
12685 PR gcov-profile/84137
12686 * doc/gcov.texi: Fix typo in documentation.
12688 2018-02-05 Martin Liska <mliska@suse.cz>
12690 PR gcov-profile/83879
12691 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12692 using dlopen functionality.
12694 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12696 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12697 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12698 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12699 _mm_maskz_range_round_ss): New intrinsics.
12700 (__builtin_ia32_rangesd128_round)
12701 (__builtin_ia32_rangess128_round): Remove.
12702 (__builtin_ia32_rangesd128_mask_round,
12703 __builtin_ia32_rangess128_mask_round): New builtins.
12704 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12705 __builtin_ia32_rangess128_round): Remove.
12706 (__builtin_ia32_rangesd128_mask_round,
12707 __builtin_ia32_rangess128_mask_round): New builtins.
12708 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12709 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12710 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12711 "<round_saeonly_constraint>")): Changed to ...
12712 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12713 "<round_saeonly_scalar_constraint>")): ... this.
12714 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12715 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12716 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12717 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12718 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12720 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12722 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12724 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12725 Remove all values except native, 8540 and 8548.
12727 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12729 * config/i386/i386.c (ix86_output_function_return): Pass
12730 INVALID_REGNUM, instead of -1, as invalid register number to
12731 indirect_thunk_name and output_indirect_thunk.
12733 2018-02-02 Julia Koval <julia.koval@intel.com>
12735 * config.gcc: Add -march=icelake.
12736 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12737 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12738 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12739 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12740 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12741 (processor_target_table): Add icelake.
12742 (ix86_option_override_internal): Handle new PTAs.
12743 (get_builtin_code_for_version): Handle icelake.
12744 (M_INTEL_COREI7_ICELAKE): New.
12745 (fold_builtin_cpu): Handle icelake.
12746 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12747 * doc/invoke.texi: Add -march=icelake.
12749 2018-02-02 Julia Koval <julia.koval@intel.com>
12751 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12752 to wide_int_bitmask.
12753 * wide-int-bitmask.h: New.
12755 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12758 * config/i386/i386.md: Replace Pmode with word_mode in
12759 builtin_setjmp_setup and builtin_longjmp to support x32.
12761 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12765 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12767 (rs6000_supported_cpu_names): New static variable.
12768 (linux_cpu_translation_table): Likewise.
12769 (elf_platform) <cpu>: Define new static variable and use it.
12770 Translate kernel AT_PLATFORM name to canonical name if needed.
12771 Error if platform name is unknown.
12773 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12776 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12778 2018-02-01 Jeff Law <law@redhat.com>
12781 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12782 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12783 the scratch if RELEASE_VIA_POP is false.
12784 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12785 If we have to save a temporary register, decrement SIZE appropriately.
12786 Pass new arguments to release_scratch_register_on_entry.
12787 (ix86_adjust_stack_and_probe): Likewise.
12788 (ix86_emit_probe_stack_range): Pass new arguments to
12789 release_scratch_register_on_entry.
12791 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12793 PR rtl-optimization/84157
12794 * combine.c (change_zero_ext): Use REG_P predicate in
12795 front of HARD_REGISTER_P predicate.
12797 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12799 * config/avr/avr.c (avr_option_override): Move disabling of
12800 -fdelete-null-pointer-checks to...
12801 * common/config/avr/avr-common.c (avr_option_optimization_table):
12804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12806 PR tree-optimization/81635
12807 * tree-data-ref.c (split_constant_offset_1): For types that
12808 wrap on overflow, try to use range info to prove that wrapping
12811 2018-02-01 Renlin Li <renlin.li@arm.com>
12814 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12815 TAILCALL_ADDR_REGS.
12816 (aarch64_register_move_cost): Likewise.
12817 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12818 TAILCALL_ADDR_REGS.
12819 (REG_CLASS_NAMES): Likewise.
12820 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12821 TAILCALL_ADDR_REGS. Remove IP registers.
12822 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12824 2018-02-01 Richard Biener <rguenther@suse.de>
12826 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12827 for specifying RPO order and allow NULL for that.
12828 * domwalk.c (dom_walker::dom_walker): Likewise.
12829 (dom_walker::walk): Handle NULL RPO order.
12830 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12832 (rewrite_update_dom_walker): Likewise.
12833 (mark_def_dom_walker): Likewise.
12835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12837 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12838 (aarch64_maybe_expand_sve_subreg_move): Declare.
12839 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12840 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12842 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12843 that are semantically a reverse operation.
12844 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12845 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12846 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12848 (aarch64_can_change_mode_class): For big-endian, forbid changes
12849 between two SVE modes if they have different element sizes.
12851 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12853 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12854 the TImode handling for big-endian targets.
12856 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12858 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12859 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12861 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12862 Remove BSWAP handing for big-endian targets and use the form of
12863 LD1RQ appropariate for the mode.
12865 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12867 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12868 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12869 duplicated element.
12871 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12874 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12875 check for operands that need to go through aarch64_sve_reload_be.
12877 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12879 PR tree-optimization/81661
12880 PR tree-optimization/84117
12881 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12882 * tree-eh.c: Include gimplify.h.
12883 (find_trapping_overflow, replace_trapping_overflow,
12884 rewrite_to_non_trapping_overflow): New functions.
12885 * tree-vect-loop.c: Include tree-eh.h.
12886 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12887 * tree-data-ref.c: Include tree-eh.h.
12888 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12890 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12892 PR rtl-optimization/84123
12893 * combine.c (change_zero_ext): Check if hard register satisfies
12894 can_change_dest_mode before calling gen_lowpart_SUBREG.
12896 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12899 * ira.c (ira_init_register_move_cost): Remove assert.
12901 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12903 PR rtl-optimization/84071
12904 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12905 * doc/tm.texi: Regenerate.
12907 2018-01-31 Richard Biener <rguenther@suse.de>
12909 PR tree-optimization/84132
12910 * tree-data-ref.c (analyze_miv_subscript): Properly
12911 check whether evolution_function_is_affine_multivariate_p
12912 before calling gcd_of_steps_may_divide_p.
12914 2018-01-31 Julia Koval <julia.koval@intel.com>
12917 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12918 * config/i386/i386.md (rdpid_rex64) New.
12919 (rdpid): Make 32bit only.
12921 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12924 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12925 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12927 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12930 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12932 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12934 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12936 PR rtl-optimization/84071
12937 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12938 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12940 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12942 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12943 (arc_attribute_table): Add 'aux' attribute.
12944 (arc_in_small_data_p): Consider aux like variables.
12945 (arc_is_aux_reg_p): New function.
12946 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12947 (arc_get_aux_arg): New function.
12948 (prepare_move_operands): Handle aux-register access.
12949 (arc_handle_aux_attribute): New function.
12950 * doc/extend.texi (ARC Variable attributes): Add subsection.
12952 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12954 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12955 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12956 (arc_attribute_table): Add 'uncached' attribute.
12957 (arc_print_operand): Print '.di' flag for uncached memory
12959 (arc_in_small_data_p): Do not consider for small data the uncached
12961 (arc_is_uncached_mem_p): New function.
12962 * config/arc/predicates.md (compact_store_memory_operand): Check
12963 for uncached memory accesses.
12964 (nonvol_nonimm_operand): Likewise.
12965 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12967 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12970 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12971 falign-loops=): Add Optimization flag.
12973 2018-01-30 Jeff Law <law@redhat.com>
12976 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12977 INT_REGISTERS_SAVED. Check it prior to calling
12978 get_scratch_register_on_entry.
12979 (ix86_adjust_stack_and_probe): Similarly.
12980 (ix86_emit_probe_stack_range): Similarly.
12981 (ix86_expand_prologue): Corresponding changes.
12983 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12986 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12987 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12989 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12992 * lra-constraints.c (curr_insn_transform): Process AND in the
12995 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12997 PR rtl-optimization/83986
12998 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12999 dependence against last_pending_memory_flush in addition to
13000 pending_jump_insns.
13002 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13004 PR tree-optimization/81611
13005 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13008 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13011 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13014 2018-01-30 Richard Biener <rguenther@suse.de>
13015 Jakub Jelinek <jakub@redhat.com>
13017 PR tree-optimization/84111
13018 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13019 inner loops added during recursion, as they don't have up-to-date
13022 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13025 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13026 (can_inline_edge_by_limits_p): ... here.
13027 (can_early_inline_edge_p, check_callers,
13028 update_caller_keys, update_callee_keys, recursive_inlining,
13029 add_new_edges_to_heap, speculation_useful_p,
13030 inline_small_functions,
13031 inline_small_functions, flatten_function,
13032 inline_to_all_callers_1): Update.
13034 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13036 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13039 2018-01-30 Richard Biener <rguenther@suse.de>
13041 PR tree-optimization/83008
13042 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13043 invariant and constant vector uses in stmts when they need
13044 more than one stmt.
13046 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13049 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13050 * configure: Regenerate.
13052 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13054 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13056 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13057 Use gen_rtx_REG rather than gen_lowpart.
13059 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13061 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13062 rather than 0 when creating partial subregs.
13064 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13066 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13069 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13072 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13073 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13074 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13075 flags. This restores the settings used before the 2017-07-24.
13076 Turning off pre increment/decrement/modify allows IVOPTS to
13077 optimize DF/SF loops where the index is an int.
13079 2018-01-29 Richard Biener <rguenther@suse.de>
13080 Kelvin Nilsen <kelvin@gcc.gnu.org>
13083 * tree-vect-stmts.c (vectorizable_call): Don't call
13084 targetm.vectorize_builtin_md_vectorized_function if callee is
13087 2018-01-22 Carl Love <cel@us.ibm.com>
13089 * doc/extend.tex: Fix typo in second arg in
13090 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13092 2018-01-29 Richard Biener <rguenther@suse.de>
13094 PR tree-optimization/84086
13095 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13096 (flush_ssaname_freelist): When SSA names were released reset
13097 the SCEV hash table.
13099 2018-01-29 Richard Biener <rguenther@suse.de>
13101 PR tree-optimization/84057
13102 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13103 removed paths when removing edges.
13105 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13107 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13108 -mfunction-return=@var{choice}.
13110 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13112 PR diagnostic/84034
13113 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13114 Handle CR like TAB.
13115 (layout::print_source_line): Likewise.
13116 (test_get_line_width_without_trailing_whitespace): Add test cases.
13118 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13120 PR middle-end/84040
13121 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13124 2018-01-26 Jim Wilson <jimw@sifive.com>
13126 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13128 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13131 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13133 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13134 and CMP + SUB-immediate -> SUBS.
13136 2018-01-26 Martin Sebor <msebor@redhat.com>
13138 PR tree-optimization/83896
13139 * tree-ssa-strlen.c (get_string_len): Rename...
13140 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13141 Avoid assuming length is constant.
13142 (handle_char_store): Use HOST_WIDE_INT for string length.
13144 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13147 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13148 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13150 2018-01-26 Richard Biener <rguenther@suse.de>
13152 PR rtl-optimization/84003
13153 * dse.c (record_store): Only record redundant stores when
13154 the earlier store aliases at least all accesses the later one does.
13156 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13158 PR rtl-optimization/83985
13159 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13160 REG_CFA_RESTORE insns.
13161 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13162 REG_CFA_RESTORE insns here.
13165 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13166 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13168 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13170 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13171 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13172 (arc_init): Likewise.
13173 (arc_override_options): Likewise.
13174 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13176 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13178 * config/arc/arc.h (TARGET_DBNZ): Define.
13179 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13180 properly set the tune attribute.
13181 (dbnz): Use TARGET_DBNZ guard.
13182 * config/arc/arc.opt (mtune): Add core3 option.
13184 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13186 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13187 recognize new pic like addresses.
13188 (arc_delegitimize_address): Clean up.
13190 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13192 * config/arc/arc-arches.def: Option mrf16 valid for all
13194 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13195 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13196 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13197 * config/arc/arc-tables.opt: Regenerate.
13198 * config/arc/arc.c (arc_conditional_register_usage): Handle
13199 reduced register file case.
13200 (arc_file_start): Set must have build attributes.
13201 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13202 mrf16 option value.
13203 * config/arc/arc.opt (mrf16): Add new option.
13204 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13205 * config/arc/genmultilib.awk: Handle new mrf16 option.
13206 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13207 * config/arc/t-multilib: Regenerate.
13208 * doc/invoke.texi (ARC Options): Document mrf16 option.
13210 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13212 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13213 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13214 (arc_attribute_table): Add 'secure_call' attribute.
13215 (arc_print_operand): Print secure call operand.
13216 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13218 (arc_is_secure_call_p): New function. * config/arc/arc.md
13219 (call_i): Add support for sjli instruction.
13220 (call_value_i): Likewise.
13221 * config/arc/constraints.md (Csc): New constraint.
13223 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13224 John Eric Martin <John.Martin@emmicro-us.com>
13226 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13227 * config/arc/arc.c (_arc_jli_section): New struct.
13228 (arc_jli_section): New type.
13229 (rc_jli_sections): New static variable.
13230 (arc_handle_jli_attribute): New function.
13231 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13232 (arc_file_end): New function.
13233 (TARGET_ASM_FILE_END): Define.
13234 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13235 (arc_add_jli_section): New function.
13236 (jli_call_scan): Likewise.
13237 (arc_reorg): Call jli_call_scan.
13238 (arc_output_addsi): Remove 'S' from printing asm operand.
13239 (arc_is_jli_call_p): New function.
13240 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13242 (movhi_insn): Likewise.
13243 (movsi_insn): Likewise.
13244 (movsi_set_cc_insn): Likewise.
13245 (loadqi_update): Likewise.
13246 (load_zeroextendqisi_update): Likewise.
13247 (load_signextendqisi_update): Likewise.
13248 (loadhi_update): Likewise.
13249 (load_zeroextendhisi_update): Likewise.
13250 (load_signextendhisi_update): Likewise.
13251 (loadsi_update): Likewise.
13252 (loadsf_update): Likewise.
13253 (movsicc_insn): Likewise.
13254 (bset_insn): Likewise.
13255 (bxor_insn): Likewise.
13256 (bclr_insn): Likewise.
13257 (bmsk_insn): Likewise.
13258 (bicsi3_insn): Likewise.
13259 (cmpsi_cc_c_insn): Likewise.
13260 (movsi_ne): Likewise.
13261 (movsi_cond_exec): Likewise.
13262 (clrsbsi2): Likewise.
13263 (norm_f): Likewise.
13273 (call_i): Remove 'S' asm letter, add jli instruction.
13274 (call_value_i): Likewise.
13275 * config/arc/arc.op (mjli-always): New option.
13276 * config/arc/constraints.md (Cji): New constraint.
13277 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13279 (subsf3_fpx): Likewise.
13280 (mulsf3_fpx): Likewise.
13281 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13283 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13284 function attrbutes.
13285 * doc/invoke.texi (ARC): Document mjli-always option.
13287 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13289 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13290 avoid addition with 0 and use incw and decw where possible.
13292 2018-01-26 Richard Biener <rguenther@suse.de>
13294 PR tree-optimization/81082
13295 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13296 association if it requires casting to unsigned.
13297 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13298 from fold_plusminus_mult_expr to catch important cases late when
13299 range info is available.
13301 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13303 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13304 * configure.ac (hidden_linkonce): New test.
13305 * configure: Regenerate.
13306 * config.in: Regenerate.
13308 2018-01-26 Julia Koval <julia.koval@intel.com>
13310 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13311 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13312 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13313 _mm_mask_bitshuffle_epi64_mask): Fix type.
13314 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13315 USI_FTYPE_V4DI_V4DI_USI): Remove.
13316 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13317 __builtin_ia32_vpshufbitqmb256_mask,
13318 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13319 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13320 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13322 2018-01-26 Alan Modra <amodra@gmail.com>
13325 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13326 UNSPEC_VBPERMQ. Sort other unspecs.
13328 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13330 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13332 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13334 PR middle-end/83055
13335 * predict.c (drop_profile): Do not push/pop cfun; update also
13337 (handle_missing_profiles): Fix logic looking for zero profiles.
13339 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13341 PR middle-end/83977
13342 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13343 on functions with #pragma omp declare simd or functions with simd
13345 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13346 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13347 Remove trailing \n from warning_at calls.
13349 2018-01-25 Tom de Vries <tom@codesourcery.com>
13352 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13353 for neutered workers.
13355 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13358 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13359 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13361 2018-01-24 Jeff Law <law@redhat.com>
13364 * i386.c (get_probe_interval): Move to earlier point.
13365 (ix86_compute_frame_layout): If -fstack-clash-protection and
13366 the frame is larger than the probe interval, then use pushes
13367 to save registers rather than reg->mem moves.
13368 (ix86_expand_prologue): Remove conditional for int_registers_saved
13371 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13374 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13375 min/max for never referenced object.
13377 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13379 PR middle-end/83977
13380 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13382 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13383 attributes from DECL_ATTRIBUTES (decl) without affecting
13384 DECL_ATTRIBUTES (current_function_decl).
13385 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13386 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13388 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13390 PR tree-optimization/83979
13391 * fold-const.c (fold_comparison): Use constant_boolean_node
13392 instead of boolean_{true,false}_node.
13394 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13396 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13399 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13401 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13402 Simplify the clause that sets the length attribute.
13403 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13404 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13405 clause that sets the length attribute.
13406 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13408 2018-01-24 Tom de Vries <tom@codesourcery.com>
13411 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13412 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13413 Add strict parameter.
13414 (prevent_branch_around_nothing): Insert dummy insn between branch to
13415 label and label with no ptx insn inbetween.
13416 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13418 2018-01-24 Tom de Vries <tom@codesourcery.com>
13421 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13422 for neutered threads in warp.
13423 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13425 2018-01-24 Richard Biener <rguenther@suse.de>
13427 PR tree-optimization/83176
13428 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13431 2018-01-24 Richard Biener <rguenther@suse.de>
13433 PR tree-optimization/82819
13434 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13435 code generating pluses that are no-ops in the target precision.
13437 2018-01-24 Richard Biener <rguenther@suse.de>
13439 PR middle-end/84000
13440 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13442 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13444 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13445 to merge probabilities.
13446 * predict.c (probably_never_executed): Also mark as cold functions
13447 with global 0 profile and guessed local profile.
13448 * profile-count.c (profile_probability::combine_with_count): New
13450 * profile-count.h (profile_probability::operator*,
13451 profile_probability::operator*=, profile_probability::operator/,
13452 profile_probability::operator/=): Reduce precision to adjusted
13453 and set value to guessed on contradictory divisions.
13454 (profile_probability::combine_with_freq): Remove.
13455 (profile_probability::combine_wiht_count): Declare.
13456 (profile_count::force_nonzero):: Set to adjusted.
13457 (profile_count::probability_in):: Set quality to adjusted.
13458 * tree-ssa-tail-merge.c (replace_block_by): Use
13459 combine_with_count.
13461 2018-01-23 Andrew Waterman <andrew@sifive.com>
13462 Jim Wilson <jimw@sifive.com>
13464 * config/riscv/riscv.c (riscv_stack_boundary): New.
13465 (riscv_option_override): Set riscv_stack_boundary. Handle
13466 riscv_preferred_stack_boundary_arg.
13467 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13468 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13469 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13470 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13471 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13472 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13474 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13477 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13478 of struct ix86_frame.
13479 (ix86_expand_epilogue): Likewise. Add a local variable for
13480 the reg_save_offset field in struct ix86_frame.
13482 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13484 PR tree-optimization/82604
13485 * tree-loop-distribution.c (enum partition_kind): New enum item
13486 PKIND_PARTIAL_MEMSET.
13487 (partition_builtin_p): Support above new enum item.
13488 (generate_code_for_partition): Ditto.
13489 (compute_access_range): Differentiate cases that equality can be
13490 proven at all loops, the innermost loops or no loops.
13491 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13492 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13493 (finalize_partitions, distribute_loop): Don't fuse partition of
13494 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13495 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13496 parloop is enabled.
13498 2018-01-23 Martin Liska <mliska@suse.cz>
13500 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13501 order to ignore the predictor.
13502 (PRED_POLYMORPHIC_CALL): Likewise.
13503 (PRED_RECURSIVE_CALL): Likewise.
13505 2018-01-23 Martin Liska <mliska@suse.cz>
13507 * tree-profile.c (tree_profiling): Print function header to
13508 aware reader which function we are working on.
13509 * value-prof.c (gimple_find_values_to_profile): Do not print
13510 not interesting value histograms.
13512 2018-01-23 Martin Liska <mliska@suse.cz>
13514 * profile-count.h (enum profile_quality): Add
13515 profile_uninitialized as the first value. Do not number values
13516 as they are zero based.
13517 (profile_count::verify): Update sanity check.
13518 (profile_probability::verify): Likewise.
13520 2018-01-23 Nathan Sidwell <nathan@acm.org>
13522 * doc/invoke.texi (ffor-scope): Deprecate.
13524 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13526 PR tree-optimization/83510
13527 * domwalk.c (set_all_edges_as_executable): New function.
13528 (dom_walker::dom_walker): Convert bool param
13529 "skip_unreachable_blocks" to enum reachability. Move setup of
13530 edge flags to set_all_edges_as_executable and only do it when
13531 reachability is REACHABLE_BLOCKS.
13532 * domwalk.h (enum dom_walker::reachability): New enum.
13533 (dom_walker::dom_walker): Convert bool param
13534 "skip_unreachable_blocks" to enum reachability.
13535 (set_all_edges_as_executable): New decl.
13536 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13537 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13539 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13540 but converting true to REACHABLE_BLOCKS.
13541 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13543 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13544 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13545 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13547 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13548 if check_all_array_refs will be called.
13550 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13552 * tree.c (selftest::test_location_wrappers): Add more test
13555 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13557 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13558 (selftest::test_bit_in_range): Likewise.
13560 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13563 * doc/sourcebuild.texi (vect_float): Say that the selector
13564 only describes the situation when -funsafe-math-optimizations is on.
13565 (vect_float_strict): Document.
13567 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13569 PR tree-optimization/83965
13570 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13571 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13572 instead of checking only for a reduction.
13573 (vect_recog_widen_sum_pattern): Likewise.
13575 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13577 * predict.c (probably_never_executed): Only use precise profile info.
13578 (compute_function_frequency): Skip after inlining hack since we now
13579 have quality checking.
13581 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13583 * profile-count.h (profile_probability::very_unlikely,
13584 profile_probability::unlikely, profile_probability::even): Set
13585 precision to guessed.
13587 2018-01-23 Richard Biener <rguenther@suse.de>
13589 PR tree-optimization/83963
13590 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13591 Properly terminate dominator walk when crossing the exit edge not
13592 when visiting its source block.
13594 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13597 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13598 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13600 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13602 PR tree-optimization/83957
13603 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13604 semicolon after for body surrounded by braces.
13606 PR tree-optimization/83081
13607 * profile-count.h (profile_probability::split): New method.
13608 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13609 Use profile_probability::split.
13610 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13611 when splitting a single conditional jump into 2.
13613 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13615 PR tree-optimization/69452
13616 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13619 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13621 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13622 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13623 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13625 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13627 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13629 * config/rl78/rl78.md (movdi): New define_expand.
13630 * config/rl78/rl78.c (rl78_split_movdi): New function.
13632 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13635 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13637 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13638 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13639 128-bit to produce an UNSPEC move to get the double word with the
13640 signbit and then a shift directly to do signbit.
13641 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13642 implementation with a new version that just does either a direct
13643 move or a regular move. Move memory interface to separate insns.
13644 Move insns so they are next to the expander.
13645 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13646 with signbit move. Split big and little endian case.
13647 (signbit<mode>2_dm_mem_le): Likewise.
13648 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13649 (signbit<mode>2_dm2): Likewise.
13651 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13653 * config/rl78/rl78.md (anddi3): New define_expand.
13655 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13657 * config/rl78/rl78.md (umindi3): New define_expand.
13659 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13661 * config/rl78/rl78.md (smindi3): New define_expand.
13663 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13665 * config/rl78/rl78.md (smaxdi3): New define_expand.
13667 2018-01-22 Carl Love <cel@us.ibm.com>
13669 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13670 LVX_V1TI): Add macro expansion.
13671 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13672 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13673 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13674 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13675 Change check to determine if the instruction is a byte reversing
13676 entry. Fix typo in comment.
13677 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13678 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13679 Add def_builtin calls for new builtins.
13680 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13681 Add define_insn expansion.
13683 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13685 * config/rl78/rl78.md (umaxdi3): New define_expand.
13687 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13689 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13690 for non-QImode registers.
13692 2018-01-22 Richard Biener <rguenther@suse.de>
13694 PR tree-optimization/83963
13695 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13696 including the loop exit block.
13697 (scop_detection::merge_sese): Likewise.
13698 (scop_detection::add_scop): Do it here instead.
13700 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13702 * doc/sourcebuild.texi (arm_softfloat): Document.
13704 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13707 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13708 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13709 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13711 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13712 David Edelsohn <dje.gcc@gmail.com>
13715 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13716 Change "crset eq" to "crset 2".
13717 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13718 (*call_indirect_aix<mode>_nospec): Likewise.
13719 (*call_value_indirect_aix<mode>_nospec): Likewise.
13720 (*call_indirect_elfv2<mode>_nospec): Likewise.
13721 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13722 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13723 change assembly output from . to $.
13724 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13725 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13726 (*tablejump<mode>_internal1_nospec): Likewise.
13728 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13731 * config/sh/sh_optimize_sett_clrt.cc:
13732 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13734 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13736 PR tree-optimization/83940
13737 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13738 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13739 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13740 use it to pass back the definition type.
13741 (vect_check_store_rhs): Likewise rhs_dt_out.
13742 (vect_build_gather_load_calls): Add a mask_dt argument and use
13743 it instead of a call to vect_is_simple_use.
13744 (vectorizable_store): Update calls to vect_check_load_store_mask
13745 and vect_check_store_rhs. Use the dt returned by the latter instead
13746 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13747 instead of calls to vect_is_simple_use. Pass the scalar rather
13748 than the vector operand to vect_is_simple_use when handling
13749 second and subsequent copies of an rhs value.
13750 (vectorizable_load): Update calls to vect_check_load_store_mask
13751 and vect_build_gather_load_calls. Use the cached mask_dt and
13752 gs_info.offset_dt instead of calls to vect_is_simple_use.
13754 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13756 PR middle-end/83945
13757 * tree-emutls.c: Include gimplify.h.
13758 (lower_emutls_2): New function.
13759 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13760 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13761 it before further processing.
13764 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13765 UINTVAL (trueop1) instead of INTVAL (op1).
13767 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13771 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13772 INCOMING_FRAME_SP_OFFSET if not defined.
13773 (scan_trace): Add ENTRY argument. If true and
13774 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13775 emit a note to adjust the CFA offset.
13776 (create_cfi_notes): Adjust scan_trace callers.
13777 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13778 INCOMING_FRAME_SP_OFFSET in the CIE.
13779 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13780 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13782 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13783 * doc/tm.texi: Regenerated.
13785 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13787 PR rtl-optimization/83147
13788 * lra-constraints.c (remove_inheritance_pseudos): Use
13789 lra_substitute_pseudo_within_insn.
13791 2018-01-19 Tom de Vries <tom@codesourcery.com>
13792 Cesar Philippidis <cesar@codesourcery.com>
13795 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13797 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13800 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13801 spaces for function labels.
13803 2018-01-19 Martin Liska <mliska@suse.cz>
13805 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13806 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13807 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13808 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13809 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13810 (PRED_CONST_RETURN): Change from 69 to 65.
13811 (PRED_NULL_RETURN): Change from 91 to 71.
13812 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13813 (PRED_LOOP_GUARD): Change from 66 to 73.
13815 2018-01-19 Martin Liska <mliska@suse.cz>
13817 * predict.c (predict_insn_def): Add new assert.
13818 (struct branch_predictor): Change type to signed integer.
13819 (test_prediction_value_range): Amend test to cover
13820 PROB_UNINITIALIZED.
13821 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13822 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13823 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13824 (PRED_LOOP_IV_COMPARE): Likewise.
13825 * predict.h (PROB_UNINITIALIZED): Define new constant.
13827 2018-01-19 Martin Liska <mliska@suse.cz>
13829 * predict.c (dump_prediction): Add new format for
13830 analyze_brprob.py script which is enabled with -details
13832 * profile-count.h (precise_p): New function.
13834 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13836 PR tree-optimization/83922
13837 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13838 there are no statements that need masking.
13839 (vect_active_double_reduction_p): New function.
13840 (vect_analyze_loop_operations): Use it when handling phis that
13841 are not in the loop header.
13843 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13845 PR tree-optimization/83914
13846 * tree-vect-loop.c (vectorizable_induction): Don't convert
13847 init_expr or apply the peeling adjustment for inductions
13848 that are nested within the vectorized loop.
13850 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13852 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13855 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13859 * function.h (gimplify_parameters): Add gimple_seq * argument.
13860 * function.c: Include gimple.h and options.h.
13861 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13862 for the added local temporaries if needed.
13863 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13864 if there are any parameter cleanups, wrap whole body into a
13865 try/finally with the cleanups.
13867 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13870 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13871 Use GET_MODE_CLASS for scalar floating point.
13873 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13877 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13878 Fix call of call_cgraph_insertion_hooks.
13880 2018-01-18 Martin Sebor <msebor@redhat.com>
13882 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13884 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13887 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13890 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13893 * common.opt: (-ffile-prefix-map): New option.
13894 * opts.c (common_handle_option): Defer it.
13895 * opts-global.c (handle_common_deferred_options): Handle it.
13896 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13897 * file-prefix-map.h: New file.
13898 (remap_debug_filename, add_debug_prefix_map): ...here.
13899 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13900 * final.c (debug_prefix_map, add_debug_prefix_map
13901 remap_debug_filename): Move to...
13902 * file-prefix-map.c: New file.
13903 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13904 generalize, get rid of alloca(), use strrchr() instead of strchr().
13905 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13906 Implement in terms of add_prefix_map().
13907 (remap_macro_filename, remap_debug_filename): Implement in term of
13909 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13910 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13911 * dbxout.c: Include file-prefix-map.h.
13912 * varasm.c: Likewise.
13913 * vmsdbgout.c: Likewise.
13914 * xcoffout.c: Likewise.
13915 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13916 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13917 * doc/invoke.texi (-ffile-prefix-map): Document.
13918 (-fdebug-prefix-map): Update description.
13920 2018-01-18 Martin Liska <mliska@suse.cz>
13922 * config/i386/i386.c (indirect_thunk_name): Document that also
13924 (output_indirect_thunk): Document why both instructions
13925 (pause and lfence) are generated.
13927 2018-01-18 Richard Biener <rguenther@suse.de>
13929 PR tree-optimization/83887
13930 * graphite-scop-detection.c
13931 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13932 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13933 (scop_detection::merge_sese): Re-implement with a flood-fill
13934 algorithm that properly finds a SESE region if it exists.
13936 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13939 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13940 pointer_diff optimizations use view_convert instead of convert.
13942 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13944 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13945 Generate different code for -mno-speculate-indirect-jumps.
13946 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13947 (*call_indirect_aix<mode>): Disable for
13948 -mno-speculate-indirect-jumps.
13949 (*call_indirect_aix<mode>_nospec): New define_insn.
13950 (*call_value_indirect_aix<mode>): Disable for
13951 -mno-speculate-indirect-jumps.
13952 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13953 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13954 -mno-speculate-indirect-jumps.
13955 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13957 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13959 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13960 long double type, set the flags for noting the default long double
13961 type, even if we don't pass or return a long double type.
13963 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13966 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13969 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13971 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13972 support for merge[hl].
13973 (fold_mergehl_helper): New helper function.
13974 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13975 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13976 (altivec_vmrglw_direct): Add xxmrglw insn.
13978 2018-01-17 Andrew Waterman <andrew@sifive.com>
13980 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13981 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13983 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13986 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13987 call the lto_location_cache before reading the
13988 DECL_SOURCE_LOCATION of the types.
13990 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13991 Richard Sandiford <richard.sandiford@linaro.org>
13993 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13994 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13995 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13996 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13997 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13999 * config/aarch64/constraints.md (aarch64_movti_operand):
14001 * config/aarch64/predicates.md (Uti): Add new constraint.
14003 2018-01-17 Carl Love <cel@us.ibm.com>
14005 * config/rs6000/vsx.md (define_expand xl_len_r,
14006 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14007 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14009 (define_expand, define_insn): Move the shift left from the
14010 define_insn to the define_expand for lxvl and stxvl instructions.
14011 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14012 and XL_LEN_R definitions to PURE.
14014 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14016 * config/i386/i386.c (indirect_thunk_name): Declare regno
14017 as unsigned int. Compare regno with INVALID_REGNUM.
14018 (output_indirect_thunk): Ditto.
14019 (output_indirect_thunk_function): Ditto.
14020 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14021 in the call to output_indirect_thunk_function.
14023 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14025 PR middle-end/83884
14026 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14027 rather than the size of inner_type to determine the stack slot size
14028 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14030 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14033 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14036 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14038 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14039 endian Linux systems to optionally enable multilibs for selecting
14040 the long double type if the user configured an explicit type.
14041 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14042 have no long double multilibs if not defined.
14043 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14044 warn if the user used -mabi={ieee,ibm}longdouble and we built
14045 multilibs for long double.
14046 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14047 appropriate multilib option.
14048 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14050 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14051 for building long double multilibs.
14052 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14054 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14056 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14059 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14061 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14064 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14067 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14070 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14072 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14073 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14075 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14077 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14078 different rtl trees depending on TARGET_64BIT.
14079 (rs6000_gen_lvx): Likewise.
14081 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14083 * config/visium/visium.md (nop): Tweak comment.
14084 (hazard_nop): Likewise.
14086 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14088 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14089 -mspeculate-indirect-jumps.
14090 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14091 for -mno-speculate-indirect-jumps.
14092 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14093 (*call_value_indirect_elfv2<mode>): Disable for
14094 -mno-speculate-indirect-jumps.
14095 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14096 (indirect_jump): Emit different RTL for
14097 -mno-speculate-indirect-jumps.
14098 (*indirect_jump<mode>): Disable for
14099 -mno-speculate-indirect-jumps.
14100 (*indirect_jump<mode>_nospec): New define_insn.
14101 (tablejump): Emit different RTL for
14102 -mno-speculate-indirect-jumps.
14103 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14104 (tablejumpsi_nospec): New define_expand.
14105 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14106 (tablejumpdi_nospec): New define_expand.
14107 (*tablejump<mode>_internal1): Disable for
14108 -mno-speculate-indirect-jumps.
14109 (*tablejump<mode>_internal1_nospec): New define_insn.
14110 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14113 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14115 * caller-save.c (insert_save): Drop unnecessary parameter. All
14118 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14119 Richard Biener <rguenth@suse.de>
14122 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14123 return early, inline manually is_gimple_sizepos. Make sure if we
14124 call gimplify_expr we don't end up with a gimple constant.
14125 * tree.c (variably_modified_type_p): Don't return true for
14126 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14127 * gimplify.h (is_gimple_sizepos): Remove.
14129 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14131 PR tree-optimization/83857
14132 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14133 vectorizable_live_operation for pure SLP statements.
14134 (vectorizable_live_operation): Handle PHIs.
14136 2018-01-16 Richard Biener <rguenther@suse.de>
14138 PR tree-optimization/83867
14139 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14140 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14142 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14145 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14146 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14147 If off is not INTEGER_CST, issue a may not be aligned warning
14148 rather than isn't aligned. Use isn%'t rather than isn't.
14149 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14151 <case MULT_EXPR>: Improve the case when bottom and one of the
14152 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14153 operand, in that case check if the other operand is multiple of
14154 bottom divided by the INTEGER_CST operand.
14156 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14159 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14160 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14161 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14162 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14163 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14164 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14165 (pa_function_arg_size): New function.
14167 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14169 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14170 in a separate statement.
14172 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14174 PR tree-optimization/83847
14175 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14176 group gathers and scatters.
14178 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14180 PR rtl-optimization/86620
14181 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14183 PR rtl-optimization/83213
14184 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14185 to last if both are JUMP_INSNs.
14187 PR tree-optimization/83843
14188 * gimple-ssa-store-merging.c
14189 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14190 store_immediate_info for bswap/nop orig_stores.
14192 2018-01-15 Andrew Waterman <andrew@sifive.com>
14194 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14196 <UDIV>: Increase cost if !TARGET_DIV.
14198 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14200 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14201 (define_attr "cr_logical_3op"): New.
14202 (cceq_ior_compare): Adjust.
14203 (cceq_ior_compare_complement): Adjust.
14204 (*cceq_rev_compare): Adjust.
14205 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14206 (is_cracked_insn): Adjust.
14207 (insn_must_be_first_in_group): Adjust.
14208 * config/rs6000/40x.md: Adjust.
14209 * config/rs6000/440.md: Adjust.
14210 * config/rs6000/476.md: Adjust.
14211 * config/rs6000/601.md: Adjust.
14212 * config/rs6000/603.md: Adjust.
14213 * config/rs6000/6xx.md: Adjust.
14214 * config/rs6000/7450.md: Adjust.
14215 * config/rs6000/7xx.md: Adjust.
14216 * config/rs6000/8540.md: Adjust.
14217 * config/rs6000/cell.md: Adjust.
14218 * config/rs6000/e300c2c3.md: Adjust.
14219 * config/rs6000/e500mc.md: Adjust.
14220 * config/rs6000/e500mc64.md: Adjust.
14221 * config/rs6000/e5500.md: Adjust.
14222 * config/rs6000/e6500.md: Adjust.
14223 * config/rs6000/mpc.md: Adjust.
14224 * config/rs6000/power4.md: Adjust.
14225 * config/rs6000/power5.md: Adjust.
14226 * config/rs6000/power6.md: Adjust.
14227 * config/rs6000/power7.md: Adjust.
14228 * config/rs6000/power8.md: Adjust.
14229 * config/rs6000/power9.md: Adjust.
14230 * config/rs6000/rs64.md: Adjust.
14231 * config/rs6000/titan.md: Adjust.
14233 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14235 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14236 ix86_indirect_branch_register logic.
14238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14240 * config/i386/constraints.md (Bs): Update
14241 ix86_indirect_branch_register check. Don't check
14242 ix86_indirect_branch_register with GOT_memory_operand.
14244 * config/i386/predicates.md (GOT_memory_operand): Don't check
14245 ix86_indirect_branch_register here.
14246 (GOT32_symbol_operand): Likewise.
14248 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14250 * config/i386/predicates.md (constant_call_address_operand):
14251 Rewrite ix86_indirect_branch_register logic.
14252 (sibcall_insn_operand): Likewise.
14254 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14256 * config/i386/constraints.md (Bs): Replace
14257 ix86_indirect_branch_thunk_register with
14258 ix86_indirect_branch_register.
14260 * config/i386/i386.md (indirect_jump): Likewise.
14261 (tablejump): Likewise.
14262 (*sibcall_memory): Likewise.
14263 (*sibcall_value_memory): Likewise.
14264 Peepholes of indirect call and jump via memory: Likewise.
14265 * config/i386/i386.opt: Likewise.
14266 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14267 (GOT_memory_operand): Likewise.
14268 (call_insn_operand): Likewise.
14269 (sibcall_insn_operand): Likewise.
14270 (GOT32_symbol_operand): Likewise.
14272 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14274 PR middle-end/83837
14275 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14276 type rather than type addr's type points to.
14277 (expand_omp_atomic_mutex): Likewise.
14278 (expand_omp_atomic): Likewise.
14280 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14283 * config/i386/i386.c (output_indirect_thunk_function): Use
14284 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14285 for __x86_return_thunk.
14287 2018-01-15 Richard Biener <rguenther@suse.de>
14289 PR middle-end/83850
14290 * expmed.c (extract_bit_field_1): Fix typo.
14292 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14295 * config/arm/iterators.md (VF): New mode iterator.
14296 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14297 Remove integer-related logic from pattern.
14298 (neon_vabd<mode>_3): Likewise.
14300 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14302 PR middle-end/82694
14303 * common.opt (fstrict-overflow): No longer an alias.
14304 (fwrapv-pointer): New option.
14305 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14306 also for pointer types based on flag_wrapv_pointer.
14307 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14308 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14309 opts->x_flag_wrapv got set.
14310 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14311 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14312 POINTER_TYPE_OVERFLOW_UNDEFINED.
14313 * match.pd: Likewise in address comparison pattern.
14314 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14316 2018-01-15 Richard Biener <rguenther@suse.de>
14319 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14320 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14321 Reset type names to their identifier if their TYPE_DECL doesn't
14322 have linkage (and thus is used for ODR and devirt).
14323 (save_debug_info_for_decl): Remove.
14324 (save_debug_info_for_type): Likewise.
14325 (add_tree_to_fld_list): Adjust.
14326 * tree-pretty-print.c (dump_generic_node): Make dumping of
14327 type names more robust.
14329 2018-01-15 Richard Biener <rguenther@suse.de>
14331 * BASE-VER: Bump to 8.0.1.
14333 2018-01-14 Martin Sebor <msebor@redhat.com>
14336 * builtins.c (check_access): Avoid warning when the no-warning bit
14339 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14341 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14342 * ira-color (allocno_hard_regs_compare): Likewise.
14344 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14347 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14348 Use .pushsection/.popsection.
14350 2018-01-14 Martin Sebor <msebor@redhat.com>
14353 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14355 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14357 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14358 entry from extra_headers.
14359 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14360 extra_headers, make the list bitwise identical to the i?86-*-* one.
14362 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14364 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14365 -mcmodel=large with -mindirect-branch=thunk,
14366 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14367 -mfunction-return=thunk-extern.
14368 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14369 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14370 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14372 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14374 * config/i386/i386.c (print_reg): Print the name of the full
14375 integer register without '%'.
14376 (ix86_print_operand): Handle 'V'.
14377 * doc/extend.texi: Document 'V' modifier.
14379 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14381 * config/i386/constraints.md (Bs): Disallow memory operand for
14382 -mindirect-branch-register.
14384 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14385 (GOT_memory_operand): Likewise.
14386 (call_insn_operand): Likewise.
14387 (sibcall_insn_operand): Likewise.
14388 (GOT32_symbol_operand): Likewise.
14389 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14390 for -mindirect-branch-register.
14391 (tablejump): Likewise.
14392 (*sibcall_memory): Likewise.
14393 (*sibcall_value_memory): Likewise.
14394 Disallow peepholes of indirect call and jump via memory for
14395 -mindirect-branch-register.
14396 (*call_pop): Replace m with Bw.
14397 (*call_value_pop): Likewise.
14398 (*sibcall_pop_memory): Replace m with Bs.
14399 * config/i386/i386.opt (mindirect-branch-register): New option.
14400 * doc/invoke.texi: Document -mindirect-branch-register option.
14402 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14404 * config/i386/i386-protos.h (ix86_output_function_return): New.
14405 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14406 set function_return_type.
14407 (indirect_thunk_name): Add ret_p to indicate thunk for function
14409 (output_indirect_thunk_function): Pass false to
14410 indirect_thunk_name.
14411 (ix86_output_indirect_branch_via_reg): Likewise.
14412 (ix86_output_indirect_branch_via_push): Likewise.
14413 (output_indirect_thunk_function): Create alias for function
14414 return thunk if regno < 0.
14415 (ix86_output_function_return): New function.
14416 (ix86_handle_fndecl_attribute): Handle function_return.
14417 (ix86_attribute_table): Add function_return.
14418 * config/i386/i386.h (machine_function): Add
14419 function_return_type.
14420 * config/i386/i386.md (simple_return_internal): Use
14421 ix86_output_function_return.
14422 (simple_return_internal_long): Likewise.
14423 * config/i386/i386.opt (mfunction-return=): New option.
14424 (indirect_branch): Mention -mfunction-return=.
14425 * doc/extend.texi: Document function_return function attribute.
14426 * doc/invoke.texi: Document -mfunction-return= option.
14428 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14430 * config/i386/i386-opts.h (indirect_branch): New.
14431 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14432 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14433 with local indirect jump when converting indirect call and jump.
14434 (ix86_set_indirect_branch_type): New.
14435 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14436 (indirectlabelno): New.
14437 (indirect_thunk_needed): Likewise.
14438 (indirect_thunk_bnd_needed): Likewise.
14439 (indirect_thunks_used): Likewise.
14440 (indirect_thunks_bnd_used): Likewise.
14441 (INDIRECT_LABEL): Likewise.
14442 (indirect_thunk_name): Likewise.
14443 (output_indirect_thunk): Likewise.
14444 (output_indirect_thunk_function): Likewise.
14445 (ix86_output_indirect_branch_via_reg): Likewise.
14446 (ix86_output_indirect_branch_via_push): Likewise.
14447 (ix86_output_indirect_branch): Likewise.
14448 (ix86_output_indirect_jmp): Likewise.
14449 (ix86_code_end): Call output_indirect_thunk_function if needed.
14450 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14452 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14453 (ix86_attribute_table): Add indirect_branch.
14454 * config/i386/i386.h (machine_function): Add indirect_branch_type
14455 and has_local_indirect_jump.
14456 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14458 (tablejump): Likewise.
14459 (*indirect_jump): Use ix86_output_indirect_jmp.
14460 (*tablejump_1): Likewise.
14461 (simple_return_indirect_internal): Likewise.
14462 * config/i386/i386.opt (mindirect-branch=): New option.
14463 (indirect_branch): New.
14466 (thunk-inline): Likewise.
14467 (thunk-extern): Likewise.
14468 * doc/extend.texi: Document indirect_branch function attribute.
14469 * doc/invoke.texi: Document -mindirect-branch= option.
14471 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14474 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14476 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14478 * ipa-inline.c (want_inline_small_function_p): Return false if
14479 inlining has already failed with CIF_FINAL_ERROR.
14480 (update_caller_keys): Call want_inline_small_function_p before
14482 (update_callee_keys): Likewise.
14484 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14486 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14488 (rs6000_quadword_masked_address_p): Likewise.
14489 (quad_aligned_load_p): Likewise.
14490 (quad_aligned_store_p): Likewise.
14491 (const_load_sequence_p): Add comment to describe the outer-most loop.
14492 (mimic_memory_attributes_and_flags): New function.
14493 (rs6000_gen_stvx): Likewise.
14494 (replace_swapped_aligned_store): Likewise.
14495 (rs6000_gen_lvx): Likewise.
14496 (replace_swapped_aligned_load): Likewise.
14497 (replace_swapped_load_constant): Capitalize argument name in
14498 comment describing this function.
14499 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14500 and stores that access quad-word aligned addresses and replace
14501 with stvx or lvx instructions when appropriate.
14502 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14503 New function prototype.
14504 (rs6000_quadword_masked_address_p): Likewise.
14505 (rs6000_gen_lvx): Likewise.
14506 (rs6000_gen_stvx): Likewise.
14507 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14508 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14509 when memory address is aligned.
14510 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14511 this split to select lvx instruction when memory address is aligned.
14512 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14513 instruction when memory address is aligned.
14514 (*vsx_le_perm_load_v16qi): Likewise.
14515 (four unnamed splitters): Modify to select the stvx instruction
14516 when memory is aligned.
14518 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14520 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14521 which appears in the queue multiple times.
14523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14524 Alan Hayward <alan.hayward@arm.com>
14525 David Sherwood <david.sherwood@arm.com>
14527 * tree-vectorizer.h (vec_lower_bound): New structure.
14528 (_loop_vec_info): Add check_nonzero and lower_bounds.
14529 (LOOP_VINFO_CHECK_NONZERO): New macro.
14530 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14531 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14532 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14533 fields. Make seg_len the distance travelled, not including the
14535 (dr_direction_indicator): Declare.
14536 (dr_zero_step_indicator): Likewise.
14537 (dr_known_forward_stride_p): Likewise.
14538 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14540 (runtime_alias_check_p): Allow runtime alias checks with
14542 (operator ==): Compare access_size and align.
14543 (prune_runtime_alias_test_list): Rework for new distinction between
14544 the access_size and seg_len.
14545 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14547 (get_segment_min_max): New function.
14548 (create_intersect_range_checks): Use it.
14549 (dr_step_indicator): New function.
14550 (dr_direction_indicator): Likewise.
14551 (dr_zero_step_indicator): Likewise.
14552 (dr_known_forward_stride_p): Likewise.
14553 * tree-loop-distribution.c (data_ref_segment_size): Return
14554 DR_STEP * (niters - 1).
14555 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14557 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14558 (vect_preserves_scalar_order_p): New function, split out from...
14559 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14560 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14561 (vect_vfa_access_size): New function.
14562 (vect_vfa_align): Likewise.
14563 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14564 (dump_lower_bound): New function.
14565 (vect_check_lower_bound): Likewise.
14566 (vect_small_gap_p): Likewise.
14567 (vectorizable_with_step_bound_p): Likewise.
14568 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14569 depencies if the vectorization factor is 1. Convert the checks
14570 for nonzero steps into checks on the bounds of DR_STEP. Try using
14571 a bunds check for variable steps if the minimum required step is
14572 relatively small. Update calls to the dr_with_seg_len
14573 constructor and to vect_compile_time_alias.
14574 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14576 (vect_loop_versioning): Call it.
14577 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14579 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14582 Alan Hayward <alan.hayward@arm.com>
14583 David Sherwood <david.sherwood@arm.com>
14585 * doc/sourcebuild.texi (vect_scatter_store): Document.
14586 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14588 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14590 * genopinit.c (main): Add supports_vec_scatter_store and
14591 supports_vec_scatter_store_cached to target_optabs.
14592 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14593 IFN_MASK_SCATTER_STORE.
14594 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14596 * internal-fn.h (internal_store_fn_p): Declare.
14597 (internal_fn_stored_value_index): Likewise.
14598 * internal-fn.c (scatter_store_direct): New macro.
14599 (expand_scatter_store_optab_fn): New function.
14600 (direct_scatter_store_optab_supported_p): New macro.
14601 (internal_store_fn_p): New function.
14602 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14603 IFN_MASK_SCATTER_STORE.
14604 (internal_fn_mask_index): Likewise.
14605 (internal_fn_stored_value_index): New function.
14606 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14607 for scatter stores.
14608 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14609 * optabs-query.c (supports_vec_scatter_store_p): New function.
14610 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14611 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14612 true for scatter stores.
14613 (vect_gather_scatter_fn_p): Handle scatter stores too.
14614 (vect_check_gather_scatter): Consider using scatter stores if
14615 supports_vec_scatter_store_p.
14616 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14617 scatter stores too.
14618 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14619 internal_fn_stored_value_index.
14620 (check_load_store_masking): Handle scatter stores too.
14621 (vect_get_store_rhs): Make public.
14622 (vectorizable_call): Use internal_store_fn_p.
14623 (vectorizable_store): Handle scatter store internal functions.
14624 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14625 when deciding whether the end of the group has been reached.
14626 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14627 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14628 (mask_scatter_store<mode>): New insns.
14630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14631 Alan Hayward <alan.hayward@arm.com>
14632 David Sherwood <david.sherwood@arm.com>
14634 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14635 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14636 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14638 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14639 Use vect_truncate_gather_scatter_offset if we can't treat the
14640 operation as a normal gather load or scatter store.
14641 (get_group_load_store_type): Take the gather_scatter_info
14642 as argument. Try using a gather load or scatter store for
14643 single-element groups.
14644 (get_load_store_type): Update calls to get_group_load_store_type
14645 and vect_use_strided_gather_scatters_p.
14647 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14648 Alan Hayward <alan.hayward@arm.com>
14649 David Sherwood <david.sherwood@arm.com>
14651 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14652 optional tree argument.
14653 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14655 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14656 but continue to use the current value as a fallback.
14657 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14658 to compare the updates.
14659 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14660 (get_load_store_type): Use it when handling a strided access.
14661 (vect_get_strided_load_store_ops): New function.
14662 (vect_get_data_ptr_increment): Likewise.
14663 (vectorizable_load): Handle strided gather loads. Always pass
14664 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14667 Alan Hayward <alan.hayward@arm.com>
14668 David Sherwood <david.sherwood@arm.com>
14670 * doc/md.texi (gather_load@var{m}): Document.
14671 (mask_gather_load@var{m}): Likewise.
14672 * genopinit.c (main): Add supports_vec_gather_load and
14673 supports_vec_gather_load_cached to target_optabs.
14674 * optabs-tree.c (init_tree_optimization_optabs): Use
14675 ggc_cleared_alloc to allocate target_optabs.
14676 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14677 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14679 * internal-fn.h (internal_load_fn_p): Declare.
14680 (internal_gather_scatter_fn_p): Likewise.
14681 (internal_fn_mask_index): Likewise.
14682 (internal_gather_scatter_fn_supported_p): Likewise.
14683 * internal-fn.c (gather_load_direct): New macro.
14684 (expand_gather_load_optab_fn): New function.
14685 (direct_gather_load_optab_supported_p): New macro.
14686 (direct_internal_fn_optab): New function.
14687 (internal_load_fn_p): Likewise.
14688 (internal_gather_scatter_fn_p): Likewise.
14689 (internal_fn_mask_index): Likewise.
14690 (internal_gather_scatter_fn_supported_p): Likewise.
14691 * optabs-query.c (supports_at_least_one_mode_p): New function.
14692 (supports_vec_gather_load_p): Likewise.
14693 * optabs-query.h (supports_vec_gather_load_p): Declare.
14694 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14695 and memory_type field.
14696 (NUM_PATTERNS): Bump to 15.
14697 * tree-vect-data-refs.c: Include internal-fn.h.
14698 (vect_gather_scatter_fn_p): New function.
14699 (vect_describe_gather_scatter_call): Likewise.
14700 (vect_check_gather_scatter): Try using internal functions for
14701 gather loads. Recognize existing calls to a gather load function.
14702 (vect_analyze_data_refs): Consider using gather loads if
14703 supports_vec_gather_load_p.
14704 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14705 (vect_get_gather_scatter_offset_type): Likewise.
14706 (vect_convert_mask_for_vectype): Likewise.
14707 (vect_add_conversion_to_patterm): Likewise.
14708 (vect_try_gather_scatter_pattern): Likewise.
14709 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14710 (vect_vect_recog_func_ptrs): Add it.
14711 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14712 internal_fn_mask_index and internal_gather_scatter_fn_p.
14713 (check_load_store_masking): Take the gather_scatter_info as an
14714 argument and handle gather loads.
14715 (vect_get_gather_scatter_ops): New function.
14716 (vectorizable_call): Check internal_load_fn_p.
14717 (vectorizable_load): Likewise. Handle gather load internal
14719 (vectorizable_store): Update call to check_load_store_masking.
14720 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14721 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14722 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14723 (aarch64_gather_scale_operand_d): New predicates.
14724 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14725 (mask_gather_load<mode>): New insns.
14727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14728 Alan Hayward <alan.hayward@arm.com>
14729 David Sherwood <david.sherwood@arm.com>
14731 * optabs.def (fold_left_plus_optab): New optab.
14732 * doc/md.texi (fold_left_plus_@var{m}): Document.
14733 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14734 * internal-fn.c (fold_left_direct): Define.
14735 (expand_fold_left_optab_fn): Likewise.
14736 (direct_fold_left_optab_supported_p): Likewise.
14737 * fold-const-call.c (fold_const_fold_left): New function.
14738 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14739 * tree-parloops.c (valid_reduction_p): New function.
14740 (gather_scalar_reductions): Use it.
14741 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14742 (vect_finish_replace_stmt): Declare.
14743 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14744 (needs_fold_left_reduction_p): New function, split out from...
14745 (vect_is_simple_reduction): ...here. Accept reductions that
14746 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14747 (vect_force_simple_reduction): Also store the reduction type in
14748 the assignment's STMT_VINFO_REDUC_TYPE.
14749 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14750 (merge_with_identity): New function.
14751 (vect_expand_fold_left): Likewise.
14752 (vectorize_fold_left_reduction): Likewise.
14753 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14754 scalar phi in place for it. Check for target support and reject
14755 cases that would reassociate the operation. Defer the transform
14756 phase to vectorize_fold_left_reduction.
14757 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14758 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14759 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14763 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14764 call to ifc_temp_var.
14766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14767 Alan Hayward <alan.hayward@arm.com>
14768 David Sherwood <david.sherwood@arm.com>
14770 * target.def (legitimize_address_displacement): Take the original
14771 offset as a poly_int.
14772 * targhooks.h (default_legitimize_address_displacement): Update
14774 * targhooks.c (default_legitimize_address_displacement): Likewise.
14775 * doc/tm.texi: Regenerate.
14776 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14777 as an argument, moving assert of ad->disp == ad->disp_term to...
14778 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14779 Try calling targetm.legitimize_address_displacement before expanding
14780 the address rather than afterwards, and adjust for the new interface.
14781 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14782 Match the new hook interface. Handle SVE addresses.
14783 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14784 new hook interface.
14786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14788 * Makefile.in (OBJS): Add early-remat.o.
14789 * target.def (select_early_remat_modes): New hook.
14790 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14791 * doc/tm.texi: Regenerate.
14792 * targhooks.h (default_select_early_remat_modes): Declare.
14793 * targhooks.c (default_select_early_remat_modes): New function.
14794 * timevar.def (TV_EARLY_REMAT): New timevar.
14795 * passes.def (pass_early_remat): New pass.
14796 * tree-pass.h (make_pass_early_remat): Declare.
14797 * early-remat.c: New file.
14798 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14800 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14803 Alan Hayward <alan.hayward@arm.com>
14804 David Sherwood <david.sherwood@arm.com>
14806 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14807 vfm1 with a bound_epilog parameter.
14808 (vect_do_peeling): Update calls accordingly, and move the prologue
14809 call earlier in the function. Treat the base bound_epilog as 0 for
14810 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14811 this base when peeling for gaps.
14812 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14813 with fully-masked loops.
14814 (vect_estimate_min_profitable_iters): Handle the single peeled
14815 iteration in that case.
14817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14818 Alan Hayward <alan.hayward@arm.com>
14819 David Sherwood <david.sherwood@arm.com>
14821 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14822 single-element interleaving even if the size is not a power of 2.
14823 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14824 accesses for single-element interleaving if the group size is
14827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14828 Alan Hayward <alan.hayward@arm.com>
14829 David Sherwood <david.sherwood@arm.com>
14831 * doc/md.texi (fold_extract_last_@var{m}): Document.
14832 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14833 * optabs.def (fold_extract_last_optab): New optab.
14834 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14835 * internal-fn.c (fold_extract_direct): New macro.
14836 (expand_fold_extract_optab_fn): Likewise.
14837 (direct_fold_extract_optab_supported_p): Likewise.
14838 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14839 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14840 EXTRACT_LAST_REDUCTION.
14841 (get_initial_def_for_reduction): Do not create an initial vector
14842 for EXTRACT_LAST_REDUCTION reductions.
14843 (vectorizable_reduction): Leave the scalar phi in place for
14844 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14845 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14846 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14847 transform phase to vectorizable_condition.
14848 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14850 (vect_finish_stmt_generation): ...here.
14851 (vect_finish_replace_stmt): New function.
14852 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14853 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14855 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 * doc/md.texi (extract_last_@var{m}): Document.
14862 * optabs.def (extract_last_optab): New optab.
14863 * internal-fn.def (EXTRACT_LAST): New internal function.
14864 * internal-fn.c (cond_unary_direct): New macro.
14865 (expand_cond_unary_optab_fn): Likewise.
14866 (direct_cond_unary_optab_supported_p): Likewise.
14867 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14868 loops using EXTRACT_LAST.
14869 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14870 (extract_last_<mode>): ...this optab.
14871 (vec_extract<mode><Vel>): Update accordingly.
14873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14874 Alan Hayward <alan.hayward@arm.com>
14875 David Sherwood <david.sherwood@arm.com>
14877 * target.def (empty_mask_is_expensive): New hook.
14878 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14879 * doc/tm.texi: Regenerate.
14880 * targhooks.h (default_empty_mask_is_expensive): Declare.
14881 * targhooks.c (default_empty_mask_is_expensive): New function.
14882 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14883 if the target says that empty masks are expensive.
14884 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14886 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14889 Alan Hayward <alan.hayward@arm.com>
14890 David Sherwood <david.sherwood@arm.com>
14892 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14893 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14894 (vect_use_loop_mask_for_alignment_p): New function.
14895 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14896 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14897 niters_skip argument. Make sure that the first niters_skip elements
14898 of the first iteration are inactive.
14899 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14900 Update call to vect_set_loop_masks_directly.
14901 (get_misalign_in_elems): New function, split out from...
14902 (vect_gen_prolog_loop_niters): ...here.
14903 (vect_update_init_of_dr): Take a code argument that specifies whether
14904 the adjustment should be added or subtracted.
14905 (vect_update_init_of_drs): Likewise.
14906 (vect_prepare_for_masked_peels): New function.
14907 (vect_do_peeling): Skip prologue peeling if we're using a mask
14908 instead. Update call to vect_update_inits_of_drs.
14909 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14911 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14912 alignment. Do not include the number of peeled iterations in
14913 the minimum threshold in that case.
14914 (vectorizable_induction): Adjust the start value down by
14915 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14916 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14917 Take the number of skipped iterations into account when calculating
14919 * tree-vect-stmts.c (vect_gen_while_not): New function.
14921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14922 Alan Hayward <alan.hayward@arm.com>
14923 David Sherwood <david.sherwood@arm.com>
14925 * doc/sourcebuild.texi (vect_fully_masked): Document.
14926 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14927 default value to 0.
14928 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14930 (vect_analyze_loop_2): ...here. Don't check the vectorization
14931 factor against the number of loop iterations if the loop is
14934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14935 Alan Hayward <alan.hayward@arm.com>
14936 David Sherwood <david.sherwood@arm.com>
14938 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14939 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14940 (dump_groups): Update accordingly.
14941 (iv_use::mem_type): New member variable.
14942 (address_p): New function.
14943 (record_use): Add a mem_type argument and initialize the new
14945 (record_group_use): Add a mem_type argument. Use address_p.
14946 Remove obsolete null checks of base_object. Update call to record_use.
14947 (find_interesting_uses_op): Update call to record_group_use.
14948 (find_interesting_uses_cond): Likewise.
14949 (find_interesting_uses_address): Likewise.
14950 (get_mem_type_for_internal_fn): New function.
14951 (find_address_like_use): Likewise.
14952 (find_interesting_uses_stmt): Try find_address_like_use before
14953 calling find_interesting_uses_op.
14954 (addr_offset_valid_p): Use the iv mem_type field as the type
14955 of the addressed memory.
14956 (add_autoinc_candidates): Likewise.
14957 (get_address_cost): Likewise.
14958 (split_small_address_groups_p): Use address_p.
14959 (split_address_groups): Likewise.
14960 (add_iv_candidate_for_use): Likewise.
14961 (autoinc_possible_for_pair): Likewise.
14962 (rewrite_groups): Likewise.
14963 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14964 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14965 (get_alias_ptr_type_for_ptr_address): New function.
14966 (rewrite_use_address): Rewrite address uses in calls that were
14967 identified by find_address_like_use.
14969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14970 Alan Hayward <alan.hayward@arm.com>
14971 David Sherwood <david.sherwood@arm.com>
14973 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14975 * gimple-expr.h (is_gimple_addressable: Likewise.
14976 * gimple-expr.c (is_gimple_address): Likewise.
14977 * internal-fn.c (expand_call_mem_ref): New function.
14978 (expand_mask_load_optab_fn): Use it.
14979 (expand_mask_store_optab_fn): Likewise.
14981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14982 Alan Hayward <alan.hayward@arm.com>
14983 David Sherwood <david.sherwood@arm.com>
14985 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14986 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14987 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14988 (cond_umax@var{mode}): Document.
14989 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14990 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14991 (cond_umin_optab, cond_umax_optab): New optabs.
14992 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14993 (COND_IOR, COND_XOR): New internal functions.
14994 * internal-fn.h (get_conditional_internal_fn): Declare.
14995 * internal-fn.c (cond_binary_direct): New macro.
14996 (expand_cond_binary_optab_fn): Likewise.
14997 (direct_cond_binary_optab_supported_p): Likewise.
14998 (get_conditional_internal_fn): New function.
14999 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15000 Cope with reduction statements that are vectorized as calls rather
15002 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15003 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15004 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15005 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15006 (UNSPEC_COND_EOR): New unspecs.
15007 (optab): Add mappings for them.
15008 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15009 (sve_int_op, sve_fp_op): New int attributes.
15011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15012 Alan Hayward <alan.hayward@arm.com>
15013 David Sherwood <david.sherwood@arm.com>
15015 * optabs.def (while_ult_optab): New optab.
15016 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15017 * internal-fn.def (WHILE_ULT): New internal function.
15018 * internal-fn.h (direct_internal_fn_supported_p): New override
15019 that takes two types as argument.
15020 * internal-fn.c (while_direct): New macro.
15021 (expand_while_optab_fn): New function.
15022 (convert_optab_supported_p): Likewise.
15023 (direct_while_optab_supported_p): New macro.
15024 * wide-int.h (wi::udiv_ceil): New function.
15025 * tree-vectorizer.h (rgroup_masks): New structure.
15026 (vec_loop_masks): New typedef.
15027 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15028 and fully_masked_p.
15029 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15030 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15031 (vect_max_vf): New function.
15032 (slpeel_make_loop_iterate_ntimes): Delete.
15033 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15034 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15035 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15036 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15037 internal-fn.h, stor-layout.h and optabs-query.h.
15038 (vect_set_loop_mask): New function.
15039 (add_preheader_seq): Likewise.
15040 (add_header_seq): Likewise.
15041 (interleave_supported_p): Likewise.
15042 (vect_maybe_permute_loop_masks): Likewise.
15043 (vect_set_loop_masks_directly): Likewise.
15044 (vect_set_loop_condition_masked): Likewise.
15045 (vect_set_loop_condition_unmasked): New function, split out from
15046 slpeel_make_loop_iterate_ntimes.
15047 (slpeel_make_loop_iterate_ntimes): Rename to..
15048 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15049 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15050 (vect_do_peeling): Update call accordingly.
15051 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15053 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15054 mask_compare_type, can_fully_mask_p and fully_masked_p.
15055 (release_vec_loop_masks): New function.
15056 (_loop_vec_info): Use it to free the loop masks.
15057 (can_produce_all_loop_masks_p): New function.
15058 (vect_get_max_nscalars_per_iter): Likewise.
15059 (vect_verify_full_masking): Likewise.
15060 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15061 retries, and free the mask rgroups before retrying. Check loop-wide
15062 reasons for disallowing fully-masked loops. Make the final decision
15063 about whether use a fully-masked loop or not.
15064 (vect_estimate_min_profitable_iters): Do not assume that peeling
15065 for the number of iterations will be needed for fully-masked loops.
15066 (vectorizable_reduction): Disable fully-masked loops.
15067 (vectorizable_live_operation): Likewise.
15068 (vect_halve_mask_nunits): New function.
15069 (vect_double_mask_nunits): Likewise.
15070 (vect_record_loop_mask): Likewise.
15071 (vect_get_loop_mask): Likewise.
15072 (vect_transform_loop): Handle the case in which the final loop
15073 iteration might handle a partial vector. Call vect_set_loop_condition
15074 instead of slpeel_make_loop_iterate_ntimes.
15075 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15076 (check_load_store_masking): New function.
15077 (prepare_load_store_mask): Likewise.
15078 (vectorizable_store): Handle fully-masked loops.
15079 (vectorizable_load): Likewise.
15080 (supportable_widening_operation): Use vect_halve_mask_nunits for
15082 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15083 (vect_gen_while): New function.
15084 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15085 (aarch64_uqdec<mode>): New insn.
15087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15088 Alan Hayward <alan.hayward@arm.com>
15089 David Sherwood <david.sherwood@arm.com>
15091 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15092 (reduc_xor_scal_optab): New optabs.
15093 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15094 (reduc_xor_scal_@var{m}): Document.
15095 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15096 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15097 internal functions.
15098 * fold-const-call.c (fold_const_call): Handle them.
15099 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15100 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15101 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15102 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15103 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15104 (UNSPEC_XORV): New unspecs.
15105 (optab): Add entries for them.
15106 (BITWISEV): New int iterator.
15107 (bit_reduc_op): New int attributes.
15109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15110 Alan Hayward <alan.hayward@arm.com>
15111 David Sherwood <david.sherwood@arm.com>
15113 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15114 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15115 * optabs.def (vec_shl_insert_optab): New optab.
15116 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15117 (duplicate_and_interleave): Likewise.
15118 * tree-vect-loop.c: Include internal-fn.h.
15119 (neutral_op_for_slp_reduction): New function, split out from
15120 get_initial_defs_for_reduction.
15121 (get_initial_def_for_reduction): Handle option 2 for variable-length
15122 vectors by loading the neutral value into a vector and then shifting
15123 the initial value into element 0.
15124 (get_initial_defs_for_reduction): Replace the code argument with
15125 the neutral value calculated by neutral_op_for_slp_reduction.
15126 Use gimple_build_vector for constant-length vectors.
15127 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15128 but the first group_size elements have a neutral value.
15129 Use duplicate_and_interleave otherwise.
15130 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15131 Update call to get_initial_defs_for_reduction. Handle SLP
15132 reductions for variable-length vectors by creating one vector
15133 result for each scalar result, with the elements associated
15134 with other scalar results stubbed out with the neutral value.
15135 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15136 Require IFN_VEC_SHL_INSERT for double reductions on
15137 variable-length vectors, or SLP reductions that have
15138 a neutral value. Require can_duplicate_and_interleave_p
15139 support for variable-length unchained SLP reductions if there
15140 is no neutral value, such as for MIN/MAX reductions. Also require
15141 the number of vector elements to be a multiple of the number of
15142 SLP statements when doing variable-length unchained SLP reductions.
15143 Update call to vect_create_epilog_for_reduction.
15144 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15145 and remove initial values.
15146 (duplicate_and_interleave): Make public.
15147 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15148 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15151 Alan Hayward <alan.hayward@arm.com>
15152 David Sherwood <david.sherwood@arm.com>
15154 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15155 (can_duplicate_and_interleave_p): New function.
15156 (vect_get_and_check_slp_defs): Take the vector of statements
15157 rather than just the current one. Remove excess parentheses.
15158 Restriction rejectinon of vect_constant_def and vect_external_def
15159 for variable-length vectors to boolean types, or types for which
15160 can_duplicate_and_interleave_p is false.
15161 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15162 (duplicate_and_interleave): New function.
15163 (vect_get_constant_vectors): Use gimple_build_vector for
15164 constant-length vectors and suitable variable-length constant
15165 vectors. Use duplicate_and_interleave for other variable-length
15166 vectors. Don't defer the update when inserting new statements.
15168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15169 Alan Hayward <alan.hayward@arm.com>
15170 David Sherwood <david.sherwood@arm.com>
15172 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15173 min_profitable_iters doesn't go negative.
15175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15176 Alan Hayward <alan.hayward@arm.com>
15177 David Sherwood <david.sherwood@arm.com>
15179 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15180 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15181 * optabs.def (vec_mask_load_lanes_optab): New optab.
15182 (vec_mask_store_lanes_optab): Likewise.
15183 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15184 (MASK_STORE_LANES): Likewise.
15185 * internal-fn.c (mask_load_lanes_direct): New macro.
15186 (mask_store_lanes_direct): Likewise.
15187 (expand_mask_load_optab_fn): Handle masked operations.
15188 (expand_mask_load_lanes_optab_fn): New macro.
15189 (expand_mask_store_optab_fn): Handle masked operations.
15190 (expand_mask_store_lanes_optab_fn): New macro.
15191 (direct_mask_load_lanes_optab_supported_p): Likewise.
15192 (direct_mask_store_lanes_optab_supported_p): Likewise.
15193 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15195 (vect_load_lanes_supported): Likewise.
15196 * tree-vect-data-refs.c (strip_conversion): New function.
15197 (can_group_stmts_p): Likewise.
15198 (vect_analyze_data_ref_accesses): Use it instead of checking
15199 for a pair of assignments.
15200 (vect_store_lanes_supported): Take a masked_p parameter.
15201 (vect_load_lanes_supported): Likewise.
15202 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15203 vect_store_lanes_supported and vect_load_lanes_supported.
15204 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15205 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15206 parameter. Don't allow gaps for masked accesses.
15207 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15208 and vect_load_lanes_supported.
15209 (get_load_store_type): Take a masked_p parameter and update
15210 call to get_group_load_store_type.
15211 (vectorizable_store): Update call to get_load_store_type.
15212 Handle IFN_MASK_STORE_LANES.
15213 (vectorizable_load): Update call to get_load_store_type.
15214 Handle IFN_MASK_LOAD_LANES.
15216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15217 Alan Hayward <alan.hayward@arm.com>
15218 David Sherwood <david.sherwood@arm.com>
15220 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15222 * config/aarch64/aarch64-protos.h
15223 (aarch64_sve_struct_memory_operand_p): Declare.
15224 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15225 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15226 (VPRED, vpred): Handle SVE structure modes.
15227 * config/aarch64/constraints.md (Utx): New constraint.
15228 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15229 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15230 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15231 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15232 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15233 structure modes. Split into pieces after RA.
15234 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15235 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15237 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15238 SVE structure modes.
15239 (aarch64_classify_address): Likewise.
15240 (sizetochar): Move earlier in file.
15241 (aarch64_print_operand): Handle SVE register lists.
15242 (aarch64_array_mode): New function.
15243 (aarch64_sve_struct_memory_operand_p): Likewise.
15244 (TARGET_ARRAY_MODE): Redefine.
15246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15247 Alan Hayward <alan.hayward@arm.com>
15248 David Sherwood <david.sherwood@arm.com>
15250 * target.def (array_mode): New target hook.
15251 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15252 * doc/tm.texi: Regenerate.
15253 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15254 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15255 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15256 targetm.array_mode.
15257 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15261 Alan Hayward <alan.hayward@arm.com>
15262 David Sherwood <david.sherwood@arm.com>
15264 * fold-const.c (fold_binary_loc): Check the argument types
15265 rather than the result type when testing for a vector operation.
15267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15269 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15270 * doc/tm.texi: Regenerate.
15272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15273 Alan Hayward <alan.hayward@arm.com>
15274 David Sherwood <david.sherwood@arm.com>
15276 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15277 (sve): Document new AArch64 extension.
15278 * doc/md.texi (w): Extend the description of the AArch64
15279 constraint to include SVE vectors.
15280 (Upl, Upa): Document new AArch64 predicate constraints.
15281 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15283 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15284 (msve-vector-bits=): New option.
15285 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15286 SVE when these are disabled.
15287 (sve): New extension.
15288 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15289 modes. Adjust their number of units based on aarch64_sve_vg.
15290 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15291 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15292 aarch64_addr_query_type.
15293 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15294 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15295 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15296 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15297 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15298 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15299 (aarch64_simd_imm_zero_p): Delete.
15300 (aarch64_check_zero_based_sve_index_immediate): Declare.
15301 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15302 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15303 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15304 (aarch64_sve_float_mul_immediate_p): Likewise.
15305 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15306 rather than an rtx.
15307 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15308 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15309 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15310 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15311 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15312 (aarch64_regmode_natural_size): Likewise.
15313 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15314 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15316 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15317 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15318 for VG and the SVE predicate registers.
15319 (V_ALIASES): Add a "z"-prefixed alias.
15320 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15321 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15322 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15323 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15324 (REG_CLASS_NAMES): Add entries for them.
15325 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15326 and the predicate registers.
15327 (aarch64_sve_vg): Declare.
15328 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15329 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15330 (REGMODE_NATURAL_SIZE): Define.
15331 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15333 * config/aarch64/aarch64.c: Include cfgrtl.h.
15334 (simd_immediate_info): Add a constructor for series vectors,
15335 and an associated step field.
15336 (aarch64_sve_vg): New variable.
15337 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15338 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15339 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15340 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15341 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15342 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15343 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15344 (aarch64_get_mask_mode): New functions.
15345 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15346 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15347 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15348 predicate modes and predicate registers. Explicitly restrict
15349 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15350 to store a vector mode if it is recognized by
15351 aarch64_classify_vector_mode.
15352 (aarch64_regmode_natural_size): New function.
15353 (aarch64_hard_regno_caller_save_mode): Return the original mode
15355 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15356 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15357 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15358 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15360 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15361 does not overlap dest if the function is frame-related. Handle
15363 (aarch64_split_add_offset): New function.
15364 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15365 them aarch64_add_offset.
15366 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15367 and update call to aarch64_sub_sp.
15368 (aarch64_add_cfa_expression): New function.
15369 (aarch64_expand_prologue): Pass extra temporary registers to the
15370 functions above. Handle the case in which we need to emit new
15371 DW_CFA_expressions for registers that were originally saved
15372 relative to the stack pointer, but now have to be expressed
15373 relative to the frame pointer.
15374 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15376 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15377 IP0 and IP1 values for SVE frames.
15378 (aarch64_expand_vec_series): New function.
15379 (aarch64_expand_sve_widened_duplicate): Likewise.
15380 (aarch64_expand_sve_const_vector): Likewise.
15381 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15382 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15383 into the register, rather than emitting a SET directly.
15384 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15385 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15386 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15387 (offset_9bit_signed_scaled_p): New functions.
15388 (aarch64_replicate_bitmask_imm): New function.
15389 (aarch64_bitmask_imm): Use it.
15390 (aarch64_cannot_force_const_mem): Reject expressions involving
15391 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15392 (aarch64_classify_index): Handle SVE indices, by requiring
15393 a plain register index with a scale that matches the element size.
15394 (aarch64_classify_address): Handle SVE addresses. Assert that
15395 the mode of the address is VOIDmode or an integer mode.
15396 Update call to aarch64_classify_symbol.
15397 (aarch64_classify_symbolic_expression): Update call to
15398 aarch64_classify_symbol.
15399 (aarch64_const_vec_all_in_range_p): New function.
15400 (aarch64_print_vector_float_operand): Likewise.
15401 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15402 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15403 and the FP immediates 1.0 and 0.5.
15404 (aarch64_print_address_internal): Handle SVE addresses.
15405 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15406 (aarch64_regno_regclass): Handle predicate registers.
15407 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15409 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15410 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15411 (aarch64_convert_sve_vector_bits): New function.
15412 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15413 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15414 rather than an rtx.
15415 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15416 Handle SVE vector and predicate modes. Accept VL-based constants
15417 that need only one temporary register, and VL offsets that require
15418 no temporary registers.
15419 (aarch64_conditional_register_usage): Mark the predicate registers
15420 as fixed if SVE isn't available.
15421 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15422 Return true for SVE vector and predicate modes.
15423 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15424 rather than an unsigned int. Handle SVE modes.
15425 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15427 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15429 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15430 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15431 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15432 (aarch64_sve_float_mul_immediate_p): New functions.
15433 (aarch64_sve_valid_immediate): New function.
15434 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15435 Explicitly reject structure modes. Check for INDEX constants.
15436 Handle PTRUE and PFALSE constants.
15437 (aarch64_check_zero_based_sve_index_immediate): New function.
15438 (aarch64_simd_imm_zero_p): Delete.
15439 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15440 vector modes. Accept constants in the range of CNT[BHWD].
15441 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15442 ask for an Advanced SIMD mode.
15443 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15444 (aarch64_simd_vector_alignment): Handle SVE predicates.
15445 (aarch64_vectorize_preferred_vector_alignment): New function.
15446 (aarch64_simd_vector_alignment_reachable): Use it instead of
15448 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15449 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15451 (MAX_VECT_LEN): Delete.
15452 (expand_vec_perm_d): Add a vec_flags field.
15453 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15454 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15455 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15457 (aarch64_evpc_rev): Rename to...
15458 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15459 (aarch64_evpc_rev_global): New function.
15460 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15461 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15463 (aarch64_evpc_sve_tbl): New function.
15464 (aarch64_expand_vec_perm_const_1): Update after rename of
15465 aarch64_evpc_rev. Handle SVE permutes too, trying
15466 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15467 than aarch64_evpc_tbl.
15468 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15469 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15470 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15471 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15472 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15473 (aarch64_expand_sve_vcond): New functions.
15474 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15475 of aarch64_vector_mode_p.
15476 (aarch64_dwarf_poly_indeterminate_value): New function.
15477 (aarch64_compute_pressure_classes): Likewise.
15478 (aarch64_can_change_mode_class): Likewise.
15479 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15480 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15481 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15482 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15483 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15484 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15485 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15486 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15488 (Dn, Dl, Dr): Accept const as well as const_vector.
15489 (Dz): Likewise. Compare against CONST0_RTX.
15490 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15491 of "vector" where appropriate.
15492 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15493 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15494 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15495 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15496 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15497 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15498 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15499 (v_int_equiv): Extend to SVE modes.
15500 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15502 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15503 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15504 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15505 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15506 (SVE_COND_FP_CMP): New int iterators.
15507 (perm_hilo): Handle the new unpack unspecs.
15508 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15510 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15511 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15512 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15513 (aarch64_equality_operator, aarch64_constant_vector_operand)
15514 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15515 (aarch64_sve_nonimmediate_operand): Likewise.
15516 (aarch64_sve_general_operand): Likewise.
15517 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15518 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15519 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15520 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15521 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15522 (aarch64_sve_float_arith_immediate): Likewise.
15523 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15524 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15525 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15526 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15527 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15528 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15529 (aarch64_sve_float_arith_operand): Likewise.
15530 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15531 (aarch64_sve_float_mul_operand): Likewise.
15532 (aarch64_sve_vec_perm_operand): Likewise.
15533 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15534 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15535 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15536 as well as const_vector.
15537 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15538 in file. Use CONST0_RTX and CONSTM1_RTX.
15539 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15540 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15541 Use aarch64_simd_imm_zero.
15542 * config/aarch64/aarch64-sve.md: New file.
15543 * config/aarch64/aarch64.md: Include it.
15544 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15545 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15546 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15547 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15548 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15549 (sve): New attribute.
15550 (enabled): Disable instructions with the sve attribute unless
15552 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15553 aarch64_expand_mov_immediate.
15554 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15555 CNT[BHSD] immediates.
15556 (movti): Split CONST_POLY_INT moves into two halves.
15557 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15558 Split additions that need a temporary here if the destination
15559 is the stack pointer.
15560 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15561 (*add<mode>3_poly_1): New instruction.
15562 (set_clobber_cc): New expander.
15564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15566 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15567 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15568 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15569 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15570 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15571 Change innermode from fixed_mode_size to machine_mode.
15572 (simplify_subreg): Update call accordingly. Handle a constant-sized
15573 subreg of a variable-length CONST_VECTOR.
15575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15576 Alan Hayward <alan.hayward@arm.com>
15577 David Sherwood <david.sherwood@arm.com>
15579 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15580 (add_offset_to_base): New function, split out from...
15581 (create_mem_ref): ...here. When handling a scale other than 1,
15582 check first whether the address is valid without the offset.
15583 Add it into the base if so, leaving the index and scale as-is.
15585 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15588 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15589 fold_for_warn before checking if arg2 is INTEGER_CST.
15591 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15593 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15594 (store_multiple_operation): Delete.
15595 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15596 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15597 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15598 guarded by TARGET_STRING.
15599 (rs6000_output_load_multiple): Delete.
15600 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15601 OPTION_MASK_STRING / TARGET_STRING handling.
15602 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15603 (const rs6000_opt_masks) <"string">: Change mask to 0.
15604 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15605 (MASK_STRING): Delete.
15606 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15608 (load_multiple): Delete.
15615 (store_multiple): Delete.
15622 (movmemsi_8reg): Delete.
15623 (corresponding unnamed define_insn): Delete.
15624 (movmemsi_6reg): Delete.
15625 (corresponding unnamed define_insn): Delete.
15626 (movmemsi_4reg): Delete.
15627 (corresponding unnamed define_insn): Delete.
15628 (movmemsi_2reg): Delete.
15629 (corresponding unnamed define_insn): Delete.
15630 (movmemsi_1reg): Delete.
15631 (corresponding unnamed define_insn): Delete.
15632 * config/rs6000/rs6000.opt (mno-string): New.
15633 (mstring): Replace by deprecation warning stub.
15634 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15636 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15638 * regrename.c (regrename_do_replace): If replacing the same
15639 reg multiple times, try to reuse last created gen_raw_REG.
15642 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15643 main to workaround a bug in GDB.
15645 2018-01-12 Tom de Vries <tom@codesourcery.com>
15648 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15650 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15652 PR rtl-optimization/80481
15653 * ira-color.c (get_cap_member): New function.
15654 (allocnos_conflict_by_live_ranges_p): Use it.
15655 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15656 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15658 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15661 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15662 (*saddl_se_1): Ditto.
15663 (*ssubsi_1): Ditto.
15664 (*ssubl_se_1): Ditto.
15666 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15668 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15669 rather than wi::to_widest for DR_INITs.
15670 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15671 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15672 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15674 (vect_analyze_group_access_1): Note that here.
15676 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15678 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15679 polynomial type sizes.
15681 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15683 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15684 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15685 (gimple_add_tmp_var): Likewise.
15687 2018-01-12 Martin Liska <mliska@suse.cz>
15689 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15690 (gimple_alloc_sizes): Likewise.
15691 (dump_gimple_statistics): Use PRIu64 in printf format.
15692 * gimple.h: Change uint64_t to int.
15694 2018-01-12 Martin Liska <mliska@suse.cz>
15696 * tree-core.h: Use uint64_t instead of int.
15697 * tree.c (tree_node_counts): Likewise.
15698 (tree_node_sizes): Likewise.
15699 (dump_tree_statistics): Use PRIu64 in printf format.
15701 2018-01-12 Martin Liska <mliska@suse.cz>
15703 * Makefile.in: As qsort_chk is implemented in vec.c, add
15704 vec.o to linkage of gencfn-macros.
15705 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15706 passing the info to record_node_allocation_statistics.
15707 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15709 * ggc-common.c (struct ggc_usage): Add operator== and use
15710 it in operator< and compare function.
15711 * mem-stats.h (struct mem_usage): Likewise.
15712 * vec.c (struct vec_usage): Remove operator< and compare
15713 function. Can be simply inherited.
15715 2018-01-12 Martin Jambor <mjambor@suse.cz>
15718 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15719 * tree-ssa-math-opts.c: Include domwalk.h.
15720 (convert_mult_to_fma_1): New function.
15721 (fma_transformation_info): New type.
15722 (fma_deferring_state): Likewise.
15723 (cancel_fma_deferring): New function.
15724 (result_of_phi): Likewise.
15725 (last_fma_candidate_feeds_initial_phi): Likewise.
15726 (convert_mult_to_fma): Added deferring logic, split actual
15727 transformation to convert_mult_to_fma_1.
15728 (math_opts_dom_walker): New type.
15729 (math_opts_dom_walker::after_dom_children): New method, body moved
15730 here from pass_optimize_widening_mul::execute, added deferring logic
15732 (pass_optimize_widening_mul::execute): Moved most of code to
15733 math_opts_dom_walker::after_dom_children.
15734 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15735 * config/i386/i386.c (ix86_option_override_internal): Added
15736 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15738 2018-01-12 Richard Biener <rguenther@suse.de>
15741 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15742 inline instance vars.
15744 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15747 * config/rx/rx.c (rx_is_restricted_memory_address):
15748 Handle SUBREG case.
15750 2018-01-12 Richard Biener <rguenther@suse.de>
15752 PR tree-optimization/80846
15753 * target.def (split_reduction): New target hook.
15754 * targhooks.c (default_split_reduction): New function.
15755 * targhooks.h (default_split_reduction): Declare.
15756 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15757 target requests first reduce vectors by combining low and high
15759 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15760 (get_vectype_for_scalar_type_and_size): Export.
15761 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15762 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15763 * doc/tm.texi: Regenerate.
15764 * config/i386/i386.c (ix86_split_reduction): Implement
15765 TARGET_VECTORIZE_SPLIT_REDUCTION.
15767 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15770 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15771 in PIC mode except for TARGET_VXWORKS_RTP.
15772 * config/sparc/sparc.c: Include cfgrtl.h.
15773 (TARGET_INIT_PIC_REG): Define.
15774 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15775 (sparc_pic_register_p): New predicate.
15776 (sparc_legitimate_address_p): Use it.
15777 (sparc_legitimize_pic_address): Likewise.
15778 (sparc_delegitimize_address): Likewise.
15779 (sparc_mode_dependent_address_p): Likewise.
15780 (gen_load_pcrel_sym): Remove 4th parameter.
15781 (load_got_register): Adjust call to above. Remove obsolete stuff.
15782 (sparc_expand_prologue): Do not call load_got_register here.
15783 (sparc_flat_expand_prologue): Likewise.
15784 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15785 (sparc_use_pseudo_pic_reg): New function.
15786 (sparc_init_pic_reg): Likewise.
15787 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15788 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15790 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15792 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15793 Add item for branch_cost.
15795 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15797 PR rtl-optimization/83565
15798 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15799 not extend the result to a larger mode for rotate operations.
15800 (num_sign_bit_copies1): Likewise.
15802 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15805 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15807 Use values-Xc.o for -pedantic.
15808 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15810 2018-01-12 Martin Liska <mliska@suse.cz>
15813 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15815 (possible_polymorphic_call_targets): Use it.
15816 (ipa_devirt): Likewise.
15818 2018-01-12 Martin Liska <mliska@suse.cz>
15820 * profile-count.h (enum profile_quality): Use 0 as invalid
15821 enum value of profile_quality.
15823 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15825 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15826 -mext-string options.
15828 2018-01-12 Richard Biener <rguenther@suse.de>
15830 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15831 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15832 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15834 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15836 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15838 * configure.ac (--with-long-double-format): Add support for the
15839 configuration option to change the default long double format on
15841 * config.gcc (powerpc*-linux*-*): Likewise.
15842 * configure: Regenerate.
15843 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15844 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15845 used without modification.
15847 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15849 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15850 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15851 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15852 MISC_BUILTIN_SPEC_BARRIER.
15853 (rs6000_init_builtins): Likewise.
15854 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15856 (speculation_barrier): New define_insn.
15857 * doc/extend.texi: Document __builtin_speculation_barrier.
15859 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15862 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15863 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15864 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15866 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15867 integral modes instead of "ss" and "sd".
15868 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15869 vectors with 32-bit and 64-bit elements.
15870 (vecdupssescalarmodesuffix): New mode attribute.
15871 (vec_dup<mode>): Use it.
15873 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15876 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15877 frame if argument is passed on stack.
15879 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15882 * ree.c (combine_reaching_defs): Optimize also
15883 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15884 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15886 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15888 PR middle-end/83189
15889 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15891 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15893 PR middle-end/83718
15894 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15895 after they are computed.
15897 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15899 PR tree-optimization/83695
15900 * gimple-loop-linterchange.cc
15901 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15902 reset cached scev information after interchange.
15903 (pass_linterchange::execute): Remove call to scev_reset_htab.
15905 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15907 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15908 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15909 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15910 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15911 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15912 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15913 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15914 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15915 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15916 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15917 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15918 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15919 (V_lane_reg): Likewise.
15920 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15922 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15923 (vfmal_lane_low<mode>_intrinsic,
15924 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15925 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15926 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15927 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15928 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15929 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15931 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15933 * config/arm/arm-cpus.in (fp16fml): New feature.
15934 (ALL_SIMD): Add fp16fml.
15935 (armv8.2-a): Add fp16fml as an option.
15936 (armv8.3-a): Likewise.
15937 (armv8.4-a): Add fp16fml as part of fp16.
15938 * config/arm/arm.h (TARGET_FP16FML): Define.
15939 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15941 * config/arm/arm-modes.def (V2HF): Define.
15942 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15943 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15944 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15945 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15946 vfmsl_low, vfmsl_high): New set of builtins.
15947 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15948 (vfml_op): New code attribute.
15949 (VFMLHALVES): New int iterator.
15950 (VFML, VFMLSEL): New mode attributes.
15951 (V_reg): Define mapping for V2HF.
15952 (V_hi, V_lo): New mode attributes.
15953 (VF_constraint): Likewise.
15954 (vfml_half, vfml_half_selector): New int attributes.
15955 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15957 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15958 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15960 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15961 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15962 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15963 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15965 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15966 Document new effective target and option set.
15968 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15970 * config/arm/arm-cpus.in (armv8_4): New feature.
15971 (ARMv8_4a): New fgroup.
15972 (armv8.4-a): New arch.
15973 * config/arm/arm-tables.opt: Regenerate.
15974 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15975 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15976 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15977 Add matching rules for -march=armv8.4-a and extensions.
15978 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15980 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15983 * config/rx/rx.md (BW): New mode attribute.
15984 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15986 2018-01-11 Richard Biener <rguenther@suse.de>
15988 PR tree-optimization/83435
15989 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15990 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15991 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15993 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15994 Alan Hayward <alan.hayward@arm.com>
15995 David Sherwood <david.sherwood@arm.com>
15997 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15999 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16000 (aarch64_print_address_internal): Use it to check for a zero offset.
16002 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16003 Alan Hayward <alan.hayward@arm.com>
16004 David Sherwood <david.sherwood@arm.com>
16006 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16007 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16008 Return a poly_int64 rather than a HOST_WIDE_INT.
16009 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16010 rather than a HOST_WIDE_INT.
16011 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16012 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16013 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16014 final_offset from HOST_WIDE_INT to poly_int64.
16015 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16016 to_constant when getting the number of units in an Advanced SIMD
16018 (aarch64_builtin_vectorized_function): Check for a constant number
16020 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16022 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16023 attribute instead of GET_MODE_NUNITS.
16024 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16025 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16026 GET_MODE_SIZE for fixed-size registers.
16027 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16028 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16029 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16030 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16031 (aarch64_print_operand, aarch64_print_address_internal)
16032 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16033 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16034 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16035 Handle polynomial GET_MODE_SIZE.
16036 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16037 wider than SImode without modification.
16038 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16039 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16040 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16041 passing and returning SVE modes.
16042 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16043 rather than GEN_INT.
16044 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16045 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16046 (aarch64_allocate_and_probe_stack_space): Likewise.
16047 (aarch64_layout_frame): Cope with polynomial offsets.
16048 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16049 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16050 polynomial offsets.
16051 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16052 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16053 poly_int64 rather than a HOST_WIDE_INT.
16054 (aarch64_get_separate_components, aarch64_process_components)
16055 (aarch64_expand_prologue, aarch64_expand_epilogue)
16056 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16057 (aarch64_anchor_offset): New function, split out from...
16058 (aarch64_legitimize_address): ...here.
16059 (aarch64_builtin_vectorization_cost): Handle polynomial
16060 TYPE_VECTOR_SUBPARTS.
16061 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16063 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16064 number of elements from the PARALLEL rather than the mode.
16065 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16066 rather than GET_MODE_BITSIZE.
16067 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16068 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16069 (aarch64_expand_vec_perm_const_1): Handle polynomial
16070 d->perm.length () and d->perm elements.
16071 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16072 Apply to_constant to d->perm elements.
16073 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16074 polynomial CONST_VECTOR_NUNITS.
16075 (aarch64_move_pointer): Take amount as a poly_int64 rather
16077 (aarch64_progress_pointer): Avoid temporary variable.
16078 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16079 the mode attribute instead of GET_MODE.
16081 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16082 Alan Hayward <alan.hayward@arm.com>
16083 David Sherwood <david.sherwood@arm.com>
16085 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16086 x exists before using it.
16087 (aarch64_add_constant_internal): Rename to...
16088 (aarch64_add_offset_1): ...this. Replace regnum with separate
16089 src and dest rtxes. Handle the case in which they're different,
16090 including when the offset is zero. Replace scratchreg with an rtx.
16091 Use 2 additions if there is no spare register into which we can
16092 move a 16-bit constant.
16093 (aarch64_add_constant): Delete.
16094 (aarch64_add_offset): Replace reg with separate src and dest
16095 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16096 Use aarch64_add_offset_1.
16097 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16098 an rtx rather than an int. Take the delta as a poly_int64
16099 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16100 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16101 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16102 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16103 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16104 and aarch64_add_sp.
16105 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16106 aarch64_add_constant.
16108 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16110 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16111 Use scalar_float_mode.
16113 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16115 * config/aarch64/aarch64-simd.md
16116 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16117 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16118 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16119 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16120 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16121 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16122 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16123 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16124 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16125 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16127 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16130 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16131 targ_options->x_arm_arch_string is non NULL.
16133 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16135 * config/aarch64/aarch64.h
16136 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16138 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16141 * expmed.c (emit_store_flag_force): Swap if const op0
16142 and change VOIDmode to mode of op0.
16144 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16146 PR rtl-optimization/83761
16147 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16148 than bytes to mode_for_size.
16150 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16152 PR middle-end/83189
16153 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16154 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16157 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16159 PR middle-end/83575
16160 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16161 when in layout mode.
16162 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16163 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16166 2018-01-10 Michael Collison <michael.collison@arm.com>
16168 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16169 * config/aarch64/aarch64-option-extension.def: Add
16170 AARCH64_OPT_EXTENSION of 'fp16fml'.
16171 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16172 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16173 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16174 * config/aarch64/constraints.md (Ui7): New constraint.
16175 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16176 (VFMLA_SEL_W): Ditto.
16179 (VFMLA16_LOW): New int iterator.
16180 (VFMLA16_HIGH): Ditto.
16181 (UNSPEC_FMLAL): New unspec.
16182 (UNSPEC_FMLSL): Ditto.
16183 (UNSPEC_FMLAL2): Ditto.
16184 (UNSPEC_FMLSL2): Ditto.
16185 (f16mac): New code attribute.
16186 * config/aarch64/aarch64-simd-builtins.def
16187 (aarch64_fmlal_lowv2sf): Ditto.
16188 (aarch64_fmlsl_lowv2sf): Ditto.
16189 (aarch64_fmlalq_lowv4sf): Ditto.
16190 (aarch64_fmlslq_lowv4sf): Ditto.
16191 (aarch64_fmlal_highv2sf): Ditto.
16192 (aarch64_fmlsl_highv2sf): Ditto.
16193 (aarch64_fmlalq_highv4sf): Ditto.
16194 (aarch64_fmlslq_highv4sf): Ditto.
16195 (aarch64_fmlal_lane_lowv2sf): Ditto.
16196 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16197 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16198 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16199 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16200 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16201 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16202 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16203 (aarch64_fmlal_lane_highv2sf): Ditto.
16204 (aarch64_fmlsl_lane_highv2sf): Ditto.
16205 (aarch64_fmlal_laneq_highv2sf): Ditto.
16206 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16207 (aarch64_fmlalq_lane_highv4sf): Ditto.
16208 (aarch64_fmlsl_lane_highv4sf): Ditto.
16209 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16210 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16211 * config/aarch64/aarch64-simd.md:
16212 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16213 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16214 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16215 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16216 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16217 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16218 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16219 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16220 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16221 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16222 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16223 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16224 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16225 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16226 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16227 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16228 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16229 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16230 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16231 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16232 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16233 (vfmlsl_low_u32): Ditto.
16234 (vfmlalq_low_u32): Ditto.
16235 (vfmlslq_low_u32): Ditto.
16236 (vfmlal_high_u32): Ditto.
16237 (vfmlsl_high_u32): Ditto.
16238 (vfmlalq_high_u32): Ditto.
16239 (vfmlslq_high_u32): Ditto.
16240 (vfmlal_lane_low_u32): Ditto.
16241 (vfmlsl_lane_low_u32): Ditto.
16242 (vfmlal_laneq_low_u32): Ditto.
16243 (vfmlsl_laneq_low_u32): Ditto.
16244 (vfmlalq_lane_low_u32): Ditto.
16245 (vfmlslq_lane_low_u32): Ditto.
16246 (vfmlalq_laneq_low_u32): Ditto.
16247 (vfmlslq_laneq_low_u32): Ditto.
16248 (vfmlal_lane_high_u32): Ditto.
16249 (vfmlsl_lane_high_u32): Ditto.
16250 (vfmlal_laneq_high_u32): Ditto.
16251 (vfmlsl_laneq_high_u32): Ditto.
16252 (vfmlalq_lane_high_u32): Ditto.
16253 (vfmlslq_lane_high_u32): Ditto.
16254 (vfmlalq_laneq_high_u32): Ditto.
16255 (vfmlslq_laneq_high_u32): Ditto.
16256 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16257 (AARCH64_FL_FOR_ARCH8_4): New.
16258 (AARCH64_ISA_F16FML): New ISA flag.
16259 (TARGET_F16FML): New feature flag for fp16fml.
16260 (doc/invoke.texi): Document new fp16fml option.
16262 2018-01-10 Michael Collison <michael.collison@arm.com>
16264 * config/aarch64/aarch64-builtins.c:
16265 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16266 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16267 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16268 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16269 (AARCH64_ISA_SHA3): New ISA flag.
16270 (TARGET_SHA3): New feature flag for sha3.
16271 * config/aarch64/iterators.md (sha512_op): New int attribute.
16272 (CRYPTO_SHA512): New int iterator.
16273 (UNSPEC_SHA512H): New unspec.
16274 (UNSPEC_SHA512H2): Ditto.
16275 (UNSPEC_SHA512SU0): Ditto.
16276 (UNSPEC_SHA512SU1): Ditto.
16277 * config/aarch64/aarch64-simd-builtins.def
16278 (aarch64_crypto_sha512hqv2di): New builtin.
16279 (aarch64_crypto_sha512h2qv2di): Ditto.
16280 (aarch64_crypto_sha512su0qv2di): Ditto.
16281 (aarch64_crypto_sha512su1qv2di): Ditto.
16282 (aarch64_eor3qv8hi): Ditto.
16283 (aarch64_rax1qv2di): Ditto.
16284 (aarch64_xarqv2di): Ditto.
16285 (aarch64_bcaxqv8hi): Ditto.
16286 * config/aarch64/aarch64-simd.md:
16287 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16288 (aarch64_crypto_sha512su0qv2di): Ditto.
16289 (aarch64_crypto_sha512su1qv2di): Ditto.
16290 (aarch64_eor3qv8hi): Ditto.
16291 (aarch64_rax1qv2di): Ditto.
16292 (aarch64_xarqv2di): Ditto.
16293 (aarch64_bcaxqv8hi): Ditto.
16294 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16295 (vsha512h2q_u64): Ditto.
16296 (vsha512su0q_u64): Ditto.
16297 (vsha512su1q_u64): Ditto.
16298 (veor3q_u16): Ditto.
16299 (vrax1q_u64): Ditto.
16300 (vxarq_u64): Ditto.
16301 (vbcaxq_u16): Ditto.
16302 * config/arm/types.md (crypto_sha512): New type attribute.
16303 (crypto_sha3): Ditto.
16304 (doc/invoke.texi): Document new sha3 option.
16306 2018-01-10 Michael Collison <michael.collison@arm.com>
16308 * config/aarch64/aarch64-builtins.c:
16309 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16310 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16311 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16312 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16313 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16314 (AARCH64_ISA_SM4): New ISA flag.
16315 (TARGET_SM4): New feature flag for sm4.
16316 * config/aarch64/aarch64-simd-builtins.def
16317 (aarch64_sm3ss1qv4si): Ditto.
16318 (aarch64_sm3tt1aq4si): Ditto.
16319 (aarch64_sm3tt1bq4si): Ditto.
16320 (aarch64_sm3tt2aq4si): Ditto.
16321 (aarch64_sm3tt2bq4si): Ditto.
16322 (aarch64_sm3partw1qv4si): Ditto.
16323 (aarch64_sm3partw2qv4si): Ditto.
16324 (aarch64_sm4eqv4si): Ditto.
16325 (aarch64_sm4ekeyqv4si): Ditto.
16326 * config/aarch64/aarch64-simd.md:
16327 (aarch64_sm3ss1qv4si): Ditto.
16328 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16329 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16330 (aarch64_sm4eqv4si): Ditto.
16331 (aarch64_sm4ekeyqv4si): Ditto.
16332 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16333 (sm3part_op): Ditto.
16334 (CRYPTO_SM3TT): Ditto.
16335 (CRYPTO_SM3PART): Ditto.
16336 (UNSPEC_SM3SS1): New unspec.
16337 (UNSPEC_SM3TT1A): Ditto.
16338 (UNSPEC_SM3TT1B): Ditto.
16339 (UNSPEC_SM3TT2A): Ditto.
16340 (UNSPEC_SM3TT2B): Ditto.
16341 (UNSPEC_SM3PARTW1): Ditto.
16342 (UNSPEC_SM3PARTW2): Ditto.
16343 (UNSPEC_SM4E): Ditto.
16344 (UNSPEC_SM4EKEY): Ditto.
16345 * config/aarch64/constraints.md (Ui2): New constraint.
16346 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16347 * config/arm/types.md (crypto_sm3): New type attribute.
16348 (crypto_sm4): Ditto.
16349 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16350 (vsm3tt1aq_u32): Ditto.
16351 (vsm3tt1bq_u32): Ditto.
16352 (vsm3tt2aq_u32): Ditto.
16353 (vsm3tt2bq_u32): Ditto.
16354 (vsm3partw1q_u32): Ditto.
16355 (vsm3partw2q_u32): Ditto.
16356 (vsm4eq_u32): Ditto.
16357 (vsm4ekeyq_u32): Ditto.
16358 (doc/invoke.texi): Document new sm4 option.
16360 2018-01-10 Michael Collison <michael.collison@arm.com>
16362 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16363 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16364 (AARCH64_FL_FOR_ARCH8_4): New.
16365 (AARCH64_FL_V8_4): New flag.
16366 (doc/invoke.texi): Document new armv8.4-a option.
16368 2018-01-10 Michael Collison <michael.collison@arm.com>
16370 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16371 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16372 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16373 * config/aarch64/aarch64-option-extension.def: Add
16374 AARCH64_OPT_EXTENSION of 'sha2'.
16375 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16376 (crypto): Disable sha2 and aes if crypto disabled.
16377 (crypto): Enable aes and sha2 if enabled.
16378 (simd): Disable sha2 and aes if simd disabled.
16379 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16381 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16382 (TARGET_SHA2): New feature flag for sha2.
16383 (TARGET_AES): New feature flag for aes.
16384 * config/aarch64/aarch64-simd.md:
16385 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16386 conditional on TARGET_AES.
16387 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16388 (aarch64_crypto_sha1hsi): Make pattern conditional
16390 (aarch64_crypto_sha1hv4si): Ditto.
16391 (aarch64_be_crypto_sha1hv4si): Ditto.
16392 (aarch64_crypto_sha1su1v4si): Ditto.
16393 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16394 (aarch64_crypto_sha1su0v4si): Ditto.
16395 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16396 (aarch64_crypto_sha256su0v4si): Ditto.
16397 (aarch64_crypto_sha256su1v4si): Ditto.
16398 (doc/invoke.texi): Document new aes and sha2 options.
16400 2018-01-10 Martin Sebor <msebor@redhat.com>
16402 PR tree-optimization/83781
16403 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16406 2018-01-11 Martin Sebor <msebor@gmail.com>
16407 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16409 PR tree-optimization/83501
16410 PR tree-optimization/81703
16412 * tree-ssa-strlen.c (get_string_cst): Rename...
16413 (get_string_len): ...to this. Handle global constants.
16414 (handle_char_store): Adjust.
16416 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16417 Jim Wilson <jimw@sifive.com>
16419 * config/riscv/riscv-protos.h (riscv_output_return): New.
16420 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16421 (riscv_attribute_table, riscv_output_return),
16422 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16423 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16424 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16426 (riscv_expand_prologue): Add early return for naked function.
16427 (riscv_expand_epilogue): Likewise.
16428 (riscv_function_ok_for_sibcall): Return false for naked function.
16429 (riscv_set_current_function): New.
16430 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16431 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16432 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16433 * doc/extend.texi (RISC-V Function Attributes): New.
16435 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16437 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16438 check for 128-bit long double before checking TCmode.
16439 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16440 128-bit long doubles before checking TFmode or TCmode.
16441 (FLOAT128_IBM_P): Likewise.
16443 2018-01-10 Martin Sebor <msebor@redhat.com>
16445 PR tree-optimization/83671
16446 * builtins.c (c_strlen): Unconditionally return zero for the empty
16448 Use -Warray-bounds for warnings.
16449 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16450 for non-constant array indices with COMPONENT_REF, arrays of
16451 arrays, and pointers to arrays.
16452 (gimple_fold_builtin_strlen): Determine and set length range for
16453 non-constant character arrays.
16455 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16457 PR middle-end/81897
16458 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16461 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16463 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16465 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16468 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16469 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16470 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16471 indexed_or_indirect_operand predicate.
16472 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16473 (*vsx_le_perm_load_v8hi): Likewise.
16474 (*vsx_le_perm_load_v16qi): Likewise.
16475 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16476 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16477 (*vsx_le_perm_store_v8hi): Likewise.
16478 (*vsx_le_perm_store_v16qi): Likewise.
16479 (eight unnamed splitters): Likewise.
16481 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16483 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16484 * config/rs6000/emmintrin.h: Likewise.
16485 * config/rs6000/mmintrin.h: Likewise.
16486 * config/rs6000/xmmintrin.h: Likewise.
16488 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16491 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16493 * tree.c (tree_nop_conversion): Return true for location wrapper
16495 (maybe_wrap_with_location): New function.
16496 (selftest::check_strip_nops): New function.
16497 (selftest::test_location_wrappers): New function.
16498 (selftest::tree_c_tests): Call it.
16499 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16500 (maybe_wrap_with_location): New decl.
16501 (EXPR_LOCATION_WRAPPER_P): New macro.
16502 (location_wrapper_p): New inline function.
16503 (tree_strip_any_location_wrapper): New inline function.
16505 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16508 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16509 stack_realign_offset for the largest alignment of stack slot
16511 (ix86_find_max_used_stack_alignment): New function.
16512 (ix86_finalize_stack_frame_flags): Use it. Set
16513 max_used_stack_alignment if we don't realign stack.
16514 * config/i386/i386.h (machine_function): Add
16515 max_used_stack_alignment.
16517 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16519 * config/arm/arm.opt (-mbranch-cost): New option.
16520 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16523 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16526 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16527 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16529 2018-01-10 Richard Biener <rguenther@suse.de>
16532 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16533 early out so it also covers the case where we have a non-NULL
16536 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16538 PR tree-optimization/83753
16539 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16540 for non-strided grouped accesses if the number of elements is 1.
16542 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16545 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16546 * i386.h (TARGET_USE_GATHER): Define.
16547 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16549 2018-01-10 Martin Liska <mliska@suse.cz>
16552 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16553 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16555 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16556 CLEANUP_NO_PARTITIONING is not set.
16558 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16560 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16561 for vectors, as a partial revert of r254296.
16562 * rtl.h (const_vec_p): Delete.
16563 (const_vec_duplicate_p): Don't test for vector CONSTs.
16564 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16565 * expmed.c (make_tree): Likewise.
16568 * common.md (E, F): Use CONSTANT_P instead of checking for
16570 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16571 checking for CONST_VECTOR.
16573 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16575 PR middle-end/83575
16576 * predict.c (force_edge_cold): Handle in more sane way edges
16577 with no prediction.
16579 2018-01-09 Carl Love <cel@us.ibm.com>
16581 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16583 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16584 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16585 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16586 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16587 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16588 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16589 * config/rs6000/rs6000-protos.h: Add extern defition for
16590 rs6000_generate_float2_double_code.
16591 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16593 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16594 (float2_v2df): Add define_expand.
16596 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16599 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16600 op_mode in the force_to_mode call.
16602 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16604 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16605 instead of checking each element individually.
16606 (aarch64_evpc_uzp): Likewise.
16607 (aarch64_evpc_zip): Likewise.
16608 (aarch64_evpc_ext): Likewise.
16609 (aarch64_evpc_rev): Likewise.
16610 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16611 instead of checking each element individually. Return true without
16613 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16614 whether all selected elements come from the same input, instead of
16615 checking each element individually. Remove calls to gen_rtx_REG,
16616 start_sequence and end_sequence and instead assert that no rtl is
16619 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16621 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16622 order of HIGH and CONST checks.
16624 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16626 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16627 if the destination isn't an SSA_NAME.
16629 2018-01-09 Richard Biener <rguenther@suse.de>
16631 PR tree-optimization/83668
16632 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16634 (canonicalize_loop_form): ... here, renamed from ...
16635 (canonicalize_loop_closed_ssa_form): ... this and amended to
16636 swap successor edges for loop exit blocks to make us use
16637 the RPO order we need for initial schedule generation.
16639 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16641 PR tree-optimization/64811
16642 * match.pd: When optimizing comparisons with Inf, avoid
16643 introducing or losing exceptions from comparisons with NaN.
16645 2018-01-09 Martin Liska <mliska@suse.cz>
16648 * asan.c (shadow_mem_size): Add gcc_assert.
16650 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16652 Don't save registers in main().
16655 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16656 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16657 * config/avr/avr.c (avr_set_current_function): Don't error if
16658 naked, OS_task or OS_main are specified at the same time.
16659 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16661 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16663 * common/config/avr/avr-common.c (avr_option_optimization_table):
16664 Switch on -mmain-is-OS_task for optimizing compilations.
16666 2018-01-09 Richard Biener <rguenther@suse.de>
16668 PR tree-optimization/83572
16669 * graphite.c: Include cfganal.h.
16670 (graphite_transform_loops): Connect infinite loops to exit
16671 and remove fake edges at the end.
16673 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16675 * ipa-inline.c (edge_badness): Revert accidental checkin.
16677 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16680 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16681 symbols; not inline clones.
16683 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16686 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16687 hard registers. Formatting fixes.
16689 PR preprocessor/83722
16690 * gcc.c (try_generate_repro): Pass
16691 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16692 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16695 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16696 Kito Cheng <kito.cheng@gmail.com>
16698 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16699 (riscv_leaf_function_p): Delete.
16700 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16702 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16704 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16706 (do_ifelse): New function.
16707 (do_isel): New function.
16708 (do_sub3): New function.
16709 (do_add3): New function.
16710 (do_load_mask_compare): New function.
16711 (do_overlap_load_compare): New function.
16712 (expand_compare_loop): New function.
16713 (expand_block_compare): Call expand_compare_loop() when appropriate.
16714 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16715 option description.
16716 (-mblock-compare-inline-loop-limit): New option.
16718 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16721 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16722 Reverse order of second and third operands in first alternative.
16723 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16724 of first and second elements in UNSPEC_VPERMR vector.
16725 (altivec_expand_vec_perm_le): Likewise.
16727 2018-01-08 Jeff Law <law@redhat.com>
16729 PR rtl-optimizatin/81308
16730 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16731 (process_switch): If group_case_labels makes a change, then set
16733 (pass_convert_switch::execute): If a switch is converted, then
16734 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16736 PR rtl-optimization/81308
16737 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16740 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16742 PR target/83663 - Revert r255946
16743 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16744 generation for cases where splatting a value is not useful.
16745 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16746 across a vec_duplicate and a paradoxical subreg forming a vector
16747 mode to a vec_concat.
16749 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16751 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16752 -march=armv8.3-a variants.
16753 * config/arm/t-multilib: Likewise.
16754 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16756 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16758 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16760 (cceq_ior_compare_complement): Give it a name so I can use it, and
16761 change boolean_or_operator predicate to boolean_operator so it can
16762 be used to generate a crand.
16763 (eqne): New code iterator.
16764 (bd/bd_neg): New code_attrs.
16765 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16766 a single define_insn.
16767 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16768 decrement (bdnzt/bdnzf/bdzt/bdzf).
16769 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16770 with the new names of the branch decrement patterns, and added the
16771 names of the branch decrement conditional patterns.
16773 2018-01-08 Richard Biener <rguenther@suse.de>
16775 PR tree-optimization/83563
16776 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16779 2018-01-08 Richard Biener <rguenther@suse.de>
16781 PR middle-end/83713
16782 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16784 2018-01-08 Richard Biener <rguenther@suse.de>
16786 PR tree-optimization/83685
16787 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16788 references to abnormals.
16790 2018-01-08 Richard Biener <rguenther@suse.de>
16793 * dwarf2out.c (output_indirect_strings): Handle empty
16794 skeleton_debug_str_hash.
16795 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16797 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16799 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16800 (emit_store_direct): Likewise.
16801 (arc_trampoline_adjust_address): Likewise.
16802 (arc_asm_trampoline_template): New function.
16803 (arc_initialize_trampoline): Use asm_trampoline_template.
16804 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16805 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16806 * config/arc/arc.md (flush_icache): Delete pattern.
16808 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16810 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16811 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16814 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16817 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16818 by not USED_FOR_TARGET.
16819 (make_pass_resolve_sw_modes): Likewise.
16821 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16823 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16826 2018-01-08 Richard Biener <rguenther@suse.de>
16828 PR middle-end/83580
16829 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16831 2018-01-08 Richard Biener <rguenther@suse.de>
16833 PR middle-end/83517
16834 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16836 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16838 PR middle-end/81897
16839 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16840 basic blocks with a small number of successors.
16841 (convert_control_dep_chain_into_preds): Improve handling of
16843 (dump_predicates): Split apart into...
16844 (dump_pred_chain): ...here...
16845 (dump_pred_info): ...and here.
16846 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16847 (can_chain_union_be_invalidated_p): Improve check for invalidation
16849 (uninit_uses_cannot_happen): Avoid unnecessary if
16850 convert_control_dep_chain_into_preds yielded nothing.
16852 2018-01-06 Martin Sebor <msebor@redhat.com>
16854 PR tree-optimization/83640
16855 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16856 subtracting negative offset from size.
16857 (builtin_access::overlap): Adjust offset bounds of the access to fall
16858 within the size of the object if possible.
16860 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16862 PR rtl-optimization/83699
16863 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16864 extract_bit_field_as_subreg to cases in which the extracted
16865 value is also a vector.
16867 * lra-constraints.c (process_alt_operands): Test for the equivalence
16868 substitutions when detecting a possible reload cycle.
16870 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16873 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16874 by default if flag_selective_schedling{,2}. Formatting fixes.
16876 PR rtl-optimization/83682
16877 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16878 if it has non-VECTOR_MODE element mode.
16879 (vec_duplicate_p): Likewise.
16881 PR middle-end/83694
16882 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16883 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16885 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16888 * config/i386/i386-builtin.def
16889 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16890 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16891 Require also OPTION_MASK_ISA_AVX512F in addition to
16892 OPTION_MASK_ISA_GFNI.
16893 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16894 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16895 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16896 to OPTION_MASK_ISA_GFNI.
16897 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16898 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16899 OPTION_MASK_ISA_AVX512BW.
16900 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16901 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16902 addition to OPTION_MASK_ISA_GFNI.
16903 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16904 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16905 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16906 to OPTION_MASK_ISA_GFNI.
16907 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16908 a requirement for all ISAs rather than any of them with a few
16910 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16912 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16913 bitmasks to be enabled with 3 exceptions, instead of requiring any
16914 enabled ISA with lots of exceptions.
16915 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16916 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16917 Change avx512bw in isa attribute to avx512f.
16918 * config/i386/sgxintrin.h: Add license boilerplate.
16919 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16920 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16921 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16922 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16924 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16925 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16926 temporarily sse2 rather than sse if not enabled already.
16929 * config/i386/sse.md (VI248_VLBW): Rename to ...
16930 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16931 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16932 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16933 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16934 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16935 mode iterator instead of VI248_VLBW.
16937 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16939 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16940 (record_modified): Skip clobbers; add debug output.
16941 (param_change_prob): Use sreal frequencies.
16943 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16945 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16946 punt for user-aligned variables.
16948 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16950 * tree-chrec.c (chrec_contains_symbols): Return true for
16953 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16956 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16957 of (x|y) == x for BICS pattern.
16959 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16961 PR tree-optimization/83605
16962 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16963 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16966 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16968 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16969 * config/epiphany/rtems.h: New file.
16971 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16972 Uros Bizjak <ubizjak@gmail.com>
16975 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16976 QIreg_operand instead of register_operand predicate.
16977 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16978 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16979 comments instead of -fmitigate[-_]rop.
16981 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16984 * cgraphunit.c (symbol_table::compile): Switch to text_section
16985 before calling assembly_start debug hook.
16986 * run-rtl-passes.c (run_rtl_passes): Likewise.
16989 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16991 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16992 range_int_cst_p rather than !symbolic_range_p before calling
16993 extract_range_from_multiplicative_op_1.
16995 2018-01-04 Jeff Law <law@redhat.com>
16997 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16998 redundant test in assertion.
17000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17002 * doc/rtl.texi: Document machine_mode wrapper classes.
17004 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17006 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17007 using tree_to_uhwi.
17009 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17011 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17012 the VEC_PERM_EXPR fold to fail.
17014 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17017 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17018 to switched_sections.
17020 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17023 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17024 test for d.testing.
17026 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17029 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17030 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17032 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17035 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17036 is BLKmode and bitpos not zero or mode change is needed.
17038 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17041 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17044 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17047 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17048 instead of MULT rtx. Update all corresponding splitters.
17049 (*saddl_se): Ditto.
17050 (*ssub<modesuffix>): Ditto.
17051 (*ssubl_se): Ditto.
17052 (*cmp_sadd_di): Update split patterns.
17053 (*cmp_sadd_si): Ditto.
17054 (*cmp_sadd_sidi): Ditto.
17055 (*cmp_ssub_di): Ditto.
17056 (*cmp_ssub_si): Ditto.
17057 (*cmp_ssub_sidi): Ditto.
17058 * config/alpha/predicates.md (const23_operand): New predicate.
17059 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17060 Look for ASHIFT, not MULT inner operand.
17061 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17063 2018-01-04 Martin Liska <mliska@suse.cz>
17065 PR gcov-profile/83669
17066 * gcov.c (output_intermediate_file): Add version to intermediate
17068 * doc/gcov.texi: Document new field 'version' in intermediate
17069 file format. Fix location of '-k' option of gcov command.
17071 2018-01-04 Martin Liska <mliska@suse.cz>
17074 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17076 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17078 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17080 2018-01-03 Martin Sebor <msebor@redhat.com>
17082 PR tree-optimization/83655
17083 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17084 checking calls with invalid arguments.
17086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17088 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17089 (vectorizable_mask_load_store): Delete.
17090 (vectorizable_call): Return false for masked loads and stores.
17091 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17092 instead of gimple_assign_rhs1.
17093 (vectorizable_load): Handle IFN_MASK_LOAD.
17094 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17098 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17100 (vectorizable_mask_load_store): ...here.
17101 (vectorizable_load): ...and here.
17103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17105 * tree-vect-stmts.c (vect_build_all_ones_mask)
17106 (vect_build_zero_merge_argument): New functions, split out from...
17107 (vectorizable_load): ...here.
17109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17111 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17113 (vectorizable_mask_load_store): ...here.
17114 (vectorizable_store): ...and here.
17116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17118 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17120 (vectorizable_mask_load_store): ...here.
17122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17124 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17125 (vect_model_store_cost): Take a vec_load_store_type instead of a
17127 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17128 (vect_model_store_cost): Take a vec_load_store_type instead of a
17130 (vectorizable_mask_load_store): Update accordingly.
17131 (vectorizable_store): Likewise.
17132 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17136 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17137 IFN_MASK_LOAD calls here rather than...
17138 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17141 Alan Hayward <alan.hayward@arm.com>
17142 David Sherwood <david.sherwood@arm.com>
17144 * expmed.c (extract_bit_field_1): For vector extracts,
17145 fall back to extract_bit_field_as_subreg if vec_extract
17148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17149 Alan Hayward <alan.hayward@arm.com>
17150 David Sherwood <david.sherwood@arm.com>
17152 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17153 they are variable or constant sized.
17154 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17155 slots for constant-sized data.
17157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17158 Alan Hayward <alan.hayward@arm.com>
17159 David Sherwood <david.sherwood@arm.com>
17161 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17162 handling COND_EXPRs with boolean comparisons, try to find a better
17163 basis for the mask type than the boolean itself.
17165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17167 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17168 is calculated and how it can be overridden.
17169 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17170 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17172 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17176 Alan Hayward <alan.hayward@arm.com>
17177 David Sherwood <david.sherwood@arm.com>
17179 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17180 Remove the mode argument.
17181 (aarch64_simd_valid_immediate): Remove the mode and inverse
17183 * config/aarch64/iterators.md (bitsize): New iterator.
17184 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17185 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17186 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17187 aarch64_simd_valid_immediate.
17188 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17189 (aarch64_reg_or_bic_imm): Likewise.
17190 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17191 with an insn_type enum and msl with a modifier_type enum.
17192 Replace element_width with a scalar_mode. Change the shift
17193 to unsigned int. Add constructors for scalar_float_mode and
17194 scalar_int_mode elements.
17195 (aarch64_vect_float_const_representable_p): Delete.
17196 (aarch64_can_const_movi_rtx_p)
17197 (aarch64_simd_scalar_immediate_valid_for_move)
17198 (aarch64_simd_make_constant): Update call to
17199 aarch64_simd_valid_immediate.
17200 (aarch64_advsimd_valid_immediate_hs): New function.
17201 (aarch64_advsimd_valid_immediate): Likewise.
17202 (aarch64_simd_valid_immediate): Remove mode and inverse
17203 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17204 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17205 and aarch64_float_const_representable_p on the result.
17206 (aarch64_output_simd_mov_immediate): Remove mode argument.
17207 Update call to aarch64_simd_valid_immediate and use of
17208 simd_immediate_info.
17209 (aarch64_output_scalar_simd_mov_immediate): Update call
17212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17213 Alan Hayward <alan.hayward@arm.com>
17214 David Sherwood <david.sherwood@arm.com>
17216 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17217 (mode_nunits): Likewise CONST_MODE_NUNITS.
17218 * machmode.def (ADJUST_NUNITS): Document.
17219 * genmodes.c (mode_data::need_nunits_adj): New field.
17220 (blank_mode): Update accordingly.
17221 (adj_nunits): New variable.
17222 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17224 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17225 listed in adj_nunits.
17226 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17227 listed in adj_nunits. Don't emit case statements for such modes.
17228 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17229 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17230 nothing if adj_nunits is nonnull.
17231 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17232 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17233 (emit_mode_fbit): Update use of print_maybe_const_decl.
17234 (emit_move_size): Likewise. Treat the array as non-const
17236 (emit_mode_adjustments): Handle adj_nunits.
17238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17240 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17241 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17242 (VECTOR_MODES): Use it.
17243 (make_vector_modes): Take the prefix as an argument.
17245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17246 Alan Hayward <alan.hayward@arm.com>
17247 David Sherwood <david.sherwood@arm.com>
17249 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17250 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17251 for MODE_VECTOR_BOOL.
17252 * machmode.def (VECTOR_BOOL_MODE): Document.
17253 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17254 (make_vector_bool_mode): New function.
17255 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17257 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17258 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17260 * stor-layout.c (int_mode_for_mode): Likewise.
17261 * tree.c (build_vector_type_for_mode): Likewise.
17262 * varasm.c (output_constant_pool_2): Likewise.
17263 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17264 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17265 for MODE_VECTOR_BOOL.
17266 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17267 of mode class checks.
17268 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17269 instead of a list of mode class checks.
17270 (expand_vector_scalar_condition): Likewise.
17271 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17274 Alan Hayward <alan.hayward@arm.com>
17275 David Sherwood <david.sherwood@arm.com>
17277 * machmode.h (mode_size): Change from unsigned short to
17279 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17280 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17281 or if measurement_type is not polynomial.
17282 (fixed_size_mode::includes_p): Check for constant-sized modes.
17283 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17284 return a poly_uint16 rather than an unsigned short.
17285 (emit_mode_size): Change the type of mode_size from unsigned short
17286 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17287 (emit_mode_adjustments): Cope with polynomial vector sizes.
17288 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17290 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17292 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17293 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17294 * caller-save.c (setup_save_areas): Likewise.
17295 (replace_reg_with_saved_mem): Likewise.
17296 * calls.c (emit_library_call_value_1): Likewise.
17297 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17298 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17299 (gen_lowpart_for_combine): Likewise.
17300 * convert.c (convert_to_integer_1): Likewise.
17301 * cse.c (equiv_constant, cse_insn): Likewise.
17302 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17303 (cselib_subst_to_values): Likewise.
17304 * dce.c (word_dce_process_block): Likewise.
17305 * df-problems.c (df_word_lr_mark_ref): Likewise.
17306 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17307 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17308 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17309 (rtl_for_decl_location): Likewise.
17310 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17311 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17312 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17313 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17314 (expand_expr_real_1): Likewise.
17315 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17316 (pad_below): Likewise.
17317 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17318 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17319 * ira.c (get_subreg_tracking_sizes): Likewise.
17320 * ira-build.c (ira_create_allocno_objects): Likewise.
17321 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17322 (ira_sort_regnos_for_alter_reg): Likewise.
17323 * ira-costs.c (record_operand_costs): Likewise.
17324 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17325 (resolve_simple_move): Likewise.
17326 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17327 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17328 (lra_constraints): Likewise.
17329 (CONST_POOL_OK_P): Reject variable-sized modes.
17330 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17331 (add_pseudo_to_slot, lra_spill): Likewise.
17332 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17333 * optabs-query.c (get_best_extraction_insn): Likewise.
17334 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17335 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17336 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17337 * recog.c (offsettable_address_addr_space_p): Likewise.
17338 * regcprop.c (maybe_mode_change): Likewise.
17339 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17340 * regrename.c (build_def_use): Likewise.
17341 * regstat.c (dump_reg_info): Likewise.
17342 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17343 (find_reloads, find_reloads_subreg_address): Likewise.
17344 * reload1.c (eliminate_regs_1): Likewise.
17345 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17346 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17347 (simplify_binary_operation_1, simplify_subreg): Likewise.
17348 * targhooks.c (default_function_arg_padding): Likewise.
17349 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17350 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17351 (verify_gimple_assign_ternary): Likewise.
17352 * tree-inline.c (estimate_move_cost): Likewise.
17353 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17354 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17355 (get_address_cost_ainc): Likewise.
17356 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17357 (vect_supportable_dr_alignment): Likewise.
17358 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17359 (vectorizable_reduction): Likewise.
17360 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17361 (vectorizable_operation, vectorizable_load): Likewise.
17362 * tree.c (build_same_sized_truth_vector_type): Likewise.
17363 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17364 * var-tracking.c (emit_note_insn_var_location): Likewise.
17365 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17366 (ADDR_VEC_ALIGN): Likewise.
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 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17374 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17375 or if measurement_type is polynomial.
17376 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17377 * combine.c (make_extraction): Likewise.
17378 * dse.c (find_shift_sequence): Likewise.
17379 * dwarf2out.c (mem_loc_descriptor): Likewise.
17380 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17381 (extract_bit_field, extract_low_bits): Likewise.
17382 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17383 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17384 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17385 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17386 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17387 * reload.c (find_reloads): Likewise.
17388 * reload1.c (alter_reg): Likewise.
17389 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17390 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17391 * tree-if-conv.c (predicate_mem_writes): Likewise.
17392 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17393 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17394 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17395 * valtrack.c (dead_debug_insert_temp): Likewise.
17396 * varasm.c (mergeable_constant_section): Likewise.
17397 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17400 Alan Hayward <alan.hayward@arm.com>
17401 David Sherwood <david.sherwood@arm.com>
17403 * expr.c (expand_assignment): Cope with polynomial mode sizes
17404 when assigning to a CONCAT.
17406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17407 Alan Hayward <alan.hayward@arm.com>
17408 David Sherwood <david.sherwood@arm.com>
17410 * machmode.h (mode_precision): Change from unsigned short to
17412 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17414 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17415 or if measurement_type is not polynomial.
17416 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17417 in which the mode is already known to be a scalar_int_mode.
17418 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17419 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17421 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17422 for GET_MODE_PRECISION.
17423 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17424 for GET_MODE_PRECISION.
17425 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17427 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17428 (expand_field_assignment, make_extraction): Likewise.
17429 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17430 (get_last_value): Likewise.
17431 * convert.c (convert_to_integer_1): Likewise.
17432 * cse.c (cse_insn): Likewise.
17433 * expr.c (expand_expr_real_1): Likewise.
17434 * lra-constraints.c (simplify_operand_subreg): Likewise.
17435 * optabs-query.c (can_atomic_load_p): Likewise.
17436 * optabs.c (expand_atomic_load): Likewise.
17437 (expand_atomic_store): Likewise.
17438 * ree.c (combine_reaching_defs): Likewise.
17439 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17440 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17441 * tree.h (type_has_mode_precision_p): Likewise.
17442 * ubsan.c (instrument_si_overflow): Likewise.
17444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17445 Alan Hayward <alan.hayward@arm.com>
17446 David Sherwood <david.sherwood@arm.com>
17448 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17449 polynomial numbers of units.
17450 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17451 (valid_vector_subparts_p): New function.
17452 (build_vector_type): Remove temporary shim and take the number
17453 of units as a poly_uint64 rather than an int.
17454 (build_opaque_vector_type): Take the number of units as a
17455 poly_uint64 rather than an int.
17456 * tree.c (build_vector_from_ctor): Handle polynomial
17457 TYPE_VECTOR_SUBPARTS.
17458 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17459 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17460 (build_vector_from_val): If the number of units is variable,
17461 use build_vec_duplicate_cst for constant operands and
17462 VEC_DUPLICATE_EXPR otherwise.
17463 (make_vector_type): Remove temporary is_constant ().
17464 (build_vector_type, build_opaque_vector_type): Take the number of
17465 units as a poly_uint64 rather than an int.
17466 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17468 * cfgexpand.c (expand_debug_expr): Likewise.
17469 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17470 (store_constructor, expand_expr_real_1): Likewise.
17471 (const_scalar_mask_from_tree): Likewise.
17472 * fold-const-call.c (fold_const_reduction): Likewise.
17473 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17474 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17475 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17476 (fold_relational_const): Likewise.
17477 (native_interpret_vector): Likewise. Change the size from an
17478 int to an unsigned int.
17479 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17480 TYPE_VECTOR_SUBPARTS.
17481 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17482 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17483 duplicating a non-constant operand into a variable-length vector.
17484 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17485 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17486 * ipa-icf.c (sem_variable::equals): Likewise.
17487 * match.pd: Likewise.
17488 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17489 * print-tree.c (print_node): Likewise.
17490 * stor-layout.c (layout_type): Likewise.
17491 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17492 * tree-cfg.c (verify_gimple_comparison): Likewise.
17493 (verify_gimple_assign_binary): Likewise.
17494 (verify_gimple_assign_ternary): Likewise.
17495 (verify_gimple_assign_single): Likewise.
17496 * tree-pretty-print.c (dump_generic_node): Likewise.
17497 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17498 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17499 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17500 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17501 (vect_shift_permute_load_chain): Likewise.
17502 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17503 (expand_vector_condition, optimize_vector_constructor): Likewise.
17504 (lower_vec_perm, get_compute_type): Likewise.
17505 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17506 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17507 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17508 (vect_recog_mask_conversion_pattern): Likewise.
17509 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17510 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17511 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17512 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17513 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17514 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17515 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17516 (supportable_widening_operation): Likewise.
17517 (supportable_narrowing_operation): Likewise.
17518 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17520 * varasm.c (output_constant): Likewise.
17522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17523 Alan Hayward <alan.hayward@arm.com>
17524 David Sherwood <david.sherwood@arm.com>
17526 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17527 so that both the length == 3 and length != 3 cases set up their
17528 own permute vectors. Add comments explaining why we know the
17529 number of elements is constant.
17530 (vect_permute_load_chain): Likewise.
17532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17533 Alan Hayward <alan.hayward@arm.com>
17534 David Sherwood <david.sherwood@arm.com>
17536 * machmode.h (mode_nunits): Change from unsigned char to
17538 (ONLY_FIXED_SIZE_MODES): New macro.
17539 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17540 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17541 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17543 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17544 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17545 or if measurement_type is not polynomial.
17546 * genmodes.c (ZERO_COEFFS): New macro.
17547 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17549 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17550 Use ZERO_COEFFS when emitting initializers.
17551 * data-streamer.h (bp_pack_poly_value): New function.
17552 (bp_unpack_poly_value): Likewise.
17553 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17554 for GET_MODE_NUNITS.
17555 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17556 for GET_MODE_NUNITS.
17557 * tree.c (make_vector_type): Remove temporary shim and make
17558 the real function take the number of units as a poly_uint64
17559 rather than an int.
17560 (build_vector_type_for_mode): Handle polynomial nunits.
17561 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17562 * emit-rtl.c (const_vec_series_p_1): Likewise.
17563 (gen_rtx_CONST_VECTOR): Likewise.
17564 * fold-const.c (test_vec_duplicate_folding): Likewise.
17565 * genrecog.c (validate_pattern): Likewise.
17566 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17567 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17568 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17569 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17570 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17571 * rtlanal.c (subreg_get_info): Likewise.
17572 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17573 (vect_grouped_load_supported): Likewise.
17574 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17575 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17576 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17577 (simplify_const_unary_operation, simplify_binary_operation_1)
17578 (simplify_const_binary_operation, simplify_ternary_operation)
17579 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17580 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17581 instead of CONST_VECTOR_NUNITS.
17582 * varasm.c (output_constant_pool_2): Likewise.
17583 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17584 explicit-encoded elements in the XVEC for variable-length vectors.
17586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17588 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17591 Alan Hayward <alan.hayward@arm.com>
17592 David Sherwood <david.sherwood@arm.com>
17594 * coretypes.h (fixed_size_mode): Declare.
17595 (fixed_size_mode_pod): New typedef.
17596 * builtins.h (target_builtins::x_apply_args_mode)
17597 (target_builtins::x_apply_result_mode): Change type to
17598 fixed_size_mode_pod.
17599 * builtins.c (apply_args_size, apply_result_size, result_vector)
17600 (expand_builtin_apply_args_1, expand_builtin_apply)
17601 (expand_builtin_return): Update accordingly.
17603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17605 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17606 * cselib.c (cselib_hash_rtx): Likewise.
17607 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17608 CONST_VECTOR encoding.
17610 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17611 Jeff Law <law@redhat.com>
17614 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17615 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17616 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17617 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17620 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17621 explicitly probe *sp in a noreturn function if there were any callee
17622 register saves or frame pointer is needed.
17624 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17627 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17628 BLKmode for ternary, binary or unary expressions.
17631 * var-tracking.c (delete_vta_debug_insn): New inline function.
17632 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17633 insns from get_insns () to NULL instead of each bb separately.
17634 Use delete_vta_debug_insn. No longer static.
17635 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17636 delete_vta_debug_insns callers.
17637 * rtl.h (delete_vta_debug_insns): Declare.
17638 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17639 instead of variable_tracking_main.
17641 2018-01-03 Martin Sebor <msebor@redhat.com>
17643 PR tree-optimization/83603
17644 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17645 arguments past the endof the argument list in functions declared
17646 without a prototype.
17647 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17648 Avoid checking when arguments are null.
17650 2018-01-03 Martin Sebor <msebor@redhat.com>
17653 * doc/extend.texi (attribute const): Fix a typo.
17654 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17655 issuing -Wsuggest-attribute for void functions.
17657 2018-01-03 Martin Sebor <msebor@redhat.com>
17659 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17660 offset_int::from instead of wide_int::to_shwi.
17661 (maybe_diag_overlap): Remove assertion.
17662 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17663 * gimple-ssa-sprintf.c (format_directive): Same.
17664 (parse_directive): Same.
17665 (sprintf_dom_walker::compute_format_length): Same.
17666 (try_substitute_return_value): Same.
17668 2018-01-03 Jeff Law <law@redhat.com>
17670 PR middle-end/83654
17671 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17672 non-constant residual for zero at runtime and avoid probing in
17673 that case. Reorganize code for trailing problem to mirror handling
17676 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17678 PR tree-optimization/83501
17679 * tree-ssa-strlen.c (get_string_cst): New.
17680 (handle_char_store): Call get_string_cst.
17682 2018-01-03 Martin Liska <mliska@suse.cz>
17684 PR tree-optimization/83593
17685 * tree-ssa-strlen.c: Include tree-cfg.h.
17686 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17687 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17688 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17690 (strlen_dom_walker::before_dom_children): Call
17691 gimple_purge_dead_eh_edges. Dump tranformation with details
17693 (strlen_dom_walker::before_dom_children): Update call by adding
17694 new argument cleanup_eh.
17695 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17697 2018-01-03 Martin Liska <mliska@suse.cz>
17700 * cif-code.def (VARIADIC_THUNK): New enum value.
17701 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17704 2018-01-03 Jan Beulich <jbeulich@suse.com>
17706 * sse.md (mov<mode>_internal): Tighten condition for when to use
17707 vmovdqu<ssescalarsize> for TI and OI modes.
17709 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17711 Update copyright years.
17713 2018-01-03 Martin Liska <mliska@suse.cz>
17716 * ipa-visibility.c (function_and_variable_visibility): Skip
17717 functions with noipa attribure.
17719 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17721 * gcc.c (process_command): Update copyright notice dates.
17722 * gcov-dump.c (print_version): Ditto.
17723 * gcov.c (print_version): Ditto.
17724 * gcov-tool.c (print_version): Ditto.
17725 * gengtype.c (create_file): Ditto.
17726 * doc/cpp.texi: Bump @copying's copyright year.
17727 * doc/cppinternals.texi: Ditto.
17728 * doc/gcc.texi: Ditto.
17729 * doc/gccint.texi: Ditto.
17730 * doc/gcov.texi: Ditto.
17731 * doc/install.texi: Ditto.
17732 * doc/invoke.texi: Ditto.
17734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17736 * vector-builder.h (vector_builder::m_full_nelts): Change from
17737 unsigned int to poly_uint64.
17738 (vector_builder::full_nelts): Update prototype accordingly.
17739 (vector_builder::new_vector): Likewise.
17740 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17741 (vector_builder::operator ==): Likewise.
17742 (vector_builder::finalize): Likewise.
17743 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17744 Take the number of elements as a poly_uint64 rather than an
17746 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17747 from unsigned int to poly_uint64.
17748 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17749 (vec_perm_indices::new_vector): Likewise.
17750 (vec_perm_indices::length): Likewise.
17751 (vec_perm_indices::nelts_per_input): Likewise.
17752 (vec_perm_indices::input_nelts): Likewise.
17753 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17754 number of elements per input as a poly_uint64 rather than an
17755 unsigned int. Use the original encoding for variable-length
17756 vectors, rather than clamping each individual element.
17757 For the second and subsequent elements in each pattern,
17758 clamp the step and base before clamping their sum.
17759 (vec_perm_indices::series_p): Handle polynomial element counts.
17760 (vec_perm_indices::all_in_range_p): Likewise.
17761 (vec_perm_indices_to_tree): Likewise.
17762 (vec_perm_indices_to_rtx): Likewise.
17763 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17764 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17765 (tree_vector_builder::new_binary_operation): Handle polynomial
17766 element counts. Return false if we need to know the number
17767 of elements at compile time.
17768 * fold-const.c (fold_vec_perm): Punt if the number of elements
17769 isn't known at compile time.
17771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17773 * vec-perm-indices.h (vec_perm_builder): Change element type
17774 from HOST_WIDE_INT to poly_int64.
17775 (vec_perm_indices::element_type): Update accordingly.
17776 (vec_perm_indices::clamp): Handle polynomial element_types.
17777 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17778 (vec_perm_indices::all_in_range_p): Likewise.
17779 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17781 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17782 polynomial vec_perm_indices element types.
17783 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17784 * fold-const.c (fold_vec_perm): Likewise.
17785 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17786 * tree-vect-generic.c (lower_vec_perm): Likewise.
17787 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17788 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17789 element type to HOST_WIDE_INT.
17791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17792 Alan Hayward <alan.hayward@arm.com>
17793 David Sherwood <david.sherwood@arm.com>
17795 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17796 rather than an int. Use plus_constant.
17797 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17798 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17801 Alan Hayward <alan.hayward@arm.com>
17802 David Sherwood <david.sherwood@arm.com>
17804 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17805 a HOST_WIDE_INT to a poly_int64.
17807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17808 Alan Hayward <alan.hayward@arm.com>
17809 David Sherwood <david.sherwood@arm.com>
17811 * calls.c (load_register_parameters): Cope with polynomial
17812 mode sizes. Require a constant size for BLKmode parameters
17813 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17814 forces a parameter to be padded at the lsb end in order to
17815 fill a complete number of words, require the parameter size
17816 to be ordered wrt UNITS_PER_WORD.
17818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17819 Alan Hayward <alan.hayward@arm.com>
17820 David Sherwood <david.sherwood@arm.com>
17822 * reload1.c (spill_stack_slot_width): Change element type
17823 from unsigned int to poly_uint64_pod.
17824 (alter_reg): Treat mode sizes as polynomial.
17826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17827 Alan Hayward <alan.hayward@arm.com>
17828 David Sherwood <david.sherwood@arm.com>
17830 * reload.c (complex_word_subreg_p): New function.
17831 (reload_inner_reg_of_subreg, push_reload): Use it.
17833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17834 Alan Hayward <alan.hayward@arm.com>
17835 David Sherwood <david.sherwood@arm.com>
17837 * lra-constraints.c (process_alt_operands): Reject matched
17838 operands whose sizes aren't ordered.
17839 (match_reload): Refer to this check here.
17841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17842 Alan Hayward <alan.hayward@arm.com>
17843 David Sherwood <david.sherwood@arm.com>
17845 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17846 that the mode size is in the set {1, 2, 4, 8, 16}.
17848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17849 Alan Hayward <alan.hayward@arm.com>
17850 David Sherwood <david.sherwood@arm.com>
17852 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17853 Use plus_constant instead of gen_rtx_PLUS.
17855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17856 Alan Hayward <alan.hayward@arm.com>
17857 David Sherwood <david.sherwood@arm.com>
17859 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17860 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17861 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17862 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17863 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17864 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17865 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17866 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17867 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17868 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17870 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17871 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17872 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17873 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17874 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17875 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17876 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17877 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17878 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17879 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17881 * expr.c (emit_move_resolve_push): Treat the input and result
17882 of PUSH_ROUNDING as a poly_int64.
17883 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17884 (emit_push_insn): Likewise.
17885 * lra-eliminations.c (mark_not_eliminable): Likewise.
17886 * recog.c (push_operand): Likewise.
17887 * reload1.c (elimination_effects): Likewise.
17888 * rtlanal.c (nonzero_bits1): Likewise.
17889 * calls.c (store_one_arg): Likewise. Require the padding to be
17890 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17897 Use plus_constant instead of gen_rtx_PLUS.
17899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17900 Alan Hayward <alan.hayward@arm.com>
17901 David Sherwood <david.sherwood@arm.com>
17903 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17904 rather than an int.
17906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17907 Alan Hayward <alan.hayward@arm.com>
17908 David Sherwood <david.sherwood@arm.com>
17910 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17911 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17912 via stack temporaries. Treat the mode size as polynomial too.
17914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17915 Alan Hayward <alan.hayward@arm.com>
17916 David Sherwood <david.sherwood@arm.com>
17918 * expr.c (expand_expr_real_2): When handling conversions involving
17919 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17920 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17921 as a poly_uint64 too.
17923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17924 Alan Hayward <alan.hayward@arm.com>
17925 David Sherwood <david.sherwood@arm.com>
17927 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17930 Alan Hayward <alan.hayward@arm.com>
17931 David Sherwood <david.sherwood@arm.com>
17933 * combine.c (can_change_dest_mode): Handle polynomial
17934 REGMODE_NATURAL_SIZE.
17935 * expmed.c (store_bit_field_1): Likewise.
17936 * expr.c (store_constructor): Likewise.
17937 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17938 and polynomial REGMODE_NATURAL_SIZE.
17939 (gen_lowpart_common): Likewise.
17940 * reginfo.c (record_subregs_of_mode): Likewise.
17941 * rtlanal.c (read_modify_subreg_p): Likewise.
17943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17944 Alan Hayward <alan.hayward@arm.com>
17945 David Sherwood <david.sherwood@arm.com>
17947 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17948 numbers of elements.
17950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17951 Alan Hayward <alan.hayward@arm.com>
17952 David Sherwood <david.sherwood@arm.com>
17954 * match.pd: Cope with polynomial numbers of vector elements.
17956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17957 Alan Hayward <alan.hayward@arm.com>
17958 David Sherwood <david.sherwood@arm.com>
17960 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17961 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
17968 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17969 (ipa_simd_modify_function_body): 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-generic.c (nunits_for_known_piecewise_op): New function.
17976 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17977 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17978 (expand_vector_condition, vector_element): Likewise.
17979 (subparts_gt): New function.
17980 (get_compute_type): Use subparts_gt.
17981 (count_type_subparts): Delete.
17982 (expand_vector_operations_1): Use subparts_gt instead of
17983 count_type_subparts.
17985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17986 Alan Hayward <alan.hayward@arm.com>
17987 David Sherwood <david.sherwood@arm.com>
17989 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17990 (vect_compile_time_alias): ...this new function. Do the calculation
17991 on poly_ints rather than trees.
17992 (vect_prune_runtime_alias_test_list): Update call accordingly.
17994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17995 Alan Hayward <alan.hayward@arm.com>
17996 David Sherwood <david.sherwood@arm.com>
17998 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18000 (vect_schedule_slp_instance): Likewise.
18002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18003 Alan Hayward <alan.hayward@arm.com>
18004 David Sherwood <david.sherwood@arm.com>
18006 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18007 constant and extern definitions for variable-length vectors.
18008 (vect_get_constant_vectors): Note that the number of units
18009 is known to be constant.
18011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18012 Alan Hayward <alan.hayward@arm.com>
18013 David Sherwood <david.sherwood@arm.com>
18015 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18016 of units as polynomial. Choose between WIDE and NARROW based
18019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18020 Alan Hayward <alan.hayward@arm.com>
18021 David Sherwood <david.sherwood@arm.com>
18023 * tree-vect-stmts.c (simd_clone_subparts): New function.
18024 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18027 Alan Hayward <alan.hayward@arm.com>
18028 David Sherwood <david.sherwood@arm.com>
18030 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18031 vectors as polynomial. Use build_index_vector for
18032 IFN_GOMP_SIMD_LANE.
18034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18035 Alan Hayward <alan.hayward@arm.com>
18036 David Sherwood <david.sherwood@arm.com>
18038 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18039 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18040 for variable-length vectors.
18041 (vectorizable_mask_load_store): Treat the number of units as
18042 polynomial, asserting that it is constant if the condition has
18043 already been enforced.
18044 (vectorizable_store, vectorizable_load): Likewise.
18046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18047 Alan Hayward <alan.hayward@arm.com>
18048 David Sherwood <david.sherwood@arm.com>
18050 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18051 of units as polynomial. Punt if we can't tell at compile time
18052 which vector contains the final result.
18054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18055 Alan Hayward <alan.hayward@arm.com>
18056 David Sherwood <david.sherwood@arm.com>
18058 * tree-vect-loop.c (vectorizable_induction): Treat the number
18059 of units as polynomial. Punt on SLP inductions. Use an integer
18060 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18061 cast of such a series for variable-length floating-point
18064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18065 Alan Hayward <alan.hayward@arm.com>
18066 David Sherwood <david.sherwood@arm.com>
18068 * tree.h (build_index_vector): Declare.
18069 * tree.c (build_index_vector): New function.
18070 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18071 of units as polynomial, forcibly converting it to a constant if
18072 vectorizable_reduction has already enforced the condition.
18073 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18074 to create a {1,2,3,...} vector.
18075 (vectorizable_reduction): Treat the number of units as polynomial.
18076 Choose vectype_in based on the largest scalar element size rather
18077 than the smallest number of units. Enforce the restrictions
18080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18081 Alan Hayward <alan.hayward@arm.com>
18082 David Sherwood <david.sherwood@arm.com>
18084 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18085 number of units as polynomial.
18087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18088 Alan Hayward <alan.hayward@arm.com>
18089 David Sherwood <david.sherwood@arm.com>
18091 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18092 * target.def (autovectorize_vector_sizes): Return the vector sizes
18093 by pointer, using vector_sizes rather than a bitmask.
18094 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18095 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18096 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18098 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18099 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18100 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18101 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18102 * omp-general.c (omp_max_vf): Likewise.
18103 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18104 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18105 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18106 * tree-vect-slp.c (vect_slp_bb): Likewise.
18107 * doc/tm.texi: Regenerate.
18108 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18110 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18111 the vector size as a poly_uint64 rather than an unsigned int.
18112 (current_vector_size): Change from an unsigned int to a poly_uint64.
18113 (get_vectype_for_scalar_type): Update accordingly.
18114 * tree.h (build_truth_vector_type): Take the size and number of
18115 units as a poly_uint64 rather than an unsigned int.
18116 (build_vector_type): Add a temporary overload that takes
18117 the number of units as a poly_uint64 rather than an unsigned int.
18118 * tree.c (make_vector_type): Likewise.
18119 (build_truth_vector_type): Take the number of units as a poly_uint64
18120 rather than an unsigned int.
18122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18123 Alan Hayward <alan.hayward@arm.com>
18124 David Sherwood <david.sherwood@arm.com>
18126 * target.def (get_mask_mode): Take the number of units and length
18127 as poly_uint64s rather than unsigned ints.
18128 * targhooks.h (default_get_mask_mode): Update accordingly.
18129 * targhooks.c (default_get_mask_mode): Likewise.
18130 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18131 * doc/tm.texi: Regenerate.
18133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18134 Alan Hayward <alan.hayward@arm.com>
18135 David Sherwood <david.sherwood@arm.com>
18137 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18138 * omp-general.c (omp_max_vf): Likewise.
18139 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18140 (expand_omp_simd): Handle polynomial safelen.
18141 * omp-low.c (omplow_simd_context): Add a default constructor.
18142 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18143 (lower_rec_simd_input_clauses): Update accordingly.
18144 (lower_rec_input_clauses): Likewise.
18146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18147 Alan Hayward <alan.hayward@arm.com>
18148 David Sherwood <david.sherwood@arm.com>
18150 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18151 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18152 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18153 (vect_analyze_slp_cost): Likewise.
18154 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18155 (vect_model_load_cost): Likewise.
18157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18158 Alan Hayward <alan.hayward@arm.com>
18159 David Sherwood <david.sherwood@arm.com>
18161 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18162 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18163 from an unsigned int * to a poly_uint64_pod *.
18164 (calculate_unrolling_factor): New function.
18165 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18168 Alan Hayward <alan.hayward@arm.com>
18169 David Sherwood <david.sherwood@arm.com>
18171 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18172 from an unsigned int to a poly_uint64.
18173 (_loop_vec_info::slp_unrolling_factor): Likewise.
18174 (_loop_vec_info::vectorization_factor): Change from an int
18176 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18177 (vect_get_num_vectors): New function.
18178 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18179 (vect_get_num_copies): Use vect_get_num_vectors.
18180 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18181 to an unsigned int *.
18182 (vect_analyze_data_refs): Change min_vf from an int * to a
18184 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18185 than an unsigned HOST_WIDE_INT.
18186 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18187 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18188 to an unsigned int *.
18189 (vect_analyze_data_ref_dependences): Likewise.
18190 (vect_compute_data_ref_alignment): Handle polynomial vf.
18191 (vect_enhance_data_refs_alignment): Likewise.
18192 (vect_prune_runtime_alias_test_list): Likewise.
18193 (vect_shift_permute_load_chain): Likewise.
18194 (vect_supportable_dr_alignment): Likewise.
18195 (dependence_distance_ge_vf): Take the vectorization factor as a
18196 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18197 (vect_analyze_data_refs): Change min_vf from an int * to a
18199 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18200 vfm1 as a poly_uint64 rather than an int. Make the same change
18201 for the returned bound_scalar.
18202 (vect_gen_vector_loop_niters): Handle polynomial vf.
18203 (vect_do_peeling): Likewise. Update call to
18204 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18205 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18207 * tree-vect-loop.c (vect_determine_vectorization_factor)
18208 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18209 (vect_get_known_peeling_cost): Likewise.
18210 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18211 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18212 (vect_transform_loop): Likewise. Use the lowest possible VF when
18213 updating the upper bounds of the loop.
18214 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18215 rather than an int.
18216 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18217 polynomial unroll factors.
18218 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18219 (vect_make_slp_decision): Likewise.
18220 (vect_supported_load_permutation_p): Likewise, and polynomial
18222 (vect_analyze_slp_cost): Handle polynomial vf.
18223 (vect_slp_analyze_node_operations): Likewise.
18224 (vect_slp_analyze_bb_1): Likewise.
18225 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18226 than an unsigned HOST_WIDE_INT.
18227 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18228 (vectorizable_load): Handle polynomial vf.
18229 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18231 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18234 Alan Hayward <alan.hayward@arm.com>
18235 David Sherwood <david.sherwood@arm.com>
18237 * match.pd: Handle bit operations involving three constants
18238 and try to fold one pair.
18240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18242 * tree-vect-loop-manip.c: Include gimple-fold.h.
18243 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18244 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18245 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18246 Add a path that uses a step of VF instead of 1, but disable it
18248 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18249 and niters_no_overflow parameters. Update calls to
18250 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18251 Create a new SSA name if the latter choses to use a ste other
18252 than zero, and return it via niters_vector_mult_vf_var.
18253 * tree-vect-loop.c (vect_transform_loop): Update calls to
18254 vect_do_peeling, vect_gen_vector_loop_niters and
18255 slpeel_make_loop_iterate_ntimes.
18256 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18257 (vect_gen_vector_loop_niters): Update declarations after above changes.
18259 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18261 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18262 128-bit round to integer instructions.
18263 (ceil<mode>2): Likewise.
18264 (btrunc<mode>2): Likewise.
18265 (round<mode>2): Likewise.
18267 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18269 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18270 unaligned VSX load/store on P8/P9.
18271 (expand_block_clear): Allow the use of unaligned VSX
18272 load/store on P8/P9.
18274 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18276 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18278 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18279 swap associated with both a load and a store.
18281 2018-01-02 Andrew Waterman <andrew@sifive.com>
18283 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18284 * config/riscv/riscv.md (clear_cache): Use it.
18286 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18288 * web.c: Remove out-of-date comment.
18290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18292 * expr.c (fixup_args_size_notes): Check that any existing
18293 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18294 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18295 (emit_single_push_insn): ...here.
18297 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18299 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18300 (const_vector_encoded_nelts): New function.
18301 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18302 (const_vector_int_elt, const_vector_elt): Declare.
18303 * emit-rtl.c (const_vector_int_elt_1): New function.
18304 (const_vector_elt): Likewise.
18305 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18306 of CONST_VECTOR_ELT.
18308 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18310 * expr.c: Include rtx-vector-builder.h.
18311 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18312 directly on the tree encoding.
18313 (const_vector_from_tree): Likewise.
18314 * optabs.c: Include rtx-vector-builder.h.
18315 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18316 sequence of "u" values.
18317 * vec-perm-indices.c: Include rtx-vector-builder.h.
18318 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18319 directly on the vec_perm_indices encoding.
18321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18323 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18324 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18325 * rtx-vector-builder.h: New file.
18326 * rtx-vector-builder.c: Likewise.
18327 * rtl.h (rtx_def::u2): Add a const_vector field.
18328 (CONST_VECTOR_NPATTERNS): New macro.
18329 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18330 (CONST_VECTOR_DUPLICATE_P): Likewise.
18331 (CONST_VECTOR_STEPPED_P): Likewise.
18332 (CONST_VECTOR_ENCODED_ELT): Likewise.
18333 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18334 (unwrap_const_vec_duplicate): Likewise.
18335 (const_vec_series_p): Check for a non-duplicated vector encoding.
18336 Say that the function only returns true for integer vectors.
18337 * emit-rtl.c: Include rtx-vector-builder.h.
18338 (gen_const_vec_duplicate_1): Delete.
18339 (gen_const_vector): Call gen_const_vec_duplicate instead of
18340 gen_const_vec_duplicate_1.
18341 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18342 (gen_const_vec_duplicate): Use rtx_vector_builder.
18343 (gen_const_vec_series): Likewise.
18344 (gen_rtx_CONST_VECTOR): Likewise.
18345 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18346 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18347 Build a new vector rather than modifying a CONST_VECTOR in-place.
18348 (handle_special_swappables): Update call accordingly.
18349 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18350 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18351 Build a new vector rather than modifying a CONST_VECTOR in-place.
18352 (handle_special_swappables): Update call accordingly.
18354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18356 * simplify-rtx.c (simplify_const_binary_operation): Use
18357 CONST_VECTOR_ELT instead of XVECEXP.
18359 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18361 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18362 the selector elements to be different from the data elements
18363 if the selector is a VECTOR_CST.
18364 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18365 ssizetype for the selector.
18367 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18369 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18370 before testing each element individually.
18371 * tree-vect-generic.c (lower_vec_perm): Likewise.
18373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18375 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18376 * selftest-run-tests.c (selftest::run_tests): Call it.
18377 * vector-builder.h (vector_builder::operator ==): New function.
18378 (vector_builder::operator !=): Likewise.
18379 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18380 (vec_perm_indices::all_from_input_p): New function.
18381 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18382 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18383 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18384 instead of reading the VECTOR_CST directly. Detect whether both
18385 vector inputs are the same before constructing the vec_perm_indices,
18386 and update the number of inputs argument accordingly. Use the
18387 utility functions added above. Only construct sel2 if we need to.
18389 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18391 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18392 the broadcast of the low byte.
18393 (expand_mult_highpart): Use an explicit encoding for the permutes.
18394 * optabs-query.c (can_mult_highpart_p): Likewise.
18395 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18396 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18397 (vectorizable_bswap): Likewise.
18398 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18399 explicit encoding for the power-of-2 permutes.
18400 (vect_permute_store_chain): Likewise.
18401 (vect_grouped_load_supported): Likewise.
18402 (vect_permute_load_chain): Likewise.
18404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18406 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18407 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18408 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18409 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18410 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18411 (vect_gen_perm_mask_any): Likewise.
18413 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18415 * int-vector-builder.h: New file.
18416 * vec-perm-indices.h: Include int-vector-builder.h.
18417 (vec_perm_indices): Redefine as an int_vector_builder.
18418 (auto_vec_perm_indices): Delete.
18419 (vec_perm_builder): Redefine as a stand-alone class.
18420 (vec_perm_indices::vec_perm_indices): New function.
18421 (vec_perm_indices::clamp): Likewise.
18422 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18423 (vec_perm_indices::new_vector): New function.
18424 (vec_perm_indices::new_expanded_vector): Update for new
18425 vec_perm_indices class.
18426 (vec_perm_indices::rotate_inputs): New function.
18427 (vec_perm_indices::all_in_range_p): Operate directly on the
18428 encoded form, without computing elided elements.
18429 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18430 encoding. Update for new vec_perm_indices class.
18431 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18432 the given vec_perm_builder.
18433 (expand_vec_perm_var): Update vec_perm_builder constructor.
18434 (expand_mult_highpart): Use vec_perm_builder instead of
18435 auto_vec_perm_indices.
18436 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18437 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18438 or double series encoding as appropriate.
18439 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18440 vec_perm_indices instead of auto_vec_perm_indices.
18441 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18442 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18443 (vect_permute_store_chain): Likewise.
18444 (vect_grouped_load_supported): Likewise.
18445 (vect_permute_load_chain): Likewise.
18446 (vect_shift_permute_load_chain): Likewise.
18447 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18448 (vect_transform_slp_perm_load): Likewise.
18449 (vect_schedule_slp_instance): Likewise.
18450 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18451 (vectorizable_mask_load_store): Likewise.
18452 (vectorizable_bswap): Likewise.
18453 (vectorizable_store): Likewise.
18454 (vectorizable_load): Likewise.
18455 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18456 vec_perm_indices instead of auto_vec_perm_indices. Use
18457 tree_to_vec_perm_builder to read the vector from a tree.
18458 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18459 vec_perm_builder instead of a vec_perm_indices.
18460 (have_whole_vector_shift): Use vec_perm_builder and
18461 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18462 truncation to calc_vec_perm_mask_for_shift.
18463 (vect_create_epilog_for_reduction): Likewise.
18464 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18465 from auto_vec_perm_indices to vec_perm_indices.
18466 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18467 instead of changing individual elements.
18468 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18469 the vector in d.perm.
18470 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18471 from auto_vec_perm_indices to vec_perm_indices.
18472 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18473 instead of changing individual elements.
18474 (arm_vectorize_vec_perm_const): Use new_vector to install
18475 the vector in d.perm.
18476 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18477 Update vec_perm_builder constructor.
18478 (rs6000_expand_interleave): Likewise.
18479 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18480 (rs6000_expand_interleave): Likewise.
18482 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18484 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18485 to qimode could truncate the indices.
18486 * optabs.c (expand_vec_perm_var): Likewise.
18488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18490 * Makefile.in (OBJS): Add vec-perm-indices.o.
18491 * vec-perm-indices.h: New file.
18492 * vec-perm-indices.c: Likewise.
18493 * target.h (vec_perm_indices): Replace with a forward class
18495 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18496 * optabs.h: Include vec-perm-indices.h.
18497 (expand_vec_perm): Delete.
18498 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18499 (expand_vec_perm_const): Declare.
18500 * target.def (vec_perm_const_ok): Replace with...
18501 (vec_perm_const): ...this new hook.
18502 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18503 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18504 * doc/tm.texi: Regenerate.
18505 * optabs.def (vec_perm_const): Delete.
18506 * doc/md.texi (vec_perm_const): Likewise.
18507 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18508 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18509 expand_vec_perm for constant permutation vectors. Assert that
18510 the mode of variable permutation vectors is the integer equivalent
18511 of the mode that is being permuted.
18512 * optabs-query.h (selector_fits_mode_p): Declare.
18513 * optabs-query.c: Include vec-perm-indices.h.
18514 (selector_fits_mode_p): New function.
18515 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18516 is defined, instead of checking whether the vec_perm_const_optab
18517 exists. Use targetm.vectorize.vec_perm_const instead of
18518 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18519 fit in the vector mode before using a variable permute.
18520 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18521 vec_perm_indices instead of an rtx.
18522 (expand_vec_perm): Replace with...
18523 (expand_vec_perm_const): ...this new function. Take the selector
18524 as a vec_perm_indices rather than an rtx. Also take the mode of
18525 the selector. Update call to shift_amt_for_vec_perm_mask.
18526 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18527 Use vec_perm_indices::new_expanded_vector to expand the original
18528 selector into bytes. Check whether the indices fit in the vector
18529 mode before using a variable permute.
18530 (expand_vec_perm_var): Make global.
18531 (expand_mult_highpart): Use expand_vec_perm_const.
18532 * fold-const.c: Includes vec-perm-indices.h.
18533 * tree-ssa-forwprop.c: Likewise.
18534 * tree-vect-data-refs.c: Likewise.
18535 * tree-vect-generic.c: Likewise.
18536 * tree-vect-loop.c: Likewise.
18537 * tree-vect-slp.c: Likewise.
18538 * tree-vect-stmts.c: Likewise.
18539 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18541 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18542 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18543 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18544 (aarch64_vectorize_vec_perm_const): ...this new function.
18545 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18546 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18547 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18548 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18549 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18550 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18551 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18553 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18554 check for NEON modes.
18555 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18556 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18557 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18558 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18560 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18561 the old VEC_PERM_CONST conditions.
18562 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18563 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18564 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18565 (ia64_vectorize_vec_perm_const_ok): Merge into...
18566 (ia64_vectorize_vec_perm_const): ...this new function.
18567 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18568 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18569 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18570 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18571 * config/mips/mips.c (mips_expand_vec_perm_const)
18572 (mips_vectorize_vec_perm_const_ok): Merge into...
18573 (mips_vectorize_vec_perm_const): ...this new function.
18574 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18575 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18576 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18577 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18578 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18579 (rs6000_expand_vec_perm_const): Delete.
18580 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18582 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18583 (altivec_expand_vec_perm_const_le): Take each operand individually.
18584 Operate on constant selectors rather than rtxes.
18585 (altivec_expand_vec_perm_const): Likewise. Update call to
18586 altivec_expand_vec_perm_const_le.
18587 (rs6000_expand_vec_perm_const): Delete.
18588 (rs6000_vectorize_vec_perm_const_ok): Delete.
18589 (rs6000_vectorize_vec_perm_const): New function.
18590 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18591 an element count and rtx array.
18592 (rs6000_expand_extract_even): Update call accordingly.
18593 (rs6000_expand_interleave): Likewise.
18594 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18595 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18596 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18597 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18598 (rs6000_expand_vec_perm_const): Delete.
18599 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18601 (altivec_expand_vec_perm_const_le): Take each operand individually.
18602 Operate on constant selectors rather than rtxes.
18603 (altivec_expand_vec_perm_const): Likewise. Update call to
18604 altivec_expand_vec_perm_const_le.
18605 (rs6000_expand_vec_perm_const): Delete.
18606 (rs6000_vectorize_vec_perm_const_ok): Delete.
18607 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18608 reference to the SPE evmerge intructions.
18609 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18610 an element count and rtx array.
18611 (rs6000_expand_extract_even): Update call accordingly.
18612 (rs6000_expand_interleave): Likewise.
18613 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18614 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18616 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18620 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18621 vector mode and that that mode matches the mode of the data
18623 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18624 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18625 directly using expand_vec_perm_1 when forcing selectors into
18627 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18631 * optabs-query.h (can_vec_perm_p): Delete.
18632 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18633 * optabs-query.c (can_vec_perm_p): Split into...
18634 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18635 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18636 particular selector is valid.
18637 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18638 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18639 (vect_grouped_load_supported): Likewise.
18640 (vect_shift_permute_load_chain): Likewise.
18641 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18642 (vect_transform_slp_perm_load): Likewise.
18643 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18644 (vectorizable_bswap): Likewise.
18645 (vect_gen_perm_mask_checked): Likewise.
18646 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18647 implementations of variable permutation vectors into account
18648 when deciding which selector to use.
18649 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18650 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18651 with a false third argument.
18652 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18653 to test whether the constant selector is valid and can_vec_perm_var_p
18654 to test whether a variable selector is valid.
18656 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18658 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18659 * optabs-query.c (can_vec_perm_p): Likewise.
18660 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18661 instead of vec_perm_indices.
18662 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18663 (vect_gen_perm_mask_checked): Likewise,
18664 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18665 (vect_gen_perm_mask_checked): Likewise,
18667 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18669 * optabs-query.h (qimode_for_vec_perm): Declare.
18670 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18671 (qimode_for_vec_perm): ...this new function.
18672 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18674 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18676 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18677 does not have a conditional at the top.
18679 2018-01-02 Richard Biener <rguenther@suse.de>
18681 * ipa-inline.c (big_speedup_p): Fix expression.
18683 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18686 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18689 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18693 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18694 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18695 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18696 cond_taken_branch_cost 3->4.
18698 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18700 PR tree-optimization/83581
18701 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18702 TODO_cleanup_cfg if any changes have been made.
18704 PR middle-end/83608
18705 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18706 convert_modes if target mode has the right side, but different mode
18709 PR middle-end/83609
18710 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18711 last argument when extracting from CONCAT. If either from_real or
18712 from_imag is NULL, use expansion through memory. If result is not
18713 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18714 the parts directly to inner mode, if even that fails, use expansion
18717 PR middle-end/83623
18718 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18719 check for bswap in mode rather than HImode and use that in expand_unop
18722 Copyright (C) 2018 Free Software Foundation, Inc.
18724 Copying and distribution of this file, with or without modification,
18725 are permitted in any medium without royalty provided the copyright
18726 notice and this notice are preserved.