1 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7 STMT_VINFO_PATTERN_DEF_SEQ to null here.
8 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10 append_pattern_def_seq instead of new_pattern_def_seq.
11 (vect_recog_divmod_pattern): Do both of the above.
12 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
15 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
18 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
19 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
20 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
21 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
22 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
23 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
24 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
25 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
26 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
27 parameter with a single stmt_vec_info.
28 (vect_recog_func_ptr): Likewise.
29 (vect_recog_gather_scatter_pattern): Likewise, folding in...
30 (vect_try_gather_scatter_pattern): ...this.
31 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
32 the stmt_vec_info of the statement to be matched. Don't clear
33 STMT_VINFO_RELATED_STMT.
34 (vect_pattern_recog): Update call accordingly.
36 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
38 PR tree-optimization/85694
39 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
40 (uavgM3_ceil): Document new optabs.
41 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
42 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
44 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
45 (savg_ceil_optab): New optabs.
46 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
47 (vect_vect_recog_func_ptrs): Add it.
48 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
49 constant directly from the associated lhs.
51 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
53 * tree-vect-patterns.c (vect_split_statement): New function.
54 (vect_convert_input): Use it to try to split an existing cast.
56 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
58 * poly-int.h (print_hex): New function.
59 * dumpfile.h (dump_dec, dump_hex): Declare.
60 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
61 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
62 min_input_precision, operation_precision and operation_sign.
63 * tree-vect-patterns.c (vect_get_range_info): New function.
64 (vect_same_loop_or_bb_p, vect_single_imm_use)
65 (vect_operation_fits_smaller_type): Delete.
66 (vect_look_through_possible_promotion): Add an optional
67 single_use_p parameter.
68 (vect_recog_over_widening_pattern): Rewrite to use new
69 stmt_vec_info infomration. Handle one operation at a time.
70 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
71 (vect_truncatable_operation_p, vect_set_operation_type)
72 (vect_set_min_input_precision): New functions.
73 (vect_determine_min_output_precision_1): Likewise.
74 (vect_determine_min_output_precision): Likewise.
75 (vect_determine_precisions_from_range): Likewise.
76 (vect_determine_precisions_from_users): Likewise.
77 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
78 (vect_vect_recog_func_ptrs): Put over_widening first.
80 (vect_pattern_recog): Call vect_determine_precisions.
82 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
84 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
85 statements that have been replaced by further pattern statements.
86 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
88 2018-07-03 Richard Biener <rguenther@suse.de>
90 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
91 always set *dt. Dump vectype in vectype overload.
92 * dumpfile.h (dump_gimple_expr): New function.
93 (dump_gimple_expr_loc): Likewise.
94 * dumpfile.c (dump_gimple_expr): New function.
95 (dump_gimple_expr_loc): Likewise.
97 2018-07-02 Jeff Law <law@redhat.com>
99 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
100 the H8/300, H8/300H and H8/S variants into a single pattern.
101 (movhi_h8300, movqi_h8300hs): Similarly.
102 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
103 (QHI mode iterator): New.
105 * config/h8300/h8300.md: Remove trailing whitespace.
107 2018-07-02 Jim Wilson <jimw@sifive.com>
109 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
110 instead of emit_insn for interrupt returns.
111 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
112 (riscv_sret, riscv_uret): Likewise.
114 2018-07-02 David Malcolm <dmalcolm@redhat.com>
116 * pretty-print.c (selftest::test_pp_format): Move save and restore
117 of quotes to class auto_fix_quotes, and add an instance.
118 * selftest.c: Include "intl.h".
119 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
120 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
121 * selftest.h (selftest::auto_fix_quotes): New class.
123 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
125 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
126 (aarch64_sve_prepare_conditional_op): Remove.
127 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
128 Allow aarch64_simd_reg_or_zero as select operand; remove
129 the aarch64_sve_prepare_conditional_op call.
130 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
131 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
132 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
133 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
134 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
135 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
136 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
137 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
138 and a splitters to match all of the *_any patterns.
139 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
141 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
142 (SVE_COND_FP_BINARY_REV): Remove.
143 (sve_int_op_rev, sve_fp_op_rev): New.
144 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
145 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
146 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
147 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
148 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
149 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
150 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
151 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
152 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
154 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
155 Remove match_dup 1 from the inner unspec.
156 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
158 * config/aarch64/aarch64.md (movprfx): New attr.
159 (length): Default movprfx to 8.
160 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
161 (*madd<SVE_I>, *msub<SVE_I): Likewise.
162 (*<su>mul<SVE_I>3_highpart): Likewise.
163 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
164 (*v<ASHIFT><SVE_I>3): Likewise.
165 (*<su><MAXMIN><SVE_I>3): Likewise.
166 (*<su><MAXMIN><SVE_F>3): Likewise.
167 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
168 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
169 (*div<SVE_F>4): Likewise.
171 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
173 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
176 2018-07-02 Richard Biener <rguenther@suse.de>
178 PR tree-optimization/86363
179 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
180 memset argument refers to a non-variable address.
182 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
184 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
185 {PLUS,MINUS}_EXPR code to...
186 (adjust_symbolic_bound): ...here,
187 (combine_bound): ...here,
188 (set_value_range_with_overflow): ...and here.
190 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
192 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
194 (extract_range_from_abs_expr): ...here.
196 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
198 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
199 -fno-omit-frame-pointer when not optimizing.
201 2018-07-02 Martin Liska <mliska@suse.cz>
204 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
205 (propagate_nothrow): Likewise.
207 2018-07-02 Martin Liska <mliska@suse.cz>
210 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
212 2018-07-02 David Malcolm <dmalcolm@redhat.com>
214 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
215 function in r262149, changing "loc" param from source_location to
216 const dump_location_t &.
217 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
218 declaration, as above.
220 2018-07-01 Paul Koning <ni1d@arrl.net>
222 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
223 -munit-asm, -mgnu-asm, -mdec-asm.
224 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
225 (pdp11_output_labelref): New.
226 (pdp11_output_def): New.
227 (pdp11_output_addr_vec_elt): New.
228 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
229 %# and %@ format codes.
230 (pdp11_option_override): New.
231 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
232 (pdp11_output_ident): New.
233 (pdp11_asm_named_section): New.
234 (pdp11_asm_init_sections): New.
235 (pdp11_file_start): New.
236 (pdp11_file_end): New.
237 (output_ascii): Use .ascii/.asciz for -mdec-asm.
238 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
239 %o, like %c but octal.
240 (pdp11_option_override): New.
241 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
243 (DATA_SECTION_ASM_OP): Ditto.
244 (READONLY_DATA_SECTION_ASM_OP): New.
245 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
246 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
247 (ASM_OUTPUT_LABELREF): Ditto.
248 (ASM_OUTPUT_DEF): Ditto.
249 (ASM_OUTPUT_EXTERNAL): New.
250 (ASM_OUTPUT_SOURCE_FILENAME): New.
251 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
252 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
253 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
254 %# and %@ format codes.
255 * config/pdp11/pdp11.opt (mgnu-asm): New.
256 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
257 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
258 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
260 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
262 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
263 dereferencing path[] beyond its length.
265 (debug_all_paths): New.
266 (rewire_first_differing_edge): New.
267 (adjust_paths_after_duplication): New.
268 (duplicate_thread_path): Call adjust_paths_after_duplication.
270 (thread_through_all_blocks): Add new argument to
271 duplicate_thread_path.
273 2018-06-30 Jim Wilson <jimw@sifive.com>
275 * config/riscv/predicates.md (p2m1_shift_operand): New.
276 (high_mask_shift_operand): New.
277 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
278 pattern using p2m1_shift_operand.
279 (lshsi3_zero_extend_3+2): New combiner pattern using
280 high_mask_shift_operand.
282 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
284 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
286 (vect_recog_rotate_pattern): ...here.
287 (vect_convert_input): Try to insert casts of invariants in the
289 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
290 preheader to be empty.
292 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
294 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
295 vector type. If given, install it in the new statement's
297 (vect_element_precision): New function.
298 (vect_unpromoted_value): New struct.
299 (vect_unpromoted_value::vect_unpromoted_value): New function.
300 (vect_unpromoted_value::set_op): Likewise.
301 (vect_look_through_possible_promotion): Likewise.
302 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
303 (vect_widened_op_tree, vect_convert_input): Likewise.
304 (vect_convert_inputs, vect_convert_output): Likewise.
305 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
306 to handle the optional cast of the multiplication result and
307 vect_widened_op_tree to detect the widened multiplication itself.
308 Do not require the input and output of promotion casts to have
309 the same sign, but base the signedness of the operation on the
310 input rather than the result. If the pattern includes two
311 promotions, check that those promotions have the same sign.
312 Do not restrict the MULT_EXPR handling to a double-width result;
313 handle quadruple-width results and wider. Use vect_convert_inputs
314 to convert the inputs to the common type.
315 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
316 to handle the optional cast of the ABS result. Also allow a sign
317 change or a sign extension between the ABS and MINUS.
318 Use vect_widened_op_tree to detect the widened subtraction and use
319 vect_convert_inputs to convert the inputs to the common type.
320 (vect_handle_widen_op_by_const): Delete.
321 (vect_recog_widen_op_pattern): New function.
322 (vect_recog_widen_mult_pattern): Use it.
323 (vect_recog_widen_shift_pattern): Likewise.
324 (vect_recog_widen_sum_pattern): Use
325 vect_look_through_possible_promotion to handle the promoted
328 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
330 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
331 the containing gimple_seq *.
332 * gimple-iterator.h (gsi_for_stmt): Declare it.
333 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
334 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
335 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
336 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
337 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
339 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
341 (vect_mark_pattern_stmts): ...here. Handle cases in which the
342 statement being replaced is part of an existing pattern
343 definition sequence, inserting the new pattern statements before
345 (vect_pattern_recog_1): Don't return a bool. If the statement
346 is already part of a pattern, instead apply pattern matching
347 to the pattern definition statements. Don't clear the
348 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
349 (vect_pattern_recog): Don't break after the first match;
350 continue processing the pattern definition statements instead.
351 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
353 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
355 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
356 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
357 (vect_recog_widen_sum_pattern): Use it.
359 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
361 * tree-vect-loop.c (vectorizable_reduction): Assert that the
362 phi is not a pattern statement and has not been replaced by
364 * tree-vect-patterns.c (type_conversion_p): Don't check
365 STMT_VINFO_IN_PATTERN_P.
366 (vect_recog_vector_vector_shift_pattern): Likewise.
367 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
368 the pattern statement rather than the original statement; check
369 directly for a WIDEN_MULT_EXPR here.
370 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
371 vect_is_simple_use to return the pattern statement rather
372 than the original statement; use is_pattern_stmt_p to check
373 for such a pattern statement.
374 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
375 to return the pattern statement rather than the original statement;
376 don't do the same transformation here.
377 (vect_is_simple_use): If the defining statement has been replaced
378 by a pattern statement, return the pattern statement instead.
379 Remove the corresponding (local) transformation from the vectype
382 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
384 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
385 end and default to null.
386 * tree-vect-loop.c (vect_create_epilog_for_reduction)
387 (vectorizable_reduction): Update calls accordingly, dropping the
388 gimple ** argument if the passed-back statement isn't needed.
389 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
390 (vect_recog_rotate_pattern): Likewise.
391 (vect_recog_mask_conversion_pattern): Likewise.
392 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
393 (vect_mask_constant_operand_p): Likewise.
394 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
395 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
396 (get_group_load_store_type, get_load_store_type): Likewise.
397 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
398 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
399 (vectorizable_conversion, vectorizable_assignment): Likewise.
400 (vectorizable_shift, vectorizable_operation): Likewise.
401 (vectorizable_store, vect_is_simple_cond): Likewise.
402 (vectorizable_condition, vectorizable_comparison): Likewise.
403 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
404 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
405 and move it to the end. Cope with null def_stmt_outs.
407 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
409 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
411 2018-06-29 Jeff Law <law@redhat.com>
413 * config/v850/v850.c (v850_legitimate_address_p): Handle large
414 displacements for TARGET_V850E2V3 and newer.
415 (TARGET_LRA_P): Remove. Defaults to LRA now.
416 * config/v850/v850.md (sign23byte_load): Remove.
417 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
418 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
420 2018-06-29 Martin Liska <mliska@suse.cz>
423 * coverage.c (coverage_init): Mangle full path name.
424 * doc/invoke.texi: Document the change.
425 * gcov-io.c (mangle_path): New.
426 * gcov-io.h (mangle_path): Likewise.
427 * gcov.c (mangle_name): Use mangle_path for path mangling.
429 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
431 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
432 if starting source register is not even.
434 2018-06-29 Martin Liska <mliska@suse.cz>
436 PR tree-optimization/86263
437 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
438 Make edge redirection.
440 2018-06-29 David Malcolm <dmalcolm@redhat.com>
442 * dumpfile.c (dump_loc): Add indentation based on scope depth.
443 (dump_scope_depth): New variable.
444 (get_dump_scope_depth): New function.
445 (dump_begin_scope): New function.
446 (dump_end_scope): New function.
447 * dumpfile.h (get_dump_scope_depth): New declaration.
448 (dump_begin_scope): New declaration.
449 (dump_end_scope): New declaration.
450 (class auto_dump_scope): New class.
451 (AUTO_DUMP_SCOPE): New macro.
452 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
455 2018-06-29 Richard Biener <rguenther@suse.de>
457 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
458 compute_all_dependences succeeds.
459 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
460 exceed --param loop-max-datarefs-for-datadeps.
462 2018-06-29 Jakub Jelinek <jakub@redhat.com>
464 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
466 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
469 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
470 alternative 0 in preferred_for_speed attribute.
472 2018-06-28 Paul Koning <ni1d@arrl.net>
474 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
475 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
476 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
477 * config/pdp11/pdp11.md: Correct "length" attribute calculation
478 for shift insn patterns.
480 2018-06-28 David Malcolm <dmalcolm@redhat.com>
482 * cgraph.c (cgraph_node::get_body): Replace assignments to
483 "dump_file" with calls to set_dump_file.
484 * dumpfile.c (alt_dump_file): Make static, and group with...
485 (alt_flags): ...this definition.
486 (dumps_are_enabled): New variable.
487 (refresh_dumps_are_enabled): New function.
488 (set_dump_file): New function.
489 (set_alt_dump_file): New function.
490 (gcc::dump_manager::dump_start): Replace assignments to
491 "dump_file" and "alt_dump_file" with calls to set_dump_file and
493 (gcc::dump_manager::dump_finish): Likewise.
494 * dumpfile.h (alt_dump_file): Delete decl.
495 (dumps_are_enabled): New variable decl.
496 (set_dump_file): New function decl.
497 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
499 * tree-nested.c (lower_nested_functions): Replace assignments to
500 "dump_file" with calls to set_dump_file.
502 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
504 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
505 goto_locus of each outgoing edge of each basic block.
507 2018-06-28 Richard Biener <rguenther@suse.de>
509 * dwarf2out.c (decl_scope_table): Remove.
510 (push_decl_scope): Likewise.
511 (pop_decl_scope): Likewise.
512 (gen_type_die_for_member): Do not call push/pop_decl_scope.
513 (gen_struct_or_union_type_die): Likewise.
514 (gen_tagged_type_die): Likewise.
515 (dwarf2out_init): Do not initialize decl_scope_table.
516 (dwarf2out_c_finalize): Do not free it.
518 2018-06-28 Richard Biener <rguenther@suse.de>
520 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
521 deciding whether to not re-use a DIE.
523 2018-06-28 Richard Biener <rguenther@suse.de>
525 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
526 DW_AT_abstract_origin attribute.
528 2018-06-28 Martin Liska <mliska@suse.cz>
530 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
531 Use newly introduced constants.
532 * tree-switch-conversion.h (struct jump_table_cluster):
533 Define max_ratio_for_size and max_ratio_for_speed.
535 2018-06-28 Martin Liska <mliska@suse.cz>
537 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
538 Add new checking assert to catch invalid state.
539 (jump_table_cluster::can_be_handled): Handle single case
541 (jump_table_cluster::is_beneficial): Bail out for such case.
542 (bit_test_cluster::find_bit_tests):
543 Add new checking assert to catch invalid state.
544 (bit_test_cluster::can_be_handled): Handle single case
546 (bit_test_cluster::is_beneficial): Bail out for such case.
547 (switch_decision_tree::analyze_switch_statement):
550 2018-06-28 Martin Liska <mliska@suse.cz>
552 * common.opt: Introduce -completion option.
553 * gcc.c (driver_handle_option): Handle it.
554 (driver::main): Print completions if completion
556 * opt-suggestions.c (option_proposer::get_completions):
558 (option_proposer::suggest_completion): Likewise.
559 (option_proposer::find_param_completions): Likewise.
560 (verify_autocompletions): Likewise.
561 (test_completion_valid_options): Likewise.
562 (test_completion_valid_params): Likewise.
563 (in_completion_p): Likewise.
564 (empty_completion_p): Likewise.
565 (test_completion_partial_match): Likewise.
566 (test_completion_garbage): Likewise.
567 (opt_proposer_c_tests): Likewise.
568 * opt-suggestions.h: Declare new functions.
569 * opts.c (common_handle_option): Handle OPT__completion_.
570 * selftest-run-tests.c (selftest::run_tests): Add
571 opt_proposer_c_tests.
572 * selftest.c (assert_str_startswith): New.
573 * selftest.h (assert_str_startswith): Likewise.
574 (opt_proposer_c_tests): New.
575 (ASSERT_STR_STARTSWITH): Likewise.
577 2018-06-28 Martin Liska <mliska@suse.cz>
579 * Makefile.in: Add opt-suggestions.o.
580 * gcc-main.c: Include opt-suggestions.h.
581 * gcc.c (driver::driver): Likewise.
582 (driver::~driver): Remove m_option_suggestions.
583 (driver::build_option_suggestions): Moved to option_proposer.
584 (driver::suggest_option): Likewise.
585 (driver::handle_unrecognized_options): Use option_proposer.
586 * gcc.h (class driver): Add new memver m_option_proposer.
587 * opt-suggestions.c: New file.
588 * opt-suggestions.h: New file.
590 2018-06-28 Martin Liska <mliska@suse.cz>
592 * vec.h (class auto_string_vec): New (moved from auto_argvec).
593 (auto_string_vec::~auto_string_vec): Likewise.
595 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
597 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
598 prevent_decl_creation_for_types fields up and add reset_location field.
599 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
600 statement if id->reset_location is true.
601 (copy_edges_for_bb): Do not set goto_locus on the new edges if
602 id->reset_location is true.
603 (copy_phis_for_bb): Force input_location on the arguments if
604 id->reset_location is true.
605 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
606 is set on the function to be inlined.
608 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
610 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
612 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
614 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
616 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
617 hard registers for the clobbered pseudo.
619 2018-06-27 Paul Koning <ni1d@arrl.net>
621 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
622 mutually exclusive options.
623 * config/pdp11/constraints.md (h): New constraint.
624 (O): Update definition to match shift code generation.
626 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
628 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
630 (output_jump): Change arguments.
631 (pdp11_fixed_cc_regs): New function.
632 (pdp11_cc_mode): Ditto.
633 (pdp11_expand_shift): Ditto.
634 (pdp11_assemble_shift): Ditto.
635 (pdp11_small_shift): Ditto.
636 (pdp11_branch_cost): Remove.
637 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
639 (pdp11_register_move_cost): Update for CC registers.
640 (pdp11_rtx_costs): Add case for LSHIFTRT.
641 (pdp11_output_jump): Add CCNZ mode conditional branches.
642 (notice_update_cc_on_set): Remove.
643 (pdp11_cc_mode): New function.
644 (simple_memory_operand): Correct pre/post decrement case.
645 (no_side_effect_operand): New function.
646 (pdp11_regno_reg_class): Add CC_REGS class.
647 (pdp11_fixed_cc_regs): New function.
648 (pdp11_small_shift): New function.
649 (pdp11_expand_shift): New function to expand shift insns.
650 (pdp11_assemble_shift): New function to output shifts.
651 (pdp11_branch_cost): Remove.
652 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
653 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
655 (PTRDIFF_TYPE): Ditto.
656 (ADJUST_INSN_LENGTH): New macro.
657 (FIXED_REGISTERS): Add CC registers.
658 (CALL_USED_REGISTERS): Ditto.
660 (REG_CLASS_NAMES): Ditto.
661 (REG_CLASS_CONTENTS): Ditto.
662 (SELECT_CC_MODE): Use new function.
663 (TARGET_FLAGS_REGNUM): New macro.
664 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
665 (cc0_reg_rtx): Remove.
666 (CC_STATUS_MDEP): Remove.
667 (CC_STATUS_MDEFP_INIT): Remove.
669 (NOTICE_UPDATE_CC): Remove.
670 (REGISTER_NAMES): Add CC registers.
671 (BRANCH_COST): Change to constant 1.
672 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
674 * config/pdp11/pdp11.opt (mbcopy): Remove.
675 (mbcopy-builtin): Remove.
676 (mbranch-cheap): Remove.
677 (mbranch-expensive): Remove.
678 * config/pdp11/predicates.md (expand_shift_operand): Update to
679 match shift code generation.
680 (ccnz_operator): New predicate.
681 * doc/invoke.texi (PDP-11 Options): Remove deleted options
682 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
683 Remove non-existent option -mabshi, -mno-abshi. Document mutually
685 * doc/md.texi (PDP-11): Document new D and h constraints. Update
686 description of O constraint.
688 2018-06-27 Jeff Law <law@redhat.com>
689 Austin Law <austinklaw@gmail.com>
691 * config/v850/v850.md (addsi3_set_flags): New pattern.
692 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
693 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
694 (zero_extendhisi2_v850_set_flags): Likewise.
695 (zero_extendqisi2_v850_set_flags): Likewise.
696 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
697 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
698 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
700 * config/v850/v850-protos.h (notice_update_cc): Remove.
701 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
702 (v850_print_operand): Handle 'D' and "d".
703 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
704 Add handling of arithmetic/logical operations compared against zero.
705 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
706 Do not look at v850_compare_op, instead get mode from last argument.
707 (v850_gen_compare): Remove
708 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
709 after reload for prologue insns.
710 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
712 (construct_save_jarl): Likewise.
713 (TARGET_FLAGS_REGNUM): Define.
714 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
715 (NOTICE_UPDATE_CC): Remove.
716 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
717 than cc0. Conditionalize on reload_completed.
718 (cmpsi_insn, setfcc_insn): Likewise.
719 (tst1 splitter): Turn into define_and_split which sets the flags
721 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
722 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
723 (cstoresf4, cstoredf4): Clobber the flags.
724 (cmpsi, cmpsf, cmpdf): Remove expanders.
725 (setf_insn): Remove pattern.
726 (addsi3): Turn into define_and_split which clobbers the flags after
727 reload and a suitable pattern (addsi3_clobber_flags) for use after
729 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
730 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
731 (ashrsi3, ashrsi3_v850e2): Likewise.
732 (bins): Clobber the flags.
733 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
734 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
735 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
736 (call_value_internal_short, call_value_internal_long): Likewise.
737 (callt_save_interrupt, callt_return_interrupt): Likewise.
738 (save_interrupt, return_interrupt): Likewise.
739 (callt_save_all_interrupt, save_all_interrupt): Likewise.
740 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
741 (restore_all_interrupt, _restore_all_interrupt): Likewise.
742 (All FP comparisons): Only allow after reload has completed.
744 (divh, divhu): Tweak output template.
745 (branch_z_normal, branch_z_invert): Remove
746 (branch_nz_normal, branch_nz_invert): Likewise.
747 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
749 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
750 * config/v850/v850.c (notice_update_cc): Remove.
751 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
752 (CC_NO_CARRY): Likewise.
753 (NOTICE_UPDATE_CC): Define to nothing.
754 * config/v850/v850.md: Remove block comment on cc0 handling
755 Remove "cc" attribute from all patterns. Remove cc_status handling
756 from all patterns. Minor formatting fixes.
758 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
760 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
761 (cortex-a76.cortex-a55): Likewise.
762 * config/aarch64/aarch64-tune.md: Regenerate.
763 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
764 cortex-a76.cortex-a55.
766 2018-06-27 Jeff Law <law@redhat.com>
768 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
769 (MULTILIB_DIRNAMES): Similarly.
771 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
773 * gimple.h (gimple_return_retbnd): Delete.
774 (gimple_return_set_retbnd): Likewise.
775 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
776 gimple_return_set_retbnd.
777 * gimple-pretty-print.c (dump_gimple_return): Remove call to
778 gimple_return_retbnd and adjust.
779 * tree-inline.h (struct copy_body_data): Remove retbnd field.
780 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
781 Explicitly return NULL in a couple more cases. Move assertion
782 on debug statements and remove unreachable code.
783 (reset_debug_binding): Do not test id->retbnd.
784 (expand_call_inline): Do not set it.
786 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
788 * configure.ac: Add --disable-gcov option.
789 * configure: Regenerate.
790 * Makefile.in: Honour @enable_gcov@.
791 * doc/install.texi: Document --disable-gcov.
793 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
795 * config/arm/arm-cpus.in (cortex-a76): New entry.
796 (cortex-a76.cortex-a55): Likewise.
797 * config/arm/arm-tables.opt: Regenerate.
798 * config/arm/arm-tune.md: Likewise.
799 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
800 * doc/invoke.texi (ARM Options): Document cortex-a76 and
801 cortex-a76.cortex-a55.
803 2018-06-27 Tamar Christina <tamar.christina@arm.com>
806 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
808 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
810 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
812 (EPILOGUE_USES): Likewise.
814 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
816 * tree-inline.c (remap_location): New function extracted from...
817 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
818 (copy_phis_for_bb): ...here. Call remap_location.
819 (copy_cfg_body): Adjust call to copy_edges_for_bb.
821 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
823 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
824 unaligned vsx for 16B memset.
826 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
829 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
830 ieee128_float_type_node to long_double_type_node unless
831 TARGET_LONG_DOUBLE_128 is set.
833 2018-06-26 David Malcolm <dmalcolm@redhat.com>
835 * cfgloop.c (get_loop_location): Convert return type from
836 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
837 by implicit construction from rtx_insn *, and using
838 dump_user_location_t::from_function_decl for the fallback case.
839 * cfgloop.h (get_loop_location): Convert return type from
840 location_t to dump_user_location_t.
841 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
842 dump_printf_loc to pass in a dump_location_t rather than a
843 location_t, via the gimple stmt.
844 * coverage.c (get_coverage_counts): Update calls to
845 dump_printf_loc to pass in dump_location_t rather than a
847 * doc/optinfo.texi (Dump types): Convert example of
848 dump_printf_loc from taking "locus" to taking "insn". Update
849 description of the "_loc" calls to cover dump_location_t.
850 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
852 (dump_user_location_t::dump_user_location_t): New constructors,
853 from gimple *stmt and rtx_insn *.
854 (dump_user_location_t::from_function_decl): New function.
855 (dump_loc): Make static.
856 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
857 const dump_location_t &.
858 (dump_generic_expr_loc): Delete.
859 (dump_printf_loc): Convert param "loc" from location_t to
860 const dump_location_t &.
861 (selftest::test_impl_location): New function.
862 (selftest::dumpfile_c_tests): New function.
863 * dumpfile.h: Include "profile-count.h".
864 (class dump_user_location_t): New class.
865 (struct dump_impl_location_t): New struct.
866 (class dump_location_t): New class.
867 (dump_printf_loc): Convert 2nd param from source_location to
868 const dump_location_t &.
869 (dump_generic_expr_loc): Delete.
870 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
871 const dump_location_t &.
872 * gimple-fold.c (fold_gimple_assign): Update call to
873 dump_printf_loc to pass in a dump_location_t rather than a
874 location_t, via the gimple stmt.
875 (gimple_fold_call): Likewise.
876 * gimple-loop-interchange.cc
877 (loop_cand::analyze_iloop_reduction_var): Update for change to
878 check_reduction_path.
879 (tree_loop_interchange::interchange): Update for change to
881 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
882 change in return-type of find_loop_location.
883 (graphite_regenerate_ast_isl): Likewise.
884 * graphite-optimize-isl.c (optimize_isl): Likewise.
885 * graphite.c (graphite_transform_loops): Likewise.
886 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
887 pass in a dump_location_t rather than a location_t, via the
889 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
890 * ipa.c (walk_polymorphic_call_targets): Likewise.
891 * loop-unroll.c (report_unroll): Convert "locus" param from
892 location_t to dump_location_t.
893 (decide_unrolling): Update for change to get_loop_location's
895 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
896 location_t to dump_user_location_t.
897 (grid_find_single_omp_among_assignments_1): Updates calls to
898 dump_printf_loc to pass in a dump_location_t rather than a
899 location_t, via the gimple stmt.
900 (grid_parallel_clauses_gridifiable): Convert "tloc" from
901 location_t to dump_location_t. Updates calls to dump_printf_loc
902 to pass in a dump_location_t rather than a location_t, via the
904 (grid_inner_loop_gridifiable_p): Likewise.
905 (grid_dist_follows_simple_pattern): Likewise.
906 (grid_gfor_follows_tiling_pattern): Likewise.
907 (grid_target_follows_gridifiable_pattern): Likewise.
908 (grid_attempt_target_gridification): Convert initialization
909 of local "grid" from memset to zero-initialization; FIXME: does
910 this require C++11? Update call to dump_printf_loc to pass in a
911 optinfo_location rather than a location_t, via the gimple stmt.
912 * profile.c (read_profile_edge_counts): Updates call to
913 dump_printf_loc to pass in a dump_location_t rather than a
915 (compute_branch_probabilities): Likewise.
916 * selftest-run-tests.c (selftest::run_tests): Call
918 * selftest.h (dumpfile_c_tests): New decl.
919 * tree-loop-distribution.c (pass_loop_distribution::execute):
920 Update for change in return type of find_loop_location.
921 * tree-parloops.c (parallelize_loops): Likewise.
922 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
923 "locus" from location_t to dump_user_location_t.
924 (canonicalize_loop_induction_variables): Likewise.
925 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
926 for change in return type of find_loop_location.
927 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
928 to dump_printf_loc to pass in a dump_location_t rather than a
929 location_t, via the stmt.
930 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
932 * tree-vect-loop-manip.c (find_loop_location): Convert return
933 type from source_location to dump_user_location_t.
934 (vect_do_peeling): Update for above change.
935 (vect_loop_versioning): Update for change in type of
937 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
938 from location_t to dump_user_location_t.
939 (vect_estimate_min_profitable_iters): Update for change in type
941 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
942 location_t to dump_location_t.
943 (vect_slp_bb): Update for change in type of vect_location.
944 * tree-vectorizer.c (vect_location): Convert from source_location
945 to dump_user_location_t.
946 (try_vectorize_loop_1): Update for change in vect_location's type.
947 (vectorize_loops): Likewise.
948 (increase_alignment): Likewise.
949 * tree-vectorizer.h (vect_location): Convert from source_location
950 to dump_user_location_t.
951 (find_loop_location): Convert return type from source_location to
952 dump_user_location_t.
953 (check_reduction_path): Convert 1st param from location_t to
954 dump_user_location_t.
955 * value-prof.c (check_counter): Update call to dump_printf_loc to
956 pass in a dump_user_location_t rather than a location_t; update
957 call to error_at for change in type of "locus".
958 (check_ic_target): Update call to dump_printf_loc to
959 pass in a dump_user_location_t rather than a location_t, via the
962 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
964 * config/s390/s390.h (enum processor_flags): Do not use
965 default tune parameter when -march was specified.
967 2018-06-26 Jakub Jelinek <jakub@redhat.com>
970 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
971 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
974 2018-06-26 Richard Biener <rguenther@suse.de>
976 PR tree-optimization/86287
978 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
979 (vect_analyze_loop): Initialize n_stmts.
981 2018-06-26 Richard Biener <rguenther@suse.de>
984 * fold-const.c (fold_convertible_p): Pointer extension
987 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
990 * dwarf2out.c (loc_list_has_views): Adjust comments.
991 (dw_loc_list): Split single cross-partition range with
994 2018-06-25 Jeff Law <law@redhat.com>
996 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
997 on -mbig-switch by default.
999 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1001 (const_float_0_operand): Remove unused predicate.
1002 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1003 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1004 (recipsf2): New expander. Original pattern now called
1006 (recipdf2, recipdf2_insn): Similarly.
1007 (rsqrtsf2, rsqrtsf2_insn): Similarly
1008 (rsqrtdf2, rsqrtdf2_insn): Similarly
1010 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1012 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1013 Simplify logic for FreeBSD (twice).
1015 2018-06-25 Martin Sebor <msebor@redhat.com>
1017 PR tree-optimization/86204
1018 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1019 a strnlen result if it's less than the length of the string.
1021 2018-06-25 Martin Sebor <msebor@redhat.com>
1023 PR tree-optimization/85700
1024 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1025 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1026 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1028 2018-06-25 Martin Sebor <msebor@redhat.com>
1030 * doc/extend.texi (Zero-length arrays): Update and clarify.
1032 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1034 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1035 added IEEE/IBM long double multilib support on PowerPC little
1036 endian Linux systems.
1037 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1038 (MULTILIB_DEFAULTS): Likewise.
1039 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1041 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1042 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1043 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1045 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1048 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1049 (REORDER_45): Likewise.
1051 2018-06-25 Jeff Law <law@redhat.com>
1053 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1054 dividend to 32 bits. Adjust length.
1055 (udivmodhi4): Cleanup output template. Fix length.
1057 2018-06-25 Carl Love <cel@us.ibm.com>
1059 * config/rs6000/vsx.md: Change word selector to prefered location.
1061 2018-06-25 Richard Biener <rguenther@suse.de>
1063 PR tree-optimization/86304
1064 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1065 epilogue-if-converted loops as well.
1067 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1069 * lto-section-out.c (lto_begin_section): Do not print section
1070 name for noaddr and unnumbered dumps.
1072 2018-06-25 Richard Biener <rguenther@suse.de>
1074 * tree-vectorizer.h (struct vec_info_shared): New structure
1075 with parts split out from struct vec_info and loop_nest from
1076 struct _loop_vec_info.
1077 (struct vec_info): Adjust accordingly.
1078 (struct _loop_vec_info): Likewise.
1079 (LOOP_VINFO_LOOP_NEST): Adjust.
1080 (LOOP_VINFO_DATAREFS): Likewise.
1081 (LOOP_VINFO_DDRS): Likewise.
1082 (struct _bb_vec_info): Likewise.
1083 (BB_VINFO_DATAREFS): Likewise.
1084 (BB_VINFO_DDRS): Likewise.
1085 (struct _stmt_vec_info): Add dr_aux member.
1086 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1087 (DR_MISALIGNMENT_UNINITIALIZED): New.
1088 (set_dr_misalignment): Adjust.
1089 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1090 (vect_analyze_loop): Adjust prototype.
1091 (vect_analyze_loop_form): Likewise.
1092 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1093 Compute dependences lazily.
1094 (vect_record_base_alignments): Use shared datarefs/ddrs.
1095 (vect_verify_datarefs_alignment): Likewise.
1096 (vect_analyze_data_refs_alignment): Likewise.
1097 (vect_analyze_data_ref_accesses): Likewise.
1098 (vect_analyze_data_refs): Likewise.
1099 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1100 constructor parameter for shared part.
1101 (vect_analyze_loop_form): Pass in shared part and adjust.
1102 (vect_analyze_loop_2): Pass in storage for the number of
1103 stmts. Move loop nest finding to the caller. Compute
1105 (vect_analyze_loop): Pass in shared part.
1106 (vect_transform_loop): Verify shared datarefs are unchanged.
1107 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1108 constructor parameter for shared part.
1109 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1110 (vect_slp_bb): Verify shared datarefs are unchanged before
1112 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1114 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1115 DR_MISALIGNMENT_UNINITIALIZED.
1116 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1117 parameter for shared part.
1118 (vec_info::~vec_info): Adjust.
1119 (vec_info_shared::vec_info_shared): New.
1120 (vec_info_shared::~vec_info_shared): Likewise.
1121 (vec_info_shared::save_datarefs): Likewise.
1122 (vec_info_shared::check_datarefs): Likewise.
1123 (try_vectorize_loop_1): Construct shared part live for analyses
1124 of a single loop for multiple vector sizes.
1125 * tree-parloops.c (gather_scalar_reductions): Adjust.
1127 2018-06-25 Richard Biener <rguenther@suse.de>
1129 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1130 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1131 (vect_analyze_data_refs): Remove similar code from here and
1132 simplify accordingly.
1134 2018-06-25 Richard Biener <rguenther@suse.de>
1136 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1137 for reverse storage order accesses rather than asserting
1138 they cannot happen here.
1140 2018-06-25 Tom de Vries <tdevries@suse.de>
1143 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1144 Use data16 instead of .byte for insn prefix.
1146 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1149 * parser.c (make_char_string_pack): Pass this literal chars
1150 through cpp_interpret_string.
1151 (cp_parser_userdef_numeric_literal): Check the result of
1152 make_char_string_pack.
1154 2018-06-24 Maya Rashish <coypu@sdf.org>
1156 * ginclude/stddef.h: Simplify conditions around avoiding
1157 re-definition of __size_t.
1159 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1161 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1162 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1164 2018-06-22 Maya Rashish <coypu@sdf.org>
1166 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1167 made to OpenBSD and NetBSD through the years.
1169 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1171 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1172 behavior of vec_pack (vector double, vector double) to match
1173 behavior of vec_float2 (vector double, vector double).
1175 2018-06-22 Olivier Hainque <hainque@adacore.com>
1177 * gimplify.c (gimplify_function_tree): Prevent creation
1178 of a trampoline for the address of the current function
1179 passed to entry/exit instrumentation hooks.
1181 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1184 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1187 2018-06-22 Martin Liska <mliska@suse.cz>
1189 PR tree-optimization/86263
1190 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1191 Bail out if is_enabled is false.
1192 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1194 (jump_table_cluster::is_enabled): New function.
1196 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1198 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1199 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1200 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1201 (lto_input_ts_binfo_tree_pointers): Likewise.
1202 * tree-streamer-out.c (streamer_write_tree_bitfields,
1203 write_ts_binfo_tree_pointers): Likewise.
1204 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1206 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1208 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1210 2018-06-22 Martin Liska <mliska@suse.cz>
1212 * symbol-summary.h (get): Make it pure and inline move
1213 functionality from ::get function.
1214 (get): Remove and inline into ::get and ::get_create.
1215 (get_create): Move code from ::get function.
1217 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1220 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1221 -x assembler-with-cpp.
1223 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1225 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1226 _FILE_OFFSET_BITS=64 for C++.
1228 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1230 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1231 conversion insn that shows up when pr85657-3.c is compiled using
1232 IEEE 128-bit long double.
1233 (neg<mode>2_internal): Use the correct mode to check whether the
1234 mode is IBM extended.
1235 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1236 multiply and divide external functions from being created more
1239 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1241 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1243 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1244 the edge can be forwarded.
1245 (cfg_layout_merge_blocks): Likewise.
1247 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1249 * except.c (finish_eh_generation): Commit edge insertions only after
1250 the EH edges have been redirected from post-landing to landing pads.
1252 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1254 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1255 create_tmp_var_for to create the FRAME decl.
1256 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1258 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1260 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1261 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1262 debug statement when resetting its value.
1263 (expand_call_inline): Copy the locus of the call onto the assignment
1264 of the return value, if any. Use local variable in more cases.
1266 2018-06-21 Martin Liska <mliska@suse.cz>
1268 * ipa-pure-const.c (propagate_nothrow): Use
1269 funct_state_summaries->get.
1270 (dump_malloc_lattice): Likewise.
1271 (propagate_malloc): Likewise.
1273 2018-06-21 Richard Biener <rguenther@suse.de>
1275 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1276 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1277 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1279 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1280 BLOCK_ABSTRACT_ORIGIN unconditionally.
1282 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1284 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1286 * ipa-reference.c (ipa_reference_c_finalize): Delete
1287 ipa_ref_opt_sum_summaries and set it to NULL.
1289 2018-06-21 Tom de Vries <tdevries@suse.de>
1291 PR tree-optimization/85859
1292 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1293 test with comment from bb_no_side_effects_p.
1295 2018-06-21 Richard Biener <rguenther@suse.de>
1297 PR tree-optimization/86232
1298 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1299 max for constant niter.
1301 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1303 * config/aarch64/aarch64-simd.md
1304 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1306 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1308 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1309 Make opernads of the unspec commutative.
1311 2018-06-21 Richard Biener <rguenther@suse.de>
1313 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1314 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1315 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1316 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1318 (vect_record_base_alignments): Do not record base alignment
1319 for gathers or scatters.
1320 (vect_compute_data_ref_alignment): Drop return value that is always
1321 true. Bail out early for gathers or scatters.
1322 (vect_enhance_data_refs_alignment): Bail out early for gathers
1324 (vect_find_same_alignment_drs): Likewise.
1325 (vect_analyze_data_refs_alignment): Remove dead code.
1326 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1327 (vect_analyze_data_refs): For possible gathers or scatters do
1328 not create an alternate DR, just check their possible validity
1329 and mark them. Adjust DECL_NONALIASED handling to not rely
1331 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1332 update inits of gathers or scatters.
1333 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1334 Also copy gather/scatter flag to pattern vinfo.
1336 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1338 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1339 behavior of vec_packsu (vector unsigned long long, vector unsigned
1340 long long) to match behavior of vec_packs with same signature.
1342 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1343 Thomas Schwinge <thomas@codesourcery.com>
1344 Cesar Philippidis <cesar@codesourcery.com>
1346 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1347 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1348 (gimplify_adjust_omp_clauses): Likewise.
1349 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1350 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1351 (gimplify_omp_target_update): Update handling of acc update and
1353 * omp-low.c (install_var_field): Remove unused parameter
1354 base_pointers_restrict.
1355 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1356 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1358 (omp_target_base_pointers_restrict_p): Delete.
1359 (scan_omp_target): Update call to scan_sharing_clauses.
1360 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1362 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1363 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1364 (convert_local_omp_clauses): Likewise.
1365 * tree-pretty-print.c (dump_omp_clause): Likewise.
1366 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1368 (omp_clause_code_name): Likewise.
1370 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1373 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1376 PR tree-optimization/86231
1377 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1378 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1380 2018-06-20 Tom de Vries <tdevries@suse.de>
1382 PR tree-optimization/86097
1383 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1384 iv type if signedness of iv type is not the same as that of *nit.
1386 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1388 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1389 EDGE_EH edges, verify they are all EDGE_EH.
1391 2018-06-20 Maya Rashish <coypu@sdf.org>
1393 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1395 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1397 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1398 * config/aarch64/aarch64.c (xgene1_tunings): Add
1399 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1400 (aarch64_mode_valid_for_sched_fusion_p):
1401 Allow 16-byte modes.
1402 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1403 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1405 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1407 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1408 * config/aarch64/iterators.md (VQ2): New mode iterator.
1410 2018-06-20 Martin Liska <mliska@suse.cz>
1412 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1413 Change default ratio from 10 to 8.
1415 2018-06-20 Martin Liska <mliska@suse.cz>
1417 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1419 (bit_test_cluster::find_bit_tests): Likewise.
1420 (switch_decision_tree::analyze_switch_statement): Find clusters.
1421 * tree-switch-conversion.h (struct jump_table_cluster): Document
1424 2018-06-20 Martin Liska <mliska@suse.cz>
1426 * tree-switch-conversion.c (switch_conversion::collect):
1427 Record m_uniq property.
1428 (switch_conversion::expand): Bail out for special conditions.
1429 (group_cluster::~group_cluster): New.
1430 (group_cluster::group_cluster): Likewise.
1431 (group_cluster::dump): Likewise.
1432 (jump_table_cluster::emit): New.
1433 (switch_decision_tree::fix_phi_operands_for_edges): New.
1434 (struct case_node): Remove struct.
1435 (jump_table_cluster::can_be_handled): New.
1436 (case_values_threshold): Moved to header.
1437 (reset_out_edges_aux): Likewise.
1438 (jump_table_cluster::is_beneficial): New.
1439 (bit_test_cluster::can_be_handled): Likewise.
1440 (add_case_node): Remove.
1441 (bit_test_cluster::is_beneficial): New.
1442 (case_bit_test::cmp): New.
1443 (bit_test_cluster::emit): New.
1444 (expand_switch_as_decision_tree_p): Remove.
1445 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1446 (fix_phi_operands_for_edge): Likewise.
1447 (switch_decision_tree::analyze_switch_statement): New.
1448 (compute_cases_per_edge): Move ...
1449 (switch_decision_tree::compute_cases_per_edge): ... here.
1450 (try_switch_expansion): Likewise.
1451 (switch_decision_tree::try_switch_expansion): Likewise.
1452 (record_phi_operand_mapping): Likewise.
1453 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1454 (emit_case_decision_tree): Likewise.
1455 (switch_decision_tree::emit): Likewise.
1456 (balance_case_nodes): Likewise.
1457 (switch_decision_tree::balance_case_nodes): Likewise.
1458 (dump_case_nodes): Likewise.
1459 (switch_decision_tree::dump_case_nodes): Likewise.
1460 (emit_jump): Likewise.
1461 (switch_decision_tree::emit_jump): Likewise.
1462 (emit_cmp_and_jump_insns): Likewise.
1463 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1464 (emit_case_nodes): Likewise.
1465 (switch_decision_tree::emit_case_nodes): Likewise.
1466 (conditional_probability): Remove.
1467 * tree-switch-conversion.h (enum cluster_type): New.
1469 (struct cluster): Likewise.
1470 (cluster::cluster): Likewise.
1471 (struct simple_cluster): Likewise.
1472 (simple_cluster::simple_cluster): Likewise.
1473 (struct group_cluster): Likewise.
1474 (struct jump_table_cluster): Likewise.
1475 (struct bit_test_cluster): Likewise.
1476 (struct min_cluster_item): Likewise.
1477 (struct case_tree_node): Likewise.
1478 (case_tree_node::case_tree_node): Likewise.
1479 (jump_table_cluster::case_values_threshold): Likewise.
1480 (struct case_bit_test): Likewise.
1481 (struct switch_decision_tree): Likewise.
1482 (struct switch_conversion): Likewise.
1483 (switch_decision_tree::reset_out_edges_aux): Likewise.
1485 2018-06-20 Martin Liska <mliska@suse.cz>
1487 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1488 (hoist_edge_and_branch_if_true): Likewise.
1489 (expand_switch_using_bit_tests_p): Likewise.
1490 (struct case_bit_test): Likewise.
1491 (case_bit_test_cmp): Likewise.
1492 (emit_case_bit_tests): Likewise.
1493 (switch_conversion::switch_conversion): New class.
1494 (struct switch_conv_info): Remove old struct.
1495 (collect_switch_conv_info): More to ...
1496 (switch_conversion::collect): ... this.
1497 (check_range): Likewise.
1498 (switch_conversion::check_range): Likewise.
1499 (check_all_empty_except_final): Likewise.
1500 (switch_conversion::check_all_empty_except_final): Likewise.
1501 (check_final_bb): Likewise.
1502 (switch_conversion::check_final_bb): Likewise.
1503 (create_temp_arrays): Likewise.
1504 (switch_conversion::create_temp_arrays): Likewise.
1505 (free_temp_arrays): Likewise.
1506 (gather_default_values): Likewise.
1507 (switch_conversion::gather_default_values): Likewise.
1508 (build_constructors): Likewise.
1509 (switch_conversion::build_constructors): Likewise.
1510 (constructor_contains_same_values_p): Likewise.
1511 (switch_conversion::contains_same_values_p): Likewise.
1512 (array_value_type): Likewise.
1513 (switch_conversion::array_value_type): Likewise.
1514 (build_one_array): Likewise.
1515 (switch_conversion::build_one_array): Likewise.
1516 (build_arrays): Likewise.
1517 (switch_conversion::build_arrays): Likewise.
1518 (gen_def_assigns): Likewise.
1519 (switch_conversion::gen_def_assigns): Likewise.
1520 (prune_bbs): Likewise.
1521 (switch_conversion::prune_bbs): Likewise.
1522 (fix_phi_nodes): Likewise.
1523 (switch_conversion::fix_phi_nodes): Likewise.
1524 (gen_inbound_check): Likewise.
1525 (switch_conversion::gen_inbound_check): Likewise.
1526 (process_switch): Use the newly created class.
1527 (switch_conversion::expand): New.
1528 (switch_conversion::~switch_conversion): New.
1529 * tree-switch-conversion.h: New file.
1531 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1533 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1534 tree-vect-patterns.c.
1535 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1536 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1537 (vect_recog_sad_pattern): Likewise.
1538 (vect_recog_widen_sum_pattern): Likewise.
1539 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1540 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1541 (vect_recog_rotate_pattern): Likewise.
1542 (vect_recog_mult_pattern): Likewise.
1543 (vect_recog_vector_vector_shift_pattern): Likewise.
1544 (vect_recog_divmod_pattern): Likewise.
1545 (vect_recog_mixed_size_cond_pattern): Likewise.
1546 (vect_recog_bool_pattern): Likewise.
1547 (vect_recog_mask_conversion_pattern): Likewise.
1548 (vect_try_gather_scatter_pattern): Likewise.
1549 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1550 (vect_recog_over_widening_pattern): Likewise.
1551 (vect_recog_gather_scatter_pattern): Likewise.
1552 (vect_recog_func_ptr): Move from tree-vectorizer.h
1553 (vect_vect_recog_func_ptrs): Move further down the file.
1554 (vect_recog_func): Likewise. Remove the third argument.
1555 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1556 (vect_pattern_recog_1): Expect the pattern function to do any
1557 necessary target tests. Also expect it to provide a vector type.
1558 Remove the type_in handling.
1560 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1562 * tree-vect-patterns.c (vect_pattern_detected): New function.
1563 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1564 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1565 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1566 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1567 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1568 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1569 (vect_recog_mask_conversion_pattern)
1570 (vect_try_gather_scatter_pattern): Likewise.
1572 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1574 * tree-vect-patterns.c (vect_get_internal_def): New function.
1575 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1576 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1577 (search_type_for_mask_1): Use it.
1579 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1581 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1582 redundant WIDEN_SUM_EXPR handling.
1583 (vect_recog_sad_pattern): Likewise.
1585 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1587 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1588 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1589 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1590 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1591 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1593 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1595 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1596 use the stmt_vec_info of the original bb statement for the
1597 new zero assignment, even if the call is part of a pattern.
1599 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1601 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1602 that the sequence is attached to the original statement rather
1603 than the pattern statement.
1604 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1605 PATTERN_DEF_SEQ from the original statement rather than
1606 the main pattern statement.
1607 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1608 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1609 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1611 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1613 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1614 definition statements before the early exit for statements that aren't
1616 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1618 (vect_transform_loop): ...here. Process pattern definition
1619 statements without first checking whether the main pattern
1620 statement is live or relevant.
1622 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1624 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1625 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1627 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1629 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1630 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1631 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1632 (expand_block_compare): Change select_block_compare_mode call.
1633 (expand_strncmp_align_check): Use new functions, fix comment.
1634 (emit_final_str_compare_gpr): New function.
1635 (expand_strn_compare): Refactor and clean up code.
1636 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1638 2018-06-19 Tony Reix <tony.reix@atos.com>
1639 Damien Bergamini <damien.bergamini@atos.com>
1640 David Edelsohn <dje.gcc@gmail.com>
1642 * collect2.c (static_obj): New variable.
1643 (static_libs): New variable.
1644 (is_in_list): Uncomment declaration.
1645 (main): Track AIX libraries linked statically.
1646 (is_in_list): Uncomment definition.
1647 (scan_prog_file): Don't add AIX shared libraries initializer
1648 to constructor list if linking statically.
1650 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1652 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1654 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1656 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1658 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1661 2018-06-19 Martin Liska <mliska@suse.cz>
1663 * config/i386/i386.c (ix86_can_inline_p): Do not use
1664 ipa_fn_summaries::get_create.
1665 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1667 (devirtualization_time_bonus): Likewise.
1668 (ipcp_propagate_stage): Likewise.
1669 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1670 (edge_set_predicate): Likewise.
1671 (evaluate_conditions_for_known_args): Likewise.
1672 (evaluate_properties_for_edge): Likewise.
1673 (ipa_call_summary::reset): Tranform to ...
1674 (ipa_call_summary::~ipa_call_summary): ... this.
1675 (ipa_fn_summary::reset): Transform to ...
1676 (ipa_fn_summary::~ipa_fn_summary): ... this.
1677 (ipa_fn_summary_t::remove): Rename to ...
1678 (ipa_fn_summary_t::remove_callees): ... this.
1679 (ipa_fn_summary_t::duplicate): Use placement new
1680 instead of memory copy.
1681 (ipa_call_summary_t::duplicate): Likewise.
1682 (ipa_call_summary_t::remove): Remove.
1683 (dump_ipa_call_summary): Change get_create to get.
1684 (ipa_dump_fn_summary): Dump only when summary exists.
1685 (analyze_function_body): Use symbol_summary::get instead
1687 (compute_fn_summary): Likewise.
1688 (estimate_edge_devirt_benefit): Likewise.
1689 (estimate_edge_size_and_time): Likewise.
1690 (inline_update_callee_summaries): Likewise.
1691 (remap_edge_change_prob): Likewise.
1692 (remap_edge_summaries): Likewise.
1693 (ipa_merge_fn_summary_after_inlining): Likewise.
1694 (write_ipa_call_summary): Likewise.
1695 (ipa_fn_summary_write): Likewise.
1696 (ipa_free_fn_summary): Likewise.
1697 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1698 (struct ipa_call_summary): Likewise.
1699 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1701 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1702 (estimate_size_after_inlining): Likewise.
1703 (estimate_growth): Likewise.
1704 (growth_likely_positive): Likewise.
1705 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1706 (inline_call): Likewise.
1707 * ipa-inline.c (caller_growth_limits): Likewise.
1708 (can_inline_edge_p): Likewise.
1709 (can_inline_edge_by_limits_p): Likewise.
1710 (compute_uninlined_call_time): Likewise.
1711 (compute_inlined_call_time): Likewise.
1712 (want_inline_small_function_p): Likewise.
1713 (edge_badness): Likewise.
1714 (update_caller_keys): Likewise.
1715 (update_callee_keys): Likewise.
1716 (inline_small_functions): Likewise.
1717 (inline_to_all_callers_1): Likewise.
1718 (dump_overall_stats): Likewise.
1719 (early_inline_small_functions): Likewise.
1720 (early_inliner): Likewise.
1721 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1722 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1723 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1724 * ipa-split.c (execute_split_functions): Likewise.
1725 * symbol-summary.h: Likewise.
1726 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1728 2018-06-19 Richard Biener <rguenther@suse.de>
1730 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1731 (vectorize_loops): ... here. Fix dbgcnt handling.
1732 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1734 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1737 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1738 ieee128 argument takes up only one (vector) register, not two (floating
1741 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1743 * gimplify.c (gimplify_init_constructor): Really never clear for an
1744 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1746 2018-06-19 Richard Biener <rguenther@suse.de>
1748 PR tree-optimization/86179
1749 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1750 after failed recognition.
1752 2018-06-18 Martin Sebor <msebor@redhat.com>
1755 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1756 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1757 Handle integer subtraction.
1758 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1759 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1761 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1763 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1764 param from rtx to rtx_insn *.
1765 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1767 (frv_ifcvt_modify_insn): Likwise.
1768 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1769 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1770 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1771 as_a <rtx_insn *> cast to local "unprotected_region" once
1772 it's been established that it's not NULL or pc_rtx.
1773 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1774 param "sethi" from rtx to rtx_insn *.
1775 (nds32_group_float_insns): Likewise for param "insn".
1776 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1778 (vax_output_int_subtract): Likewise.
1779 * config/vax/vax.c (vax_output_int_add): Likewise for param
1781 (vax_output_int_subtract): Likewise.
1782 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1783 (emit_pattern_after): Likewise for param "after".
1784 (emit_insn_after): Likewise.
1785 (emit_jump_insn_after): Likewise.
1786 (emit_call_insn_after): Likewise.
1787 (emit_debug_insn_after): Likewise.
1788 (emit_pattern_before): Likewise for param "before".
1789 (emit_insn_before): Likewise.
1790 (emit_jump_insn_before): Likewise.
1791 * final.c (get_insn_template): Likewise for param "insn", removing
1793 * output.h (get_insn_template): Likewise for 2nd param.
1794 * rtl.h (emit_insn_before): Likewise.
1795 (emit_jump_insn_before): Likewise.
1796 (emit_debug_insn_before_noloc): Likewise.
1797 (emit_insn_after): Likewise.
1798 (emit_jump_insn_after): Likewise.
1799 (emit_call_insn_after): Likewise.
1800 (emit_debug_insn_after): Likewise.
1801 (set_insn_deleted): Likewise for param.
1803 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1806 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1807 floating point modes, so that IFmode is numerically greater than
1808 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1809 to declare the ordering. This prevents IFmode from being
1810 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1811 machine. Include rs6000-modes.h to share the fractional values
1812 between genmodes* and the rest of the compiler.
1816 * config/rs6000/rs6000-modes.h: New file.
1817 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1818 meaning of rs6000_long_double_size so that 126..128 selects an
1819 appropriate 128-bit floating point type.
1820 (rs6000_option_override_internal): Likewise.
1821 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1822 (TARGET_LONG_DOUBLE_128): Change the meaning of
1823 rs6000_long_double_size so that 126..128 selects an appropriate
1824 128-bit floating point type.
1825 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1826 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1827 source and destination to match the standard usage.
1828 (truncifkf2): Likewise.
1829 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1830 ISA 2.07 to use an explicit clobber, instead of passing in a
1832 (copysign<mode>3_soft): Likewise.
1834 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1836 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1837 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1838 (vect_slp_analyze_instance_dependence): Likewise.
1839 (vect_enhance_data_refs_alignment): Likewise.
1840 (vect_analyze_data_refs_alignment): Likewise.
1841 (vect_slp_analyze_and_verify_instance_alignment
1842 (vect_analyze_data_ref_accesses): Likewise.
1843 (vect_prune_runtime_alias_test_list): Likewise.
1844 (vect_analyze_data_refs): Likewise.
1845 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1846 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1847 (vect_analyze_scalar_cycles_1): Likewise.
1848 (vect_get_loop_niters): Likewise.
1849 (vect_analyze_loop_form_1): Likewise.
1850 (vect_update_vf_for_slp): Likewise.
1851 (vect_analyze_loop_operations): Likewise.
1852 (vect_analyze_loop): Likewise.
1853 (vectorizable_induction): Likewise.
1854 (vect_transform_loop): Likewise.
1855 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1856 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1857 (vect_make_slp_decision): Likewise.
1858 (vect_detect_hybrid_slp): Likewise.
1859 (vect_slp_analyze_operations): Likewise.
1860 (vect_slp_bb): Likewise.
1861 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1862 (vectorizable_bswap): Likewise.
1863 (vectorizable_call): Likewise.
1864 (vectorizable_simd_clone_call): Likewise.
1865 (vectorizable_conversion): Likewise.
1866 (vectorizable_assignment): Likewise.
1867 (vectorizable_shift): Likewise.
1868 (vectorizable_operation): Likewise.
1869 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1871 2018-06-18 Martin Sebor <msebor@redhat.com>
1873 PR tree-optimization/81384
1874 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1875 * builtins.c (expand_builtin_strnlen): New function.
1876 (expand_builtin): Call it.
1877 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1878 * builtins.def (BUILT_IN_STRNLEN): New.
1879 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1880 Warn for bounds in excess of maximum object size.
1881 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1882 single-value ranges. Handle strnlen.
1883 (handle_builtin_strlen): Handle strnlen.
1884 (strlen_check_and_optimize_stmt): Same.
1885 * doc/extend.texi (Other Builtins): Document strnlen.
1887 2018-06-18 Maya Rashish <coypu@sdf.org>
1889 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1890 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1891 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1893 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1895 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1897 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1899 * tree.c (escaped_string::escape): Replace cast to char * by
1900 const_cast<char *> (unescaped).
1902 2018-06-18 Nick Clifton <nickc@redhat.com>
1905 * tree.c (escaped_string): New class. Converts an unescaped
1906 string into its escaped equivalent.
1907 (warn_deprecated_use): Use the new class to convert the
1908 deprecation message, if present.
1909 (test_escaped_strings): New self test.
1910 (test_c_tests): Add test_escaped_strings.
1911 * doc/extend.texi (deprecated): Add a note that the
1912 deprecation message is affected by the -fmessage-length
1913 option, and that control characters will be escaped.
1914 (#pragma GCC error): Document this pragma.
1915 (#pragma GCC warning): Likewise.
1916 * doc/invoke.texi (-fmessage-length): Document this option's
1917 effect on the #warning and #error preprocessor directives and
1918 the deprecated attribute.
1920 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1922 * tree.c (decl_value_expr_lookup): Revert latest change.
1923 (decl_value_expr_insert): Likewise.
1925 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1927 * gimplify.c (nonlocal_vlas): Delete.
1928 (nonlocal_vla_vars): Likewise.
1929 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1931 (gimplify_body): Do not create and destroy nonlocal_vlas.
1932 * tree-nested.c: Include diagnostic.h.
1933 (use_pointer_in_frame): Tweak.
1934 (lookup_field_for_decl): Add assertion and declare the transformation.
1935 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1936 internal error when the reference is in a wrong context. Do not
1937 create a debug decl by default.
1938 (note_nonlocal_block_vlas): Delete.
1939 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1940 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1941 create a debug decl by default.
1942 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1943 call is in a wrong context.
1944 (fixup_vla_decls): New function.
1945 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1946 debug variables were created.
1947 * tree.c (decl_value_expr_lookup): Add checking assertion.
1948 (decl_value_expr_insert): Likewise.
1950 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1953 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1954 * tree-scalar-evolution.c (interpret_expr): Likewise.
1955 (expression_expensive_p): Likewise.
1956 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1957 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1958 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1959 (ssa_defined_by_minus_one_stmt_p): New.
1961 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1964 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1965 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1966 * dojump.c (do_jump): Likewise.
1967 * expr.c (expand_expr_real_2): Check operand type's sign.
1968 * fold-const.c (const_unop): Handle ABSU_EXPR.
1969 (fold_abs_const): Likewise.
1970 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1971 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1972 (strip_sign_op_1): Likesise.
1973 * match.pd: Add new pattern to generate ABSU_EXPR.
1974 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1975 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1976 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1977 * tree-inline.c (estimate_operator_cost): Likewise.
1978 * tree-pretty-print.c (dump_generic_node): Likewise.
1979 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1980 * tree.def (ABSU_EXPR): New.
1982 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1985 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1986 documented as preserved for backward compatibility only.
1987 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1989 PR rtl-optimization/86108
1990 * bb-reorder.c (create_forwarder_block): Renamed to ...
1991 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1992 jump from new landing pad to the second part.
1993 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1996 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1999 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2000 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2001 Only call store_expr for halves if the mode is the same.
2004 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2005 Fix up comment formatting.
2007 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2009 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2010 (typed_splay_tree::closure,
2011 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2012 (typed_splay_tree::typed_splay_tree,
2013 typed_splay_tree::operator =): Declared private.
2014 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2015 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2016 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2017 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2018 typed_splay_tree::splay_tree_splay,
2019 typed_splay_tree::splay_tree_foreach_helper,
2020 typed_splay_tree::splay_tree_insert,
2021 typed_splay_tree::splay_tree_remove,
2022 typed_splay_tree::splay_tree_lookup,
2023 typed_splay_tree::splay_tree_predecessor,
2024 typed_splay_tree::splay_tree_successor,
2025 typed_splay_tree::splay_tree_min,
2026 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2027 (typed_splay_tree::root, typed_splay_tree::comp,
2028 typed_splay_tree::delete_key,
2029 typed_splay_tree::delete_value): New data members.
2030 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2031 typed_splay_tree::remove.
2033 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2035 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2036 -mginv and -mno-ginv to the assembler.
2037 * config/mips/mips.opt (-mcrc): New option.
2039 * doc/invoke.text (-mcrc): Document.
2042 2018-06-15 Nick Clifton <nickc@redhat.com>
2045 * tree.c (escaped_string): New class. Converts an unescaped
2046 string into its escaped equivalent.
2047 (warn_deprecated_use): Use the new class to convert the
2048 deprecation message, if present.
2049 (test_escaped_strings): New self test.
2050 (test_c_tests): Add test_escaped_strings.
2051 * doc/extend.texi (deprecated): Add a note that the
2052 deprecation message is affected by the -fmessage-length
2053 option, and that control characters will be escaped.
2054 (#pragma GCC error): Document this pragma.
2055 (#pragma GCC warning): Likewise.
2056 * doc/invoke.texi (-fmessage-length): Document this option's
2057 effect on the #warning and #error preprocessor directives and
2058 the deprecated attribute.
2060 2018-06-15 Richard Biener <rguenther@suse.de>
2062 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2063 here, also noting vector size used.
2064 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2065 size used in MSG_OPTIMIZED_LOCATIONS dump.
2066 (pass_slp_vectorize::execute): Adjust.
2068 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2071 * config/arc/arc.c (arc_return_address_register): Fix
2074 2018-06-15 Richard Biener <rguenther@suse.de>
2077 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2078 leave useless conversion stripping to force_gimple_operand_gsi.
2079 (gimplify_build2): Likewise.
2080 (gimplify_build1): Likewise.
2082 2018-06-15 Richard Biener <rguenther@suse.de>
2085 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2086 before adjusting their block.
2088 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2090 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2091 multilibs for *-*-rtems*.
2092 * config/riscv/t-rtems: New file.
2094 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2097 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2098 unsigned_type_for returns NULL.
2101 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2102 subregs of multi-word pseudos unless the float mode has word size.
2104 2018-06-14 Richard Biener <rguenther@suse.de>
2107 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2108 duplicate and harmful type_hash_canon.
2109 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2111 2018-06-14 Richard Biener <rguenther@suse.de>
2114 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2117 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2119 * config/rtems.h (STDINT_LONG32): Define.
2121 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2122 Prachi Godbole <prachi.godbole@imgtec.com>
2124 * config/mips/mips-cpus.def: Define P6600.
2125 * config/mips/mips-tables.opt: Regenerate.
2126 * config/mips/mips.c (mips_ucbranch_type): New enum.
2127 (mips_rtx_cost_data): Add support for P6600.
2128 (mips_issue_rate): Likewise.
2129 (mips_multipass_dfa_lookahead): Likewise.
2130 (mips_avoid_hazard): Likewise.
2131 (mips_reorg_process_insns): Likewise.
2132 (mips_classify_branch_p6600): New function.
2133 * config/mips/mips.h (TUNE_P6600): New define.
2134 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2135 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2136 * config/mips/mips.md: Include p6600.md.
2137 (processor): Add p6600.
2138 * config/mips/p6600.md: New file.
2139 * doc/invoke.texi: Add p6600 to supported architectures.
2141 2018-06-13 Martin Sebor <msebor@redhat.com>
2143 PR tree-optimization/86114
2144 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2146 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2148 2018-06-13 Richard Biener <rguenther@suse.de>
2150 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2151 Properly set vector type of the intermediate stmt.
2152 * tree-vect-stmts.c (vectorizable_operation): The destination
2153 var always has vectype_out type.
2155 2018-06-13 Jeff Law <law@redhat.com>
2157 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2158 integer 0 for argument to print_rtl_with_bb.
2159 (rl78_reorg): Likewise.
2161 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2163 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2164 from rtx to rtx_insn *.
2165 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2167 (add_sched_insns_for_speculation): Likewise for local "target",
2168 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2169 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2170 from rtx_insn ** to rtx_code_label **.
2171 (reorg_emit_nops): Likewise.
2172 (c6x_reorg): Likewise for local "call_labels".
2173 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2175 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2176 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2177 the loops over LABEL_REFS.
2178 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2180 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2181 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2182 (split_branches): Strengthen local "olabel" from rtx to
2183 rtx_insn *, adding a safe_as_a cast.
2184 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2186 (add_insn_after): Likewise for first two params.
2187 (add_insn_before): Likewise.
2188 (remove_insn): Likewise for param.
2189 (emit_pattern_before_noloc): Likewise for second and third params.
2190 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2191 (emit_call_insn_before_noloc): Likewise.
2192 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2194 (emit_barrier_before): Likewise.
2195 (emit_label_before): Strengthen "label" param from "rtx" to
2196 "rtx_code_label *". Strengthen "before" param from "rtx" to
2198 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2200 (emit_pattern_after_noloc): Likewise.
2201 (emit_insn_after_noloc): Likewise.
2202 (emit_jump_insn_after_noloc): Likewise.
2203 (emit_call_insn_after_noloc): Likewise.
2204 (emit_debug_insn_after_noloc): Likewise.
2205 (emit_barrier_after): Likewise.
2206 (emit_label_after): Likewise for both params.
2207 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2208 "loc" param from "int" to "location_t".
2209 (emit_insn_after_setloc): Likewise.
2210 (emit_jump_insn_after_setloc): Likewise.
2211 (emit_call_insn_after_setloc): Likewise.
2212 (emit_debug_insn_after_setloc): Likewise.
2213 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2214 "loc" param from "int" to "location_t".
2215 (emit_pattern_before): Convert NULL_RTX to NULL.
2216 (emit_insn_before_setloc): Convert "loc" param from "int" to
2218 (emit_jump_insn_before_setloc): Likewise.
2219 (emit_call_insn_before_setloc): Likewise.
2220 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2221 rtx_insn *. Convert "loc" param from "int" to "location_t".
2222 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2223 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2224 Convert 3rd param from "int" to "location_t".
2225 (emit_barrier_before, emit_barrier_after, next_real_insn):
2226 Strengthen param from rtx to rtx_insn *.
2227 (emit_label_before): Strengthen 1st param from "rtx" to
2228 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2230 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2231 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2232 Strengthen 2nd param from "rtx" to "rtx_insn *".
2233 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2234 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2235 Likewise. Convert 3rd param from "int" to "location_t".
2236 (emit_label_after): Strengthen 1st param from "rtx" to
2238 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2240 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2241 from "rtx" to "rtx_insn *".
2243 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2245 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2246 bodies streamed in with -Q.
2247 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2248 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2249 * gimple-streamer-out.c: Include gimple-pretty-print.h
2250 (output_bb): Dump stmts streamed.
2251 * lto-section-out.c: Include print-tree.h
2252 (lto_begin_section): Dump sections created.
2253 (lto_output_decl_index): Dump decl encoded.
2254 * lto-streamer-out.c: Include print-tree.h
2255 (create_output_block): Dump output block created.
2256 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2257 (output_function): Dump function output.
2258 (output_constructor): Dump constructor streamed.
2259 (write_global_stream): Output indexes encoded.
2260 (produce_asm_for_decls): Dump streams encoded.
2261 * lto-streamer.c (streamer_dump_file): New global var.
2262 * lto-streamer.h (streamer_dump_file): Declare.
2263 * passes.c (ipa_write_summaries): Initialize streamer dump.
2264 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2267 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2270 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2271 offsets for register save directives. Emit a second batch of save
2272 directives, if need be, when the function accesses prior frames.
2274 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2276 * config/arc/fpu.md (fmasf4): Force operand to register.
2277 (fnmasf4): Likewise.
2279 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2281 * config/arc/arc-protos.h (arc_pad_return): Remove.
2282 * config/arc/arc.c (machine_function): Remove force_short_suffix
2284 (arc_print_operand): Adjust printing of '&'.
2285 (arc_verify_short): Remove conditional printing of short suffix.
2286 (arc_final_prescan_insn): Remove reference to size_reason.
2287 (pad_return): New function.
2288 (arc_reorg): Call pad_return.
2289 (arc_pad_return): Remove.
2290 (arc_init_machine_status): Remove reference to force_short_suffix.
2291 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2292 (attr length): When attribute iscompact is true force to 2
2293 regardless; in the case of maybe check if we want to force the
2294 instruction to have 4 bytes length.
2295 (nopv): Change it to generate 4 byte long nop as well.
2296 (blockage): New pattern.
2297 (simple_return): Remove call to arc_pad_return.
2298 (p_return_i): Likewise.
2300 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2302 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2304 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2306 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2309 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2311 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2312 for ARC700 and ARCv2.
2314 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2317 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2318 operands[2] instead of operands[1].
2321 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2323 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2324 case, check whether the outer register overlaps an unallocatable
2325 register, not just whether it fits the required class.
2327 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2329 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2331 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2332 (memrefs_conflict_p): Likewise.
2333 (init_alias_analysis): Likewise.
2334 * cfgexpand.c (expand_debug_expr): Likewise.
2335 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2336 * cse.c (fold_rtx): Likewise.
2337 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2338 * expr.c (emit_block_move_hints): Likewise.
2339 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2340 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2341 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2342 (emit_group_store): Likewise.
2343 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2344 to read the PRE/POST_MODIFY increment.
2345 * calls.c (store_one_arg): Use strip_offset.
2346 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2348 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2350 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2351 (simplify_binary_operation_1): Extend CONST_INT handling to
2353 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2354 than a HOST_WIDE_INT.
2355 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2357 (adjust_mems, add_stores): Update accodingly.
2358 (vt_canonicalize_addr): Track polynomial offsets.
2359 (emit_note_insn_var_location): Likewise.
2360 (vt_add_function_parameter): Likewise.
2361 (vt_initialize): Likewise.
2363 2018-06-12 Jeff Law <law@redhat.com>
2365 * config.gcc (alpha*-*-freebsd*): Remove.
2366 * config/alpha/freebsd.h: Remove.
2368 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2371 * spellcheck-tree.c (levenshtein_distance): Rename to...
2372 (get_edit_distance): ...this, and update for underlying renaming.
2373 * spellcheck-tree.h (levenshtein_distance): Rename to...
2374 (get_edit_distance): ...this.
2375 * spellcheck.c (levenshtein_distance): Rename to...
2376 (get_edit_distance): ...this. Convert from Levenshtein distance
2377 to Damerau-Levenshtein distance by supporting transpositions of
2378 adjacent characters. Rename "v1" to "v_next" and "v0" to
2380 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2381 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2382 update for underlying renaming.
2383 (selftest::levenshtein_distance_unit_test): Rename to...
2384 (selftest::test_get_edit_distance_unit): ...this, and update for
2385 underlying renaming.
2386 (selftest::test_find_closest_string): Add example from PR 69968
2387 where transposition helps
2388 (selftest::test_metric_conditions): Update for renaming.
2389 (selftest::test_metric_conditions): Likewise.
2390 (selftest::spellcheck_c_tests): Likewise.
2391 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2392 (get_edit_distance): ...this.
2393 (best_match::consider): Update for renaming.
2395 2018-06-12 Martin Sebor <msebor@redhat.com>
2397 PR tree-optimization/85259
2398 * builtins.c (compute_objsize): Handle constant offsets.
2399 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2400 true iff a warning has been issued.
2401 * gimple.h (gimple_nonartificial_location): New function.
2402 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2403 gimple_nonartificial_location and handle -Wno-system-headers.
2404 (handle_builtin_stxncpy): Same.
2406 2018-06-12 Martin Sebor <msebor@redhat.com>
2409 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2411 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2413 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2414 BUILTIN_VEC_XST entries for pointer to double and long long.
2416 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2419 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2421 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2424 2018-06-12 Martin Liska <mliska@suse.cz>
2426 * doc/options.texi: Document IntegerRange.
2428 2018-06-12 Martin Liska <mliska@suse.cz>
2430 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2431 * opt-functions.awk: Handle Deprecated flag.
2432 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2434 (read_cmdline_option): Report warning for a deprecated option.
2435 * opts.h (struct cl_option): Add new field cl_deprecated.
2436 (CL_ERR_DEPRECATED): New.
2438 2018-06-12 Martin Liska <mliska@suse.cz>
2440 * doc/options.texi: Document Deprecated option flag.
2442 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2444 * config/arc/arc-arch.h (arc_extras): New enum.
2445 (arc_cpu_t):Add field extra.
2446 (arc_cpu_types): Consider the extras.
2447 * config/arc/arc-cpus.def: Add extras info.
2448 * config/arc/arc-opts.h (processor_type): Consider extra field.
2449 * config/arc/arc.c (arc_override_options): Handle extra field.
2451 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2453 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2454 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2456 * config/arc/arc.c (arc_init): Update pic warning.
2457 (irq_range): Update irq range parsing warnings.
2458 (arc_override_options): Update various warning messages.
2459 (arc_handle_aux_attribute): Likewise.
2461 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2463 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2465 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2467 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2468 compiler messages directives.
2470 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2472 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2473 * config/mips/mips-tables.opt: Regenerate.
2474 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2476 * doc/invoke.texi: Document -march=i6500.
2478 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2480 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2481 (i6400_gpmul): Add cpu_unit.
2482 (i6400_gpdiv): Likewise.
2483 (i6400_msa_add_d): Update reservations.
2484 (i6400_msa_int_add) Likewise.
2485 (i6400_msa_short_logic3) Likewise.
2486 (i6400_msa_short_logic2) Likewise.
2487 (i6400_msa_short_logic) Likewise.
2488 (i6400_msa_move) Likewise.
2489 (i6400_msa_cmp) Likewise.
2490 (i6400_msa_short_float2) Likewise.
2491 (i6400_msa_div_d) Likewise.
2492 (i6400_msa_long_logic1) Likewise.
2493 (i6400_msa_long_logic2) Likewise.
2494 (i6400_msa_mult) Likewise.
2495 (i6400_msa_long_float2) Likewise.
2496 (i6400_msa_long_float4) Likewise.
2497 (i6400_msa_long_float5) Likewise.
2498 (i6400_msa_long_float8) Likewise.
2499 (i6400_fpu_fadd): Include frint type.
2500 (i6400_fpu_store): New define_insn_reservation.
2501 (i6400_fpu_load): Likewise.
2502 (i6400_fpu_move): Likewise.
2503 (i6400_fpu_fcmp): Likewise.
2504 (i6400_fpu_fmadd): Likewise.
2505 (i6400_int_mult): Include imul3nc type and update reservation.
2506 (i6400_int_div): Include idiv3 type and update reservation.
2507 (i6400_int_load): Update to check type not move_type.
2508 (i6400_int_store): Likewise.
2509 (i6400_int_prefetch): Set zero latency.
2511 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2513 * gcc.c: Document new %@{...} sequence.
2514 (LINK_COMMAND_SPEC): Use it for the -L switches.
2515 (cpp_unique_options): Use it for the -I switches.
2516 (at_file_argbuf): New global variable.
2517 (in_at_file): Likewise.
2518 (alloc_args): Create at_file_argbuf.
2519 (clear_args): Truncate at_file_argbuf.
2520 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2521 (open_at_file): New function.
2522 (close_at_file): Likewise.
2523 (create_at_file): Delete.
2524 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2527 (validate_switches_from_spec): Deal with %@{...} sequence.
2528 (validate_switches): Likewise.
2529 (driver::finalize): Call clear_args.
2531 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2533 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2535 2018-06-11 Martin Sebor <msebor@redhat.com>
2537 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2538 (Wno-class-memaccess): Add @opindex.
2539 (Wno-templates, Wno-multiple-inheritance): Same.
2540 (Wno-virtual-inheritance, Wno-namespaces): Same.
2541 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2542 (Wno-format-overflow, Wno-format-truncation): Same.
2543 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2544 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2545 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2546 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2547 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2548 (Wno-misspelled-isr): Same.
2550 2018-06-11 Martin Sebor <msebor@redhat.com>
2552 * PR tree-optimization/86083
2553 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2555 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2557 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2559 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2562 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2563 on the correct operand.
2564 (*movdi_internal64): Ditto.
2566 2018-06-11 Martin Liska <mliska@suse.cz>
2568 PR tree-optimization/86089
2569 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2571 2018-06-11 Julia Koval <julia.koval@intel.com>
2573 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2574 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2575 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2577 2018-06-11 Olivier Hainque <hainque@adacore.com>
2579 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2580 for Ada with strict dwarf2.
2582 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2585 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2588 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2590 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2592 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2594 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2597 2018-06-08 Martin Liska <mliska@suse.cz>
2599 * tree-cfg.h (debug_function): Fix argument type to match
2602 2018-06-08 Martin Liska <mliska@suse.cz>
2604 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2605 Remove usage of MPX-related (and removed) fields.
2606 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2608 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2610 * cfg.c (debug): Use TDF_NONE rather than 0.
2611 * cfghooks.c (debug): Likewise.
2612 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2613 (struct dump_option_value_info): Convert to...
2614 (struct kv_pair): ...this template type.
2615 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2617 (optinfo_verbosity_options): Likewise.
2618 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2620 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2621 than int for "optgroup_flags" param.
2622 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2624 (dump_dec): Likewise.
2625 (dump_finish): Use TDF_NONE rather than 0.
2626 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2627 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2628 than 0. Update for change to option_ptr.
2629 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2630 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2631 0. Update for changes to optinfo_verbosity_options and
2633 (opt_info_switch_p): Convert optgroup_flags from int to
2635 (dump_basic_block): Use dump_flags_t rather than int
2636 for "dump_kind" param.
2637 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2638 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2639 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2640 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2641 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2642 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2643 TDF_NONE): Convert from macros to...
2644 (enum dump_flag): ...this new enum.
2645 (dump_flags_t): Update to use enum.
2646 (operator|, operator&, operator~, operator|=, operator&=):
2647 Implement for dump_flags_t.
2648 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2649 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2650 Convert from macros to...
2651 (enum optgroup_flag): ...this new enum.
2652 (optgroup_flags_t): New typedef.
2653 (operator|, operator|=): Implement for optgroup_flags_t.
2654 (struct dump_file_info): Convert field "alt_flags" to
2655 dump_flags_t. Convert field "optgroup_flags" to
2657 (dump_basic_block): Use dump_flags_t rather than int for param.
2658 (dump_generic_expr_loc): Likewise.
2659 (dump_dec): Likewise.
2660 (dump_register): Convert param "optgroup_flags" to
2662 (opt_info_enable_passes): Likewise.
2663 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2665 * gimple-pretty-print.c (debug): Likewise.
2666 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2667 (merged_store_group::apply_stores): Likewise.
2668 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2669 * gimple.c (verify_gimple_pp): Likewise.
2670 * graphite-poly.c (print_pbb_body): Likewise.
2671 * passes.c (pass_manager::register_one_dump_file): Convert
2672 local "optgroup_flags" to optgroup_flags_t.
2673 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2675 (debug_body): Likewise.
2676 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2677 to optgroup_flags_t.
2678 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2680 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2681 (convert_mult_to_fma): Likewise.
2682 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2683 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2684 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2685 "dump_kind" to dump_flags_t.
2687 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2689 * config/rs6000/rs6000.c (min, max): Delete.
2691 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2693 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2694 -mabi=spe and -mabi=no-spe.
2696 2018-06-08 Martin Liska <mliska@suse.cz>
2698 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2699 where we expect an existing summary.
2701 2018-06-08 Martin Liska <mliska@suse.cz>
2703 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2704 * ipa-inline.h (estimate_edge_growth): Likewise.
2706 2018-06-08 Martin Liska <mliska@suse.cz>
2708 * cgraph.c (function_version_hasher::hash): Use
2709 cgraph_node::get_uid ().
2710 (function_version_hasher::equal):
2711 * cgraph.h (cgraph_node::get_uid): New method.
2712 * ipa-inline.c (update_caller_keys): Use
2713 cgraph_node::get_uid ().
2714 (update_callee_keys): Likewise.
2715 * ipa-utils.c (searchc): Likewise.
2716 (ipa_reduced_postorder): Likewise.
2717 * lto-cgraph.c (input_node): Likewise.
2718 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2719 * symbol-summary.h (symtab_insertion): Likewise.
2720 (symtab_removal): Likewise.
2721 (symtab_duplication): Likewise.
2722 * tree-pretty-print.c (dump_function_header): Likewise.
2723 * tree-sra.c (convert_callers_for_node): Likewise.
2725 2018-06-08 Martin Liska <mliska@suse.cz>
2727 * cgraph.c (symbol_table::create_edge): Always assign a new
2729 (symbol_table::free_edge): Do not recycle numbers.
2730 * cgraph.h (cgraph_edge::get): New method.
2731 * symbol-summary.h (symtab_removal): Use it.
2732 (symtab_duplication): Likewise.
2733 (call_summary::hashable_uid): Remove.
2735 2018-06-08 Martin Liska <mliska@suse.cz>
2737 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2738 (initialize_growth_caches): Remove.
2739 (free_growth_caches): Likewise.
2740 (do_estimate_edge_time): Use edge_growth_cache.
2741 (do_estimate_edge_size): Likewise.
2742 (do_estimate_edge_hints): Likewise.
2743 * ipa-inline.c (reset_edge_caches): Likewise.
2744 (recursive_inlining): Likewise.
2745 (inline_small_functions): Likewise.
2746 * ipa-inline.h (initialize_growth_caches): Remove.
2747 (estimate_edge_size): Likewise.
2748 (estimate_edge_time): Likewise.
2749 (estimate_edge_hints): Likewise.
2750 (reset_edge_growth_cache): Likewise.
2751 * symbol-summary.h (call_summary::remove): New method.
2753 2018-06-08 Martin Liska <mliska@suse.cz>
2755 * ipa-cp.c (class edge_clone_summary): New summary.
2756 (grow_edge_clone_vectors): Remove.
2757 (ipcp_edge_duplication_hook): Remove.
2758 (class edge_clone_summary_t): New call_summary class.
2759 (ipcp_edge_removal_hook): Remove.
2760 (edge_clone_summary_t::duplicate): New function.
2761 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2762 (create_specialized_node): Likewise.
2763 (ipcp_driver): Initialize edge_clone_summaries and do not
2766 2018-06-08 Martin Liska <mliska@suse.cz>
2768 * symbol-summary.h (get): New function.
2769 (call_summary::m_initialize_when_cloning): New class member.
2771 2018-06-08 Martin Liska <mliska@suse.cz>
2773 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2774 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2775 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2776 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2777 (remove_cgraph_node_from_order): Use the removed_nodes set.
2778 (do_per_function_toporder): Likwise.
2779 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2780 instead of summary_uid.
2781 (symtab_removal): Likewise.
2782 (symtab_duplication): Likewise.
2784 2018-06-08 Martin Liska <mliska@suse.cz>
2786 * ipa-cp.c (ipcp_store_bits_results): Use
2787 ipcp_transformation_sum.
2788 (ipcp_store_vr_results): Likewise.
2789 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2791 (ipcp_transformation_initialize): ... this.
2792 (ipa_set_node_agg_value_chain):
2793 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2794 (write_ipcp_transformation_info): Likewise.
2795 (read_ipcp_transformation_info): Likewise.
2796 (ipcp_update_bits): Likewise.
2797 (ipcp_update_vr): Likewise.
2798 (ipcp_transform_function): Likewise.
2799 * ipa-prop.h: Rename ipcp_transformation_summary to
2800 ipcp_transformation.
2801 (class ipcp_transformation_t): New function summary.
2802 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2803 (ipa_get_agg_replacements_for_node): Likewise.
2805 2018-06-08 Martin Liska <mliska@suse.cz>
2807 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2809 (class funct_state_summary_t): New function_summary class.
2810 (has_function_state): Remove.
2811 (get_function_state): Likewise.
2812 (set_function_state): Likewise.
2813 (add_new_function): Likewise.
2814 (funct_state_summary_t::insert): New function.
2815 (duplicate_node_data): Remove.
2816 (remove_node_data): Remove.
2817 (funct_state_summary_t::duplicate): New function.
2818 (register_hooks): Create new funct_state_summaries.
2819 (pure_const_generate_summary): Use it.
2820 (pure_const_write_summary): Likewise.
2821 (pure_const_read_summary): Likewise.
2822 (propagate_pure_const): Likewise.
2823 (propagate_nothrow): Likewise.
2824 (dump_malloc_lattice): Likewise.
2825 (propagate_malloc): Likewise.
2826 (execute): Do not register hooks, just remove summary
2828 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2831 2018-06-08 Martin Liska <mliska@suse.cz>
2833 * ipa-reference.c (remove_node_data): Remove.
2834 (duplicate_node_data): Likewise.
2835 (class ipa_ref_var_info_summary_t): New class.
2836 (class ipa_ref_opt_summary_t): Likewise.
2837 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2838 (get_reference_optimization_summary): Use
2839 ipa_ref_opt_sum_summaries.
2840 (set_reference_vars_info): Remove.
2841 (set_reference_optimization_summary): Likewise.
2842 (ipa_init): Create summaries.
2843 (init_function_info): Use function summary.
2844 (ipa_ref_opt_summary_t::duplicate): New function.
2845 (ipa_ref_opt_summary_t::remove): New function.
2846 (get_read_write_all_from_node): Fix GNU coding style.
2847 (propagate): Use function summary.
2848 (write_node_summary_p): Fix GNU coding style.
2849 (stream_out_bitmap): Likewise.
2850 (ipa_reference_read_optimization_summary): Use function summary.
2851 (ipa_reference_c_finalize): Do not release hooks.
2853 2018-06-08 Martin Liska <mliska@suse.cz>
2855 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2856 (analyze_function_body): Extract multiple calls of get_create.
2857 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2858 * ipa-inline.c (recursive_inlining): Use ::get method.
2859 * ipa-inline.h (estimate_edge_growth): Likewise.
2861 2018-06-08 Martin Liska <mliska@suse.cz>
2863 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2865 (hsa_function_summary::hsa_function_summary): Use the new enum
2867 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2868 * hsa-gen.c (hsa_get_host_function): Likewise.
2869 (get_brig_function_name): Likewise.
2870 * ipa-hsa.c (process_hsa_functions): Likewise.
2871 (ipa_hsa_write_summary): Likewise.
2872 * symbol-summary.h (symtab_duplication): Use ::get function/
2873 (get): New function.
2875 2018-06-08 Martin Liska <mliska@suse.cz>
2877 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2879 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2880 (hsa_register_kernel): Likewise.
2881 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2882 * hsa-gen.c (hsa_get_host_function): Likewise.
2883 (get_brig_function_name): Likewise.
2884 (generate_hsa): Likewise.
2885 (pass_gen_hsail::execute): Likewise.
2886 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2887 (devirtualization_time_bonus): Likewise.
2888 (ipcp_propagate_stage): Likewise.
2889 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2890 (edge_set_predicate): Likewise.
2891 (evaluate_conditions_for_known_args): Likewise.
2892 (evaluate_properties_for_edge): Likewise.
2893 (ipa_fn_summary::reset): Likewise.
2894 (ipa_fn_summary_t::duplicate): Likewise.
2895 (dump_ipa_call_summary): Likewise.
2896 (ipa_dump_fn_summary): Likewise.
2897 (analyze_function_body): Likewise.
2898 (compute_fn_summary): Likewise.
2899 (estimate_edge_devirt_benefit): Likewise.
2900 (estimate_edge_size_and_time): Likewise.
2901 (estimate_calls_size_and_time): Likewise.
2902 (estimate_node_size_and_time): Likewise.
2903 (inline_update_callee_summaries): Likewise.
2904 (remap_edge_change_prob): Likewise.
2905 (remap_edge_summaries): Likewise.
2906 (ipa_merge_fn_summary_after_inlining): Likewise.
2907 (ipa_update_overall_fn_summary): Likewise.
2908 (read_ipa_call_summary): Likewise.
2909 (inline_read_section): Likewise.
2910 (write_ipa_call_summary): Likewise.
2911 (ipa_fn_summary_write): Likewise.
2912 (ipa_free_fn_summary): Likewise.
2913 * ipa-hsa.c (process_hsa_functions): Likewise.
2914 (ipa_hsa_write_summary): Likewise.
2915 (ipa_hsa_read_section): Likewise.
2916 * ipa-icf.c (sem_function::merge): Likewise.
2917 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2918 (do_estimate_edge_time): Likewise.
2919 (estimate_size_after_inlining): Likewise.
2920 (estimate_growth): Likewise.
2921 (growth_likely_positive): Likewise.
2922 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2923 (inline_call): Likewise.
2924 * ipa-inline.c (caller_growth_limits): Likewise.
2925 (can_inline_edge_p): Likewise.
2926 (can_inline_edge_by_limits_p): Likewise.
2927 (compute_uninlined_call_time): Likewise.
2928 (compute_inlined_call_time): Likewise.
2929 (want_inline_small_function_p): Likewise.
2930 (edge_badness): Likewise.
2931 (update_caller_keys): Likewise.
2932 (update_callee_keys): Likewise.
2933 (recursive_inlining): Likewise.
2934 (inline_small_functions): Likewise.
2935 (inline_to_all_callers_1): Likewise.
2936 (dump_overall_stats): Likewise.
2937 (early_inline_small_functions): Likewise.
2938 (early_inliner): Likewise.
2939 * ipa-inline.h (estimate_edge_growth): Likewise.
2940 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2941 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2942 * ipa-prop.h (IPA_NODE_REF): Likewise.
2943 (IPA_EDGE_REF): Likewise.
2944 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2945 (propagate_malloc): Likewise.
2946 * ipa-split.c (execute_split_functions): Likewise.
2947 * symbol-summary.h: Rename get to get_create.
2949 (get_create): Likewise.
2950 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2952 2018-06-08 Martin Liska <mliska@suse.cz>
2954 * symbol-summary.h (release): Move definition out of class
2956 (symtab_removal): Likewise.
2957 (symtab_duplication): Likewise.
2959 2018-06-08 Martin Liska <mliska@suse.cz>
2961 * symbol-summary.h (function_summary): Move constructor
2962 implementation out of class declaration.
2963 (release): Likewise.
2964 (symtab_insertion): Likewise.
2965 (symtab_removal): Likewise.
2966 (symtab_duplication): Likewise.
2969 2018-06-08 Martin Liska <mliska@suse.cz>
2971 * Makefile.in: Remove support for MPX (macros, related functions,
2972 fields in cgraph_node, ...).
2973 * builtin-types.def (BT_BND): Likewise.
2974 (BT_FN_BND_CONST_PTR): Likewise.
2975 (BT_FN_CONST_PTR_BND): Likewise.
2976 (BT_FN_VOID_PTR_BND): Likewise.
2977 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2978 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2979 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2980 (expand_builtin_mempcpy_with_bounds): Likewise.
2981 (expand_builtin_memset_with_bounds): Likewise.
2982 (expand_builtin_memset_args): Likewise.
2983 (std_expand_builtin_va_start): Likewise.
2984 (expand_builtin): Likewise.
2985 (expand_builtin_with_bounds): Likewise.
2986 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2987 (DEF_LIB_BUILTIN_CHKP): Likewise.
2988 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2989 (DEF_CHKP_BUILTIN): Likewise.
2990 (BUILT_IN_MEMCPY): Likewise.
2991 (BUILT_IN_MEMMOVE): Likewise.
2992 (BUILT_IN_MEMPCPY): Likewise.
2993 (BUILT_IN_MEMSET): Likewise.
2994 (BUILT_IN_STPCPY): Likewise.
2995 (BUILT_IN_STRCAT): Likewise.
2996 (BUILT_IN_STRCHR): Likewise.
2997 (BUILT_IN_STRCPY): Likewise.
2998 (BUILT_IN_STRLEN): Likewise.
2999 (BUILT_IN_MEMCPY_CHK): Likewise.
3000 (BUILT_IN_MEMMOVE_CHK): Likewise.
3001 (BUILT_IN_MEMPCPY_CHK): Likewise.
3002 (BUILT_IN_MEMSET_CHK): Likewise.
3003 (BUILT_IN_STPCPY_CHK): Likewise.
3004 (BUILT_IN_STRCAT_CHK): Likewise.
3005 (BUILT_IN_STRCPY_CHK): Likewise.
3006 * calls.c (store_bounds): Likewise.
3007 (emit_call_1): Likewise.
3008 (special_function_p): Likewise.
3009 (maybe_warn_nonstring_arg): Likewise.
3010 (initialize_argument_information): Likewise.
3011 (finalize_must_preallocate): Likewise.
3012 (compute_argument_addresses): Likewise.
3013 (expand_call): Likewise.
3014 * cfgexpand.c (expand_call_stmt): Likewise.
3015 (expand_return): Likewise.
3016 (expand_gimple_stmt_1): Likewise.
3017 (pass_expand::execute): Likewise.
3018 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3019 (cgraph_node::remove): Likewise.
3020 (cgraph_node::dump): Likewise.
3021 (cgraph_node::verify_node): Likewise.
3022 * cgraph.h (chkp_function_instrumented_p): Likewise.
3023 (symtab_node::get_alias_target): Likewise.
3024 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3025 (cgraph_local_p): Likewise.
3026 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3027 (cgraph_edge::rebuild_references): Likewise.
3028 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3029 (walk_polymorphic_call_targets): Likewise.
3030 (cgraph_node::expand_thunk): Likewise.
3031 (symbol_table::output_weakrefs): Likewise.
3032 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3033 (ix86_handle_option): Likewise.
3034 * config/i386/constraints.md: Likewise.
3035 * config/i386/i386-builtin-types.def (BND): Likewise.
3039 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3040 (BDESC_FIRST): Likewise.
3042 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3043 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3044 * config/i386/i386.c (enum reg_class): Likewise.
3045 (ix86_target_string): Likewise.
3046 (ix86_option_override_internal): Likewise.
3047 (ix86_conditional_register_usage): Likewise.
3048 (ix86_valid_target_attribute_inner_p): Likewise.
3049 (ix86_set_indirect_branch_type): Likewise.
3050 (ix86_set_current_function): Likewise.
3051 (ix86_function_arg_regno_p): Likewise.
3052 (init_cumulative_args): Likewise.
3053 (ix86_function_arg_advance): Likewise.
3054 (ix86_function_arg): Likewise.
3055 (ix86_pass_by_reference): Likewise.
3056 (ix86_function_value_regno_p): Likewise.
3057 (ix86_function_value_1): Likewise.
3058 (ix86_function_value_bounds): Likewise.
3059 (ix86_return_in_memory): Likewise.
3060 (ix86_setup_incoming_vararg_bounds): Likewise.
3061 (ix86_va_start): Likewise.
3062 (indirect_thunk_need_prefix): Likewise.
3063 (print_reg): Likewise.
3064 (ix86_print_operand): Likewise.
3065 (ix86_expand_call): Likewise.
3066 (ix86_output_function_return): Likewise.
3067 (reg_encoded_number): Likewise.
3068 (BDESC_VERIFYS): Likewise.
3069 (ix86_init_mpx_builtins): Likewise.
3070 (ix86_init_builtins): Likewise.
3071 (ix86_emit_cmove): Likewise.
3072 (ix86_emit_move_max): Likewise.
3073 (ix86_expand_builtin): Likewise.
3074 (ix86_builtin_mpx_function): Likewise.
3075 (ix86_get_arg_address_for_bt): Likewise.
3076 (ix86_load_bounds): Likewise.
3077 (ix86_store_bounds): Likewise.
3078 (ix86_load_returned_bounds): Likewise.
3079 (ix86_store_returned_bounds): Likewise.
3080 (ix86_class_likely_spilled_p): Likewise.
3081 (ix86_hard_regno_mode_ok): Likewise.
3082 (x86_order_regs_for_local_alloc): Likewise.
3083 (ix86_mitigate_rop): Likewise.
3084 (ix86_bnd_prefixed_insn_p): Likewise.
3085 (ix86_mpx_bound_mode): Likewise.
3086 (ix86_make_bounds_constant): Likewise.
3087 (ix86_initialize_bounds): Likewise.
3088 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3089 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3090 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3091 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3092 (TARGET_CHKP_BOUND_MODE): Likewise.
3093 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3094 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3095 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3096 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3097 * config/i386/i386.h (TARGET_MPX): Likewise.
3098 (TARGET_MPX_P): Likewise.
3099 (VALID_BND_REG_MODE): Likewise.
3100 (FIRST_BND_REG): Likewise.
3101 (LAST_BND_REG): Likewise.
3102 (enum reg_class): Likewise.
3103 (BND_REG_P): Likewise.
3104 (BND_REGNO_P): Likewise.
3105 (BNDmode): Likewise.
3106 (ADJUST_INSN_LENGTH): Likewise.
3107 * config/i386/i386.md: Likewise.
3108 * config/i386/i386.opt: Likewise.
3109 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3110 (defined): Likewise.
3111 (LINK_MPX): Likewise.
3112 (MPX_SPEC): Likewise.
3113 (LIBMPX_SPEC): Likewise.
3114 (LIBMPXWRAPPERS_SPEC): Likewise.
3115 (CHKP_SPEC): Likewise.
3116 * config/i386/predicates.md: Likewise.
3117 * dbxout.c (dbxout_type): Likewise.
3118 * doc/extend.texi: Likewise.
3119 * doc/invoke.texi: Likewise.
3120 * doc/md.texi: Likewise.
3121 * doc/tm.texi: Likewise.
3122 * doc/tm.texi.in: Likewise.
3123 * dwarf2out.c (is_base_type): Likewise.
3124 (gen_formal_types_die): Likewise.
3125 (gen_subprogram_die): Likewise.
3126 (gen_type_die_with_usage): Likewise.
3127 (gen_decl_die): Likewise.
3128 (dwarf2out_late_global_decl): Likewise.
3129 * expr.c (expand_assignment): Likewise.
3130 (emit_storent_insn): Likewise.
3131 (store_expr_with_bounds): Likewise.
3132 (store_expr): Likewise.
3133 (expand_expr_real_1): Likewise.
3134 * expr.h (store_expr_with_bounds): Likewise.
3135 * function.c (use_register_for_decl): Likewise.
3136 (struct bounds_parm_data): Likewise.
3137 (assign_parms_augmented_arg_list): Likewise.
3138 (assign_parm_find_entry_rtl): Likewise.
3139 (assign_parm_is_stack_parm): Likewise.
3140 (assign_parm_load_bounds): Likewise.
3141 (assign_bounds): Likewise.
3142 (assign_parms): Likewise.
3143 (expand_function_start): Likewise.
3144 * gcc.c (CHKP_SPEC): Likewise.
3145 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3146 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3147 (wrestrict_dom_walker::check_call): Likewise.
3148 * gimple.c (gimple_build_call_from_tree): Likewise.
3149 * gimple.h (enum gf_mask): Likewise.
3150 (gimple_call_with_bounds_p): Likewise.
3151 (gimple_call_set_with_bounds): Likewise.
3152 * gimplify.c (gimplify_init_constructor): Likewise.
3153 * ipa-cp.c (initialize_node_lattices): Likewise.
3154 (propagate_constants_across_call): Likewise.
3155 (find_more_scalar_values_for_callers_subset): Likewise.
3156 * ipa-hsa.c (process_hsa_functions): Likewise.
3157 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3158 * ipa-icf.c (sem_function::merge): Likewise.
3159 * ipa-inline.c (early_inliner): Likewise.
3160 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3161 (warn_function_cold): Likewise.
3162 (propagate_pure_const): Likewise.
3163 * ipa-ref.h (enum GTY): Likewise.
3164 * ipa-split.c (find_retbnd): Likewise.
3165 (consider_split): Likewise.
3166 (split_function): Likewise.
3167 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3168 * ipa.c (walk_polymorphic_call_targets): Likewise.
3169 (symbol_table::remove_unreachable_nodes): Likewise.
3170 (process_references): Likewise.
3171 (cgraph_build_static_cdtor_1): Likewise.
3172 * lto-cgraph.c (lto_output_node): Likewise.
3173 (output_refs): Likewise.
3174 (compute_ltrans_boundary): Likewise.
3175 (input_overwrite_node): Likewise.
3176 (input_node): Likewise.
3177 (input_cgraph_1): Likewise.
3178 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3179 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3180 (class pass_chkp_instrumentation_passes): Likewise.
3181 (make_pass_chkp_instrumentation_passes): Likewise.
3182 * passes.def: Likewise.
3183 * rtl.h (struct GTY): Likewise.
3184 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3185 * stor-layout.c (layout_type): Likewise.
3186 * symtab.c: Likewise.
3187 * target.def: Likewise.
3188 * targhooks.c (default_chkp_bound_type): Likewise.
3189 (default_chkp_bound_mode): Likewise.
3190 (default_builtin_chkp_function): Likewise.
3191 (default_chkp_function_value_bounds): Likewise.
3192 (default_chkp_make_bounds_constant): Likewise.
3193 (default_chkp_initialize_bounds): Likewise.
3194 * targhooks.h (default_chkp_bound_type): Likewise.
3195 (default_chkp_bound_mode): Likewise.
3196 (default_builtin_chkp_function): Likewise.
3197 (default_chkp_function_value_bounds): Likewise.
3198 (default_chkp_make_bounds_constant): Likewise.
3199 (default_chkp_initialize_bounds): Likewise.
3200 * toplev.c (compile_file): Likewise.
3201 (process_options): Likewise.
3202 * tree-core.h (DEF_BUILTIN): Likewise.
3203 (DEF_BUILTIN_CHKP): Likewise.
3204 * tree-inline.c (declare_return_variable): Likewise.
3205 (remap_gimple_stmt): Likewise.
3206 (copy_bb): Likewise.
3207 (initialize_inlined_parameters): Likewise.
3208 (expand_call_inline): Likewise.
3209 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3210 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3211 (make_pass_ipa_chkp_produce_thunks): Likewise.
3212 (make_pass_chkp): Likewise.
3213 (make_pass_chkp_opt): Likewise.
3214 (make_pass_chkp_instrumentation_passes): Likewise.
3215 * tree-pretty-print.c (dump_generic_node): Likewise.
3216 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3217 * tree-ssa-dce.c (propagate_necessity): Likewise.
3218 (eliminate_unnecessary_stmts): Likewise.
3219 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3220 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3221 * tree-ssa-sccvn.h: Likewise.
3222 * tree-ssa-strlen.c (get_string_length): Likewise.
3223 (valid_builtin_call): Likewise.
3224 (adjust_last_stmt): Likewise.
3225 (handle_builtin_strchr): Likewise.
3226 (handle_builtin_strcpy): Likewise.
3227 (handle_builtin_stxncpy): Likewise.
3228 (handle_builtin_memcpy): Likewise.
3229 (handle_builtin_strcat): Likewise.
3230 (strlen_check_and_optimize_stmt): Likewise.
3231 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3232 * tree-streamer-in.c: Likewise.
3233 * tree-streamer.c (record_common_node): Likewise.
3234 * tree.c (tree_code_size): Likewise.
3235 (wide_int_to_tree_1): Likewise.
3236 (type_contains_placeholder_1): Likewise.
3237 (build_common_tree_nodes): Likewise.
3238 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3239 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3240 (POINTER_BOUNDS_P): Likewise.
3241 (BOUNDED_TYPE_P): Likewise.
3242 (BOUNDED_P): Likewise.
3243 (CALL_WITH_BOUNDS_P): Likewise.
3244 (pointer_bounds_type_node): Likewise.
3245 * value-prof.c (gimple_ic): Likewise.
3246 * var-tracking.c (vt_add_function_parameters): Likewise.
3247 * varasm.c (make_decl_rtl): Likewise.
3248 (assemble_start_function): Likewise.
3249 (output_constant): Likewise.
3250 (maybe_assemble_visibility): Likewise.
3251 * varpool.c (ctor_for_folding): Likewise.
3252 * chkp-builtins.def: Remove.
3253 * ipa-chkp.c: Remove.
3254 * ipa-chkp.h: Remove.
3255 * rtl-chkp.c: Remove.
3256 * rtl-chkp.h: Remove.
3257 * tree-chkp-opt.c: Remove.
3258 * tree-chkp.c: Remove.
3259 * tree-chkp.h: Remove.
3261 2018-06-07 Carl Love <cel@us.ibm.com>
3263 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3264 vextract_fp_from_shortl): Add BE support.
3266 2018-06-07 Paul Koning <ni1d@arrl.net>
3268 * compare-elim.c (try_merge_compare): Don't merge compare if
3269 address contains a side effect.
3270 (try_eliminate_compare): Likewise.
3272 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3274 * config.gcc: Support "tremont".
3275 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3276 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3278 * config/i386/i386.c (m_TREMONT): Define.
3279 (processor_target_table): Add "tremont".
3280 (PTA_TREMONT): Define.
3281 (ix86_lea_outperforms): Add TARGET_TREMONT.
3282 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3283 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3284 and M_INTEL_GOLDMONT_PLUS.
3285 (fold_builtin_cpu): Add "tremont".
3286 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3287 (ix86_option_override_internal): Add "tremont".
3288 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3289 (processor_type): Add PROCESSOR_TREMONT.
3290 * config/i386/x86-tune.def: Add m_TREMONT.
3291 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3293 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3295 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3296 symbol defined for msp430i* devices to be lower case.
3298 gcc/testsuite/gcc.target/msp430/
3299 * msp430i-device-symbol.c: New test.
3300 * msp430f-device-symbol.c: New test.
3301 * msp430.h: New test header file.
3303 2018-06-07 Richard Biener <rguenther@suse.de>
3305 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3306 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3307 Properly wrap signed arithmetic if overflow wraps.
3309 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3311 PR tree-optimization/69615
3312 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3313 of a cast from a same precision integral SSA_NAME in a bb dominated
3314 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3315 cast to utype if rhs2 has already a compatible type.
3317 2018-06-07 Richard Biener <rguenther@suse.de>
3319 PR tree-optimization/85935
3320 * graphite-scop-detection.c (find_params_in_bb): Analyze
3321 condition operands with respect to the correct loop. Assert
3322 the analysis doesn't fail.
3324 2018-06-04 Carl Love <cel@us.ibm.com>
3326 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3327 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3328 as it is slightly cheaper.
3329 (first_match_or_eos_index_<mode>):
3330 Calculate index using natural element order.
3331 (first_match_index_<mode>):
3332 Calculate index using natural element order.
3333 (first_match_or_eos_index_<mode>):
3334 Calculate index using natural order.
3335 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3336 for BE and LE modes.
3337 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3338 P9V_BUILTIN_VCLZLSBB_V16QI.
3339 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3342 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3344 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3345 indentation and line wrap for many prototypes. Add missing
3346 @smallexample directives around block of prototypes for vec_xl and
3349 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3351 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3352 track if we pass or return IEEE 128-bit floating point.
3353 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3354 C++ mangling that is compatible with GCC 8.1.
3355 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3356 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3357 floating point types.
3358 (rs6000_function_arg_advance_1): Likewise.
3359 (rs6000_mangle_type): Optionally generate mangled names that match
3360 what GCC 8.1 generated for IEEE 128-bit floating point types.
3361 (rs6000_globalize_decl_name): If we have an external function that
3362 passes or returns IEEE 128-bit types, generate a weak reference
3363 from the mangled name used in GCC 8.1 to the current mangled
3365 (rs6000_init_builtins): Make __ibm128 use the long double type if
3366 long double is IBM extended double. Make __float128 use the long
3367 double type if long double is IEEE 128-bit.
3370 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3371 macro for __ibm128 built-in functions.
3372 (PACK_IF): Add __ibm128 pack/unpack functions.
3373 (UNPACK_IF): Likewise.
3374 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3375 enable long double built-in functions if long double is IEEE
3376 128-bit floating point.
3377 (rs6000_invalid_builtin): Update long double built-in function
3379 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3380 functions, adjust the built-in function to use the long double
3381 built-in function if __ibm128 and long double are the same type.
3382 * doc/extend.texi (PowerPC builtins): Update documention for
3383 __builtin_{,un}pack_longdouble. Add documentation for
3384 __builtin_{,un}pack_ibm128.
3386 2018-06-06 Jim Wilson <jimw@sifive.com>
3388 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3389 (struct machine_function): New field interrupt_mode.
3390 (riscv_handle_type_attribute): New function. Add forward declaration.
3391 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3392 (riscv_expand_epilogue): Check interrupt_mode field.
3393 (riscv_set_current_function): Check interrupt attribute args and
3394 set interrupt_mode field.
3395 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3396 (riscv_sret, riscv_uret): New.
3397 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3398 new arguments to interrupt attribute.
3400 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3403 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3404 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3406 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3408 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3409 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3410 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3411 several redundant entries.
3413 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3415 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3416 type from "rtx" to "rtx_insn *".
3417 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3418 for local "call_insn", removing cast.
3419 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3421 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3423 PR tree-optimization/86066
3424 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3425 for BIT_INSERT_EXPR stores.
3427 2018-06-06 Richard Biener <rguenther@suse.de>
3429 PR tree-optimization/86062
3430 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3431 component refs ontop
3432 of to be offsetted base.
3434 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3436 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3437 to be static and remove check on interrupt attribute name.
3439 gcc/testsuite/gcc.target/msp430/
3440 * function-attributes-4.c: New test.
3441 * static-interrupts.c: New test.
3443 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3445 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3446 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3448 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3451 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3453 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3454 Remove second argument, change how error is called.
3455 (aarch64_layout_arg): Remove second argument from
3456 aarch64_err_no_fpadvsimd call.
3457 (aarch64_init_cumulative_args): Ditto.
3458 (aarch64_gimplify_va_arg_expr): Ditto.
3459 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3461 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3463 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3464 (*simple_return_indirect_internal<mode>): Rename from
3465 simple_return_indirect_internal. Use W mode iterator.
3466 (rstorssp): New expander.
3467 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3468 (clrssbsy): New expander.
3469 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3471 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3473 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3475 (cmse_check_pointed_object): Likewise.
3477 2018-06-05 Martin Liska <mliska@suse.cz>
3479 PR gcov-profile/47618
3480 * doc/invoke.texi: Document how -fprofile-dir format
3483 2018-06-05 Richard Biener <rguenther@suse.de>
3485 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3486 removal pretend DOM info isn't available so we do not update
3487 it and only remove edges, not dominated blocks. Actually free
3488 DOM info in case we removed something. Remove unreachable blocks.
3489 (mfb_keep_latches): Work with either DOM info or marked backedges.
3490 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3491 first. Mark backedges if DOM info isn't available.
3492 (Re-)compute DOM info after cleanup_control_flow_pre.
3494 2018-06-05 Richard Biener <rguenther@suse.de>
3496 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3497 (locus_discrim_hasher::hash): Adjust.
3498 (locus_discrim_hasher::equal): Likewise.
3499 (next_discriminator_for_locus): Work on line directly.
3500 (same_line_p): Pass in expanded locus1 as well.
3501 (assign_discriminators): Avoid redundant location expansions.
3503 2018-06-05 Richard Biener <rguenther@suse.de>
3505 PR tree-optimization/86046
3506 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3507 if required after clearing TREE_ADDRESSABLE.
3509 2018-06-05 Richard Biener <rguenther@suse.de>
3511 PR tree-optimization/86047
3512 * tree-ssa-loop.c (for_each_index): Glob handling of all
3513 decls and constants and really handle all of them.
3515 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3518 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3519 qualifier_void_pointer and qualifier_const_void_pointer.
3520 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3521 (arm_init_builtins): Handle the above.
3522 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3523 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3524 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3527 2018-06-05 Martin Liska <mliska@suse.cz>
3529 * auto-profile.c (read_autofdo_file): Do not use
3530 gcov_ctr_summary struct.
3531 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3532 * coverage.c (struct counts_entry): Likewise.
3533 (read_counts_file): Read just single summary entry.
3534 (get_coverage_counts): Use gcov_summary struct.
3535 * coverage.h (get_coverage_counts): Likewise.
3536 * gcov-dump.c (dump_working_sets): Likewise.
3537 (tag_summary): Dump just single summary.
3538 * gcov-io.c (gcov_write_summary): Write just histogram
3540 (gcov_read_summary): Read just single summary.
3541 (compute_working_sets): Use gcov_summary struct.
3542 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3543 of GCOV_COUNTERS_SUMMABLE.
3544 (GCOV_COUNTERS_SUMMABLE): Remove.
3545 (GCOV_FIRST_VALUE_COUNTER): Replace with
3546 GCOV_COUNTER_V_INTERVAL.
3547 (struct gcov_ctr_summary): Remove.
3548 (struct gcov_summary): Directly use fields of former
3550 (compute_working_sets): Use gcov_summary struct.
3551 * gcov.c (read_count_file): Do not use ctrs fields.
3552 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3554 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3556 * profile.c: Likewise.
3557 * profile.h: Likewise.
3559 2018-06-05 Martin Liska <mliska@suse.cz>
3561 PR gcov-profile/84846
3562 * gcov.c (output_lines): Print working directory only
3563 in intermediate format.
3565 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3567 * config/s390/s390-builtin-types.def: Add void function type.
3568 * config/s390/s390-builtins.def: Use the function type for the
3571 2018-06-04 Jim Wilson <jimw@sifive.com>
3573 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3575 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3576 and maybe_eh_return. Change regno to unsigned int. Use new args to
3577 handle EH_RETURN_DATA_REGNO registers properly.
3578 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3579 (riscv_expand_epilogue): Update comment. Change argument name and
3580 type. Update code to use new name and type. Pass new args to
3581 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3583 * config/riscv/riscv.md (NORMAL_RETURN): New.
3584 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3585 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3586 (eh_return): Call gen_eh_return_internal and emit barrier.
3587 (eh_return_internal): Call riscv_expand_epilogue.
3589 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3591 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3592 bit_insertion field and declare can_be_merged_into method.
3593 (merged_store_group::can_be_merged_into): New method.
3594 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3595 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3596 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3598 2018-06-04 Richard Biener <rguenther@suse.de>
3600 PR tree-optimization/85955
3601 * builtins.c (fold_builtin_sincos): Convert pointers to
3602 destination to appropriate type before dereferencing.
3604 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3606 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3608 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3610 * expr.c (expand_expr_real_1): Force the operand into memory if
3611 its TYPE_MODE is BLKmode and if there is no integer mode for
3612 the number of bits being extracted.
3614 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3618 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3619 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3621 2018-06-04 Richard Biener <rguenther@suse.de>
3623 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3624 (cleanup_tree_cfg_noloop): ... single caller. Do
3625 start_recording_case_labels later.
3627 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3629 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3630 to _IMMINTRIN_H_INCLUDED.
3631 * config/i386/pconfigintrin.h: Ditto.
3632 * config/i386/waitpkgintrin.h: Ditto.
3633 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3634 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3635 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3636 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3637 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3638 waitpkgintrin.h and cldemoteintrin.h.
3640 2018-06-04 Richard Biener <rguenther@suse.de>
3642 PR tree-optimization/86038
3643 * tracer.c (find_best_successor): Check probability for
3644 being initialized, bail out if not.
3646 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3649 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3650 of bits to ignore when comparing architectures.
3652 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3654 PR tree-optimization/69615
3655 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3656 maximum or minimum of the type, try to merge it also as if
3657 range1 is + [-, x - 1] or + [x + 1, -].
3660 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3662 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3664 PR tree-optimization/86034
3665 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3666 the unsigned bitfield type in a bit insertion sequence if it does not
3667 have a larger precision than the bitfield size.
3668 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3670 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3672 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3674 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3676 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3677 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3678 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3679 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3681 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3683 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3684 Disable -fdelete-null-pointer-checks for ELF toolchain.
3686 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3687 Kito Cheng <kito.cheng@gmail.com>
3689 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3690 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3691 (nds32*-*-*): Add glibc and uclibc conditions.
3692 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3693 (TARGET_EXCEPT_UNWIND_INFO): Define.
3694 * config/nds32/elf.h: New file.
3695 * config/nds32/linux.h: New file.
3696 * config/nds32/nds32-elf.opt: New file.
3697 * config/nds32/nds32-linux.opt: New file.
3698 * config/nds32/nds32-fp-as-gp.c
3699 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3700 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3702 (nds32_asm_file_end): Ditto.
3703 (nds32_print_operand): Ditto.
3704 (nds32_insert_attributes): Ditto.
3705 (nds32_init_libfuncs): New function.
3706 (TARGET_HAVE_TLS): Define.
3707 (TARGET_INIT_LIBFUNCS): Define.
3708 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3710 (TARGET_ELF): Apply different mcmodel setting.
3711 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3712 been migrated into elf.h and linux.h files.
3713 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3714 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3715 (mcmodel): The content has been migrated into nds32-elf.opt and
3716 nds32-linux.opt files.
3717 * config/nds32/t-elf: New file.
3718 * config/nds32/t-linux: New file.
3720 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3721 Shiva Chen <shiva0217@gmail.com>
3723 * config/nds32/constants.md (unspec_volatile_element): Add
3724 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3725 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3727 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3728 (make_pass_nds32_fp_as_gp): Declare.
3729 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3731 (nds32_asm_function_end_prologue): Remove unused asm output.
3732 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3733 (nds32_asm_file_start): Output necessary fp_as_gp information.
3734 (nds32_option_override): Adjust register usage.
3735 (nds32_expand_prologue): Consider fp_as_gp situation.
3736 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3737 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3740 (simple_return): Ditto.
3741 (omit_fp_begin): Output special directive for fp_as_gp.
3742 (omit_fp_end): Output special directive for fp_as_gp.
3743 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3744 mforbid-fp-as-gp): New options.
3746 2018-06-01 Mark Wielaard <mark@klomp.org>
3748 * dwarf2out.c (dwarf2out_finish): Remove generation of
3749 DW_AT_loclists_base.
3751 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3753 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3754 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3755 (struct merged_store_group): Add bit_insertion field.
3756 (dump_char_array): Use standard hexadecimal format.
3757 (merged_store_group::merged_store_group): Set bit_insertion to false.
3758 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3759 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3760 also print the mask in the dump file.
3761 (pass_store_merging::gate): Minor tweak.
3762 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3763 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3764 stores with INTEGER_CST stores.
3765 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3766 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3767 and use it throughout. Generate bit insertion sequences if need be.
3768 (pass_store_merging::process_store): Remove redundant condition.
3769 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3771 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3773 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3774 the 128-bit floating point types. Fix function comment.
3776 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3778 * config/aarch64/aarch64-simd.md
3779 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3781 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3784 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3786 PR tree-optimization/85989
3787 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3789 (backprop::intersect_uses): Check it when deciding whether this
3790 is a backedge reference.
3791 (backprop::process_block): Add each phi to m_visited_phis
3792 after visiting it, then clear it at the end.
3794 2018-06-01 Richard Biener <rguenther@suse.de>
3796 * tree-vectorizer.h (vect_dr_stmt): New function.
3797 (vect_get_load_cost): Adjust.
3798 (vect_get_store_cost): Likewise.
3799 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3800 Use vect_dr_stmt instead of DR_SMTT.
3801 (vect_record_base_alignments): Likewise.
3802 (vect_calculate_target_alignment): Likewise.
3803 (vect_compute_data_ref_alignment): Likewise and make static.
3804 (vect_update_misalignment_for_peel): Likewise.
3805 (vect_verify_datarefs_alignment): Likewise.
3806 (vector_alignment_reachable_p): Likewise.
3807 (vect_get_data_access_cost): Likewise. Pass down
3808 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3809 (vect_get_peeling_costs_all_drs): Likewise.
3810 (vect_peeling_hash_get_lowest_cost): Likewise.
3811 (vect_enhance_data_refs_alignment): Likewise.
3812 (vect_find_same_alignment_drs): Likewise.
3813 (vect_analyze_data_refs_alignment): Likewise.
3814 (vect_analyze_group_access_1): Likewise.
3815 (vect_analyze_group_access): Likewise.
3816 (vect_analyze_data_ref_access): Likewise.
3817 (vect_analyze_data_ref_accesses): Likewise.
3818 (vect_vfa_segment_size): Likewise.
3819 (vect_small_gap_p): Likewise.
3820 (vectorizable_with_step_bound_p): Likewise.
3821 (vect_prune_runtime_alias_test_list): Likewise.
3822 (vect_analyze_data_refs): Likewise.
3823 (vect_supportable_dr_alignment): Likewise.
3824 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3825 (vect_gen_prolog_loop_niters): Likewise.
3826 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3827 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3829 (vect_recog_mask_conversion_pattern): Likewise.
3830 (vect_try_gather_scatter_pattern): Likewise.
3831 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3832 to vect_get_store_cost.
3833 (vect_get_store_cost): Get stmt_info instead of DR.
3834 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3835 (vect_get_load_cost): Get stmt_info instead of DR.
3837 2018-06-01 Richard Biener <rguenther@suse.de>
3840 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3841 vars wrapped in MEM_REFs.
3843 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3845 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3846 Fix subreg tests so that we only return a choice between
3847 GENERAL_REGS and FP_REGS if the original classes included both.
3849 2018-06-01 Richard Biener <rguenther@suse.de>
3852 * tree-ssa-structalias.c (get_function_part_constraint):
3853 Handle NULL fi->decl.
3854 (find_func_aliases_for_call): Properly handle indirect
3855 fi from direct call.
3856 (find_func_clobbers): Likewise.
3857 (ipa_pta_execute): Likewise.
3858 (create_variable_info_for): For functions that are ifunc_resolver
3859 resolve to a varinfo that contains the result of the resolver call.
3860 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3863 2018-05-31 Michael Collison <michael.collison@arm.com>
3865 * config/aarch64/aarch64.md:
3866 (*fix_to_zero_extenddfdi2): New pattern.
3867 * gcc.target/aarch64/fix_extend1.c: New testcase.
3869 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3873 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3874 and BUILT_IN_STRNCMP_EQ.
3875 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3876 BUILT_IN_STRNCMP_EQ.
3877 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3878 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3879 (gimple_fold_builtin): Likewise.
3880 * tree-ssa-strlen.c (compute_string_length): New function.
3881 (determine_min_obsize): New function.
3882 (handle_builtin_string_cmp): New function to handle calls to
3883 string compare functions.
3884 (strlen_optimize_stmt): Add handling to builtin string compare
3886 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3887 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3888 * tree.c (build_common_builtin_nodes): Add new defines of
3889 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3891 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3894 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3895 functions with naked attribute.
3897 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3899 * config/i386/sse.md (avx_vec_concat<mode>):
3900 Substitute concat_tg_mode mode attribute with xtg_mode.
3901 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3902 (concat_tg_mode): Remove mode attribute.
3904 2018-05-31 Martin Sebor <msebor@redhat.com>
3907 * calls.c (alloc_max_size): Correct a logic error/typo.
3908 Treat excessive arguments as infinite. Warn for invalid arguments.
3909 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3911 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3914 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3915 and movx for Haswell.
3917 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3918 Cesar Philippidis <cesar@codesourcery.com>
3921 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3922 when emitting error on private/firstprivate reductions.
3923 * omp-low.c (lower_omp_target): Avoid reference-type processing
3924 on pointers for firstprivate clause.
3926 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3928 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3931 * config/aarch64/aarch64-simd.md
3932 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3933 (aarch64_ld1_x3_<mode>): Likewise
3934 (aarch64_st1x2<VALLDIF:mode>): Likewise
3935 (aarch64_st1_x2_<mode>): Likewise
3936 (aarch64_st1x3<VALLDIF:mode>): Likewise
3937 (aarch64_st1_x3_<mode>): Likewise
3938 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3939 (vld1_s8_x3): Likewise.
3940 (vld1_u16_x3): Likewise.
3941 (vld1_s16_x3): Likewise.
3942 (vld1_u32_x3): Likewise.
3943 (vld1_s32_x3): Likewise.
3944 (vld1_u64_x3): Likewise.
3945 (vld1_s64_x3): Likewise.
3946 (vld1_f16_x3): Likewise.
3947 (vld1_f32_x3): Likewise.
3948 (vld1_f64_x3): Likewise.
3949 (vld1_p8_x3): Likewise.
3950 (vld1_p16_x3): Likewise.
3951 (vld1_p64_x3): Likewise.
3952 (vld1q_u8_x3): Likewise.
3953 (vld1q_s8_x3): Likewise.
3954 (vld1q_u16_x3): Likewise.
3955 (vld1q_s16_x3): Likewise.
3956 (vld1q_u32_x3): Likewise.
3957 (vld1q_s32_x3): Likewise.
3958 (vld1q_u64_x3): Likewise.
3959 (vld1q_s64_x3): Likewise.
3960 (vld1q_f16_x3): Likewise.
3961 (vld1q_f32_x3): Likewise.
3962 (vld1q_f64_x3): Likewise.
3963 (vld1q_p8_x3): Likewise.
3964 (vld1q_p16_x3): Likewise.
3965 (vld1q_p64_x3): Likewise.
3966 (vst1_s64_x2): Likewise.
3967 (vst1_u64_x2): Likewise.
3968 (vst1_f64_x2): Likewise.
3969 (vst1_s8_x2): Likewise.
3970 (vst1_p8_x2): Likewise.
3971 (vst1_s16_x2): Likewise.
3972 (vst1_p16_x2): Likewise.
3973 (vst1_s32_x2): Likewise.
3974 (vst1_u8_x2): Likewise.
3975 (vst1_u16_x2): Likewise.
3976 (vst1_u32_x2): Likewise.
3977 (vst1_f16_x2): Likewise.
3978 (vst1_f32_x2): Likewise.
3979 (vst1_p64_x2): Likewise.
3980 (vst1q_s8_x2): Likewise.
3981 (vst1q_p8_x2): Likewise.
3982 (vst1q_s16_x2): Likewise.
3983 (vst1q_p16_x2): Likewise.
3984 (vst1q_s32_x2): Likewise.
3985 (vst1q_s64_x2): Likewise.
3986 (vst1q_u8_x2): Likewise.
3987 (vst1q_u16_x2): Likewise.
3988 (vst1q_u32_x2): Likewise.
3989 (vst1q_u64_x2): Likewise.
3990 (vst1q_f16_x2): Likewise.
3991 (vst1q_f32_x2): Likewise.
3992 (vst1q_f64_x2): Likewise.
3993 (vst1q_p64_x2): Likewise.
3994 (vst1_s64_x3): Likewise.
3995 (vst1_u64_x3): Likewise.
3996 (vst1_f64_x3): Likewise.
3997 (vst1_s8_x3): Likewise.
3998 (vst1_p8_x3): Likewise.
3999 (vst1_s16_x3): Likewise.
4000 (vst1_p16_x3): Likewise.
4001 (vst1_s32_x3): Likewise.
4002 (vst1_u8_x3): Likewise.
4003 (vst1_u16_x3): Likewise.
4004 (vst1_u32_x3): Likewise.
4005 (vst1_f16_x3): Likewise.
4006 (vst1_f32_x3): Likewise.
4007 (vst1_p64_x3): Likewise.
4008 (vst1q_s8_x3): Likewise.
4009 (vst1q_p8_x3): Likewise.
4010 (vst1q_s16_x3): Likewise.
4011 (vst1q_p16_x3): Likewise.
4012 (vst1q_s32_x3): Likewise.
4013 (vst1q_s64_x3): Likewise.
4014 (vst1q_u8_x3): Likewise.
4015 (vst1q_u16_x3): Likewise.
4016 (vst1q_u32_x3): Likewise.
4017 (vst1q_u64_x3): Likewise.
4018 (vst1q_f16_x3): Likewise.
4019 (vst1q_f32_x3): Likewise.
4020 (vst1q_f64_x3): Likewise.
4021 (vst1q_p64_x3): Likewise.
4023 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4025 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4026 user_label_prefix to name.
4028 * tree-core.h: Update comment about the format of NAME string
4029 passed to handler in attribute_spec.
4031 * config/msp430/msp430.md: Remove erroneous subreg expression from
4032 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4033 zero_extend{q,h}isi2.
4035 2018-05-30 Borislav Petkov <bp@suse.de>
4037 * doc/extend.texi: Document some architecture specific
4038 constraints and sort entries.
4040 2018-05-30 Martin Sebor <msebor@redhat.com>
4043 * builtins.c (expand_builtin_stpcpy_1): New function.
4044 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4045 only if the former succeeds.
4047 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4049 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4052 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4054 * doc/invoke.texi (-flinker-output): Document
4056 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4058 * passes.c (ipa_write_summaries): Only modify statements if body
4060 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4061 incrementally linking.
4062 (ipa_passes): Likewise.
4063 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4064 pass down resolution info.
4065 * common.opt (flag_incremental_link): Update info.
4066 * gcc.c (plugin specs): Turn flinker-output=* to
4067 -plugin-opt=-linker-output-known
4068 * toplev.c (compile_file): Also cut compilation when doing incremental
4070 * flag-types. (enum lto_partition_model): Add
4071 LTO_LINKER_OUTPUT_NOLTOREL.
4072 (invoke.texi): Add -flinker-output docs.
4073 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4074 link same way as WPA; do not stream in dead initializers.
4076 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4077 darf2out_register_external_decl): Support incremental link.
4079 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4081 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4083 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4085 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4086 it down to simple_object_copy_lto_debug_sections.
4087 (run_gcc): Determine incremental LTO link time and configure
4088 lto1 into non-wpa mode, disable renaming of debug sections.
4090 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4092 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4093 descriptions of various incorrectly documented functions.
4095 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4098 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4099 address check not strict.
4101 2018-05-30 Richard Biener <rguenther@suse.de>
4103 PR tree-optimization/85964
4104 * tracer.c (better_p): Drop initialized count check, we only
4105 call the function with initialized counts now.
4106 (find_best_successor): Do find a best edge if one
4107 has uninitialized count.
4108 (find_best_predecessor): Likewise. Do BB frequency check only
4109 if count is initialized.
4111 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4113 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4114 (aarch64_ldrstr_offset_compare): New.
4115 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4116 load/store orderings.
4117 (aarch64_gen_adjusted_ldpstp): Likewise.
4119 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4121 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4122 Check for subset of GENERAL_REGS and FP_REGS.
4123 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4126 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4128 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4129 and wi::to_poly_offset. Add the current offset and then check
4130 whether the sum fits, rather than using an unchecked addition of
4131 a checked term. Check for a shwi rather than a uhwi.
4132 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4133 (store_constructor): Use poly_int_tree_p.
4134 (expand_expr_real_1): Likewise.
4135 * function.c (assign_temp): Likewise.
4136 * fold-const.c (const_binop): Use poly_int_tree_p and
4138 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4140 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4141 to_poly_offset for MEM offsets.
4142 * ipa-icf.c (sem_variable::equals): Likewise.
4143 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4144 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4145 wi::to_poly_offset for BIT_FIELD_REF offsets.
4146 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4148 * var-tracking.c (emit_note_insn_var_location): Use
4149 tree_to_poly_uint64.
4151 2018-05-29 Jim Wilson <jimw@sifive.com>
4153 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4155 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4158 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4159 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4161 (sse4_1_round<mode>2): Use nonimmediate_operand
4162 for operand 1 predicate.
4164 2018-05-29 Martin Sebor <msebor@redhat.com>
4165 Richard Biener <rguenther@suse.de>
4168 * calls.c (get_size_range): Call determine_value_range instead
4169 of get_value_range..
4170 * tree-vrp.h (determine_value_range): Declared new function.
4171 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4173 2018-05-29 Richard Biener <rguenther@suse.de>
4175 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4176 sure to use non-pattern stmts for get_earlier_stmt arguments.
4177 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4178 called on pattern stmts.
4179 (get_later_stmt): Likewise.
4181 2018-05-29 Martin Liska <mliska@suse.cz>
4183 PR gcov-profile/85759
4184 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4187 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4189 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4191 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4195 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4196 VEC_PACK_FLOAT_EXPR): New tree codes.
4197 * tree-pretty-print.c (op_code_prio): Handle
4198 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4199 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4200 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4201 * tree-inline.c (estimate_operator_cost): Likewise.
4202 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4203 * fold-const.c (const_binop): Likewise.
4204 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4205 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4206 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4207 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4208 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4209 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4210 * expr.c (expand_expr_real_2): Likewise.
4211 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4212 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4213 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4215 * optabs.c (expand_widen_pattern_expr): For
4216 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4217 sign from result type rather than operand's type.
4218 (expand_binop_directly): For vec_packu_float_optab and
4219 vec_packs_float_optab allow result type to be different from operand's
4221 * optabs-tree.c (optab_for_tree_code): Handle
4222 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4223 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4224 * tree-vect-generic.c (expand_vector_operations_1): Handle
4225 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4226 VEC_PACK_FLOAT_EXPR.
4227 * tree-vect-stmts.c (supportable_widening_operation): Handle
4229 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4230 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4231 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4232 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4233 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4235 (vec_pack<floatprefix>_float_<mode>): New expander.
4236 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4238 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4239 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4240 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4241 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4242 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4244 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4245 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4246 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4247 VEC_PACK_FLOAT_EXPR): Document.
4249 2018-05-29 Richard Biener <rguenther@suse.de>
4251 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4253 (stmt_vec_info_vec): Make pointer.
4254 (init_stmt_vec_info_vec): Remove.
4255 (free_stmt_vec_info_vec): Likewise.
4256 (set_stmt_vec_info_vec): New function.
4257 (free_stmt_vec_infos): Likewise.
4258 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4259 (set_vinfo_for_stmt): Likewise.
4260 (get_earlier_stmt): Likewise.
4261 (get_later_stmt): Likewise.
4262 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4263 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4264 (vec_info::~vec_info): Free stmt_vec_infos.
4265 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4266 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4267 (pass_slp_vectorize::execute): Likewise.
4268 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4269 (free_stmt_vec_info_vec): Likewise.
4270 (set_stmt_vec_info_vec): New function.
4271 (free_stmt_vec_infos): Likewise.
4272 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4273 the global stmt_vec_info_vec.
4274 * tree-parloops.c (gather_scalar_reductions): Use
4275 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4278 2018-05-29 Richard Biener <rguenther@suse.de>
4280 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4282 2018-05-29 Martin Liska <mliska@suse.cz>
4283 David Malcolm <dmalcolm@redhat.com>
4285 * vec.c (test_reverse): New.
4286 (vec_c_tests): Add new test.
4287 * vec.h (vl_ptr>::reverse): New function.
4289 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4291 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4293 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4296 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4298 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4300 2018-05-28 Richard Biener <rguenther@suse.de>
4302 PR tree-optimization/85933
4303 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4304 look at stmts marked as vectorizable.
4306 2018-05-28 Richard Biener <rguenther@suse.de>
4308 PR tree-optimization/85934
4309 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4310 vector boolean check before scalar optimization.
4312 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4314 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4317 2018-05-28 Mark Wielaard <mark@klomp.org>
4319 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4320 if it is an expression containing a minus sign.
4322 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4324 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4326 2018-05-27 Paul Koning <ni1d@arrl.net>
4328 * config/pdp11/pdp11.md (truncsihi2): Remove.
4330 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4331 Chung-Ju Wu <jasonwucj@gmail.com>
4333 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4335 (unaligned_store_dw): Ditto.
4336 * config/nds32/nds32-memory-manipulation.c
4337 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4338 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4339 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4340 (emit_setmem_word_loop): Rename to ...
4341 (emit_setmem_doubleword_loop): ... this.
4342 (nds32_gen_dup_4_byte_to_word_value): New function.
4343 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4344 (nds32_expand_setmem_loop): Refine implementation.
4345 (nds32_expand_setmem_loop_v3m): Ditto.
4346 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4349 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4351 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4353 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4355 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4356 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4357 machine->attr_no_prologue_p.
4358 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4359 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4360 (nds32_expand_epilogue): Consider attr_naked_p.
4361 (nds32_expand_epilogue_v3pop): Likewise.
4362 (nds32_can_use_return_insn): Likewise.
4363 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4364 attr_no_prologue_p fields.
4365 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4367 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4370 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4372 * config/i386/sse.md
4373 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4375 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4377 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4379 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4381 (*<floatsuffix>floatv2div2sf2): Rename to ...
4382 (*float<floatunssuffix>v2div2sf2): ... this.
4383 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4384 (float<floatunssuffix>v2div2sf2_mask): ... this.
4385 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4386 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4387 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4389 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4391 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4393 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4395 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4397 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4399 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4400 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4401 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4402 gen_ufix_truncv8dfv8si2.
4403 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4404 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4405 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4406 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4407 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4408 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4409 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4410 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4412 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4416 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4418 2018-05-25 Jim Wilson <jimw@sifive.com>
4420 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4421 * config/riscv/riscv.c (struct machine_function): Add
4422 interrupt_handler_p and attribute_checked_p fields.
4423 (riscv_attribute_table): Add interrupt.
4424 (riscv_interrupt_type_p): New.
4425 (riscv_save_reg_p): Save extra regs for interrupt handler.
4426 (riscv_use_save_libcall): Return false for interrupt handler.
4427 (riscv_first_stack_step): Add forward declaration.
4428 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4429 for interrupt handler with large frame. Use it for saved reg list.
4430 (riscv_expand_prologue): Move flag_stack_usage_info support to
4431 eliminate duplication.
4432 (riscv_expand_epilogue): Generate mret for interrupt handler.
4433 (riscv_epilogue_uses): New.
4434 (riscv_can_use_return_insn): Return false for interrupt handler.
4435 (riscv_function_ok_for_sibcall): Likewise.
4436 (riscv_set_current_function): Add interrupt handler support.
4437 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4438 * config/riscv/riscv.md (UNSPECV_MRET): New.
4440 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4442 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4444 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4446 PR tree-optimization/85712
4447 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4448 this candidate has already been replaced in-situ by a copy.
4450 2018-05-25 Jason Merrill <jason@redhat.com>
4452 PR c++/80485 - inline function non-zero address.
4453 * symtab.c (nonzero_address): Check DECL_COMDAT.
4455 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4458 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4459 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4460 extension of SImode operation. Use const123_operand predicate.
4461 (*saddsi_1): Remove.
4462 (*saddl_se_1): Ditto.
4464 (*ssubl_se_1): Ditto.
4465 * config/alpha/predicates.md (const123_operand): New predicate.
4466 * config/alpha/constraints.md (P): Use IN_RANGE.
4468 2018-05-25 Richard Biener <rguenther@suse.de>
4470 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4472 (ref_maybe_used_by_stmt_p): Likewise.
4473 (stmt_may_clobber_ref_p): Likewise.
4474 (stmt_may_clobber_ref_p_1): Likewise.
4475 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4477 (ref_maybe_used_by_stmt_p): Likewise.
4478 (stmt_may_clobber_ref_p): Likewise.
4479 (stmt_may_clobber_ref_p_1): Likewise.
4480 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4481 the alias oracle to disambiguate DRs with stmts DR analysis
4483 (vect_analyze_data_refs): Do not give up on not analyzable
4484 DRs for BB vectorization. Remove code truncating the dataref
4487 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4490 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4491 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4492 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4494 2018-05-25 Richard Biener <rguenther@suse.de>
4496 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4497 function, combining stmt data ref gathering and fatal analysis
4499 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4500 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4501 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4502 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4503 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4505 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4507 PR tree-optimization/85720
4508 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4509 SCC if all partitions are builtins.
4510 (version_loop_by_alias_check): New parameter. Generate cancelable
4511 runtime alias check if all partitions are builtins.
4512 (distribute_loop): Update call to above function.
4514 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4516 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4517 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4518 (parm_default_def_partition_arg): Ditto.
4519 (set_parm_default_def_partition): Ditto.
4520 (get_parm_default_def_partitions): Ditto and make it static.
4521 (get_undefined_value_partitions): Ditto and make it static.
4522 (remove_ssa_form): Refactor call to init_var_map here.
4523 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4524 computation for loop region.
4525 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4526 (register_default_def): Delete.
4527 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4528 (parm_default_def_partition_arg): Ditto.
4529 (set_parm_default_def_partition): Ditto.
4530 (get_parm_default_def_partitions): Ditto and make it static.
4531 (get_undefined_value_partitions): Ditto and make it static.
4532 (coalesce_with_default, coalesce_with_default): Update comment.
4533 (create_coalesce_list_for_region): New func factored out from
4534 create_outofssa_var_map.
4535 (populate_coalesce_list_for_outofssa): New func factored out from
4536 create_outofssa_var_map and coalesce_ssa_name.
4537 (create_outofssa_var_map): Delete.
4538 (coalesce_ssa_name): Refactor to support live range computation.
4539 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4540 (get_parm_default_def_partitions): Delete.
4541 (get_undefined_value_partitions): Ditto.
4542 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4543 computation for loop region.
4544 (new_tree_live_info, loe_visit_block): Ditto.
4545 (live_worklist, set_var_live_on_entry): Ditto.
4546 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4547 * tree-ssa-live.h (struct _var_map): New fields.
4548 (init_var_map): Change decl.
4549 (region_contains_p): New.
4551 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4553 * tree-ssa-live.h (live_merge_and_clear): Delete.
4555 2018-05-25 Richard Biener <rguenther@suse.de>
4558 * tree-dump.c (dequeue_and_dump): Remove access to removed
4559 operand 2 of a SWITCH_EXPR.
4561 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4563 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4564 multiplication and division.
4565 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4566 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4567 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4568 (cond_udiv_optab, cond_umod_optab): New optabs.
4569 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4570 (IFN_COND_RDIV): New internal functions.
4571 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4572 TRUNC_MOD_EXPR and RDIV_EXPR.
4573 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4574 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4576 (SVE_INT_BINARY): Include mult.
4577 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4578 (optab, sve_int_op): Handle mult.
4579 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4581 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4582 for SVE_INT_BINARY_SD.
4584 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4586 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4587 (optab, sve_int_op): Handle div and udiv.
4588 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4589 for SVE_INT_BINARY_SD.
4590 (*<optab><mode>3): New insn for the same.
4592 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4594 * tree-vect-patterns.c: Include predict.h.
4595 (vect_recog_divmod_pattern): Restrict check for division support
4596 to when optimizing for size.
4598 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4600 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4601 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4602 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4603 (gimple_match_op::set_op): Likewise.
4604 (gimple_resimplify4): Declare.
4605 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4606 (expr::gen_transform): Likewise.
4607 (decision_tree::gen): Generate a simplification routine for 4 operands.
4608 * gimple-match-head.c (gimple_simplify): Add an overload for
4609 4 operands. In the top-level function, handle up to 4 call
4610 arguments and call gimple_resimplify4.
4611 (gimple_resimplify4): New function.
4612 (build_call_internal): Pass a fourth operand.
4613 (maybe_push_to_seq): Likewise.
4614 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4615 Fold VEC_COND_EXPRs of an operation and a default value into
4616 an IFN_COND_* function if possible.
4617 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4619 (SVE_COND_FP_BINARY): Include them.
4620 (optab, sve_fp_op): Handle them.
4621 (SVE_INT_BINARY_REV): New code iterator.
4622 (SVE_COND_FP_BINARY_REV): New int iterator.
4623 (commutative): New int attribute.
4624 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4626 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4628 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4629 (*cond_<optab><mode>): New patterns for reversed operands.
4631 2018-05-25 Richard Biener <rguenther@suse.de>
4633 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4634 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4635 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4636 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4637 * tree-vect-data-refs.c (everywhere): Adjust users.
4638 * tree-vect-loop.c (everywhere): Likewise.
4639 * tree-vect-slp.c (everywhere): Likewise.
4640 * tree-vect-stmts.c (everywhere): Likewise.
4641 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4643 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4645 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4647 (gcc_cv_as_section_exclude): ... this.
4648 Try Solaris as #exclude syntax.
4649 * configure: Regenerate.
4650 * config.in: Regenerate.
4651 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4653 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4654 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4656 * varasm.c (default_elf_asm_named_section): Don't check if
4657 HAVE_GAS_SECTION_EXCLUDE is defined.
4659 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4661 * doc/md.texi: Update the documentation of the cond_* optabs
4662 to mention the new final operand. Fix GET_MODE_NUNITS call.
4663 Describe the scalar case too.
4664 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4665 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4667 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4668 (get_conditional_internal_fn): Update comment.
4669 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4670 accumulator value as a final argument to conditional functions.
4671 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4672 a define_expand and add an "else" operand. Assert for now that
4673 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4674 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4675 (*cond_<optab><mode>): New patterns.
4676 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4677 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4678 (UNSPEC_COND_EOR): Delete.
4679 (optab): Remove associated mappings.
4680 (SVE_INT_BINARY): New code iterator.
4681 (sve_int_op): Remove int attribute and add "minus" to the code
4683 (SVE_COND_INT_OP): Delete.
4684 (SVE_COND_FP_OP): Rename to...
4685 (SVE_COND_FP_BINARY): ...this.
4687 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4689 * optabs.c (can_reuse_operands_p): New function.
4690 (maybe_legitimize_operands): Try to reuse the results for
4693 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4695 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4696 Add {q} suffix to insn mnemonic.
4698 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4700 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4701 (msp430_warn_func_return): New.
4703 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4705 * fold-const.c (tree_nonzero_bits): New function.
4706 * fold-const.h (tree_nonzero_bits): Likewise.
4707 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4708 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4710 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4714 * varasm.c (assemble_alias): Check ifunc_resolver only on
4717 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4720 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4721 when memory input operand is handled.
4723 2018-05-24 Luis Machado <luis.machado@linaro.org>
4725 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4727 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4729 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4731 * match.pd: Delay FMA folds until after vectorization.
4733 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4736 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4737 address check not strict.
4739 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4741 * gimple-match.h (gimple_match_op): New class.
4742 (mprts_hook): Replace parameters with a gimple_match_op *.
4743 (maybe_build_generic_op): Likewise.
4744 (gimple_simplified_result_is_gimple_val): Replace parameters with
4745 a const gimple_match_op *.
4746 (gimple_simplify): Replace code_helper * and tree * parameters with
4747 a gimple_match_op * parameter.
4748 (gimple_resimplify1): Replace code_helper *, tree and tree *
4749 parameters with a gimple_match_op * parameter.
4750 (gimple_resimplify2): Likewise.
4751 (gimple_resimplify3): Likewise.
4752 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4753 parameters with a gimple_match_op * parameter.
4754 * gimple-match-head.c (gimple_simplify): Change prototypes of
4755 auto-generated functions to take a gimple_match_op * instead of
4756 separate code_helper * and tree * parameters. Make the same
4757 change in the top-level overload and update calls to the
4758 gimple_resimplify routines. Update calls to the auto-generated
4759 functions and to maybe_push_res_to_seq in the publicly-facing
4760 operation-specific gimple_simplify overloads.
4761 (gimple_match_op::MAX_NUM_OPS): Define.
4762 (gimple_resimplify1): Replace rcode and ops with a single res_op
4763 parameter. Update call to gimple_simplify.
4764 (gimple_resimplify2): Likewise.
4765 (gimple_resimplify3): Likewise.
4766 (mprts_hook): Replace parameters with a gimple_match_op *.
4767 (maybe_build_generic_op): Likewise.
4768 (build_call_internal): Replace type, nargs and ops with
4769 a gimple_match_op *.
4770 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4771 with a single gimple_match_op *. Update calls to mprts_hook,
4772 build_call_internal and gimple_simplified_result_is_gimple_val.
4773 Factor out code that is common to the tree_code and combined_fn cases.
4774 * genmatch.c (expr::gen_transform): Replace tem_code and
4775 tem_ops with a gimple_match_op called tem_op. Update calls
4776 to the gimple_resimplify functions and maybe_push_res_to_seq.
4777 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4778 res_ops. Update call to the gimple_resimplify functions.
4779 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4780 (decision_tree::gen): Make the functions take a gimple_match_op *
4781 called res_op instead of separate res_code and res_ops parameters.
4782 Update call accordingly.
4783 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4784 and ops with a single res_op parameter. Update calls to
4785 maybe_build_generic_op and maybe_push_res_to_seq.
4786 (fold_stmt_1): Update calls to gimple_simplify and
4787 replace_stmt_with_simplification.
4788 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4789 and gimple_simplified_result_is_gimple_val.
4790 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4792 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4793 with a gimple_match_op *.
4794 (vn_nary_build_or_lookup): Likewise. Update call to
4795 vn_nary_build_or_lookup_1.
4796 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4797 gimple_match_op *. Update calls to the gimple_resimplify routines
4798 and to gimple_simplified_result_is_gimple_val.
4799 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4800 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4801 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4802 (visit_nary_op): Likewise.
4803 (visit_reference_op_load): Likewise.
4805 2018-05-23 Luis Machado <luis.machado@linaro.org>
4807 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4808 modifier for printing the step amount.
4810 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4813 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4816 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4818 * doc/sourcebuild.texi (Endianness): New subsubsection.
4820 2018-05-23 Luis Machado <luis.machado@linaro.org>
4822 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4823 <prefetch_dynamic_strides>: New const bool field.
4824 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4825 prefetch_dynamic_strides.
4826 (exynosm1_prefetch_tune): Likewise.
4827 (thunderxt88_prefetch_tune): Likewise.
4828 (thunderx_prefetch_tune): Likewise.
4829 (thunderx2t99_prefetch_tune): Likewise.
4830 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4832 (aarch64_override_options_internal): Update to set
4833 PARAM_PREFETCH_DYNAMIC_STRIDES.
4834 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4835 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4836 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4837 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4838 prefetch-dynamic-strides setting.
4840 2018-05-23 Luis Machado <luis.machado@linaro.org>
4842 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4843 <minimum_stride>: New const int field.
4844 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4845 minimum_stride field defaulting to -1.
4846 (exynosm1_prefetch_tune): Likewise.
4847 (thunderxt88_prefetch_tune): Likewise.
4848 (thunderx_prefetch_tune): Likewise.
4849 (thunderx2t99_prefetch_tune): Likewise.
4850 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4851 <default_opt_level>: Set to 3.
4852 (aarch64_override_options_internal): Update to set
4853 PARAM_PREFETCH_MINIMUM_STRIDE.
4854 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4855 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4856 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4857 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4858 stride is constant and is below the minimum stride threshold.
4860 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4862 * config/arm/arm-cpus.in (mode26): Delete.
4863 (armv4): Delete mode26 reference.
4864 * config/arm/arm.c (arm_configure_build_target): Delete use of
4867 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4869 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4871 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4872 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4874 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4875 (floatunsdidf2): Ditto.
4877 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4879 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4880 (fixuns_trunc<mode>si2_avx512f): Ditto.
4881 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4882 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4883 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4885 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4887 PR rtl-optimization/79985
4888 * df-scan.c (df_insn_refs_collect): Remove special case for
4889 global registers and asm statements.
4891 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4893 * extend.texi (Global Register Variables): Rewrite the bullet list.
4894 Note that the register is available for allocation. Note that access
4895 via inline asm must use constraints. Add note about async-signal
4896 handlers. Remove paragraph about automagic register selection.
4898 2018-05-23 Richard Biener <rguenther@suse.de>
4900 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4901 of fixed offset from memset VN.
4903 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4905 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4907 (alloc_cand_and_find_basis): Initialize first_interp field.
4908 (slsr_process_mul): Modify first_interp field.
4909 (slsr_process_add): Likewise.
4910 (slsr_process_cast): Modify first_interp field for each new
4912 (slsr_process_copy): Likewise.
4913 (dump_candidate): Dump first_interp field.
4914 (replace_mult_candidate): Process all interpretations, not just
4916 (replace_rhs_if_not_dup): Likewise.
4917 (replace_one_candidate): Likewise.
4919 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4921 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4923 (aarch64_needs_frame_chain): New function.
4924 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4926 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4929 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4930 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4931 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4932 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4933 as true for strict-align.
4934 (aarch64_can_inline_p): Perform checks even when callee has no
4935 attributes to check for strict alignment.
4936 * doc/extend.texi (AArch64 Function Attributes): Document
4938 * doc/invoke.texi: (AArch64 Options): Likewise.
4940 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4942 PR tree-optimization/85853
4943 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4944 the handling of the root of the node to...
4945 (vect_slp_analyze_node_operations_1): ...this new function,
4946 and run the whole thing with the child nodes' def types
4947 set according to their SLP node's def type.
4949 2018-05-23 Richard Biener <rguenther@suse.de>
4952 * tree-data-ref.c (create_runtime_alias_checks): Defer
4953 and ignore overflow warnings.
4955 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4957 PR tree-optimization/85822
4958 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4961 2018-05-23 Richard Biener <rguenther@suse.de>
4963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4964 memset constants via native_interpret_expr.
4966 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4969 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4971 (cgraph_node::create_alias): Likewise.
4972 (cgraph_node::get_availability): Check ifunc_resolver instead
4973 of looking up ifunc attribute.
4974 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4975 * varasm.c (do_assemble_alias): Likewise.
4976 (assemble_alias): Likewise.
4977 (default_binds_local_p_3): Likewise.
4978 * cgraph.h (cgraph_node): Add ifunc_resolver.
4979 (cgraph_node::only_called_directly_or_aliased_p): Return false
4981 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4983 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4984 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4985 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4986 instead of looking up ifunc attribute.
4988 2018-05-22 Luis Machado <luis.machado@linaro.org>
4990 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4992 2018-05-22 Martin Sebor <msebor@redhat.com>
4995 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4996 only when expasion succeeds.
4997 (expand_builtin_strcmp): Same.
4998 (expand_builtin_strncmp): Same.
5000 2018-05-22 Martin Sebor <msebor@redhat.com>
5002 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5004 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5005 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5007 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5008 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5010 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5011 Add check for legitimate_address.
5012 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5013 (aarch64_swap_ldrstr_operands): New.
5014 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5017 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5018 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5020 * config/aarch64/aarch64.md: New patterns to generate stp
5022 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5023 single words and double words.
5024 (load_pair_sw, load_pair_dw): Likewise.
5025 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5027 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5029 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5030 for different mode ldpstp and add peephole for merged zero stores.
5032 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5034 (aarch64_gen_store_pair): Rename calls to match new patterns.
5035 (aarch64_gen_load_pair): Rename calls to match new patterns.
5036 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5037 (load_pair<DREG:mode><DREG2:mode>): ... This.
5038 (store_pair<mode>): Rename to...
5039 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5040 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5042 (V_INT_EQUIV): Handle SImode.
5043 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5046 2018-05-22 Martin Sebor <msebor@redhat.com>
5049 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5050 or ajust the presumed bound on an operation to avoid unnecessary
5053 2018-05-22 Martin Sebor <msebor@redhat.com>
5055 PR tree-optimization/85826
5056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5057 assuming that a DECL necesarily has a constant size.
5059 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5062 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5064 2018-05-22 Richard Biener <rguenther@suse.de>
5066 PR tree-optimization/85834
5067 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5068 non-constant and non-zero memset arguments.
5070 2018-05-22 Martin Liska <mliska@suse.cz>
5073 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5075 2018-05-22 Richard Biener <rguenther@suse.de>
5077 PR tree-optimization/85863
5078 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5079 comparisons when vectype is specified.
5080 (vectorizable_condition): Do not specify vectype for
5081 vect_is_simple_cond when SLP vectorizing.
5083 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5086 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5087 define __ibm128 as long double.
5088 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5089 as a distinct type when IEEE 128-bit support is enabled.
5090 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5091 128-bit types to use the correct functions.
5092 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5093 convert between 128-bit floating point types that have different
5094 modes but the same representation, instead of using gen_lowpart to
5096 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5098 (IFKF_reg): New attributes to give the register constraints for
5100 (extend<mode>tf2_internal): New insns to mark an explicit
5101 conversion between 128-bit floating point types that have a
5102 different mode but share the same representation.
5104 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5106 PR tree-optimization/85814
5107 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5108 a null return from get_strinfo when unsharing the next
5109 strinfo in the chain.
5111 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5114 * varasm.c (weak_finish): Clean up weak_decls.
5116 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5118 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5119 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5120 UNSPEC_UADALP values.
5121 * config/aarch64/iterators.md (ABAL): New int iterator.
5124 (sur): Add mappings for the above.
5125 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5127 (aarch64_<sur>abal<mode>_4): Likewise.
5128 (aarch64_<sur>adalp<mode>_3): Likewise.
5129 (<sur>sadv16qi): New define_expand.
5131 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5133 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5134 (*movdf_internal): Ditto.
5135 (*rcpsf2_sse): Ditto.
5136 (*rsqrtsf2_sse): Ditto.
5137 (*sqrt<mode>2_sse): Ditto.
5139 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5141 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5143 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5144 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5145 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5146 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5148 * config/aarch64/arm_neon.h: Likewise.
5149 * config/aarch64/iterators.md (VQ_I): New.
5151 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5153 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5155 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5158 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5160 * config/nds32/constraints.md (S): New constraint.
5161 * config/nds32/nds32.md (call_internal): Use constraint S.
5162 (call_value_internal): Likewise.
5163 (sibcall_internal): Likewise.
5164 (sibcall_value_internal): Likewise.
5166 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5167 Chung-Ju Wu <jasonwucj@gmail.com>
5169 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5172 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5173 Chung-Ju Wu <jasonwucj@gmail.com>
5175 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5176 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5177 (nds32_rtx_costs_impl): Simplify.
5178 (nds32_address_cost_impl): Simplify.
5179 (nds32_init_rtx_costs): New function.
5180 (nds32_rtx_costs_speed_prefer): Likewise.
5181 (nds32_rtx_costs_size_prefer): Likewise.
5182 (nds32_address_cost_speed_prefer): Likewise.
5183 (nds32_address_cost_speed_fwprop): Likewise.
5184 (nds32_address_cost_size_prefer): Likewise.
5185 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5186 * config/nds32/nds32.c (nds32_option_override): Use
5187 nds32_init_rtx_costs function.
5189 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5191 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5192 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5193 (TARGET_PIPELINE_N8): Likewise.
5194 (TARGET_PIPELINE_N10): Likewise.
5195 (TARGET_PIPELINE_N13): Likewise.
5196 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5198 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5200 * config/nds32/nds32-fpu.md: Update copyright year.
5202 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5204 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5206 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5208 * config/nds32/nds32.c
5209 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5210 * config/nds32/nds32.opt (minline-asm-r15): New option.
5212 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5214 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5216 * config/nds32/nds32.md (abssi2): New pattern.
5218 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5220 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5221 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5222 Merge insn pattern from sse_cvtsi2ss<round_name> and
5223 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5224 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5225 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5226 using SWI48 mode iterator.
5227 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5228 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5229 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5230 pattern from sse_cvttss2si<round_saeonly_name>
5231 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5232 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5233 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5234 using SWI48 mode iterator.
5235 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5236 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5237 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5238 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5239 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5240 using SWI48 mode iterator.
5241 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5242 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5243 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5244 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5245 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5246 SWI48 mode iterator.
5247 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5248 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5249 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5250 pattern from sse_cvttsd2si<round_saeonly_name>
5251 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5253 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5255 * config/nds32/nds32-md-auxiliary.c
5256 (nds32_valid_smw_lwm_base_p): Refine.
5257 (nds32_output_smw_single_word): Refine.
5258 (nds32_output_smw_double_word): New.
5259 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5261 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5263 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5264 (nds32_output_stack_pop): Refine.
5265 (nds32_expand_unaligned_load): Refine.
5266 (nds32_expand_unaligned_store): Refine.
5268 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5269 Chung-Ju Wu <jasonwucj@gmail.com>
5271 * config/nds32/constants.md: Add TP_REGNUM constant.
5272 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5273 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5275 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5276 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5277 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5278 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5279 and PIC code generation.
5280 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5282 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5284 * config/nds32/nds32.md: Support TLS and PIC.
5285 * config/nds32/nds32.c: Support TLS and PIC.
5286 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5287 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5290 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5292 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5293 mode with E_ prefix.
5295 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5296 Chung-Ju Wu <jasonwucj@gmail.com>
5298 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5299 * config/nds32/nds32-md-auxiliary.c
5300 (symbolic_reference_mentioned_p): New.
5301 (nds32_legitimize_ict_address): New.
5302 (nds32_expand_ict_move): New.
5303 (nds32_indirect_call_referenced_p): New.
5304 (nds32_symbol_binds_local_p): Delete.
5305 (nds32_long_call_p): Modify.
5306 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5307 * config/nds32/nds32-protos.h
5308 (symbolic_reference_mentioned_p): Declare.
5309 (nds32_legitimize_ict_address): Declare.
5310 (nds32_expand_ict_move): Declare.
5311 (nds32_indirect_call_referenced_p): Declare.
5312 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5313 (nds32_relax_group): Use nds32_ict_const_p as condition.
5314 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5315 (nds32_asm_file_start): Output ict_model directive in asm code.
5316 (nds32_legitimate_address_p): Consider indirect call.
5317 (nds32_print_operand): Consider indirect call.
5318 (nds32_print_operand_address): Consider indirect call.
5319 (nds32_insert_attributes): Handle "indirect_call" attribute.
5320 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5321 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5322 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5323 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5324 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5325 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5326 (TARGET_ICT_MODEL_SMALL): Define.
5327 (TARGET_ICT_MODEL_LARGE): Define.
5328 * config/nds32/nds32.md (movsi): Consider ict model.
5329 (call, call_value): Consider ict model.
5330 (sibcall, sibcall_value): Consider ict model.
5331 * config/nds32/nds32.opt (mict-model): New option.
5332 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5335 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5336 Monk Chiang <sh.chiang04@gmail.com>
5337 Jim Wilson <jimw@sifive.com>
5339 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5340 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5341 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5342 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5343 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5344 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5345 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5346 compute save_libcall_adjustment properly.
5347 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5348 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5349 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5350 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5351 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5352 (ABI_SPEC): Handle mabi=ilp32e.
5353 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5354 (RVE): Add RVE mask.
5355 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5356 <-march>: Add rv32e as an example.
5358 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5361 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5362 (intra_create_variable_infos): Handle C++ constructors.
5364 2018-05-18 Martin Liska <mliska@suse.cz>
5366 * passes.def: Remove a redundant pass.
5368 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5371 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5373 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5375 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5377 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5379 (armv2, armv2a, armv3, armv3m): Delete architectures.
5380 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5381 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5382 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5384 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5385 (*mulsidi3adddi): Likewise.
5386 (mulsidi3): Likewise.
5387 (*mulsidi3_nov6): Likewise.
5388 (umulsidi3): Likewise.
5389 (umulsidi3_nov6): Likewise.
5390 (umaddsidi4): Likewise.
5391 (*umulsidi3adddi): Likewise.
5392 (smulsi3_highpart): Likewise.
5393 (*smulsi3_highpart_nov6): Likewise.
5394 (umulsi3_highpart): Likewise.
5395 (*umulsi3_highpart_nov6): Likewise.
5396 * config/arm/arm.h (arm_arch3m): Delete.
5397 * config/arm/arm.c (arm_arch3m): Delete.
5398 (arm_option_override_internal): Update armv3-related comment.
5399 (arm_configure_build_target): Delete use of isa_bit_mode32.
5400 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5401 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5402 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5404 (mulusa3): Likewise.
5405 * config/arm/arm-protos.h (arm_arch3m): Delete.
5406 * config/arm/arm-tables.opt: Regenerate.
5407 * config/arm/arm-tune.md: Likewise.
5408 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5409 deleted architectures.
5411 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5413 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5414 (armv5t, armv5te): New features.
5415 (ARMv5, ARMv5e): Delete fgroups.
5416 (ARMv5t, ARMv5te): Adjust for above changes.
5418 (armv5, armv5e): Delete arches.
5419 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5421 (*call_reg_arm): Likewise.
5422 (*call_value_reg_armv5): Likewise.
5423 (*call_value_reg_arm): Likewise.
5424 (*call_symbol): Likewise.
5425 (*call_value_symbol): Likewise.
5426 (*sibcall_insn): Likewise.
5427 (*sibcall_value_insn): Likewise.
5429 (prefetch): Likewise.
5430 (define_split and define_peephole2 dependent on arm_arch5):
5432 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5434 (TARGET_ARM_QBIT): Likewise.
5435 (TARGET_DSP_MULTIPLY): Likewise.
5436 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5437 (arm_arch5, arm_arch5e): Delete.
5438 (arm_arch5t, arm_arch5te): Declare.
5439 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5440 (arm_arch5t): Declare.
5441 (arm_option_reconfigure_globals): Update for the above.
5442 (arm_options_perform_arch_sanity_checks): Update comment, replace
5443 use of arm_arch5 with arm_arch5t.
5444 (use_return_insn): Likewise.
5445 (arm_emit_call_insn): Likewise.
5446 (output_return_instruction): Likewise.
5447 (arm_final_prescan_insn): Likewise.
5448 (arm_coproc_builtin_available): Likewise.
5449 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5450 arm_arch5e with arm_arch5t and arm_arch5te.
5451 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5452 (arm_arch5t, arm_arch5te): Declare.
5453 * config/arm/arm-tables.opt: Regenerate.
5454 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5455 * config/arm/t-multilib: Likewise.
5456 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5457 instead of arm_arch5.
5458 (*call_reg_thumb1): Likewise.
5459 (*call_value_reg_thumb1_v5): Likewise.
5460 (*call_value_reg_thumb1): Likewise.
5461 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5463 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5465 2018-05-18 Martin Liska <mliska@suse.cz>
5467 PR gcov-profile/84846
5468 * doc/gcov.texi: Document -t option of gcov tool.
5470 2018-05-18 Martin Liska <mliska@suse.cz>
5472 PR gcov-profile/84846
5473 * gcov.c (print_usage): Add new -t option.
5474 (process_args): Handle the option.
5475 (generate_results): Use stdout as output when requested by
5478 2018-05-18 Martin Liska <mliska@suse.cz>
5480 PR gcov-profile/84846
5481 * coverage.c (coverage_init): Write PWD to .gcno file.
5482 * doc/gcov.texi: Document how working directory is printed.
5483 * gcov-dump.c (dump_gcov_file): Print PWD.
5484 * gcov.c (output_intermediate_file): Likewise.
5485 (read_graph_file): Read PWD string.
5486 (output_lines): Print PWD.
5488 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5491 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5492 for retval and return false if all args to phi are zero.
5494 2018-05-18 Richard Biener <rguenther@suse.de>
5496 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5498 (evrp_dom_walker::before_dom_children): Call it.
5500 2018-05-18 Richard Biener <rguenther@suse.de>
5502 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5503 results when processing array refs with variable index.
5505 2018-05-18 Toon Moene <toon@moene.org>
5507 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5508 directly after that of -floop-interchange. Indicate that both
5509 options are enabled by default when specifying -O3.
5511 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5513 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5514 iterator. Delete separate integer-mode vec_set<mode> expander.
5515 (aarch64_simd_vec_setv2di): Delete.
5516 (vec_setv2di): Delete.
5517 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5518 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5519 the "w, r" alternative.
5521 2018-05-18 Martin Liska <mliska@suse.cz>
5523 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5524 * tree-pass.h (make_pass_lower_switch_O0): New function.
5525 * tree-switch-conversion.c (node_has_low_bound): Remove.
5526 (node_has_high_bound): Likewise.
5527 (node_is_bounded): Likewise.
5528 (class pass_lower_switch): Make it a template type and create
5530 (pass_lower_switch::execute): Add template argument.
5531 (make_pass_lower_switch): New function.
5532 (make_pass_lower_switch_O0): New function.
5533 (do_jump_if_equal): Remove.
5534 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5535 all the hard work to tree optimization passes.
5537 2018-05-18 Martin Liska <mliska@suse.cz>
5539 * dbgcnt.c (limit_low): Renamed from limit.
5540 (limit_high): New variable.
5541 (dbg_cnt_is_enabled): Check for upper limit.
5542 (dbg_cnt): Adjust dumping.
5543 (dbg_cnt_set_limit_by_index): Add new argument for high
5545 (dbg_cnt_set_limit_by_name): Likewise.
5546 (dbg_cnt_process_single_pair): Parse new format.
5547 (dbg_cnt_process_opt): Use strtok.
5548 (dbg_cnt_list_all_counters): Remove 'value' and add
5550 * doc/invoke.texi: Document changes.
5552 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5554 * doc/sourcebuild.texi (scalar_all_fma): Document.
5555 * tree.def (FMA_EXPR): Delete.
5556 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5557 * internal-fn.c (ternary_direct): New macro.
5558 (expand_ternary_optab_fn): Likewise.
5559 (direct_ternary_optab_supported_p): Likewise.
5560 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5561 * builtins.c (fold_builtin_fma): Delete.
5562 (fold_builtin_3): Don't call it.
5563 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5564 * expr.c (expand_expr_real_2): Likewise.
5565 * fold-const.c (operand_equal_p): Likewise.
5566 (fold_ternary_loc): Likewise.
5567 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5568 * gimple.c (DEFTREECODE): Likewise.
5569 * gimplify.c (gimplify_expr): Likewise.
5570 * optabs-tree.c (optab_for_tree_code): Likewise.
5571 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5572 * tree-eh.c (operation_could_trap_p): Likewise.
5573 (stmt_could_throw_1_p): Likewise.
5574 * tree-inline.c (estimate_operator_cost): Likewise.
5575 * tree-pretty-print.c (dump_generic_node): Likewise.
5576 (op_code_prio): Likewise.
5577 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5578 * tree-ssa-operands.c (get_expr_operands): Likewise.
5579 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5580 * fold-const-call.h (fold_fma): Delete.
5581 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5582 CFN_FNMA and CFN_FNMS.
5584 * genmatch.c (combined_fn): New enum.
5585 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5586 (commutative_op): New function.
5587 (commutate): Use it. Handle more than 2 operands.
5588 (dt_operand::gen_gimple_expr): Use commutative_op.
5589 (parser::parse_expr): Allow :c to be used with non-binary
5590 operators if the commutative operand is known.
5591 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5592 CFN_FMS, CFN_FNMA and CFN_FNMS.
5593 (backprop::process_assign_use): Remove FMA_EXPR handling.
5594 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5595 (gen_hsa_fma): New function.
5596 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5597 IFN_FNMA and IFN_FNMS.
5598 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5599 * gimple-fold.h (follow_all_ssa_edges): Declare.
5600 * gimple-fold.c (follow_all_ssa_edges): New function.
5601 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5602 gimple_build interface and use follow_all_ssa_edges to fold the result.
5603 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5604 instead of checking for optabs directly.
5605 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5606 rather than FMA_EXPRs.
5607 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5608 call to IFN_FMA instead of an FMA_EXPR.
5610 2018-05-17 Jim Wilson <jimw@sifive.com>
5612 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5613 sign extended promoted subreg, and the range does not have the sign bit
5614 set, then do a sign extend.
5616 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5617 test, check for sign extended subreg and/or constant operands, and
5618 do a sign extend in that case.
5620 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5622 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5623 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5625 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5626 Change logics_shift_reg to logics_shift_imm.
5627 (thunderx2t99_fp_loadpair_basic): Delete.
5628 (thunderx2t99_fp_storepair_basic): Delete.
5629 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5630 (thunderx2t99_asimd_polynomial): Delete.
5631 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5632 and neon_fp_mul_d_scalar_q.
5633 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5634 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5635 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5636 (thunderx2t99_asimd_lut): Add missing tbl types.
5637 (thunderx2t99_asimd_ext): Delete.
5638 (thunderx2t99_asimd_load1_1_mult): Delete.
5639 (thunderx2t99_asimd_load1_2_mult): Delete.
5640 (thunderx2t99_asimd_load1_ldp): New.
5641 (thunderx2t99_asimd_load1): New.
5642 (thunderx2t99_asimd_load2): Add missing *load2* types.
5643 (thunderx2t99_asimd_load3): New.
5644 (thunderx2t99_asimd_load4): New.
5645 (thunderx2t99_asimd_store1_1_mult): Delete.
5646 (thunderx2t99_asimd_store1_2_mult): Delete.
5647 (thunderx2t99_asimd_store2_mult): Delete.
5648 (thunderx2t99_asimd_store2_onelane): Delete.
5649 (thunderx2t99_asimd_store_stp): New.
5650 (thunderx2t99_asimd_store1): New.
5651 (thunderx2t99_asimd_store2): New.
5652 (thunderx2t99_asimd_store3): New.
5653 (thunderx2t99_asimd_store4): New.
5655 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5657 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5658 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5660 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5661 Segher Boessenkool <segher@kernel.crashing.org>
5664 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5667 2018-05-17 Richard Biener <rguenther@suse.de>
5669 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5670 for pruning loop and prune defs feeding only already visited PHIs.
5672 2018-05-17 Richard Biener <rguenther@suse.de>
5674 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5676 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5677 Richard Biener <rguenther@suse.de>
5679 PR tree-optimization/85793
5680 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5681 for VMAT_ELEMENTWISE.
5683 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5685 * internal-fn.h (lookup_internal_fn): Declare
5686 * internal-fn.c (lookup_internal_fn): New function.
5687 * gimple.c (gimple_build_call_from_tree): Handle calls to
5689 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5690 internal function names.
5691 * tree-pretty-print.c (dump_generic_node): Likewise.
5692 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5694 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5696 * gimple-fold.h (gimple_build): Make the function forms take
5697 combined_fn rather than built_in_function.
5698 (gimple_simplify): Likewise.
5699 * gimple-match-head.c (gimple_simplify): Likewise.
5700 * gimple-fold.c (gimple_build): Likewise.
5701 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5702 rather than gimple_build_call_internal.
5703 (get_initial_defs_for_reduction): Likewise.
5704 (vect_create_epilog_for_reduction): Likewise.
5705 (vectorizable_live_operation): Likewise.
5707 2018-05-17 Martin Liska <mliska@suse.cz>
5709 * gimple-ssa-sprintf.c (format_directive): Do not use
5710 space in between 'G_' and '('.
5712 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5715 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5716 even if the mask is not all ones.
5719 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5721 (ix86_gimple_fold_builtin): Likewise.
5724 * config/i386/i386.c: Include tree-vector-builder.h.
5725 (ix86_vector_shift_count): New function.
5726 (ix86_fold_builtin): Fold shift builtins by scalar count.
5727 (ix86_gimple_fold_builtin): Likewise.
5729 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5730 _mm512_setzero): New intrinsics.
5732 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5733 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5735 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5736 code generation for cases where splatting a value is not useful.
5737 * simplify-rtx.c (simplify_ternary_operation): Simplify
5738 vec_merge across a vec_duplicate and a paradoxical subreg forming
5739 a vector mode to a vec_concat.
5741 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5743 * config.gcc: Support "goldmont-plus".
5744 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5746 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5747 PROCESSOR_GOLDMONT_PLUS.
5748 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5749 (processor_target_table): Add "goldmont-plus".
5750 (PTA_GOLDMONT_PLUS): Define.
5751 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5752 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5753 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5754 (fold_builtin_cpu): Add "goldmont-plus".
5755 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5756 (ix86_option_override_internal): Add "goldmont-plus".
5757 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5758 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5759 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5760 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5762 2018-05-17 Richard Biener <rguenther@suse.de>
5764 PR tree-optimization/85757
5765 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5766 remove defs that only feed that PHI from further processing.
5768 2018-05-16 Jim Wilson <jimw@sifive.com>
5770 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5772 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5774 2018-05-16 Mark Wielaard <mark@klomp.org>
5776 * dwarf2out.c (count_index_strings): New function.
5777 (output_indirect_strings): Call count_index_strings and generate
5778 header for dwarf_version >= 5.
5780 2018-05-16 Mark Wielaard <mark@klomp.org>
5782 * dwarf2out.c (dwarf_FORM): New function.
5783 (set_indirect_string): Use dwarf_FORM.
5784 (reset_indirect_string): Likewise.
5785 (size_of_die): Likewise.
5786 (value_format): Likewise.
5787 (output_die): Likewise.
5788 (add_skeleton_AT_string): Likewise.
5789 (output_macinfo_op): Likewise.
5790 (index_string): Likewise.
5791 (output_index_string_offset): Likewise.
5792 (output_index_string): Likewise.
5793 (count_index_strings): Likewise.
5795 2018-05-16 Carl Love <cel@us.ibm.com>
5797 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5798 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5800 2018-05-16 Martin Jambor <mjambor@suse.cz>
5802 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5803 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5805 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5807 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5808 (fnma<mode>4): Likewise.
5809 (fms<mode>4): Likewise.
5810 (fnms<mode>4): Likewise.
5811 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5812 (aarch64_fnma<mode>4): Likewise.
5813 (aarch64_fms<mode>4): Likewise.
5814 (aarch64_fnms<mode>4): Likewise.
5815 (aarch64_fnmadd<mode>4): Likewise.
5817 2018-05-16 Jason Merrill <jason@redhat.com>
5819 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5821 2018-05-16 Richard Biener <rguenther@suse.de>
5823 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5824 (dump_stmt_cost): Declare.
5825 (add_stmt_cost): Dump cost we add.
5826 (add_stmt_costs): New function.
5827 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5829 (vect_analyze_stmt): Adjust prototype.
5830 (vectorizable_condition): Likewise.
5831 (vectorizable_live_operation): Likewise.
5832 (vectorizable_reduction): Likewise.
5833 (vectorizable_induction): Likewise.
5834 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5835 cost vector to pass to vectorizable_ and record afterwards.
5836 (vect_model_reduction_cost): Take cost vector argument and adjust.
5837 (vect_model_induction_cost): Likewise.
5838 (vectorizable_reduction): Likewise.
5839 (vectorizable_induction): Likewise.
5840 (vectorizable_live_operation): Likewise.
5841 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5842 SLP_TREE_NUMBER_OF_VEC_STMTS.
5843 (vect_analyze_slp_cost_1): Remove.
5844 (vect_analyze_slp_cost): Likewise.
5845 (vect_slp_analyze_node_operations): Take visited args and
5846 a target cost vector. Avoid processing already visited stmt sets.
5847 (vect_slp_analyze_operations): Use a local cost vector to gather
5848 costs and register those of non-discarded instances.
5849 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5850 (vect_schedule_slp_instance): Remove copying of
5851 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5853 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5854 adding cost. Record cost entry location.
5855 (vect_prologue_cost_for_slp_op): Function to compute cost of
5856 a constant or invariant generated for SLP vect in the prologue,
5857 split out from vect_analyze_slp_cost_1.
5858 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5859 (vect_model_promotion_demotion_cost): Likewise.
5860 (vect_model_store_cost): Likewise, make static.
5861 (vect_model_load_cost): Likewise.
5862 (vectorizable_bswap): Add cost vector arg and adjust.
5863 (vectorizable_call): Likewise.
5864 (vectorizable_simd_clone_call): Likewise.
5865 (vectorizable_conversion): Likewise.
5866 (vectorizable_assignment): Likewise.
5867 (vectorizable_shift): Likewise.
5868 (vectorizable_operation): Likewise.
5869 (vectorizable_store): Likewise.
5870 (vectorizable_load): Likewise.
5871 (vectorizable_condition): Likewise.
5872 (vectorizable_comparison): Likewise.
5873 (can_vectorize_live_stmts): Likewise.
5874 (vect_analyze_stmt): Likewise.
5875 (vect_transform_stmt): Adjust calls to vectorizable_*.
5876 * tree-vectorizer.c: Include gimple-pretty-print.h.
5877 (dump_stmt_cost): New function.
5879 2018-05-16 Richard Biener <rguenther@suse.de>
5881 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5882 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5883 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5884 (check_name): New callback.
5885 (dse_classify_store): Track cycles via a visited bitmap of PHI
5886 defs and simplify handling of in-loop and across loop dead stores
5887 and properly fail for loop-variant refs. Handle byte-tracking with
5888 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5891 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5893 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5894 (vect_get_mask_type_for_stmt): Likewise.
5895 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5897 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5898 to determine the statement's vector type and the vector type that
5899 should be used for calculating nunits. Deal with cases in which
5900 the type has to be deferred.
5901 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5902 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5903 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5904 (vect_determine_vf_for_stmt): New functions, split out from...
5905 (vect_determine_vectorization_factor): ...here.
5906 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5907 (vect_get_mask_type_for_stmt): New functions, split out from
5908 vect_determine_vectorization_factor.
5910 2018-05-16 Richard Biener <rguenther@suse.de>
5912 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5913 verify the [VEC_]COND_EXPR embedded comparison.
5915 2018-05-15 Martin Sebor <msebor@redhat.com>
5917 PR tree-optimization/85753
5918 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5919 RECORD_TYPE in addition to ARRAY_TYPE.
5921 2018-05-15 Martin Sebor <msebor@redhat.com>
5924 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5926 2018-05-15 Richard Biener <rguenther@suse.de>
5928 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5929 add by_clobber_p one. Change algorithm to collect all defs
5930 representing uses we need to walk and try reducing them to
5931 a single one before failing.
5932 (dse_dom_walker::dse_optimize_stmt): Adjust.
5934 2018-05-13 Mark Wielaard <mark@klomp.org>
5936 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5937 (size_of_loc_descr): Likewise.
5938 (output_loc_operands): Likewise.
5939 (output_loc_operands_raw): Likewise.
5940 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5941 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5942 (hash_loc_operands): Likewise.
5943 (compare_loc_operands): Likewise.
5945 2018-05-14 Mark Wielaard <mark@klomp.org>
5947 * dwarf2out.c (count_index_addrs): New function.
5948 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5950 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5952 PR tree-optimization/83648
5953 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5954 return value as malloc candidate.
5956 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5959 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5960 param as true in call to suggest_attribute.
5962 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5964 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5965 -mreadonly-in-sdata.
5967 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5969 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5971 (aarch64_crypto_aesd_fused): Likewise.
5973 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5975 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5976 (movsi_aarch64): Likewise.
5977 (load_pairsi): Likewise.
5978 (load_pairdi): Likewise.
5979 (store_pairsi): Likewise.
5980 (store_pairdi): Likewise.
5981 (load_pairsf): Likewise.
5982 (load_pairdf): Likewise.
5983 (store_pairsf): Likewise.
5984 (store_pairdf): Likewise.
5985 (zero_extend): Likewise.
5986 (trunc): Swap alternatives.
5987 (fcvt_target): Add '?' to prefer w over r.
5989 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5992 * config/i386/i386.md: Disallow non-commutative arithmetics in
5993 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5994 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5995 in the peephole2 before it.
5997 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5999 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6000 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6001 (ix86_handle_option): Handle -mcldemote.
6002 * config.gcc: New header.
6003 * config/i386/cldemoteintrin.h: New file.
6004 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6005 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6007 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6008 OPTION_MASK_ISA_CLDEMOTE.
6009 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6010 (ix86_valid_target_attribute_inner_p): Ditto.
6011 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6012 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6013 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6014 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6015 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6017 * config/i386/i386.opt: Add -mcldemote.
6018 * config/i386/x86intrin.h: New header.
6019 * doc/invoke.texi: Add -mcldemote.
6021 2018-05-14 Richard Biener <rguenther@suse.de>
6023 * doc/match-and-simplify.texi: Adjust :s documentation.
6025 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6027 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6028 intended memcpy size.
6029 (REORDER_45): Likewise.
6031 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6033 * sort.cc: New file.
6034 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6035 * vec.c (qsort_chk): Use gcc_qsort.
6036 * Makefile.in (OBJS-libcommon): Add sort.o.
6037 (build/sort.o): New target. Use it...
6038 (BUILD_RTL): ... here, and...
6039 (build/gencfn-macros): ... here, and...
6040 (build/genmatch): ... here.
6042 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6043 Chung-Ju Wu <jasonwucj@gmail.com>
6045 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6046 * config/nds32/nds32-graywolf.md: New file.
6047 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6048 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6050 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6051 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6052 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6053 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6054 * config/nds32/pipelines.md: Include n15 settings.
6056 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6057 Chung-Ju Wu <jasonwucj@gmail.com>
6059 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6060 * config/nds32/nds32-n13.md: New file.
6061 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6062 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6064 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6065 * config/nds32/nds32.md (pipeline_model): Add n13.
6066 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6067 * config/nds32/pipelines.md: Include n13 settings.
6069 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6070 Chung-Ju Wu <jasonwucj@gmail.com>
6072 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6073 * config/nds32/nds32-n10.md: New file.
6074 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6075 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6077 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6078 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6079 * config/nds32/nds32.md (pipeline_model): Add n10.
6080 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6081 * config/nds32/pipelines.md: Include n10 settings.
6083 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6084 Kito Cheng <kito.cheng@gmail.com>
6085 Chung-Ju Wu <jasonwucj@gmail.com>
6087 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6088 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6089 Add enum values for DSP extension instructions.
6090 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6092 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6093 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6095 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6096 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6097 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6098 * config/nds32/nds32-intrinsic.md: Likewise.
6099 * config/nds32/nds32_intrinsic.h: Likewise.
6100 * config/nds32/nds32-md-auxiliary.c: Likewise.
6101 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6102 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6103 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6104 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6105 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6106 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6107 TYPE_DMAC in switch statement.
6108 * config/nds32/nds32.c: New checking and implementation for DSP
6109 extension instructions.
6110 * config/nds32/nds32.h: Likewise.
6111 * config/nds32/nds32.md: Likewise.
6112 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6113 * config/nds32/predicates.md: Implement new predicates for DSP
6116 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6118 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6119 Reformat alternatives and attributes so it is easier to identify
6120 which constraints/attributes go with which instruction.
6121 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6122 (mov<mode>_softfloat32, FMOVE64): Likewise.
6123 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6124 (mov<mode>_softfloat64, FMOVE64): Likewise.
6126 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6128 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6130 (Basic PowerPC Built-in Functions): The new name of the
6131 subsection previously known as "PowerPC Built-in Functions".
6132 (Basic PowerPC Built-in Functions Available on all Configurations):
6134 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6135 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6136 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6137 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6139 2018-05-11 Martin Jambor <mjambor@suse.cz>
6142 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6145 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6148 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6150 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6152 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6153 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6154 (ix86_handle_option): Handle -mwaitpkg.
6155 * config.gcc: New header.
6156 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6157 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6158 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6160 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6161 OPTION_MASK_ISA_WAITPKG.
6162 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6163 (ix86_option_override_internal): Add PTA_WAITPKG.
6164 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6165 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6166 IX86_BUILTIN_TPAUSE.
6167 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6168 __builtin_ia32_umwait and __builtin_ia32_tpause.
6169 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6170 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6171 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6172 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6173 UNSPECV_TPAUSE): New.
6174 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6175 * config/i386/i386.opt: Add -mwaitpkg.
6176 * config/i386/waitpkgintrin.h: New file.
6177 * config/i386/x86intrin.h: New header.
6178 * doc/invoke.texi: Add -mwaitpkg.
6180 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6183 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6185 (cortex-m0): Use armv6s-m isa.
6186 (cortex-m0plus): Likewise.
6187 (cortex-m1): Likewise.
6188 (cortex-m0.small-multiply): Likewise.
6189 (cortex-m0plus.small-multiply): Likewise.
6190 (cortex-m1.small-multiply): Likewise.
6192 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6193 Jakub Jelinek <jakub@redhat.com>
6195 PR tree-optimization/85692
6196 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6197 source permute as well.
6199 2018-05-11 Martin Liska <mliska@suse.cz>
6202 * doc/extend.texi: Document LLVM style format for no_sanitize
6205 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6207 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6208 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6209 (mode_supports_vsx_dform_quad): Likewise.
6210 (mode_supports_vmx_dform): Move these functions to be next to the
6211 other mode_supports functions.
6212 (mode_supports_dq_form): Likewise.
6213 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6214 mode_supports_dq_form.
6215 (reg_offset_addressing_ok_p): Likewise.
6216 (offsettable_ok_by_alignment): Likewise.
6217 (rs6000_legitimate_offset_address_p): Likewise.
6218 (legitimate_lo_sum_address_p): Likewise.
6219 (rs6000_legitimize_address): Likewise.
6220 (rs6000_legitimize_reload_address): Likewise.
6221 (rs6000_secondary_reload_inner): Likewise.
6222 (rs6000_preferred_reload_class): Likewise.
6223 (rs6000_output_move_128bit): Likewise.
6225 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6227 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6228 Generate SImode target register for null target.
6229 <case IX86_BUILTIN_XGETBV>: Ditto.
6230 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6231 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6233 2018-05-10 Carl Love <cel@us.ibm.com>
6235 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6236 dcbtt and dcbtstt if operands[2] is 0.
6238 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6241 * config/i386/sse.md (usadv64qi): New expander.
6243 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6245 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6246 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6247 -maltivec=be support.
6248 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6249 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6250 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6251 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6252 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6253 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6254 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6255 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6256 altivec_vsumsws): Adjust.
6257 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6258 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6259 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6261 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6262 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6263 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6264 (altivec_lve<VI_char>x): Delete expand.
6265 (*altivec_lve<VI_char>x_internal): Rename to...
6266 (altivec_lve<VI_char>x): ... this.
6267 (altivec_lvxl_<mode>): Delete expand.
6268 (*altivec_lvxl_<mode>_internal): Rename to ...
6269 (altivec_lvxl_<mode>): ... this.
6270 (altivec_stvxl_<mode>): Delete expand.
6271 (*altivec_stvxl_<mode>_internal): Rename to ...
6272 (altivec_stvxl_<mode>): ... this.
6273 (altivec_stve<VI_char>x): Delete expand.
6274 (*altivec_stve<VI_char>x_internal): Rename to ...
6275 (altivec_stve<VI_char>x): ... this.
6276 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6277 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6278 reduc_plus_scal_<mode>): Adjust.
6279 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6281 (rs6000_cpu_cpp_builtins): Adjust.
6282 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6283 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6284 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6285 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6286 -maltivec=be support.
6287 (rs6000_split_vec_extract_var): Adjust.
6288 (rs6000_split_v4si_init): Adjust.
6289 (swap_selector_for_mode): Delete.
6290 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6291 altivec_expand_stvex_be): Delete.
6292 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6293 -maltivec=be support.
6294 (rs6000_gimple_fold_builtin): Ditto.
6295 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6297 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6298 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6299 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6300 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6301 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6302 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6303 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6304 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6305 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6306 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6307 anonymous split): Adjust.
6308 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6309 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6311 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6313 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6314 when --with-gxx-include-dir is also specified.
6315 * configure: Regenerate.
6317 2018-05-09 Jim Wilson <jimw@sifive.com>
6320 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6321 * config/riscv/t-withmultilib: New.
6322 * config/riscv/withmultilib.h: New.
6323 * doc/install.texi: Document RISC-V --with-multilib-list support.
6325 2018-05-09 Richard Biener <rguenther@suse.de>
6327 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6329 (vect_bb_vectorization_profitable_p): Adjust. Compute
6330 actual scalar cost using the cost vector and the add_stmt_cost
6333 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6335 PR rtl-optimization/85645
6336 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6337 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6339 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6341 PR rtl-optimization/85645
6342 * shrink-wrap.c (spread_components): Return a boolean saying if
6343 anything was changed.
6344 (try_shrink_wrapping_separate): Iterate spread_components until
6345 nothing changes anymore.
6347 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6349 PR rtl-optimization/85645
6350 * regrename.c (build_def_use): Also kill the chains that include the
6351 destination of a REG_CFA_REGISTER note.
6353 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6355 PR rtl-optimization/85645
6356 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6357 insn that has a REG_CFA_REGISTER note.
6359 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6361 * cfgexpand.c (expand_clobber): New function.
6362 (expand_gimple_stmt_1): Use it.
6363 * tree-vect-stmts.c (vect_clobber_variable): New function,
6365 (vectorizable_simd_clone_call): ...here.
6366 (vectorizable_store): Emit a clobber either side of an
6367 IFN_STORE_LANES sequence.
6368 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6370 2018-05-09 Tom de Vries <tom@codesourcery.com>
6373 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6374 (define_insn "trap_if_false"): Add exit after trap.
6376 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6378 PR rtl-optimization/85638
6379 * bb-reorder.c: Include common/common-target.h.
6380 (create_forwarder_block): New function extracted from...
6381 (fix_up_crossing_landing_pad): ...here. Rename into...
6382 (dw2_fix_up_crossing_landing_pad): ...this.
6383 (sjlj_fix_up_crossing_landing_pad): New function.
6384 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6385 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6386 from both partitions and exit the loop after one iteration.
6388 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6391 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6393 (Basic PowerPC Built-in Functions): The new name of the
6394 subsection previously known as "PowerPC Built-in Functions".
6395 (Basic PowerPC Built-in Functions Available on all Configurations):
6397 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6399 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6401 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6403 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6406 2018-05-08 Jim Wilson <jimw@sifive.com>
6408 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6409 (LD_EMUL_SUFFIX): New.
6410 (LINK_SPEC): Use it.
6412 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6414 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6416 (Basic PowerPC Built-in Functions): The new name of the
6417 subsection previously known as "PowerPC Built-in Functions".
6418 (Basic PowerPC Built-in Functions Available on all Configurations):
6420 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6422 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6424 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6426 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6429 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6432 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6433 after cmpelim optimization.
6435 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6437 * config.gcc: Support "goldmont".
6438 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6439 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6441 * config/i386/i386.c (m_GOLDMONT): Define.
6442 (processor_target_table): Add "goldmont".
6443 (PTA_GOLDMONT): Define.
6444 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6445 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6446 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6447 (fold_builtin_cpu): Add "goldmont".
6448 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6449 (ix86_option_override_internal): Add "goldmont".
6450 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6451 (processor_type): Add PROCESSOR_GOLDMONT.
6452 * config/i386/i386.md: Add CPU "glm".
6453 * config/i386/glm.md: New file.
6454 * config/i386/x86-tune.def: Add m_GOLDMONT.
6455 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6457 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6460 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6462 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6463 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6464 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6467 * config/i386/i386.c (ix86_fold_builtin): Handle
6468 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6471 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6472 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6474 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6477 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6478 (check_arch): Likewise.
6479 (check_fpu): Return the result rather than printing it.
6480 (end arch): Fix operator precedence.
6481 (end cpu): Likewise.
6482 (END): Print the result from check_fpu.
6484 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6485 Alan Hayward <alan.hayward@arm.com>
6486 David Sherwood <david.sherwood@arm.com>
6488 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6489 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6490 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6491 (*fcmuo<mode>_and): New patterns.
6493 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6495 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6496 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6497 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6498 (cmp_op, sve_imm_con): New code attributes.
6499 (SVE_COND_INT_CMP, imm_con): Delete.
6500 (cmp_op): Remove above unspecs from int attribute.
6501 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6503 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6504 comparison-specific unspecs.
6505 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6506 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6507 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6508 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6509 (*vec_fcm<cmp_op><mode>): Rename to...
6510 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6511 (*vec_fcmuo<mode>): Rename to...
6512 (*fcmuo<mode>): ...this and adjust likewise.
6513 (*pred_fcm<cmp_op><mode>): New pattern.
6514 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6515 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6517 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6519 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6520 (aarch64_emit_sve_predicated_cond): New function.
6521 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6522 (aarch64_emit_unspec_cond_or): Replace with...
6523 (aarch64_emit_sve_or_conds): ...this new function. Use
6524 aarch64_emit_sve_ptrue_op for the individual comparisons and
6525 aarch64_emit_binop to OR them together.
6526 (aarch64_emit_inverted_unspec_cond): Replace with...
6527 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6528 aarch64_emit_sve_ptrue_op for the comparison and
6529 aarch64_emit_unop to invert the result.
6530 (aarch64_expand_sve_vec_cmp_float): Update after the above
6531 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6533 2018-05-07 Nathan Sidwell <nathan@acm.org>
6535 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6536 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6537 (Backwards Compatibility): Likewise.
6539 2018-05-07 Luis Machado <luis.machado@linaro.org>
6543 2018-05-07 Luis Machado <luis.machado@linaro.org>
6545 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6546 <prefetch_dynamic_strides>: New const bool field.
6547 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6548 prefetch_dynamic_strides.
6549 (exynosm1_prefetch_tune): Likewise.
6550 (thunderxt88_prefetch_tune): Likewise.
6551 (thunderx_prefetch_tune): Likewise.
6552 (thunderx2t99_prefetch_tune): Likewise.
6553 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6555 (aarch64_override_options_internal): Update to set
6556 PARAM_PREFETCH_DYNAMIC_STRIDES.
6557 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6558 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6559 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6560 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6561 prefetch-dynamic-strides setting.
6563 2018-05-07 Luis Machado <luis.machado@linaro.org>
6565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6566 <minimum_stride>: New const int field.
6567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6568 minimum_stride field.
6569 (exynosm1_prefetch_tune): Likewise.
6570 (thunderxt88_prefetch_tune): Likewise.
6571 (thunderx_prefetch_tune): Likewise.
6572 (thunderx2t99_prefetch_tune): Likewise.
6573 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6574 (aarch64_override_options_internal): Update to set
6575 PARAM_PREFETCH_MINIMUM_STRIDE.
6576 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6577 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6578 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6579 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6580 stride is constant and is below the minimum stride threshold.
6582 2018-05-07 Luis Machado <luis.machado@linaro.org>
6584 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6587 2018-05-07 Luis Machado <luis.machado@linaro.org>
6589 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6590 <prefetch_dynamic_strides>: New const bool field.
6591 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6592 prefetch_dynamic_strides.
6593 (exynosm1_prefetch_tune): Likewise.
6594 (thunderxt88_prefetch_tune): Likewise.
6595 (thunderx_prefetch_tune): Likewise.
6596 (thunderx2t99_prefetch_tune): Likewise.
6597 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6599 (aarch64_override_options_internal): Update to set
6600 PARAM_PREFETCH_DYNAMIC_STRIDES.
6601 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6602 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6603 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6604 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6605 prefetch-dynamic-strides setting.
6607 2018-05-07 Luis Machado <luis.machado@linaro.org>
6609 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6610 <minimum_stride>: New const int field.
6611 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6612 minimum_stride field.
6613 (exynosm1_prefetch_tune): Likewise.
6614 (thunderxt88_prefetch_tune): Likewise.
6615 (thunderx_prefetch_tune): Likewise.
6616 (thunderx2t99_prefetch_tune): Likewise.
6617 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6618 (aarch64_override_options_internal): Update to set
6619 PARAM_PREFETCH_MINIMUM_STRIDE.
6620 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6621 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6622 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6623 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6624 stride is constant and is below the minimum stride threshold.
6626 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6629 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6630 the type is addressable. Don't force op into register if it has
6633 2018-05-05 Roland McGrath <mcgrathr@google.com>
6636 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6637 any section for which we don't know a specific type it should have,
6638 regardless of name. Previously this was done only for the exact
6639 names ".init_array", ".fini_array", and ".preinit_array".
6640 (default_elf_asm_named_section): Add comment about
6641 relationship with default_section_type_flags and SECTION_NOTYPE.
6642 (get_section): Don't consider it a type conflict if one side has
6643 SECTION_NOTYPE and the other doesn't, as long as neither has the
6644 SECTION_BSS et al used in the default_section_type_flags logic.
6646 2018-05-05 Tom de Vries <tom@codesourcery.com>
6649 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6650 (workaround_barsyncs): New function.
6651 (nvptx_reorg): Use workaround_barsyncs.
6652 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6653 (define_expand "nvptx_membar_cta"): New define_expand.
6654 (define_insn "*nvptx_membar_cta"): New insn.
6656 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6658 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6659 To improve optimization opportunities.
6660 * builtin-types.def: The new needed builtin types for the above.
6662 2018-05-04 Richard Biener <rguenther@suse.de>
6664 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6665 * gimple-ssa-store-merging.c
6666 (imm_store_chain_info::output_merged_store): Remove redundant create,
6667 release split_store vector contents on failure.
6668 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6669 scalar stmt vector on cache hit.
6671 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6673 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6675 * config.gcc (powerpc-xilinx-eabi*): Remove.
6676 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6678 (fusion_addis_mem_combo_load): Ditto.
6679 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6681 (rs6000_cpu_cpp_builtins): Ditto.
6682 * config/rs6000/rs6000-linux.c
6683 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6684 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6685 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6687 (rs6000_setup_reg_addr_masks): Ditto.
6688 (rs6000_init_hard_regno_mode_ok): Ditto.
6689 (rs6000_option_override_internal): Ditto.
6690 (legitimate_lo_sum_address_p): Ditto.
6691 (rs6000_legitimize_address): Ditto.
6692 (rs6000_legitimize_reload_address): Ditto.
6693 (rs6000_legitimate_address_p): Ditto.
6694 (abi_v4_pass_in_fpr): Ditto.
6695 (setup_incoming_varargs): Ditto.
6696 (rs6000_gimplify_va_arg): Ditto.
6697 (rs6000_split_multireg_move): Ditto.
6698 (rs6000_savres_strategy): Ditto.
6699 (rs6000_emit_prologue_components): Ditto.
6700 (rs6000_emit_epilogue_components): Ditto.
6701 (rs6000_emit_prologue): Ditto.
6702 (rs6000_emit_epilogue): Ditto.
6703 (rs6000_elf_file_end): Ditto.
6704 (rs6000_function_value): Ditto.
6705 (rs6000_libcall_value): Ditto.
6706 * config/rs6000/rs6000.h: Ditto.
6707 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6708 (TARGET_MINMAX): ... this. New.
6709 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6710 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6711 (*movsi_internal1_single): Delete.
6712 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6713 mfpu=, mxilinx-fpu): Delete.
6714 * config/rs6000/singlefp.h: Delete.
6715 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6716 * config/rs6000/t-rs6000: Ditto.
6717 * config/rs6000/t-xilinx: Delete.
6718 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6719 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6720 (VStype_simple): Delete.
6721 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6722 * config/rs6000/xfpu.h: Delete.
6723 * config/rs6000/xfpu.md: Delete.
6724 * config/rs6000/xilinx.h: Delete.
6725 * config/rs6000/xilinx.opt: Delete.
6726 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6727 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6729 2018-05-04 Tom de Vries <tom@codesourcery.com>
6732 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6735 2018-05-04 Richard Biener <rguenther@suse.de>
6738 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6739 (expand_complex_div_wide): Likewise.
6740 (expand_complex_operations_1): Likewise.
6741 (expand_complex_libcall): Preserve EH info of the original stmt.
6742 (tree_lower_complex): Handle removed blocks.
6743 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6744 on complex multiplication and division libcall builtins.
6746 2018-05-04 Richard Biener <rguenther@suse.de>
6749 * fold-const.c (negate_expr_p): Restrict negation of operand
6750 zero of a division to when we know that can happen without
6752 (fold_negate_expr_1): Likewise.
6754 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6757 * real.h (real_nextafter): Declare.
6758 * real.c (real_nextafter): New function.
6759 * fold-const-call.c (fold_const_nextafter): New function.
6760 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6761 CASE_CFN_NEXTTOWARD.
6762 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6763 even when arg1_mode is different from arg0_mode.
6765 2018-05-03 Nathan Sidwell <nathan@acm.org>
6767 * doc/extend.texi (Deprecated Features): Remove
6769 (Backwards Compatibility): Likewise.
6770 * doc/invoke.texi (C++ Language Options): Likewise.
6771 (C++ Dialect Options): Likewise.
6773 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6776 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6777 _mm512_mask_mullox_epi64): New intrinsics.
6779 2018-05-03 Tom de Vries <tom@codesourcery.com>
6782 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6783 dump files): Add offload-tree.
6785 2018-05-03 Richard Biener <rguenther@suse.de>
6787 PR tree-optimization/85615
6788 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6789 to loops not nested in BBs loop father to avoid creating multi-entry
6792 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6794 PR tree-optimization/70291
6795 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6796 arguments. Change return type to tree. Emit libcall as a new
6797 statement rather than replacing existing one when inplace_p is true.
6798 (expand_complex_multiplication_components): New function.
6799 (expand_complex_multiplication): Expand floating-point complex
6800 multiplication using the above.
6801 (expand_complex_division): Rename inner_type parameter to type.
6802 Update expand_complex_libcall call-site.
6803 (expand_complex_operations_1): Update expand_complex_multiplication
6804 and expand_complex_division call-sites.
6806 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6809 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6810 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6811 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6812 the highest significant bit of the shift count mask is clear. In
6813 check whether and[sq]i3 is needed verify that all significant bits
6814 of the shift count other than the highest are set.
6816 2018-05-02 Tom de Vries <tom@codesourcery.com>
6819 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6820 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6821 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6822 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6823 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6824 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6825 __builtin_goacc_parlevel_size.
6827 2018-05-02 Richard Biener <rguenther@suse.de>
6829 PR tree-optimization/85597
6830 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6831 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6834 2018-05-02 Tom de Vries <tom@codesourcery.com>
6837 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6838 dump files): Add ltrans-tree.
6840 2018-05-02 Tom de Vries <tom@codesourcery.com>
6843 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6844 dump files): Add wpa-ipa.
6846 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6848 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6849 powerpc*-*-linux*paired* target.
6850 * config/rs6000/750cl.h: Delete.
6851 * config/rs6000/paired.h: Delete.
6852 * config/rs6000/paired.md: Delete.
6853 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6855 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6856 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6857 comment. Remove paired float support.
6858 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6859 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6861 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6862 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6864 * config/rs6000/rs6000.c: Remove paired float support.
6865 (paired_expand_vector_init, paired_expand_vector_move,
6866 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6867 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6868 paired_expand_builtin, paired_expand_predicate_builtin,
6869 paired_init_builtins): Delete.
6870 * config/rs6000/rs6000.h: Remove paired float support.
6871 * config/rs6000/rs6000.md: Remove paired float support.
6872 (move_from_CR_ov_bit): Delete.
6873 * config/rs6000/rs6000.opt (mpaired): Delete.
6874 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6875 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6877 2018-05-02 Richard Biener <rguenther@suse.de>
6880 * gimplify.c (gimplify_save_expr): When in SSA form allow
6881 SAVE_EXPRs to compute to SSA vars.
6883 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6886 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6887 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6888 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6889 clobber operands[2], instead use a new pseudo. Formatting fixes.
6891 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6893 PR tree-optimization/85586
6894 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6895 exit early for statements in the same group if the accesses are
6898 2018-05-02 Tom de Vries <tom@codesourcery.com>
6901 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6904 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6906 PR tree-optimization/85143
6907 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6909 2018-05-01 Tom de Vries <tom@codesourcery.com>
6912 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6913 not found" error message.
6915 2018-05-01 Tom de Vries <tom@codesourcery.com>
6918 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6919 * vec.c (test_ordered_remove_if): New function.
6920 (vec_c_tests): Call test_ordered_remove_if.
6921 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6922 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6923 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6924 VEC_ORDERED_REMOVE_IF.
6926 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6928 PR tree-optimization/82665
6929 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6930 pointer subtraction where arguments come from a memchr call.
6932 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6934 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6935 --push-state --as-needed and --pop-state instead of --as-needed and
6936 --no-as-needed if ld supports it.
6937 * configure: Regenerated.
6940 * doc/install.texi2html: Replace _002d with - and _002a with * in
6941 generated html files using sed.
6943 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6946 * gcc-rich-location.c (blank_line_before_p): New function.
6947 (use_new_line): New function.
6948 (gcc_rich_location::add_fixit_insert_formatted): New function.
6949 * gcc-rich-location.h
6950 (gcc_rich_location::add_fixit_insert_formatted): New function.
6952 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6954 * selftest.c (assert_streq): Rename "expected" and "actual" to
6955 "val1" and "val2". Extend NULL-handling to cover both inputs
6956 symmetrically, while still requiring both to be non-NULL for a pass.
6957 * selftest.h (assert_streq): Rename "expected" and "actual" to
6959 (ASSERT_EQ): Likewise.
6960 (ASSERT_EQ_AT): Likewise.
6961 (ASSERT_KNOWN_EQ): Likewise.
6962 (ASSERT_KNOWN_EQ_AT): Likewise.
6963 (ASSERT_NE): Likewise.
6964 (ASSERT_MAYBE_NE): Likewise.
6965 (ASSERT_MAYBE_NE_AT): Likewise.
6966 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6967 the assertion to pass.
6968 (ASSERT_STREQ_AT): Likewise.
6970 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6972 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6973 interaction with -pie.
6975 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6977 * selftest.h: Fix alphabetization of per-source-file selftest
6980 2018-04-30 Jason Merrill <jason@redhat.com>
6982 PR c++/61982 - dead stores to destroyed objects.
6983 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6986 2018-04-30 Jason Merrill <jason@redhat.com>
6988 * tree.c (build_clobber): New.
6989 * tree.h: Declare it.
6990 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6992 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6994 * diagnostic-show-locus.c (layout::layout): Update for
6995 location_get_source_line returning a char_span.
6996 (struct char_span): Move to input.h.
6997 (struct correction): Update for fields in char_span becoming
6999 (struct source_line): Update for location_get_source_line
7000 returning a char_span.
7001 (layout::print_line): Likewise.
7002 * edit-context.c (edited_file::print_content): Likewise.
7003 (edited_file::print_diff_hunk): Likewise.
7004 (edited_file::print_run_of_changed_lines): Likewise.
7005 (edited_file::get_num_lines): Likewise.
7006 (edited_line::edited_line): Likewise.
7007 * final.c (asm_show_source): Likewise.
7008 * input.c (location_get_source_line): Convert return type
7009 from const char * to char_span, losing the final "line_len"
7011 (dump_location_info): Update for the above.
7012 (get_substring_ranges_for_loc): Likewise. Use a char_span
7013 when handling the literal within the line.
7014 (test_reading_source_line): Update for location_get_source_line
7015 returning a char_span.
7016 * input.h (class char_span): Move here from
7017 diagnostic-show-locus.c, converting from a struct to a class.
7018 Make data members private.
7019 (char_span::operator bool): New.
7020 (char_span::length): New.
7021 (char_span::get_buffer): New.
7022 (char_span::operator[]): New.
7023 (char_span::subspan): Make const.
7024 (char_span::xstrdup): New.
7025 (location_get_source_line): Convert return type from const char *
7026 to char_span, losing the final "line_size" param.
7028 2018-04-30 Jan Hubicka <jh@suse.cz>
7030 * lto-wrapper.c (ltrans_priorities): New static var.
7031 (cmp_priority): New.
7032 (run_gcc): Read priorities and if doing parallel build order
7033 the Makefile by them.
7035 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7037 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7039 2018-04-30 Richard Biener <rguenther@suse.de>
7041 * tree-cfg.c (verify_address): Remove base argument, add
7042 flag whether to check TREE_ADDRESSABLE and do that.
7043 (verify_expr): Remove.
7044 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7045 (verify_gimple_assign_single): Likewise.
7046 (verify_gimple_switch): Likewise.
7047 (verify_expr_location_1): Dereference tp once. Add (disabled)
7048 piece from verify_expr.
7049 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7051 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7053 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7055 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7057 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7058 (small_data_pattern): Likewise.
7059 (arc_rewrite_small_data): Likewise.
7060 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7061 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7062 (get_symbol_alignment): New function.
7063 (legitimate_small_data_address_p): Likewise.
7064 (legitimate_scaled_address): Update, call
7065 legitimate_small_data_address_p.
7066 (output_sdata): New static variable.
7067 (arc_print_operand): Update how we handle small data operands.
7068 (arc_print_operand_address): Likewise.
7069 (arc_legitimate_address_p): Update, use
7070 legitimate_small_data_address_p.
7071 (arc_rewrite_small_data_p): Remove.
7072 (arc_rewrite_small_data_1): Likewise.
7073 (arc_rewrite_small_data): Likewise.
7074 (small_data_pattern): Likewise.
7075 (compact_sda_memory_operand): Update to use
7076 legitimate_small_data_address_p and get_symbol_alignment.
7077 (prepare_move_operands): Don't rewite sdata pattern.
7078 (prepare_extend_operands): Remove.
7079 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7081 (zero_extendqisi2): Likewise.
7082 (zero_extendhisi2): Likewise.
7083 (extendqihi2): Likewise.
7084 (extendqisi2): Likewise.
7085 (extendhisi2): Likewise.
7089 * config/arc/constraints.md (Usd): Change it to memory constraint.
7091 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7093 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7094 as source of std instructions.
7095 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7096 allow 6-bit constants as source for store instructions.
7097 (movdi_insn): Update instruction pattern to allow 6-bit constants
7098 as source for store instructions.
7100 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7102 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7104 2018-04-30 Nathan Sidwell <nathan@acm.org>
7105 Sandra Loosemore <sandra@codesourcery.com>
7107 * dumpfile.c (dump_open): Allow '-' for stdout.
7108 * doc/invoke.texi (Developer Options): Document dump filename
7109 determination early. Document stdin/stdout selection.
7111 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7113 Microblaze Target: PIC data text relative
7115 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7116 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7118 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7119 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7120 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7121 New addressing mode for data-text relative position indepenedent code.
7122 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7123 'ADDRESS_SYMBOLIC_TXT_REL'.
7124 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7125 (microblaze_legitimate_pic_operand): Exclude function calls from
7126 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7127 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7129 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7130 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7131 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7132 for 'address + offset'.
7133 (microblaze_expand_prologue): Add new function prologue call for
7135 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7136 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7137 table in case of TARGET_PIC_DATA_TEXT_REL.
7138 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7139 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7140 Add new macros 'UNSPEC_TEXT',
7141 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7142 + exclude function calls from 'UNSPEC_PLT' in case of data text
7144 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7145 new target hook for generating address diff vector tables in case of
7147 * doc/tm.texi : Regenerate.
7148 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7149 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7150 of addr diff vector generation.
7151 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7152 target hook definition.
7153 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7154 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7155 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7156 option for data text relative.
7158 2018-04-30 Richard Biener <rguenther@suse.de>
7160 * tree-chrec.h (evolution_function_is_constant_p): Remove
7162 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7164 2018-04-30 Richard Biener <rguenther@suse.de>
7167 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7169 2018-04-30 Richard Biener <rguenther@suse.de>
7171 PR tree-optimization/28364
7172 PR tree-optimization/85275
7173 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7174 copying first exit test.
7176 2018-04-28 Mark Wielaard <mark@klomp.org>
7178 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7180 (dwarf_AT): Handle DW_AT_addr_base.
7181 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7183 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7186 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7187 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7188 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7189 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7191 2018-04-28 Richard Biener <rguenther@suse.de>
7193 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7194 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7195 to reflect use. Only add interesting stmts.
7197 2018-04-27 Martin Jambor <mjambor@suse.cz>
7200 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7201 the jump function allows for passing through aggregate values.
7203 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7205 * input.h (in_system_header_at): Convert from macro to inline
7207 (from_macro_expansion_at): Likewise.
7208 (from_macro_definition_at): Likewise.
7210 2018-04-27 Jeff Law <law@redhat.com>
7212 * config.gcc: Mark tile* targets as deprecated/obsolete.
7214 2018-04-27 Richard Biener <rguenther@suse.de>
7216 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7219 2018-04-27 Richard Biener <rguenther@suse.de>
7221 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7223 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7225 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7226 with Yd constraint. Set "preferred_for_speed" attribute from
7227 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7229 (*movdi_internal): Ditto.
7230 (movti_interunit splitters): Remove
7231 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7232 (movdi_interunit splitters): Ditto.
7233 * config/i386/constraints.md (Ye): Remove.
7234 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7236 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7239 * config/aarch64/constraints.md (Usg): Limit to 31.
7242 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7244 PR tree-optimization/85529
7245 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7246 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7247 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7248 zero extension or masking of the MSB bit.
7249 (optimize_range_tests): Add FIRST_BB argument, pass it through
7250 to optimize_range_tests_var_bound.
7251 (maybe_optimize_range_tests, reassociate_bb): Adjust
7252 optimize_range_tests callers.
7254 2018-04-26 Richard Biener <rguenther@suse.de>
7255 Jakub Jelinek <jakub@redhat.com>
7257 * cgraph.h (symbol_table): Just declare debug method here.
7258 * symtab.c (symbol_table::debug): Define.
7260 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7262 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7264 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7266 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7267 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7268 (*movdi_internal): Substitute Yi and Yj constraint with x
7269 and Ym and Yn constraint with y constraint. Update "isa"
7270 attribute and set "preferred_for_speed" attribute from
7271 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7272 (*movsi_internal): Ditto.
7273 (*movdf_internal): Ditto.
7274 (*movsf_internal): Ditto.
7275 (*zero_extendsidi2): Ditto.
7276 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7277 (sse2_loadld): Ditto.
7278 (*vec_extract<ssevecmodelower>_0): Ditto.
7279 (*vec_extractv4si_0_zext_sse4): Ditto.
7280 (vec_concatv2di): Ditto.
7281 (*vec_dup<mode>): Ditto.
7282 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7283 * config/i386/constraints.md (Yi): Remove.
7288 2018-04-26 Nathan Sidwell <nathan@acm.org>
7290 * dumpfile.c (dump_open): New.
7291 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7292 (dump_finish): Detect stdio/stderr by value not name.
7294 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7296 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7298 2018-04-26 Tom de Vries <tom@codesourcery.com>
7301 * config/nvptx/nvptx.c (verify_neutering_jumps)
7302 (verify_neutering_labels): New function
7303 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7305 2018-04-26 Tom de Vries <tom@codesourcery.com>
7308 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7309 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7312 2018-04-26 Richard Biener <rguenther@suse.de>
7313 Tom de Vries <tom@codesourcery.com>
7316 * lto-streamer-out.c (output_function): Fixup loops if required to match
7317 discovery done in the reader.
7319 2018-04-26 Richard Biener <rguenther@suse.de>
7321 PR tree-optimization/85116
7322 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7323 have a loop exit from the single latch predecessor. Remove
7324 case of header with just condition.
7325 (ch_base::copy_headers): Exclude infinite loops from any
7327 (pass_ch::execute): Record exits.
7329 2018-04-26 Richard Biener <rguenther@suse.de>
7331 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7332 prologue cost vector and pass it to vect_get_load_cost.
7333 (vect_get_peeling_costs_all_drs): Likewise.
7334 (vect_peeling_hash_get_lowest_cost): Likewise.
7335 (vect_enhance_data_refs_alignment): Likewise.
7337 2018-04-26 Richard Biener <rguenther@suse.de>
7340 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7341 checking of integer<->pointer conversions.
7342 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7343 sign-/zero-extending pointer types.
7344 (expand_omp_for_static_chunk): Likewise.
7346 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7347 Jean Lee <xiaoyur347@gmail.com>
7349 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7350 (TARGET_ASAN_SHADOW_OFFSET): Define.
7351 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7352 true for -fsanitize=address.
7354 2018-04-25 Mark Wielaard <mark@klomp.org>
7356 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7359 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7361 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7362 than "alu", remove explicit "memory" and "imm_disp" attributes.
7363 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7366 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7367 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7368 gen_lowpart_no_emit.
7370 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7373 * config/i386/i386.c (ix86_expand_builtin): Change memory
7374 operand to XI, extend p0 to Pmode.
7375 * config/i386/i386.md: Change unspec volatile and operand
7376 1 mode to XI, change operand 0 mode to P.
7378 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7380 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7381 GET_MODE_MASK before any checking.
7382 (nds32_can_use_bset_p): Likewise.
7383 (nds32_can_use_btgl_p): Likewise.
7385 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7387 * config/nds32/nds32-doubleword.md: New define_split pattern for
7388 illegal register number.
7390 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7392 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7394 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7396 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7398 2018-04-25 Richard Biener <rguenther@suse.de>
7400 * lto-streamer.h (LTO_major_version): Bump to 8.
7402 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7404 * BASE-VER: Set to 9.0.0.
7406 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7408 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7409 in __abskf2 and __powikf2.
7411 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7414 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7415 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7416 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7417 Use the above on operand 2. Reindent.
7418 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7420 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7423 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7425 * config/i386/i386.opt (mcet): Removed.
7426 * doc/install.texi: Remove -mcet documentation.
7427 * doc/invoke.texi: Likewise.
7429 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7432 * doc/install.texi: Remove -mcet from bootstrap-cet.
7434 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7437 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7438 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7442 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7443 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7444 containing a CONST_VECTOR.
7446 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7448 * doc/install.texi: Update newlib dependency for nvptx.
7450 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7453 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7454 instead of INTVAL when shifting x left.
7456 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7458 PR tree-optimization/85478
7459 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7460 vect_grouped_store_supported for single element vectors.
7462 2018-04-24 Richard Biener <rguenther@suse.de>
7465 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7466 load cost increase to the case of non-constant step.
7468 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7471 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7472 destination if any_malformed_asm.
7474 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7477 * expr.c (store_field): In the bitfield case, if the value comes from
7478 a function call and is returned in registers by means of a PARALLEL,
7479 do not change the mode of the temporary unless BLKmode and VOIDmode.
7481 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7483 PR rtl-optimization/85423
7484 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7485 dependencies to debug insns when the previous insn is non-debug.
7487 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7489 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7490 enums into a single definition.
7491 (fls): Fix predicates and printing.
7494 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7496 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7497 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7498 and short u6 immediate.
7499 (check_if_valid_sleep_operand): Remove.
7500 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7502 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7504 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7505 flag_always_save_lp condition.
7506 * config/nds32/nds32.opt (malways-save-lp): New option.
7508 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7510 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7511 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7512 * config/nds32/nds32.h
7513 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7514 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7516 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7518 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7519 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7521 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7522 Chung-Ju Wu <jasonwucj@gmail.com>
7524 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7526 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7527 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7529 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7531 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7533 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7535 * config/nds32/nds32-protos.h (nds32_data_alignment,
7536 nds32_local_alignment): Declare.
7537 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7538 nds32_local_alignment): New functions.
7539 (TARGET_CONSTANT_ALIGNMENT): Define.
7540 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7542 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7544 * config/nds32/nds32.c
7545 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7546 (TARGET_MODES_TIEABLE_P): Likewise.
7548 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7550 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7553 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7554 Chung-Ju Wu <jasonwucj@gmail.com>
7556 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7557 for unaligned access.
7558 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7560 * config/nds32/nds32-intrinsic.md: Likewise.
7561 * config/nds32/nds32_intrinsic.h: Likewise.
7562 * config/nds32/nds32.h (nds32_builtins): Likewise.
7563 * config/nds32/nds32.opt (munaligned-access): New option.
7564 * config/nds32/nds32.c (nds32_asm_file_start): Display
7565 flag_unaligned_access status.
7567 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7569 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7570 -mno-relax is present.
7571 * config/riscv/linux.h (LINK_SPEC): Ditto.
7573 2018-04-20 Martin Sebor <msebor@redhat.com>
7576 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7578 (gimple_fold_builtin_stxcpy_chk): Same.
7579 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7581 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7584 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7585 __powikf2 when long double is IEEE 128-bit.
7587 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7589 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7590 step to make sure stack always aligned.
7592 2018-04-20 Carl Love <cel@us.ibm.com>
7595 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7596 size check for arg0.
7598 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7599 Tom de Vries <tom@codesourcery.com>
7602 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7603 Emit insns for calls too.
7604 (nvptx_find_par): Always look for worker-level predecessor insn.
7605 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7607 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7608 (nvptx_process_pars): Propagate frames for calls.
7610 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7613 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7615 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7616 (ix86_handle_option): Don't handle OPT_mibt.
7617 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7619 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7621 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7622 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7623 (ix86_target_macros): Define __CET__ with flag_cf_protection
7624 for -fcf-protection.
7625 * config/i386/i386.c (isa2_opts): Remove -mibt.
7626 * config/i386/i386.h (TARGET_IBT): Removed.
7627 (TARGET_IBT_P): Likewise.
7628 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7629 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7630 * config/i386/i386.opt (mcet): Update help message.
7633 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7634 -mcet as an alias for -mshstk.
7636 2018-04-20 Richard Biener <rguenther@suse.de>
7639 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7640 complexity by forcing a single use of the multiply operand.
7642 2018-04-20 Martin Jambor <mjambor@suse.cz>
7645 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7646 recursion dependency to only apply to non-clones.
7648 2018-04-20 Martin Jambor <mjambor@suse.cz>
7651 * ipa-cp.c (create_specialized_node): Check that clones of
7652 self-recursive edges exist during IPA-CP.
7654 2018-04-19 Toon Moene <toon@moene.org>
7656 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7659 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7661 PR tree-optimization/85467
7662 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7663 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7664 VECTOR_CST element to type.
7666 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7669 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7670 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7671 (builtin_longjmp): Likewise.
7672 (save_stack_nonlocal): New pattern.
7673 (restore_stack_nonlocal): Likewise.
7675 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7678 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7679 Replace ASM_OUTPUT_LABEL with fprintf.
7681 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7684 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7685 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7686 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7687 define __IBT__ and __SHSTK__ for -fcf-protection.
7688 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7690 (ix86_trampoline_init): Likewise.
7691 (x86_output_mi_thunk): Likewise.
7692 (ix86_notrack_prefixed_insn_p): Likewise.
7693 (ix86_option_override_internal): Don't disallow -fcf-protection.
7694 * config/i386/i386.md (rdssp<mode>): Also enable for
7696 (incssp<mode>): Likewise.
7697 (nop_endbr): Likewise.
7698 * config/i386/i386.opt (mcet): Change help message to built-in
7702 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7703 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7704 enable CET built-in functions.
7706 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7708 * common/config/i386/i386-common.c
7709 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7710 OPTION_MASK_ISA_MOVDIRI_UNSET,
7711 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7712 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7713 * config.gcc (movdirintrin.h): New header.
7714 * config/i386/cpuid.h (bit_MOVDIRI,
7715 bit_MOVDIR64B): New bits.
7716 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7718 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7719 (VOID, PVOID, PCVOID)): New function types.
7720 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7721 __builtin_ia32_directstoreu_u64,
7722 __builtin_ia32_movdir64b): New builtins.
7723 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7724 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7726 (ix86_valid_target_attribute_inner_p): Ditto.
7727 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7728 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7729 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7730 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7731 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7732 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7733 (movdiri<mode>, movdir64b_<mode>): New.
7734 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7735 * config/i386/immintrin.h: Include movdirintrin.h.
7736 * config/i386/movdirintrin.h: New file.
7737 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7739 2018-04-19 Richard Biener <rguenther@suse.de>
7742 * cfg.c (clear_bb_flags): When loop state says we have
7743 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7745 2018-04-19 Richard Biener <rguenther@suse.de>
7747 PR tree-optimization/84737
7748 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7749 copying restrict info.
7750 (vect_setup_realignment): Use it.
7751 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7752 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7753 the first DR to all generated stores.
7754 (vectorizable_load): Likewise for loads.
7756 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7758 PR tree-optimization/85446
7759 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7760 the integral and pointer types to have the same precision.
7762 * doc/install.texi: Document --disable-cet being the default and
7765 2018-04-18 Martin Liska <mliska@suse.cz>
7767 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7770 2018-04-18 Martin Liska <mliska@suse.cz>
7773 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7776 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7777 arguments if they are comparable.
7779 2018-04-18 Martin Liska <mliska@suse.cz>
7782 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7785 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7788 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7791 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7792 ENDBR after calling __morestack.
7794 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7797 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7798 by using gcc_base_ver to generate a gcc_driver_version, and use
7799 it when generating GCC_DRIVER_NAME.
7800 * configure: Regenerate.
7802 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7805 * config.gcc: Obsolete powerpc*-*-*spe*.
7807 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7810 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7811 (stabstr_D): Change type of unum from unsigned int to
7812 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7815 2018-04-17 Jim Wilson <jimw@sifive.com>
7818 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7819 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7820 Set arg_pointer_offset after using pretend_args_size.
7822 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7824 PR rtl-optimization/85431
7825 * dse.c (record_store): Ignore zero width stores.
7828 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7829 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7830 __builtin_stack_restore rather than after it.
7831 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7832 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7833 argument instead of virtual_dynamic_stack_rtx.
7835 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7837 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7839 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7840 Add note to error message to explain internal mapping of overloaded
7841 built-in function name to non-overloaded built-in function name.
7842 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7845 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7848 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7849 where the inputs overlap with the output.
7851 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7853 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7854 (=v, v) alternative and explicit "memory" attribute.
7855 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7856 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7858 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7859 "sselog1" type instead of "sselog".
7860 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7861 "sselog". Remove explicit "memory" attribute.
7862 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7863 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7865 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7866 "sselog1" type instead of "sselog". Remove explicit "memory"
7868 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7869 use "sselog1" type instead of "sselog". Remove explicit "memory"
7871 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7872 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7874 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7875 "sselog1" type instead of "sselog". Remove explicit "memory"
7877 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7878 use "sselog1" type instead of "sselog". Remove explicit "memory"
7882 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7885 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7888 2018-04-17 Martin Jambor <mjambor@suse.cz>
7891 * ipa-cp.c (create_specialized_node): Call
7892 expand_all_artificial_thunks if necessary.
7894 2018-04-17 Martin Liska <mliska@suse.cz>
7897 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7898 in message, remote space in between '_G' and '('.
7900 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7903 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7904 avx512f_vmcmp<mode>3<round_saeonly_name>,
7905 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7906 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7907 avx512f_rndscale<mode><round_saeonly_name>,
7908 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7909 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7910 Use %<iptr>2 instead of %2 for -masm=intel.
7911 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7912 avx512f_vcvttss2usi<round_saeonly_name>,
7913 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7915 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7916 avx512f_vcvttsd2usi<round_saeonly_name>,
7917 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7918 Use %q1 instead of %1 for -masm=intel.
7919 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7920 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7921 of %3 for -masm=intel.
7922 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7924 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7926 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7928 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7929 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7931 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7933 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7934 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7935 %g1 and one with %0 and %1.
7936 (avx512er_vmrcp28<mode><round_saeonly_name>,
7937 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7939 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7940 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7941 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7942 of %0 and %{%4%} for -masm=intel.
7943 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7944 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7945 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7946 order of %0 and %{%5%}%{z%} for -masm=intel.
7948 2018-04-17 Jan Hubicka <jh@suse.cz>
7951 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7953 2018-04-17 Martin Liska <mliska@suse.cz>
7956 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7957 for target_clone error message. Make default implementation
7958 clone to be a local declaration.
7959 (separate_attrs): Add new argument and check for an empty
7961 (expand_target_clones): Handle it.
7962 (ipa_target_clone): Make redirection just for target_clones
7965 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7966 Tom de Vries <tom@codesourcery.com>
7969 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7970 tiled basic blocks without omp continue statements.
7972 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7975 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7976 vec_extract expression as having side effects to make sure it gets
7979 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7982 * config/i386/i386.c (get_builtin_code_for_version): Check
7985 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7988 * gcc/config.gcc: Support "skylake".
7989 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7991 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7992 (processor_target_table): Add "skylake".
7993 (ix86_option_override_internal): Add "skylake".
7994 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7995 PROCESSOR_CANNONLAKE.
7996 (get_builtin_code_for_version): Fix priority for
7997 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7998 PROCESSOR_SKYLAKE-AVX512.
7999 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8000 (processor_type): Add PROCESSOR_SKYLAKE.
8002 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8003 Jason Merrill <jason@redhat.com>
8006 * convert.c (convert_to_integer_1): Use direct recursion for
8007 enumeral types and types with a precision less than the number
8008 of bits in their mode.
8010 2018-04-16 Julia Koval <julia.koval@intel.com>
8013 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8014 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8016 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8019 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8020 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8021 and -mno-direct-move.
8023 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8026 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8027 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8028 Ensure negative shifts result in {0}.
8030 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8032 PR rtl-optimization/79916
8033 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8034 regs (if any) to define how to gnerate SD moves when LRA is in
8037 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8039 PR rtl-optimization/85393
8040 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8041 * except.c (expand_dw2_landing_pad_for_region): Make static.
8042 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8043 a label and unconditional jump to old_bb, rather than
8044 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8047 PR rtl-optimization/85376
8048 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8049 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8050 instead of a specific value.
8052 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8053 Bin Cheng <bin.cheng@arm.com>
8055 PR tree-optimization/82965
8056 PR tree-optimization/83991
8057 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8058 by_profile_only parameter.
8059 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8060 information if the loop was predicted to iterate too many times.
8061 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8063 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8066 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8069 2018-04-13 Martin Liska <mliska@suse.cz>
8070 Jakub Jelinek <jakub@redhat.com>
8073 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8074 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8075 * builtins.c (expand_builtin_memory_copy_args): Use
8076 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8077 handle dest_addr == pc_rtx.
8079 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8082 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8083 asked to not generate direct moves.
8084 (fix_trunc<mode>si2_stfiwx): Similar.
8085 (fix_trunc<mode>si2_internal): Similar.
8087 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8090 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8091 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8092 lookup if dest in some wider mode is known to be const0_rtx and
8093 if so, record permanent equivalence for it to be ZERO_EXTEND of
8094 the narrower mode destination.
8096 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8098 * lto-streamer-out.c (output_function): Revert 259346.
8099 * omp-expand.c (expand_oacc_for): Likewise.
8101 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8103 PR rtl-optimization/85354
8104 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8105 * sel-sched.c (sel_global_init): ... here.
8107 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8110 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8111 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8112 mode for PE-COFF targets.
8113 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8114 (i386_pe_asm_lto_end): Likewise.
8115 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8116 (TARGET_ASM_LTO_END): Likewise.
8117 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8118 (i386_pe_asm_lto_start): New function.
8119 (i386_pe_asm_lto_end): Likewise.
8121 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8122 Richard Biener <rguenther@suse.de>
8125 * lto-streamer-out.c (output_function): Fix CFG loop state before
8127 * omp-expand.c (expand_oacc_for): Handle calls to internal
8128 functions like regular functions.
8130 2018-04-12 Richard Biener <rguenther@suse.de>
8133 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8134 for the early LTO debug to properly generate references to it
8135 during DIE emission. Do not re-use that for the skeleton for
8137 (dwarf2out_early_finish): Likewise.
8139 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8142 * config/i386/sse.md
8143 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8144 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8145 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8146 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8147 and output is a reg, avoid creating invalid lowpart subreg, but
8148 instead split into a 512-bit move. Don't split if not AVX512VL,
8149 input is xmm16+ reg and output is a mem.
8150 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8151 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8152 xmm16+ reg and output is a mem.
8154 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8156 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8157 also for flag_dwarf2_cfi_asm.
8159 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8161 PR rtl-optimization/85342
8162 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8163 a bool scalar var inside of the loop instead. Don't try to update
8164 recog_data.operand after failed apply_change_group.
8166 2018-04-12 Tom de Vries <tom@codesourcery.com>
8169 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8170 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8171 array with flexible array member as array without given dimension.
8172 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8173 argument for undefined param to true.
8175 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8178 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8179 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8180 from PowerPC section.
8181 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8182 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8183 help text that is too long.
8184 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8185 help text that is too long.
8186 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8187 help text that is too long.
8189 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8191 * config/alpha/alpha.md (stack_probe_internal): Rename
8192 from "probe_stack". Update all callers.
8194 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8196 PR rtl-optimization/84566
8197 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8198 sched_macro_fuse_insns.
8200 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8203 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8204 (compute_block_dependences): ... from here.
8206 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8208 PR tree-optimization/85331
8209 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8210 from int to HOST_WIDE_INT.
8212 2018-04-11 Martin Jambor <mjambor@suse.cz>
8215 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8216 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8217 not the same as the source val.
8218 (cgraph_edge_brings_value_p): New parameter.
8219 (gather_edges_for_value): Pass destination value to
8220 cgraph_edge_brings_value_p.
8221 (perhaps_add_new_callers): Likewise.
8222 (get_info_about_necessary_edges): Likewise and exclude values brought
8223 only by self-recursive edges.
8224 (create_specialized_node): Redirect only clones of self-calling edges.
8225 (+self_recursive_pass_through_p): New function.
8226 (find_more_scalar_values_for_callers_subset): Use it.
8227 (find_aggregate_values_for_callers_subset): Likewise.
8228 (known_aggs_to_agg_replacement_list): Removed.
8229 (decide_whether_version_node): Re-calculate known constants for all
8230 remaining context clones.
8232 2018-04-11 Richard Biener <rguenther@suse.de>
8235 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8236 from early DWARF output.
8237 (dwarf2out_early_finish): Output line info unconditionally into
8238 early DWARF and add reference to it.
8240 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8243 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8244 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8245 other than V2DFmode using iptr mode attribute.
8246 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8248 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8250 PR rtl-optimization/84659
8251 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8253 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8256 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8258 (output_loc_list): Pass address of a dummy size variable even in the
8259 locview handling loop.
8260 (index_location_lists): Add comment on why skip_loc_list_entry can't
8263 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8266 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8269 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8272 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8273 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8274 and -mstring-compare-inline-limit.
8276 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8279 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8280 for stack clash protection in a register whenever we need it to be in
8283 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8285 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8286 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8288 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8291 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8293 (mlong-double-): Ditto.
8294 * config/rs6000/sysv4.opt (msdata=): Ditto.
8295 (mtls-size=): Ditto.
8297 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8299 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8300 erroneous entries for
8301 "vector int vec_ldl (int, long int *)", and
8302 "vector unsigned int vec_ldl (int, unsigned long int *)".
8303 Add comments and entries for
8304 "vector bool char vec_ldl (int, bool char *)",
8305 "vector bool short vec_ldl (int, bool short *)",
8306 "vector bool int vec_ldl (int, bool int *)",
8307 "vector bool long long vec_ldl (int, bool long long *)",
8308 "vector pixel vec_ldl (int, pixel *)",
8309 "vector long long vec_ldl (int, long long *)",
8310 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8311 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8312 type tree bool_long_long_type_node and correct definition of
8313 bool_V2DI_type_node to make reference to this new type tree.
8314 (rs6000_mangle_type): Replace erroneous reference to
8315 bool_long_type_node with bool_long_long_type_node.
8316 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8317 comments to emphasize sign distinctions for char and int types and
8318 replace RS6000_BTI_bool_long constant with
8319 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8320 use of RS6000_BTI_pixel.
8321 (bool_long_type_node): Remove this macro definition.
8322 (bool_long_long_type_node): New macro definition
8324 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8326 PR rtl-optimization/85300
8327 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8328 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8329 simplify_unary_operation fails.
8331 2018-04-10 Martin Liska <mliska@suse.cz>
8333 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8334 cgraph_edge and ipa_ref.
8336 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8340 * config/i386/sse.md
8341 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8342 computation of the VEC_MERGE selector from mask.
8343 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8344 Fix decoding of the VEC_MERGE selector into mask.
8346 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8348 PR tree-optimization/85286
8349 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8351 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8353 * final.c (final_1): Set insn_last_address as well as
8354 insn_current_address.
8356 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8359 * explow.c (emit_stack_probe): Call validize_mem on memory location
8360 before passing it to gen_probe_stack. Create address operand and
8361 legitimize it for the probe_stack_address case.
8363 2018-04-09 Jan Hubicka <jh@suse.cz>
8366 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8367 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8368 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8370 (free_lang_data): Rebuild type inheritance graph.
8372 2018-04-09 Martin Sebor <msebor@redhat.com>
8374 * invoke.texi (-finline-small-functions): Mention other optimization
8376 (-findirect-inlining, -fpartial-inlining): Same.
8377 (-finline-functions-called-once): Same.
8378 (-freorder-blocks-and-partition): Same.
8380 2018-04-09 Jan Hubicka <jh@suse.cz>
8383 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8384 jumps; choose last target that matches the criteria (i.e.
8385 no partition changes for non-crossing jumps).
8386 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8387 support for redirecting crossing jumps to non-crossing.
8389 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8391 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8392 also for naked functions.
8394 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8396 * config/arc/arc.md (add_shift): New pattern.
8397 (add_shift2): Likewise.
8398 (sub_shift): Likewise.
8399 (sub_shift_cmp0_noout): Likewise.
8400 (compare_si_ashiftsi): Likewise.
8401 (xbfu_cmp0_noout): New combine pattern.
8402 (xbfu_cmp0"): Likewise.
8403 (movsi_set_cc_insn): Place the predicable variant first.
8404 (commutative_binary_cmp0_noout): Remove clobber.
8405 (commutative_binary_cmp0): New pattern.
8406 (noncommutative_binary_cmp0): Likewise.
8407 (noncommutative_binary_cmp0_noout): Likewise.
8408 (noncommutative_binary_comparison_result_used): Removed.
8409 (rsub_cmp0): New pattern.
8410 (rsub_cmp0_noout): Likewise.
8411 (extzvsi): Changed, keep only meaningful variants.
8412 (SQH, SEZ): New iterators.
8413 (SQH_postfix): New mode attribute.
8414 (SEZ_prefix): New code attribute.
8415 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8416 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8417 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8419 (noncommutative_operator): Check the availability of barrel
8422 2018-04-09 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/85284
8425 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8426 Only use the niter constraining form of simple_iv when the exit
8429 2018-04-09 Tom de Vries <tom@codesourcery.com>
8432 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8433 (define_expand "*memory_barrier"): New define_expand.
8434 (define_insn "memory_barrier"): New insn.
8436 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8438 PR rtl-optimization/80463
8439 PR rtl-optimization/83972
8440 PR rtl-optimization/83480
8442 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8443 correct producer for the insn.
8444 (tidy_control_flow): Fixup seqnos in case of debug insns.
8446 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8448 PR rtl-optimization/83913
8450 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8451 different sched-times when merging exprs.
8453 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8455 PR rtl-optimization/83962
8457 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8458 tidy_fallthru_edge and tidy_control_flow.
8460 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8462 PR rtl-optimization/83530
8464 * sel-sched.c (force_next_insn): New global variable.
8465 (remove_insn_for_debug): When force_next_insn is true, also leave only
8466 next insn in the ready list.
8467 (sel_sched_region): When the region wasn't scheduled, make another pass
8468 over it with force_next_insn set to 1.
8470 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8472 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8474 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8475 for interrupt control.
8476 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8477 functions for interrupt control.
8478 * config/nds32/nds32-intrinsic.md: Likewise.
8479 * config/nds32/nds32_intrinsic.h: Likewise.
8480 * config/nds32/nds32.h (nds32_builtins): Likewise.
8482 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8484 * config/nds32/nds32.c (nds32_init_machine_status,
8485 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8486 strict_aligned_p field.
8487 (nds32_expand_to_rtl_hook): New function.
8488 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8489 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8491 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8492 Chung-Ju Wu <jasonwucj@gmail.com>
8494 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8495 * config/nds32/nds32-n7.md: New file.
8496 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8497 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8499 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8500 * config/nds32/nds32.md (pipeline_model): Add n7.
8501 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8502 * config/nds32/pipelines.md: Include n7 settings.
8504 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8505 Chung-Ju Wu <jasonwucj@gmail.com>
8507 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8508 * config/nds32/nds32-e8.md: New file.
8509 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8510 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8512 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8513 * config/nds32/nds32.md (pipeline_model): Add e8.
8514 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8515 * config/nds32/pipelines.md: Include e8 settings.
8517 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8518 Chung-Ju Wu <jasonwucj@gmail.com>
8520 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8521 * config/nds32/nds32-n8.md: New file.
8522 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8523 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8525 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8526 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8527 * config/nds32/nds32.md (pipeline_model): Add n8.
8528 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8529 * config/nds32/pipelines.md: Include n8 settings.
8531 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8532 Chung-Ju Wu <jasonwucj@gmail.com>
8534 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8535 * config/nds32/nds32-n9-2r1w.md: New file.
8536 * config/nds32/nds32-n9-3r2w.md: New file.
8537 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8538 nds32_register_ports): New or modify for cpu n9.
8539 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8541 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8542 * config/nds32/nds32-utils.c: New file.
8543 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8544 TARGET_MUL_SLOW): Define.
8545 * config/nds32/nds32.md (pipeline_model): New attribute.
8546 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8547 New options that support cpu n9.
8548 * config/nds32/pipelines.md: Include n9 settings.
8549 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8551 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8553 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8554 information if necessary.
8555 (output_cond_branch_compare_zero): Likewise.
8556 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8557 (nds32_target_alignment): Refine for alignment.
8558 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8559 (FUNCTION_BOUNDARY): Modify.
8560 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8562 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8564 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8566 * config/nds32/constants.md (unspec_volatile_element): Add values for
8567 TLB operation and data prefetch.
8568 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8569 functions for TLB operation and data prefetch.
8570 * config/nds32/nds32-intrinsic.md: Likewise.
8571 * config/nds32/nds32_intrinsic.h: Likewise.
8572 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8573 (nds32_print_operand): Likewise.
8574 * config/nds32/nds32.h (nds32_builtins): Likewise.
8576 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8577 Andrew Pinski <pinsika@gcc.gnu.org>
8580 * match.pd: Use constant_boolean_node of correct type instead of
8581 boolean_true_node or boolean_false_node for simplifying
8582 pointer comparisons to zero.
8584 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8586 PR tree-optimization/80021
8587 * tree.c (verify_type_variant): Make error call in verify_variant_match
8588 translatable and remove final full stop.
8590 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8592 * config/nds32/constants.md (unspec_volatile_element): Add
8593 UNSPEC_VOLATILE_EH_RETURN.
8594 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8595 nds32_output_stack_pop): Support dwarf exception handling process.
8596 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8597 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8598 exception handling process.
8599 (nds32_compute_stack_frame): Likewise.
8600 (nds32_return_addr_rtx): Likewise.
8601 (nds32_initial_elimination_offset): Likewise.
8602 (nds32_expand_prologue): Likewise.
8603 (nds32_expand_epilogue): Likewise.
8604 (nds32_dynamic_chain_address): New function.
8605 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8607 (DYNAMIC_CHAIN_ADDRESS): Define.
8608 (EH_RETURN_DATA_REGNO): Define.
8609 (EH_RETURN_STACKADJ_RTX): Define.
8610 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8611 patterns for dwarf exception handling.
8613 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8615 * config/nds32/nds32.h: Clean up obsolete macros.
8617 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8619 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8620 Add enum values for particular instructions.
8621 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8622 particular intrinsic functions.
8623 * config/nds32/nds32-intrinsic.md: Likewise.
8624 * config/nds32/nds32_intrinsic.h: Likewise.
8625 * config/nds32/nds32.h (nds32_builtins): Likewise.
8626 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8627 (btst, ave): New patterns for particular instructions.
8629 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8631 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8632 Add enum values for atomic load/store and memory sync.
8633 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8635 * config/nds32/nds32-intrinsic.md: Likewise.
8636 * config/nds32/nds32_intrinsic.h: Likewise.
8637 * config/nds32/nds32.h (nds32_builtins): Likewise.
8639 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8641 PR tree-optimization/85257
8642 * fold-const.c (native_encode_vector): If not all elts could fit
8643 and off is -1, return 0 rather than offset.
8644 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8645 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8646 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8647 adjust buffer in native_interpret_expr call.
8649 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8651 * config/nds32/constants.md (unspec_volatile_element): Add cache
8652 control enum values.
8653 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8654 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8655 * config/nds32/nds32.c (nds32_cctl_names): New.
8656 (nds32_print_operand): Handle cache control register names.
8657 * config/nds32/nds32.h (nds32_builtins): New enum values.
8658 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8660 * config/nds32/nds32.md (type): Add mmu.
8661 * config/nds32/pipelines.md (simple_insn): Add mmu.
8663 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8665 * config/nds32/nds32.md (type): Remove call.
8666 * config/nds32/pipelines.md (simple_insn): Likewise.
8668 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8670 * config/nds32/constants.md (unspec_volatile_element): Add
8671 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8672 UNSPEC_VOLATILE_FMFCFG.
8673 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8674 description for fmfcfg and fmfcsr.
8675 (bdesc_1arg): Add fmtcsr.
8676 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8677 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8678 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8679 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8680 unspec_fmfcfg): New patterns.
8681 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8682 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8683 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8684 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8685 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8686 __nds32__fmfcfg): Define.
8688 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8690 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8691 intrinsic register names.
8692 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8693 intrinsic register enum values and macros.
8695 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8697 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8698 for load/store addressing form.
8699 (nds32_print_operand_address): Likewise.
8701 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8704 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8705 based on LABEL_REF. Remove useless assertion.
8706 (pic_address_needs_scratch): Fix formatting.
8707 (sparc_legitimize_pic_address): Minor tweaks.
8708 (sparc_delegitimize_address): Adjust assertion accordingly.
8709 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8710 into symbolic_operand.
8711 (movsi_high_pic_label_ref): Likewise.
8712 (movsi_lo_sum_pic_label_ref): Likewise.
8713 (movdi_pic_label_ref): Likewise.
8714 (movdi_high_pic_label_ref): Likewise.
8715 (movdi_lo_sum_pic_label_ref): Likewise.
8717 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8719 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8720 custom LIB_SPEC setup.
8722 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8723 Kito Cheng <kito.cheng@gmail.com>
8725 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8726 * config/riscv/freebsd.h: New.
8728 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8730 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8731 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8734 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8735 Kito Cheng <kito.cheng@gmail.com>
8737 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8738 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8739 * config/nds32/nds32-protos.h (nds32_output_call,
8740 nds32_output_return): Declare.
8741 * config/nds32/nds32.md: Refine all the call and return patterns.
8743 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8746 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8747 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8749 PR rtl-optimization/84872
8750 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8751 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8754 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8756 * expr.c (copy_blkmode_to_reg): Revert 254862.
8757 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8759 2018-04-06 Richard Biener <rguenther@suse.de>
8762 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8763 after seeing a component reference with an adjacent field. Treat
8764 refs to arrays at struct end of external decls similar to
8765 refs to unconstrained commons.
8767 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8770 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8771 look through SAVE_EXPRs with non-side-effects argument. Adjust
8773 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8776 2018-04-06 Richard Biener <rguenther@suse.de>
8779 * alias.c (find_base_term): New wrapper around find_base_term
8780 unwinding CSELIB_VAL_PTR changes.
8781 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8784 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8786 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8788 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8789 constant definitions.
8790 ("nop"): lr 0,0 -> nopr r0
8791 ("nop_lr0", "nop_lr1"): New insn definitions.
8793 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8795 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8796 NDS32_V3PUSH_AVAILABLE_P macro.
8798 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8799 Chung-Ju Wu <jasonwucj@gmail.com>
8801 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8802 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8803 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8804 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8805 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8806 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8807 * config/nds32/constraints.md: New constraints and checking for hard
8808 float configuration.
8809 * config/nds32/iterators.md: New mode iterator and attribute for hard
8810 float configuration.
8811 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8813 * config/nds32/nds32-fpu.md: New file.
8814 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8815 deal with hard float code generation.
8816 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8818 (abi_type, float_reg_number): New enum type.
8819 * config/nds32/nds32-predicates.c: New predicates for hard float.
8820 * config/nds32/nds32-protos.h: Declare functions for hard float.
8821 * config/nds32/nds32.c: Implementation for hard float configuration.
8822 * config/nds32/nds32.h: Definitions for hard float configuration.
8823 * config/nds32/nds32.md: Include hard float machine description and
8824 modify patterns for hard float configuration.
8825 * config/nds32/nds32.opt: New options for hard float configuration.
8826 * config/nds32/predicates.md: New predicates for hard float
8829 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8831 * common/config/nds32/nds32-common.c
8832 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8834 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8837 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8838 CONSTRUCTOR_ELT (ctor, ...)->value.
8840 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8843 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8845 2018-04-05 Tom de Vries <tom@codesourcery.com>
8848 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8851 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8852 Kito Cheng <kito.cheng@gmail.com>
8854 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8855 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8856 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8857 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8859 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8860 Kito Cheng <kito.cheng@gmail.com>
8862 * config/nds32/constraints.md (Ufe): New memory constraint.
8863 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8864 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8865 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8867 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8868 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8870 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8872 * config/nds32/nds32.md: Use optimize_size in the condition for
8873 alu-shift instructions.
8875 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8877 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8879 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8881 * config/nds32/nds32.md (negsi2): Refine pattern.
8883 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8884 Chung-Ju Wu <jasonwucj@gmail.com>
8886 * config/nds32/iterators.md (shift_rotate): New code iterator.
8887 (shift): New code attribute.
8888 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8889 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8890 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8891 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8892 bit-wise operations.
8893 (andsi3, *andsi3): Ditto.
8894 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8895 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8896 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8897 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8898 nds32_ior_operand, nds32_xor_operand): New predicates.
8900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8902 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8903 (addsi3, subsi3): ... this.
8905 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8907 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8909 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8911 * config/nds32/nds32.md: Adjust indention.
8913 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8915 * config/nds32/nds32.md (feature): New attribute.
8917 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8919 * config/nds32/nds32.md (subtype): New attribute.
8921 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8924 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8925 expansion to perform a bitwise AND of the argument followed by a
8926 boolean negation of the result.
8928 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8930 PR rtl-optimization/84878
8931 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8932 the basic block. Assert the use reference is not artificial and that
8933 it has an associated insn.
8935 2018-04-04 Michael Matz <matz@suse.de>
8937 * builtins.c (compute_objsize): Pass correct operand
8938 to array_at_struct_end_p.
8940 2018-04-04 Richard Biener <rguenther@suse.de>
8943 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8944 from contexts for DINFO_LEVEL_TERSE and below.
8946 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8948 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8949 resiter_operand condition.
8950 * config/nds32/nds32.md (*move<mode>): Ditto.
8952 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8953 Monk Chiang <sh.chiang04@gmail.com>
8955 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8957 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8959 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8961 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8962 Kito Cheng <kito.cheng@gmail.com>
8964 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8965 nds32_cond_code_str, output_cond_branch,
8966 output_cond_branch_compare_zero, nds32_expand_cbranch,
8967 nds32_expand_cstore, nds32_expand_movcc,
8968 nds32_output_cbranchsi4_equality_zero,
8969 nds32_output_cbranchsi4_equality_reg,
8970 nds32_output_cbranchsi4_equality_reg_or_const_int,
8971 nds32_output_cbranchsi4_greater_less_zero: New functions.
8972 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8973 nds32_expand_cstore, nds32_expand_movcc,
8974 nds32_output_cbranchsi4_equality_zero,
8975 nds32_output_cbranchsi4_equality_reg,
8976 nds32_output_cbranchsi4_equality_reg_or_const_int,
8977 nds32_output_cbranchsi4_greater_less_zero): Declare.
8978 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8979 nds32_rimm11s_operand): New predicates.
8980 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8981 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8984 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8986 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8987 * config/nds32/nds32.md: Ditto.
8988 * config/nds32/pipelines.md: Ditto.
8990 2018-04-04 Richard Biener <rguenther@suse.de>
8992 PR tree-optimization/85168
8993 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8994 propagating abnormals.
8996 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8998 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9000 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9001 Kito Cheng <kito.cheng@gmail.com>
9003 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9004 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9005 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9006 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9007 * config/nds32/nds32.md (sibcall_internal): New.
9008 (sibcall_register): Remove.
9009 (sibcall_immediate): Remove.
9010 (sibcall_value_internal): New.
9011 (sibcall_value_register): Remove.
9012 (sibcall_value_immediate): Remove.
9013 * config/nds32/predicates.md (nds32_general_register_operand): New.
9014 (nds32_call_address_operand): New.
9016 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9018 PR rtl-optimization/85167
9019 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9020 bb_defs if *split_p, instead preinitialize it to NULL.
9022 PR tree-optimization/85156
9023 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9024 evaluating the argument multiple times.
9026 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9028 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9030 (_mm_cvtpd_ps): Likewise.
9031 (_mm_cvttpd_epi32): Likewise.
9032 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9033 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9034 vector, pixel, and bool following altivec.h include.
9036 2018-04-03 Martin Sebor <msebor@redhat.com>
9038 * doc/extend.texi (Common Function Attributes): Clarify.
9039 (const attribute): Likewise.
9040 (pure attribute): Likewise.
9042 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9045 * config/i386/i386.c (ix86_expand_vector_set): Use
9046 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9048 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9050 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9051 instructions when changing rounding bits to preserve precision bits
9052 in the x87 control word.
9054 2018-04-03 Martin Liska <mliska@suse.cz>
9056 PR tree-optimization/82491
9057 * rtl.h (strip_offset_and_add): Replace += suboffset with
9058 poly_uint64 () + suboffset.
9060 2018-03-29 Martin Liska <mliska@suse.cz>
9061 Martin Jambor <mjambor@suse.cz>
9064 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9065 param_type is not an integral or pointer type.
9067 2018-04-03 Richard Biener <rguenther@suse.de>
9069 * sese.h (recompute_all_dominators): Remove.
9071 2018-04-02 Martin Sebor <msebor@redhat.com>
9073 * doc/invoke.texi (-Wrestrict): Fix typos.
9075 2018-04-02 Jim Wilson <jimw@sifive.com>
9077 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9078 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9079 (<optab>di3, <optab>si3_extend): Likewise.
9080 (<optab>si3_mask, <optab>si3_mask_1): New.
9081 (<optab>di3_mask, <optab>di3_mask_1): New.
9082 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9083 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9084 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9086 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9088 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9091 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9093 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9094 (nds32_canonicalize_comparison): New function.
9096 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9097 Kito Cheng <kito.cheng@gmail.com>
9098 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9100 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9101 * config/nds32/constants.md (unspec_volatile_element): Add
9102 UNSPEC_VOLATILE_RELAX_GROUP.
9103 * config/nds32/nds32-relax-opt.c: New file.
9104 * config/nds32/nds32-predicates.c
9105 (nds32_symbol_load_store_p): New function.
9106 * config/nds32/nds32-protos.h
9107 (nds32_symbol_load_store_p): Declare function.
9108 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9109 * config/nds32/nds32.c
9110 (nds32_register_pass): New function to register pass.
9111 (nds32_register_passes): New function to register passes.
9112 * config/nds32/nds32.md (relax_group): New pattern.
9113 * config/nds32/nds32.opt (mrelax-hint): New option.
9114 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9116 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9118 * config/nds32/t-nds32: Modify files dependency.
9120 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9122 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9123 (PROFILE_HOOK): Define its implementation.
9125 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9127 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9128 type and 32-bit size.
9130 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9133 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9134 (V_128_256): New mode iterator.
9135 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9136 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9137 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9139 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9140 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9142 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9145 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9146 NaN inputs correctly.
9148 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9151 * config/rs6000/vsx.md (??r): New mode attribute.
9152 (*vsx_mov<mode>_64bit): Use it.
9153 (*vsx_mov<mode>_32bit): Likewise.
9155 2018-03-30 Martin Sebor <msebor@redhat.com>
9157 PR tree-optimization/84818
9158 * builtins.c (check_access): Use warning_n.
9160 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9163 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9165 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9168 2018-03-30 Julia Koval <julia.koval@intel.com>
9171 * x86-tune.def (movx, partial_reg_dependency): Enable for
9174 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9177 * lra-constraints.c (process_alt_operands): Move setting
9178 this_alternative_matches below.
9180 2018-03-29 Martin Liska <mliska@suse.cz>
9183 * doc/invoke.texi: Document how LTO works with debug info.
9184 Describe auto-load support of binutils. Mention 'x86-64'
9185 as valid option value of -march option.
9187 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9189 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9192 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9193 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9194 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9197 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9200 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9201 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9202 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9203 for RS6000_BTM_POWERPC64.
9204 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9205 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9206 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9211 2018-03-28 Carl Love <cel@us.ibm.com>
9214 2017-09-27 Carl Love <cel@us.ibm.com>
9216 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9217 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9218 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9221 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9223 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9224 instead of __vector bool.
9225 (_mm_max_pu8): Likewise.
9226 (_mm_min_pi16): Likewise.
9228 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9231 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9232 (DIVWEUO): Likewise.
9234 (DIVDEUO): Likewise.
9235 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9236 DIVWEUO and DIVDEUO.
9237 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9238 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9239 (div_extend): Likewise.
9240 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9242 (__builtin_divweuo): Likewise.
9243 (__builtin_divdeo): Likewise.
9244 (__builtin_divdeuo): Likewise.
9246 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9249 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9250 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9252 PR tree-optimization/82004
9253 * gimple-match-head.c (optimize_pow_to_exp): New function.
9254 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9255 Don't fold to exp if optimize_pow_to_exp is false.
9257 2018-03-28 Martin Liska <mliska@suse.cz>
9260 * calls.c (initialize_argument_information): Fix trailing space.
9261 * common.opt: Fix typo and provide better explanation for
9262 -fsanitize-coverage option.
9263 * config/i386/i386.opt: Fix typo.
9265 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9266 Martin Liska <mliska@suse.cz>
9269 * gimplify.c (asan_poison_variable): Don't do the check for
9270 gimplify_omp_ctxp here.
9271 (gimplify_decl_expr): Do it here.
9272 (gimplify_target_expr): Likewise.
9274 2018-03-28 Martin Liska <mliska@suse.cz>
9277 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9278 chkp_type_bounds_count if MPX is not enabled.
9280 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9282 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9284 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9287 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9288 function to create the function decl for complex long double
9289 multiply and divide for -mabi=ieeelongdouble.
9290 (init_float128_ieee): Call it.
9292 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9295 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9296 -fcf-protection=branch -mibt.
9297 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9299 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9302 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9303 arm_word_relocations.
9305 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9308 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9309 extern array declarations.
9311 2018-03-27 Richard Biener <rguenther@suse.de>
9314 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9315 explicit single_use checks.
9317 2018-03-27 Richard Biener <rguenther@suse.de>
9319 PR tree-optimization/85082
9320 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9323 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9325 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9326 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9327 Turn on fasynchronous-unwind-tables and funwind-tables.
9329 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9332 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9333 (*bmi_blsr_<mode>_ccz): Ditto.
9335 2018-03-26 Tom de Vries <tom@codesourcery.com>
9337 PR tree-optimization/85063
9338 * omp-general.c (offloading_function_p): New function. Factor out
9340 * omp-offload.c (pass_omp_target_link::gate): ... here.
9341 * omp-general.h (offloading_function_p): Declare.
9342 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9343 with attribute omp declare target for offloading functions.
9345 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9347 PR tree-optimization/84005
9348 * tree-data-ref.h (get_base_for_alignment): Declare.
9349 * tree-data-ref.c (get_base_for_alignment_1): New function.
9350 (get_base_for_alignment): Likewise.
9351 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9352 get_base_for_alignment to find a suitable base object, instead
9353 of always using drb->base_address.
9355 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9358 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9359 known size by default.
9361 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9364 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9365 and non BLKmode operands.
9367 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9370 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9371 Clean up attributes.
9373 2018-03-23 Richard Biener <rguenther@suse.de>
9376 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9377 we are going to emit early debug for LTO.
9379 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9382 * function.c (match_asm_constraints_1): Don't optimize if input
9383 doesn't satisfy general_operand predicate for output's mode.
9386 * alias.c (write_dependence_p): Don't require for x_canonicalized
9387 non-VOIDmode if x has VOIDmode.
9390 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9391 just don't try to optimize it rather than assert it never happens.
9393 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9395 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9396 macro expansions for definition of ST_INTERNAL_<mode> and
9397 LD_INTERNAL_<mode> builtins.
9398 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9400 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9402 (altivec_expand_st_builtin): Likewise.
9403 (altivec_expand_builtin): Remove calls to deleted functions.
9404 (rs6000_address_for_altivec): Delete this function.
9405 * config/rs6000/vector.md: Remove expands for
9406 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9408 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9411 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9412 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9413 re-computing once computed.
9414 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9415 (arm_init_machine_status): Initialize
9416 machine->static_chain_stack_bytes.
9418 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9421 * doc/extend.texi: Add four new prototypes for vec_ld.
9422 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9423 definitions for more logical presentation.
9424 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9425 entries for V1TI variants of __builtin_altivec_ld builtin.
9426 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9427 handling of V1TI variant of LVX icode pattern.
9428 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9429 (rs6000_gimple_fold_builtin): Likewise.
9430 (altivec_init_builtins): Add code to define
9431 __builtin_altivec_lvx_v1ti function.
9433 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9436 * function.c (match_asm_constraints_1): Don't do the optimization
9437 if input isn't a REG, SUBREG, MEM or constant.
9439 2018-03-22 Tom de Vries <tom@codesourcery.com>
9441 PR tree-optimization/84956
9442 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9443 bb_has_abnormal_pred.
9445 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9448 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9449 DECL_INITIAL (decl) to decl at the end.
9450 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9453 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9455 * doc/extend.texi (__builtin_tgmath): Document when complex
9456 integer types are treated as _Complex _Float64.
9458 2018-03-21 Tom de Vries <tom@codesourcery.com>
9460 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9462 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9464 PR tree-optimization/84960
9465 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9466 if it is ENTRY block, move them into single succ of ENTRY in that case.
9468 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9470 PR tree-optimization/84811
9471 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9472 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9473 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9474 (known_subrange_p): Update accordingly. Cast each value involved
9475 in the size comparison, rather than casting the result of the
9478 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9480 PR tree-optimization/84982
9481 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9482 by flipping the least significant bit rather than all bits from
9483 bitpos to bitpos + bitsize - 1.
9485 2018-03-21 Nathan Sidwell <nathan@acm.org>
9487 * doc/extend.texi (Deprecated Features): Remove mention of
9488 long-deleted deprecations.
9490 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9493 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9494 * configure: Regenerate.
9496 2018-03-21 Tom de Vries <tom@codesourcery.com>
9498 PR tree-optimization/83126
9499 * tree-parloops.c (num_phis): New function.
9500 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9502 2018-03-21 Nathan Sidwell <nathan@acm.org>
9504 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9505 mention anon-struct/union members and trailing attributes.
9507 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9509 PR tree-optimization/84969
9510 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9511 builtin memset partitions if they set different rhs values.
9513 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9515 PR rtl-optimization/84989
9516 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9517 VEC_DUPLICATE with scalar result mode.
9519 2018-03-21 Martin Liska <mliska@suse.cz>
9522 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9523 not intended return statement.
9525 2018-03-21 Martin Liska <mliska@suse.cz>
9528 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9529 (chkp_find_bound_slots_1): Limit number of iterations.
9531 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9534 * Minor grammar fixes for x86 options.
9536 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9539 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9540 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9542 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9545 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9546 (altivec_lvx_<mode>_1op): Likewise.
9547 (altivec_stvx_<mode>_2op): Likewise.
9548 (altivec_stvx_<mode>_1op): Likewise.
9549 (altivec_lvx_<VM2:mode>): New define_expand.
9550 (altivec_stvx_<VM2:mode>): Likewise.
9551 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9552 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9553 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9554 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9555 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9556 (rs6000_gen_lvx): Likewise.
9557 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9558 (altivec_expand_stv_builtin): Likewise.
9559 (altivec_expand_builtin): Likewise.
9560 * config/rs6000/vector.md: Likewise.
9562 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9565 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9568 2018-03-20 Richard Biener <rguenther@suse.de>
9571 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9572 sign-conversions as zero, fall back to standard scalar_stmt
9575 2018-03-20 Martin Liska <mliska@suse.cz>
9578 * predict.c (rebuild_frequencies): Handle case when we have
9579 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9581 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9584 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9585 flag_section_anchors.
9586 * varasm.c (use_blocks_for_decl_p): Remove hack for
9587 dw2_force_const_mem.
9590 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9592 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9593 be created, use lowpart_subreg of operands[0] rather than operands[0]
9595 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9596 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9597 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9598 and n constraint instead of aarch64_shift_imm_di and Usd.
9599 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9600 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9602 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9605 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9606 to favor GPR over NEON registers.
9607 (<shift>di3_neon): Likewise.
9609 2018-03-20 Tom de Vries <tom@codesourcery.com>
9612 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9613 (nvptx_process_pars): Emit bar.sync asap and alap.
9615 2018-03-20 Tom de Vries <tom@codesourcery.com>
9618 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9619 seen_label if seen_label is already set.
9621 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9624 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9625 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9626 Use 1U instead of 1. Formatting fixes.
9629 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9630 instead of TREE_TYPE (s1) for the return value.
9632 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9634 PR tree-optimization/84946
9635 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9636 bitsize + bitsize in poly_uint64 rather than poly_int64.
9639 * dwarf2asm.c: Include fold-const.c.
9640 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9641 of decl rather than decl itself.
9643 PR rtl-optimization/84643
9644 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9646 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9649 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9650 calling assemble_variable.
9652 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9655 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9656 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9658 2018-03-19 Jim Wilson <jimw@sifive.com>
9661 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9662 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9663 (riscv_first_stack_step): Likewise.
9664 (riscv_option_override): Use STACK_BOUNDARY instead of
9666 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9668 (BIGGEST_ALIGNMENT): Set to 128.
9669 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9670 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9673 2018-03-19 Richard Biener <rguenther@suse.de>
9675 PR tree-optimization/84933
9676 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9677 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9679 2018-03-19 Richard Biener <rguenther@suse.de>
9681 PR tree-optimization/84859
9682 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9683 (cond_if_else_store_replacement): Perform sinking operation on
9684 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9685 Generalize what a BB with a single eligible store is.
9687 2018-03-19 Richard Biener <rguenther@suse.de>
9689 PR tree-optimization/84929
9690 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9691 chrec_is_positive against non-chrec arg.
9693 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9696 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9698 2018-03-18 Martin Liska <mliska@suse.cz>
9700 PR rtl-optimization/84635
9701 * regrename.c (build_def_use): Use matches_mode only when
9704 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9706 PR tree-optimization/84913
9707 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9708 vectorize chains of COND_EXPRs.
9710 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9712 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9714 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9716 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9718 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9720 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9722 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9723 Kito Cheng <kito.cheng@gmail.com>
9725 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9726 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9727 (nds32_adjust_reg_alloc_order): New function.
9728 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9730 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9732 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9733 nds32_print_operand, nds32_print_operand_address): Use
9734 HOST_WIDE_INT_PRINT_DEC instead.
9736 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9738 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9740 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9743 * config/i386/i386.c (initial_ix86_tune_features,
9744 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9746 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9747 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9748 rather than 1u << ix86_tune. Formatting fix.
9749 (ix86_option_override_internal): Change ix86_arch_mask from
9750 unsigned int to unsigned HOST_WIDE_INT, initialize to
9751 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9752 (ix86_function_specific_restore): Likewise.
9754 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9757 * postreload.c (reload_combine_recognize_pattern): Perform
9758 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9759 truncate_int_for_mode the result for the destination's mode.
9762 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9763 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9765 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9768 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9771 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9772 regno_allocno_class_array and sorted_pseudos.
9773 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9774 insns where regno is used.
9776 2018-03-16 Martin Liska <mliska@suse.cz>
9779 * multiple_target.c (create_dispatcher_calls): Redirect
9780 reference in the symbol table.
9782 2018-03-16 Martin Liska <mliska@suse.cz>
9785 * multiple_target.c (create_dispatcher_calls): Redirect also
9788 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9792 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9793 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9794 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9796 2018-03-16 Julia Koval <julia.koval@intel.com>
9798 * doc/invoke.texi (Skylake Server): Add CLWB.
9799 Cannonlake): Remove CLWB.
9801 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9803 PR tree-optimization/84841
9804 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9806 (FLOAT_ONE_CONST_TYPE): Define.
9807 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9808 (sort_by_operand_rank): Put entries with higher constant_type last
9809 rather than first to match comments.
9811 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9813 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9816 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9819 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9820 check and use address of target always.
9822 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9825 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9826 (indirect_thunk_bnd_needed): Likewise.
9827 (indirect_thunks_used): Likewise.
9828 (indirect_thunks_bnd_used): Likewise.
9829 (indirect_return_needed): New.
9830 (indirect_return_bnd_needed): Likewise.
9831 (output_indirect_thunk_function): Add a bool argument for
9833 (output_indirect_thunk_function): Don't generate alias for
9834 function return thunk.
9835 (ix86_code_end): Call output_indirect_thunk_function to generate
9836 function return thunks.
9837 (ix86_output_function_return): Set indirect_return_bnd_needed
9838 and indirect_return_needed instead of indirect_thunk_bnd_needed
9839 and indirect_thunk_needed.
9841 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9843 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9844 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9845 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9847 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9848 Paul Hua <paul.hua.gm@gmail.com>
9851 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9853 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9855 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9856 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9859 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9862 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9863 instead of GET_MODE_SIZE when comparing Units.
9865 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9868 * varasm.c (hash_section): Return an unchangeble hash value
9869 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9870 Return !aarch64_can_use_per_function_literal_pools_p ().
9872 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9875 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9876 rather than address of cmode as last argument to prepare_cmp_insn.
9878 2018-03-15 Julia Koval <julia.koval@intel.com>
9880 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9881 F_AVX512VNNI, F_AVX512BITALG): New.
9883 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9886 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9887 insn for floating-point loads and stores.
9889 2018-03-14 Carl Love <cel@us.ibm.com>
9891 * config/rs6000/rs6000-c.c: Add macro definitions for
9892 ALTIVEC_BUILTIN_VEC_PERMXOR.
9893 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9894 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9895 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9896 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9898 * config/doc/extend.texi: Add prototypes for vec_permxor.
9900 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9903 * diagnostic-show-locus.c (class layout_point): Convert m_line
9904 from int to linenum_type.
9905 (line_span::comparator): Use linenum "compare" function when
9906 comparing line numbers.
9907 (test_line_span): New function.
9908 (layout_range::contains_point): Convert param "row" from int to
9910 (layout_range::intersects_line_p): Likewise.
9911 (layout::will_show_line_p): Likewise.
9912 (layout::print_source_line): Likewise.
9913 (layout::should_print_annotation_line_p): Likewise.
9914 (layout::print_annotation_line): Likewise.
9915 (layout::print_leading_fixits): Likewise.
9916 (layout::annotation_line_showed_range_p): Likewise.
9917 (struct line_corrections): Likewise for field m_row.
9918 (line_corrections::line_corrections): Likewise for param "row".
9919 (layout::print_trailing_fixits): Likewise.
9920 (layout::get_state_at_point): Likewise.
9921 (layout::get_x_bound_for_row): Likewise.
9922 (layout::print_line): Likewise.
9923 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9924 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9925 * input.c (selftest::test_linenum_comparisons): New function.
9926 (selftest::input_c_tests): Call it.
9927 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9928 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9929 * selftest.h (ASSERT_GT): New macro.
9930 (ASSERT_GT_AT): New macro.
9931 (ASSERT_LT): New macro.
9932 (ASSERT_LT_AT): New macro.
9934 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9936 PR rtl-optimization/84780
9937 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9939 2018-03-14 Martin Liska <mliska@suse.cz>
9941 * tree.c (record_node_allocation_statistics): Use
9942 get_stats_node_kind.
9943 (get_stats_node_kind): New function extracted from
9944 record_node_allocation_statistics.
9945 (free_node): Use get_stats_node_kind.
9947 2018-03-14 Richard Biener <rguenther@suse.de>
9949 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9950 that the value-set of ANTIC_IN doesn't grow.
9953 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9955 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9956 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9958 2018-03-14 Julia Koval <julia.koval@intel.com>
9960 * config.gcc (icelake-client, icelake-server): New.
9962 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9963 (initial_ix86_arch_features): Ditto.
9964 (PTA_SKYLAKE): Add SGX.
9965 (PTA_ICELAKE): Remove.
9966 (PTA_ICELAKE_CLIENT): New.
9967 (PTA_ICELAKE_SERVER): New.
9968 (ix86_option_override_internal): Split up icelake on icelake client and
9970 (get_builtin_code_for_version): Ditto.
9971 (fold_builtin_cpu): Ditto.
9972 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9973 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9974 * config/i386/i386.h (processor_type): Ditto.
9975 * doc/invoke.texi: Ditto.
9977 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9980 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9981 INTEGER_CST offset, add it together with bitpos / 8 and
9982 sign extend based on POINTER_SIZE.
9986 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9989 * config/i386/constraints.md (Yc): New register constraint.
9990 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9991 Use Yc constraint for alternative 2 of operand 0. Remove
9992 preferred_for_speed attribute.
9994 2018-03-14 Richard Biener <rguenther@suse.de>
9996 PR tree-optimization/84830
9997 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9998 with the old one to avoid oscillations.
10000 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10003 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10005 (assign_by_spills): Return a flag of reload assignment failure.
10006 Do not process the reload assignment failures. Do not spill other
10007 reload pseudos if they has the same reg class. Update n if
10009 (lra_assign): Add a return arg. Set up from the result of
10010 assign_by_spills call.
10011 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10012 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10013 usage_insns if it is not NULL.
10014 (spill_hard_reg_in_range): New function.
10015 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10016 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10017 function prototypes.
10018 (lra_assign): Change prototype.
10019 * lra.c (lra): Add code to deal with fails by splitting hard reg
10022 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10024 * config/riscv/riscv.opt (mrelax): New option.
10025 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10026 "norelax" when riscv_mrelax is disabled.
10027 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10029 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10032 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10033 reassociation for int modes.
10035 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10037 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10038 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10040 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10041 * config/aarch64/aarch64-sve.md
10042 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10043 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10044 (*extend<mode><Vwide>2): Rename to...
10045 (aarch64_sve_extend<mode><Vwide>2): ...this.
10046 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10047 renaming the old pattern to...
10048 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10050 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10051 define_expand, renaming the old pattern to...
10052 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10053 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10054 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10055 account when deciding which SVE instruction the optab should use.
10056 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10058 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10060 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10061 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10062 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10063 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10064 (tlsdesc_small_<mode>): Turn a define_expand and use
10065 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10066 (tlsdesc_small_advsimd_<mode>): ...this.
10067 (tlsdesc_small_sve_<mode>): New pattern.
10069 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10071 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10072 (UNSPEC_UMUL_HIGHPART): New constants.
10073 (MUL_HIGHPART): New int iteraor.
10074 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10075 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10077 (*<su>mul<mode>3_highpart): New define_insn.
10079 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10082 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10085 2018-03-13 Martin Liska <mliska@suse.cz>
10088 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10090 (sem_item_optimizer::~sem_item_optimizer): Release it.
10091 (sem_item_optimizer::merge_classes): Register variable aliases.
10092 (sem_item_optimizer::fixup_pt_set): New function.
10093 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10094 * ipa-icf.h: Declare new variables and functions.
10096 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10098 PR middle-end/84834
10099 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10100 integer_pow2p@2 and test integer_pow2p in condition.
10101 (A < 0 ? C : 0): Similarly for @1.
10103 PR middle-end/84831
10104 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10105 characters starting at p contain '\0' character, don't look beyond
10109 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10110 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10113 * reg-stack.c (change_stack): Change update_end var from int to
10114 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10115 also call set_block_for_insn on the newly added insns and rescan.
10118 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10119 on the last operand.
10122 * tree.c (stabilize_reference_1): Return save_expr (e) for
10123 STATEMENT_LIST even if it doesn't have side-effects.
10125 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10127 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10129 2018-03-12 Renlin Li <renlin.li@arm.com>
10131 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10132 aarch64_output_scalar_simd_mov_immediate.
10134 2018-03-12 Martin Sebor <msebor@redhat.com>
10136 PR tree-optimization/83456
10137 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10138 for perfectly overlapping calls to memcpy.
10139 (gimple_fold_builtin_memory_chk): Same.
10140 (gimple_fold_builtin_strcpy): Handle no-warning.
10141 (gimple_fold_builtin_stxcpy_chk): Same.
10142 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10144 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10146 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10147 parameter. Use it for SFmode.
10148 (rs6000_function_arg_advance_1): Adjust.
10149 (rs6000_function_arg): Adjust.
10150 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10152 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10154 PR rtl-optimization/84169
10155 PR rtl-optimization/84780
10156 * combine.c (can_combine_p): Check for a 2-insn combination whether
10157 the destination register is used between the two insns, too.
10159 2018-03-12 Richard Biener <rguenther@suse.de>
10161 PR tree-optimization/84803
10162 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10163 for refs DR analysis didn't process.
10165 2018-03-12 Richard Biener <rguenther@suse.de>
10167 PR tree-optimization/84777
10168 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10169 force-vectorize loops ignore whether we are optimizing for size.
10171 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10173 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10174 (TARGET_MD_ASM_ADJUST): Define.
10176 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10177 Kito Cheng <kito.cheng@gmail.com>
10178 Chung-Ju Wu <jasonwucj@gmail.com>
10180 * config/nds32/nds32.c (nds32_compute_stack_frame,
10181 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10182 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10183 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10184 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10185 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10186 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10187 * config/nds32/nds32.md (prologue, epilogue): Use macro
10188 NDS32_V3PUSH_AVAILABLE_P to do checking.
10190 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10193 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10194 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10195 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10196 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10197 addition of most attributes on !orig_type_die or the attribute not
10198 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10200 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10201 Chung-Ju Wu <jasonwucj@gmail.com>
10203 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10204 __NDS32_VH__ macro.
10205 * config/nds32/nds32.opt (mvh): New option.
10207 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10208 Chung-Ju Wu <jasonwucj@gmail.com>
10210 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10212 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10213 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10216 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10217 Chung-Ju Wu <jasonwucj@gmail.com>
10219 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10221 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10222 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10224 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10225 Kito Cheng <kito.cheng@gmail.com>
10226 Chung-Ju Wu <jasonwucj@gmail.com>
10228 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10229 UNSPEC_FFMISM and UNSPEC_FLMISM.
10230 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10231 for ffb, ffmism and flmism.
10232 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10233 (unspec_ffmism): Ditto.
10234 (unspec_flmism): Ditto.
10235 (nds32_expand_builtin_impl): Check if string extension is available.
10236 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10237 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10239 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10242 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10245 * lra-assigns.c (assign_by_spills): Return a flag of reload
10246 assignment failure. Do not process the reload assignment
10247 failures. Do not spill other reload pseudos if they has the same
10249 (lra_assign): Add a return arg. Set up from the result of
10250 assign_by_spills call.
10251 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10252 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10253 usage_insns if it is not NULL.
10254 (spill_hard_reg_in_range): New function.
10255 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10256 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10257 function prototypes.
10258 (lra_assign): Change prototype.
10259 * lra.c (lra): Add code to deal with fails by splitting hard reg
10262 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10265 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10267 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10270 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10271 (dw_val_node): Add val_symbolic_view.
10272 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10273 (symview_upper_bound): New.
10274 (new_line_info_table): Initialize symviews_since_reset.
10275 (dwarf2out_source_line): Count symviews_since_reset and set
10276 symview_upper_bound.
10277 (dw_val_equal_p): Handle symview.
10278 (add_AT_symview): New.
10279 (print_dw_val): Handle symview.
10280 (attr_checksum, attr_checksum_ordered): Likewise.
10281 (same_dw_val_p, size_of_die): Likewise.
10282 (value_format, output_die): Likewise.
10283 (add_high_low_attributes): Use add_AT_symview for entry_view.
10284 (dwarf2out_finish): Reset symview_upper_bound, clear
10287 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10290 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10291 Add strict argument and use it.
10292 (rs6000_split_multireg_move): Update for new strict argument.
10293 (mem_operand_gpr): Disallow all non-offsettable addresses.
10294 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10296 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10299 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10300 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10301 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10304 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10305 decl, use remap_type if we want to use the type.
10307 2018-03-09 Martin Sebor <msebor@redhat.com>
10309 PR tree-optimization/84526
10310 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10312 (builtin_access::generic_overlap): Be prepared to handle non-array
10315 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10317 PR rtl-optimization/84682
10318 * lra-constraints.c (process_address_1): Check is_address flag
10319 for address constraints.
10320 (process_alt_operands): Likewise.
10321 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10322 preprocess_constraints.
10323 * recog.h (preprocess_constraints): Add oploc parameter.
10325 * recog.c (preprocess_constraints): Test address_operand for
10326 CT_ADDRESS constraints.
10328 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10331 * lra-assigns.c (assign_by_spills): Return a flag of reload
10332 assignment failure. Do not process the reload assignment
10333 failures. Do not spill other reload pseudos if they has the same
10335 (lra_assign): Add a return arg. Set up from the result of
10336 assign_by_spills call.
10337 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10338 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10339 usage_insns if it is not NULL.
10340 (spill_hard_reg_in_range): New function.
10341 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10342 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10343 function prototypes.
10344 (lra_assign): Change prototype.
10345 * lra.c (lra): Add code to deal with fails by splitting hard reg
10348 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10351 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10352 Accept complain bool parameter. Only emit errors if it is true.
10353 (arm_parse_cpu_option_name): Likewise.
10354 (arm_target_thumb_only): Adjust callers of the above.
10355 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10356 prototype to take a default true bool parameter.
10357 (arm_parse_arch_option_name): Likewise.
10359 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10360 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10364 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10365 * configure: Regenerate.
10366 * configure.ac ("linker --version-script option"): New.
10367 ("linker soname option"): New.
10369 2018-03-09 Richard Biener <rguenther@suse.de>
10371 PR tree-optimization/84775
10372 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10373 immediate uses of predicate stmts and mark them modified.
10376 PR tree-optimization/84178
10377 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10379 (version_loop_for_if_conversion): Delay update_ssa call.
10380 (tree_if_conversion): Delay update_ssa until after predicate
10383 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10386 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10387 when the function accesses prior frames.
10389 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10392 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10393 gen_llsym, otherwise call maybe_gen_llsym.
10395 PR inline-asm/84742
10396 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10397 has ',' character inside of it.
10399 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10402 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10403 as clobbering CC_REGNUM.
10405 2018-03-08 Richard Biener <rguenther@suse.de>
10407 PR middle-end/84552
10408 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10409 (follow_copies_to_constant): Do not follow SSA names registered
10412 2018-03-08 Richard Biener <rguenther@suse.de>
10414 PR tree-optimization/84178
10415 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10417 (version_loop_for_if_conversion): Delay update_ssa call.
10418 (tree_if_conversion): Delay update_ssa until after predicate
10421 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10423 PR tree-optimization/84178
10424 * tree-if-conv.c (release_bb_predicate): Remove the
10425 the assertion that the stmts have NULL use_ops.
10426 Discard the statements, asserting that they haven't
10427 yet been added to a BB.
10429 2018-03-08 Richard Biener <rguenther@suse.de>
10431 PR tree-optimization/84746
10432 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10433 (phi_translate): Pass in destination ANTIC_OUT set.
10434 (phi_translate_1): Likewise. For a simplified result lookup
10435 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10436 (phi_translate_set): Adjust.
10437 (do_pre_regular_insertion): Likewise.
10438 (do_pre_partial_partial_insertion): Likewise.
10440 2018-03-08 Martin Liska <mliska@suse.cz>
10442 PR gcov-profile/84735
10443 * doc/gcov.texi: Document usage of profile files.
10444 * gcov-io.h: Document changes in the format.
10446 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10450 * dwarf2out.c (struct dw_line_info_table): Update comments for
10452 (FORCE_RESET_NEXT_VIEW): New.
10453 (FORCE_RESETTING_VIEW_P): New.
10454 (RESETTING_VIEW_P): Check for -1 too.
10455 (ZERO_VIEW_P): Likewise.
10456 (new_line_info_table): Force-reset next view.
10457 (dwarf2out_begin_function): Likewise.
10458 (dwarf2out_source_line): Simplify zero_view_p initialization.
10459 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10460 view directly. Omit view when omitting .loc at line 0.
10462 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10464 PR tree-optimization/84740
10465 * tree-switch-conversion.c (process_switch): Call build_constructors
10466 only if info.phi_count is non-zero.
10468 PR tree-optimization/84739
10469 * tree-tailcall.c (find_tail_calls): Check call arguments against
10470 DECL_ARGUMENTS (current_function_decl) rather than
10471 DECL_ARGUMENTS (func) when checking for tail recursion.
10473 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10475 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10476 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10477 Volker Reichelt's entry and add entries for people that perform
10478 GCC fuzzy testing and report numerous bugs.
10480 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10483 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10484 readonly data in sdata, if that is disabled.
10485 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10486 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10487 -mreadonly-in-sdata option.
10489 2018-03-07 Martin Sebor <msebor@redhat.com>
10491 PR tree-optimization/84468
10492 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10493 basic block when looking for nul assignment.
10495 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10498 * except.h (output_function_exception_table): Adjust prototype.
10499 * except.c (output_function_exception_table): Remove FNNAME parameter
10500 and add SECTION parameter. Ouput one part of the table at a time.
10501 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10502 the first part of the exception table and emit unwind directives.
10503 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10504 (i386_pe_seh_cold_init): Likewise.
10505 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10506 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10507 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10508 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10509 * config/i386/winnt.c: Include except.h.
10510 (struct seh_frame_state): Add reg_offset, after_prologue and
10511 in_cold_section fields.
10512 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10513 (i386_pe_seh_cold_init): New function.
10514 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10515 to seh->in_cold_section.
10516 (seh_emit_push): Record the offset of the push.
10517 (seh_emit_save): Record the offet of the save.
10518 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10519 Test seh->after_prologue to disregard the epilogue.
10520 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10521 (i386_pe_end_cold_function): New function.
10523 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10526 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10527 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10530 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10532 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10533 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10535 PR middle-end/84723
10536 * multiple_target.c: Include tree-inline.h and intl.h.
10537 (expand_target_clones): Diagnose and fail if node->definition and
10538 !tree_versionable_function_p (node->decl).
10540 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10542 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10544 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10545 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10546 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10548 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10551 * combine.c (try_combine): Use reg_or_subregno instead of handling
10552 just paradoxical SUBREGs and REGs.
10554 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10556 * config/arc/arc.c (arc_finalize_pic): Remove function.
10557 (arc_must_save_register): We use single base PIC register, remove
10558 checks to save/restore the PIC register.
10559 (arc_expand_prologue): Likewise.
10560 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10562 (arc_verify_short): Likewise.
10563 (arc_attr_type): Likewise.
10564 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10565 (walk_stores): Likewise.
10566 (arc_address_cost): Make it static.
10567 (arc_verify_short): Likewise.
10568 (branch_dest): Likewise.
10569 (arc_attr_type): Likewise.
10570 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10571 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10572 (arc_final_prescan_insn): Remove inserting the nops due to
10573 hardware hazards. It is done in reorg step.
10574 (insn_length_variant_t): Remove.
10575 (insn_length_parameters_t): Likewise.
10576 (arc_insn_length_parameters): Likewise.
10577 (arc_get_insn_variants): Likewise.
10578 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10580 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10582 PR inline-asm/84683
10583 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10586 PR tree-optimization/84687
10587 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10589 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10591 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10593 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10594 Rename to ppc_speculation_barrier.
10595 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10596 __builtin_ppc_speculation_barrier.
10598 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10601 * combine.c (combine_simplify_rtx): Don't try to simplify if
10602 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10605 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10607 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10608 to 32 bytes when compiling for POWER9.
10610 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10613 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10614 regparm >= 3 with no arg reg available also for calls with
10615 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10618 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10620 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10622 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10625 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10627 2018-03-05 Richard Biener <rguenther@suse.de>
10629 PR tree-optimization/84486
10630 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10631 When inserting a __builtin_assume_aligned call set the LHS
10632 SSA name alignment info accordingly.
10634 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10636 PR tree-optimization/84114
10637 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10638 Avoid reassociation of FLOAT_MODE addition.
10640 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10642 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10643 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10644 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10645 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10646 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10647 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10648 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10650 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10651 __builtin_ia32_wbinvd): New builtins.
10652 (SPECIAL_ARGS2): New.
10653 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10654 (SPECIAL_ARGS2): New.
10655 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10656 (ix86_valid_target_attribute_inner_p): Ditto.
10657 (ix86_init_mmx_sse_builtins): Add special_args2.
10658 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10659 TARGET_WBNOINVD_P): New.
10660 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10661 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10662 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10663 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10664 * config/i386/pconfigintrin.h: New file.
10665 * config/i386/wbnoinvdintrin.h: Ditto.
10666 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10668 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10670 2018-03-05 Richard Biener <rguenther@suse.de>
10672 PR tree-optimization/84670
10673 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10675 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10676 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10677 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10678 doesn't grow if all successors (recursively) were visited at least
10681 2018-03-05 Richard Biener <rguenther@suse.de>
10683 PR tree-optimization/84650
10684 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10685 if executed in the loop pipeline.
10687 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10689 * doc/configfiles.texi (Configuration Files): Move info about
10690 conditionalizing $target-protos.h to...
10691 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10692 differs from $target-protos.h.
10694 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10695 Chung-Ju Wu <jasonwucj@gmail.com>
10697 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10698 * config/nds32/nds32-multiple.md (setmemsi): Define.
10699 * config/nds32/nds32-memory-manipulation.c
10700 (nds32_gen_dup_4_byte_to_word_value): New.
10701 (emit_setmem_word_loop): New.
10702 (emit_setmem_byte_loop): New.
10703 (nds32_expand_setmem_loop): New.
10704 (nds32_expand_setmem_loop_v3m): New.
10705 (nds32_expand_setmem_unroll): New.
10706 (nds32_expand_setmem): New.
10708 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10709 Chung-Ju Wu <jasonwucj@gmail.com>
10711 * config/nds32/nds32-memory-manipulation.c
10712 (nds32_emit_load_store): New.
10713 (nds32_emit_post_inc_load_store): New.
10714 (nds32_emit_mem_move): New.
10715 (nds32_emit_mem_move_block): New.
10716 (nds32_expand_movmemsi_loop_unknown_size): New.
10717 (nds32_expand_movmemsi_loop_known_size): New.
10718 (nds32_expand_movmemsi_loop): New.
10719 (nds32_expand_movmemsi_unroll): New.
10720 (nds32_expand_movmemqi): Rename ...
10721 (nds32_expand_movmemsi): ... to this.
10722 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10723 (movmemsi): ... to this.
10724 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10725 (nds32_expand_movmemsi): ... to this.
10727 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10728 Monk Chiang <sh.chiang04@gmail.com>
10729 Chung-Ju Wu <jasonwucj@gmail.com>
10731 * config/nds32/nds32-protos.h
10732 (nds32_expand_load_multiple): New arguments.
10733 (nds32_expand_store_multiple): Ditto.
10734 (nds32_valid_multiple_load_store): Rename ...
10735 (nds32_valid_multiple_load_store_p): ... to this.
10736 * config/nds32/nds32-memory-manipulation.c
10737 (nds32_expand_load_multiple): Refine implementation.
10738 (nds32_expand_store_multiple): Ditto.
10739 * config/nds32/nds32-multiple.md
10740 (load_multiple): Update nds32_expand_load_multiple interface.
10741 (store_multiple): Update nds32_expand_store_multiple interface.
10742 * config/nds32/nds32-predicates.c
10743 (nds32_valid_multiple_load_store): Rename ...
10744 (nds32_valid_multiple_load_store_p): ... to this and refine
10746 * config/nds32/predicates.md
10747 (nds32_load_multiple_and_update_address_operation): New predicate.
10748 (nds32_store_multiple_and_update_address_operation): New predicate.
10750 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10751 Chung-Ju Wu <jasonwucj@gmail.com>
10753 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10754 (combo): New attribute.
10755 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10757 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10759 * config/nds32/nds32.opt: Change -mcmodel= default value.
10761 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10762 Monk Chiang <sh.chiang04@gmail.com>
10763 Chung-Ju Wu <jasonwucj@gmail.com>
10765 * config/nds32/constants.md (unspec_element): New enum.
10766 * config/nds32/constraints.md (Umw): New constraint.
10767 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10768 * config/nds32/nds32-intrinsic.md: Likewise.
10769 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10770 (nds32_valid_smw_lwm_base_p): New.
10771 (nds32_output_smw_single_word): New.
10772 (nds32_output_lmw_single_word): New.
10773 (nds32_expand_unaligned_load): New.
10774 (nds32_expand_unaligned_store): New.
10775 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10776 (nds32_output_smw_single_word): Declare.
10777 (nds32_output_lmw_single_word): Declare.
10778 (nds32_expand_unaligned_load): Declare.
10779 (nds32_expand_unaligned_store): Declare.
10780 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10781 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10782 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10783 NDS32_BUILTIN_UASTORE_DW.
10784 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10787 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10788 Kito Cheng <kito.cheng@gmail.com>
10789 Chung-Ju Wu <jasonwucj@gmail.com>
10791 * config/nds32/nds32-intrinsic.c
10792 (nds32_expand_builtin_null_ftype_reg): Delete.
10793 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10794 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10795 (nds32_read_argument): New.
10796 (nds32_legitimize_target): Ditto.
10797 (nds32_legitimize_argument): Ditto.
10798 (nds32_check_constant_argument): Ditto.
10799 (nds32_expand_unop_builtin): Ditto.
10800 (nds32_expand_unopimm_builtin): Ditto.
10801 (nds32_expand_binop_builtin): Ditto.
10802 (nds32_builtin_decl_impl): Ditto.
10803 (builtin_description): Ditto.
10804 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10805 (nds32_init_builtins_impl): Ditto.
10806 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10807 (nds32_builtin_decl): New.
10808 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10809 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10811 2018-03-02 Jeff Law <law@redhat.com>
10813 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10814 (redundant_insn, fill_simple_delay_slots): Likewise.
10815 (fill_slots_from_thread): Likewise.
10816 * resource.c (mark_referenced_resources): Likewise.
10817 (mark_set_resources, find_dead_or_set_registers): Likewise.
10819 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10821 * substring-locations.h (format_warning_va): Formatting fix for
10822 ATTRIBUTE_GCC_DIAG.
10823 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10825 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10826 * substring-locations.c: Include intl.h.
10827 (format_warning_va): Turned into small wrapper around
10828 format_warning_n_va, renamed to ...
10829 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10830 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10832 (format_warning_at_substring_n): New function.
10833 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10834 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10835 format_warning_at_substring with just a shorter name instead of
10836 const function pointer.
10837 (fmtwarn_n): New function.
10838 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10839 appropriate, get rid of all the fmtstr temporaries, move conditionals
10840 with G_() wrapped string literals directly into fmtwarn arguments,
10841 cast dir.len to (int), formatting fixes.
10843 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10845 * doc/invoke.texi: Remove "Cilk Plus" references.
10847 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10848 Richard Biener <rguenther@suse.de>
10851 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10852 for error or warning attributes if CALL_FROM_THUNK_P is set.
10855 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10858 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10859 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10862 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10863 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10865 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10866 instead of -1U in last predictors element's probability member.
10868 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10871 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10872 arguments if they are comparable.
10874 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10876 PR tree-optimization/84634
10877 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10878 masks and masked_loop_p with a single loop_masks, making sure it's
10879 null for bb vectorization.
10881 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10883 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10884 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10885 loop->force_vectorize to check whether there is no alias.
10887 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10890 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10892 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10894 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10895 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10896 instead of a loop around prev_real_insn.
10897 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10900 PR inline-asm/84625
10901 * config/i386/i386.c (ix86_print_operand): Use conditional
10902 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10905 2018-03-02 Richard Biener <rguenther@suse.de>
10907 PR tree-optimization/84427
10908 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10909 (bitmap_set_subtract_values): Rewrite to handle multiple
10912 (prune_clobbered_mems): Likewise.
10913 (phi_translate): Take edge instead of pred/phiblock.
10914 (phi_translate_1): Likewise.
10915 (phi_translate_set): Likewise. Insert all translated
10916 exprs for a value into the set, keeping possibly multiple
10917 expressions per value.
10918 (compute_antic_aux): Adjust for phi_translate changes.
10919 When intersecting union the expressions and prune those
10920 not in the final value set, keeping possibly multiple
10921 expressions per value. Do not use value-insertion
10922 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10923 all expressions. Add verification that the value-sets
10924 only shrink during iteration.
10925 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10926 (do_pre_regular_insertion): Likewise.
10927 (do_pre_partial_partial_insertion): Likewise.
10929 2018-03-02 Richard Biener <rguenther@suse.de>
10932 * config/darwin.c (saved_debug_info_level): New static global.
10933 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10934 (darwin_asm_lto_end): Restore debug info generation settings.
10936 2018-03-01 Martin Liska <mliska@suse.cz>
10939 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10940 volatile arguments.
10942 2018-03-01 Richard Biener <rguenther@suse.de>
10945 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10946 type annotation with LTO when debug was disabled at compile-time.
10948 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10950 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10952 (mips_final_postscan_insn): Likewise.
10954 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10956 PR rtl-optimization/84528
10957 * alias.c (init_alias_target): Add commentary.
10958 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10959 a unique base value if the frame pointer is not eliminated
10960 to the stack pointer.
10962 2018-03-01 Tom de Vries <tom@codesourcery.com>
10964 PR rtl-optimization/83327
10965 * lra-int.h (hard_regs_spilled_into): Declare.
10966 * lra.c (hard_regs_spilled_into): Define.
10967 (init_reg_info): Init hard_regs_spilled_into.
10968 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10969 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10970 (process_bb_lives): Handle hard_regs_spilled_into.
10971 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10972 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10974 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10976 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10977 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10978 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10979 * config/rs6000/aix72.h: New file.
10981 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10983 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10984 instead of warning_at with conditional singular and plural messages
10988 * stor-layout.c (update_alignment_for_field): For
10989 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10990 && !DECL_PACKED (field), do the alignment update, just use
10991 only desired_align instead of MAX (type_align, desired_align)
10993 (place_field): Don't do known_align < desired_align handling
10994 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10995 is non-NULL, instead do it after rli->prev_field handling and
10996 only if not within a bitfield word. For DECL_PACKED (field)
10997 use type_align of BITS_PER_UNIT.
10999 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11001 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11002 superfluous parentheses and trailing spaces.
11004 2018-02-28 Richard Biener <rguenther@suse.de>
11006 PR tree-optimization/84584
11007 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11008 SCoPs with fake exit edge.
11010 2018-02-28 Martin Liska <mliska@suse.cz>
11013 * timevar.c (timer::print): Fix format to properly print 100%
11016 2018-02-28 Richard Biener <rguenther@suse.de>
11018 PR middle-end/84607
11019 * genmatch.c (capture_info::walk_match): Do not mark
11020 captured expressions without operands as expr_p given
11021 they act more like predicates and should be subject to
11022 "lost tail" side-effect preserving.
11024 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11026 PR rtl-optimization/81611
11027 * auto-inc-dec.c (attempt_change): Move dead note from
11028 mem_insn if it's the next use of regno
11029 (find_address): Take address use of reg holding
11030 non-incremented value. Add parm to limit search to the named
11032 (merge_in_block): Attempt to use a mem insn that is the next
11033 use of the original regno.
11035 2018-02-27 Martin Sebor <msebor@redhat.com>
11038 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11039 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11041 2018-02-27 Martin Sebor <msebor@redhat.com>
11043 PR translation/84207
11044 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11045 n argument to unsigned HOST_WIDE_INT.
11046 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11047 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11048 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11049 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11051 2018-02-27 Richard Biener <rguenther@suse.de>
11053 PR tree-optimization/84512
11054 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11055 Do not use the estimate returned from record_stmt_cost for
11056 the scalar iteration cost but sum properly using add_stmt_cost.
11058 2018-02-27 Richard Biener <rguenther@suse.de>
11060 PR tree-optimization/84466
11061 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11062 Adjust last change to less strictly validate use operands.
11064 2018-02-27 Martin Liska <mliska@suse.cz>
11066 PR gcov-profile/84548
11067 * gcov.c (process_file): Allow partial overlap and consider it
11068 also as group functions.
11069 (output_lines): Properly calculate range of lines for a group.
11071 2018-02-27 Martin Liska <mliska@suse.cz>
11073 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11074 'ggc' suffixes. Change first column width.
11075 (timer::print): Fix formatting of the column.
11077 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11079 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11080 preserve inline entry blocks for the sake of debug inline
11081 entry point markers alone.
11082 (remove_unused_locals): Suggest in comments a better place to
11083 force the preservation of inline entry blocks that are
11084 otherwise unused, but do not preserve them.
11086 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11088 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11090 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11093 * config/i386/constraints.md (Bs): Replace
11094 ix86_indirect_branch_register with
11095 TARGET_INDIRECT_BRANCH_REGISTER.
11097 * config/i386/i386.md (indirect_jump): Likewise.
11098 (tablejump): Likewise.
11099 (*sibcall_memory): Likewise.
11100 (*sibcall_value_memory): Likewise.
11101 Peepholes of indirect call and jump via memory: Likewise.
11102 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11103 (*sibcall_value_GOT_32): Likewise.
11104 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11105 (GOT_memory_operand): Likewise.
11106 (call_insn_operand): Likewise.
11107 (sibcall_insn_operand): Likewise.
11108 (GOT32_symbol_operand): Likewise.
11109 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11111 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11113 PR rtl-optimization/83496
11114 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11115 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11116 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11117 redundant insn, if any.
11118 (relax_delay_slots): Likewise.
11119 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11121 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11123 PR tree-optimization/83965
11124 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11125 that grouped statements are part of a reduction chain. Return
11126 true if the statement is not marked as a reduction itself but
11127 is part of a group.
11128 (vect_recog_dot_prod_pattern): Don't check whether the statement
11129 is part of a group here.
11130 (vect_recog_sad_pattern): Likewise.
11131 (vect_recog_widen_sum_pattern): Likewise.
11133 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11136 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11138 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11141 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11143 (ix86_output_indirect_function_return): New prototype.
11144 (ix86_split_simple_return_pop_internal): Likewise.
11145 * config/i386/i386.c (indirect_return_via_cx): New.
11146 (indirect_return_via_cx_bnd): Likewise.
11147 (indirect_thunk_name): Handle return va CX_REG.
11148 (output_indirect_thunk_function): Create alias for
11149 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11150 (ix86_output_indirect_jmp): Remove the bool argument.
11151 (ix86_output_indirect_function_return): New function.
11152 (ix86_split_simple_return_pop_internal): Likewise.
11153 * config/i386/i386.md (*indirect_jump): Don't pass false
11154 to ix86_output_indirect_jmp.
11155 (*tablejump_1): Likewise.
11156 (simple_return_pop_internal): Change it to define_insn_and_split.
11157 Call ix86_split_simple_return_pop_internal to split it for
11158 -mfunction-return=.
11159 (simple_return_indirect_internal): Call
11160 ix86_output_indirect_function_return instead of
11161 ix86_output_indirect_jmp.
11163 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11166 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11167 memset and value initialization afterwards.
11169 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11171 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11173 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11176 * common/config/aarch64/aarch64-common.c
11177 (aarch_option_optimization_table[]): Switch
11178 off fomit-frame-pointer
11180 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11181 Chung-Ju Wu <jasonwucj@gmail.com>
11183 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11185 (store_multiple): Ditto.
11187 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11189 * config.gcc: Add --with-cpu support for nds32 target.
11190 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11191 * config/nds32/nds32.opt: Add -mcpu= option.
11193 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11195 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11196 isel=yes): Warn for these deprecated options.
11198 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11200 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11201 ISA_2_5_MASKS_EMBEDDED.
11203 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11205 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11206 p->max as pointers rather than using iterative_hash_expr.
11208 2018-02-23 Carl Love <cel@us.ibm.com>
11210 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11211 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11213 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11214 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11215 P8V_BUILTIN_VEC_VUNSIGNED2.
11217 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11220 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11221 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11223 * lra-constraints.c (curr_insn_transform): Set up
11224 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11226 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11227 * lra-eliminations.c (spill_pseudos): Ditto.
11228 (process_insn_for_elimination): Ditto.
11229 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11230 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11231 LRA_NON_CLOBBERED_ALT.
11233 2018-02-22 Martin Sebor <msebor@redhat.com>
11235 PR tree-optimization/84480
11236 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11237 to maybe_diag_stxncpy_trunc. Call it.
11238 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11239 from gimple_fold_builtin_strcpy. Print inlining stack.
11240 (handle_builtin_stxncpy): Print inlining stack.
11241 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11243 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11246 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11247 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11248 and -fcheck-pointer-bounds are used together.
11249 (indirect_thunk_prefix): New enum.
11250 (indirect_thunk_need_prefix): New function.
11251 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11252 "_nt" instead of "_bnd" for NOTRACK prefix.
11253 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11254 (output_indirect_thunk_function): Likewise.
11256 (ix86_code_end): Update output_indirect_thunk_function calls.
11257 (ix86_output_indirect_branch_via_reg): Replace
11258 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11259 (ix86_output_indirect_branch_via_push): Likewise.
11260 (ix86_output_function_return): Likewise.
11261 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11262 incompatible with -fcf-protection=branch and
11263 -fcheck-pointer-bounds.
11265 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11268 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11269 Change gcc_assert call to output_operand_lossage.
11271 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11273 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11275 2018-02-22 DJ Delorie <dj@redhat.com>
11276 Sebastian Perta <sebastian.perta@renesas.com>
11277 Oleg Endo <olegendo@gcc.gnu.org>
11279 * config/rx/rx.c (rx_rtx_costs): New function.
11280 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11282 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11284 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11286 2018-02-22 Martin Liska <mliska@suse.cz>
11289 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11290 Add "native" as a possible value.
11292 2018-02-22 Martin Liska <mliska@suse.cz>
11295 * config/i386/i386.c (ix86_option_override_internal):
11296 Add "native" as a possible value for -march and -mtune.
11298 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11301 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11302 to all type variants.
11304 PR tree-optimization/84503
11305 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11306 width as info->bitpos + info->bitsize - start.
11307 (merged_store_group::merge_overlapping): Simplify width computation.
11308 (check_no_overlap): New function.
11309 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11310 start + width and last_order of the group, fail if check_no_overlap
11312 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11313 to group if check_no_overlap fails.
11315 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11317 * config/rs6000/altivec.md: Delete contraint arguments to
11318 define_expand, define_split, and define_peephole2, and in
11319 define_insn_and_split if always unused.
11320 * config/rs6000/darwin.md: Ditto.
11321 * config/rs6000/dfp.md: Ditto.
11322 * config/rs6000/rs6000.md: Ditto.
11323 * config/rs6000/sync.md: Ditto.
11324 * config/rs6000/vector.md: Ditto.
11325 * config/rs6000/vsx.md: Ditto.
11327 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11329 * config/rs6000/altivec.md: Write output control strings as braced
11330 blocks instead of double-quoted strings.
11331 * config/rs6000/darwin.md: Ditto.
11332 * config/rs6000/rs6000.md: Ditto.
11333 * config/rs6000/vector.md: Ditto.
11334 * config/rs6000/vsx.md: Ditto.
11336 2018-02-21 Jason Merrill <jason@redhat.com>
11338 PR c++/84314 - ICE with templates and fastcall attribute.
11339 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11341 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11343 * ipa-cp.c (determine_versionability): Fix comment typos.
11345 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11348 * ipa-cp.c (determine_versionability): Do not version functions caling
11351 2018-02-21 Martin Liska <mliska@suse.cz>
11354 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11355 Add "native" as a possible value.
11356 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11357 the macro when native cpu detection is available.
11359 2018-02-21 Martin Liska <mliska@suse.cz>
11362 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11363 Add "native" as a possible value.
11364 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11365 when native cpu detection is available.
11367 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11368 Martin Sebor <msebor@redhat.com>
11370 PR tree-optimization/84478
11371 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11373 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11374 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11375 support which is conservatively correct, for 2 only stay conservative
11376 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11377 argument to the 2 argument get_range_strlen, adjust 6 arg
11378 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11380 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11381 (gimple_fold_builtin_strlen): Pass true as last argument to
11384 2018-02-20 Martin Sebor <msebor@redhat.com>
11386 PR middle-end/84095
11387 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11388 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11389 (builtin_memref::builtin_memref): Factor out parts into
11390 set_base_and_offset and call it.
11392 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11394 PR middle-end/84406
11395 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11396 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11397 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11398 search at the associated MODE_INT.
11400 2018-02-20 Jeff Law <law@redhat.com>
11402 PR middle-end/82123
11403 PR tree-optimization/81592
11404 PR middle-end/79257
11405 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11406 for range data rather than using global data.
11407 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11408 range data rather than using global data.
11409 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11410 pass it to children as needed.
11411 (struct directive::fmtresult): Similarly.
11412 (struct directive::set_width): Similarly.
11413 (struct directive::set_precision): Similarly.
11414 (format_integer, format_directive, parse_directive): Similarly.
11415 (format_none): Accept unnamed vr_values parameter.
11416 (format_percent, format_floating, format_character): Similarly.
11417 (format_string, format_plain): Similarly.
11418 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11419 the EVRP range analyzer for range data rather than using global data.
11420 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11421 gimple-ssa-evrp-analyze.h
11422 (class sprintf_dom_walker): Add after_dom_children member function.
11423 Add evrp_range_analyzer member.
11424 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11425 range analyzer as needed.
11426 (sprintf_dom_walker::after_dom_children): New member function.
11427 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11429 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11430 (evrp_range_analyzer::pop_to_marker): Likewise.
11432 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11434 PR tree-optimization/84419
11435 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11436 with the required type if its current type is compatible but
11439 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11441 PR middle-end/82004
11442 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11443 after vectorization.
11445 2018-02-20 Martin Liska <mliska@suse.cz>
11448 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11449 possible values if we don't have a hint.
11451 2018-02-20 Martin Liska <mliska@suse.cz>
11455 * final.c (shorten_branches): Build align_tab array with one
11457 * opts.c (finish_options): Add alignment option limit check.
11458 (MAX_CODE_ALIGN): Likewise.
11459 (MAX_CODE_ALIGN_VALUE): Likewise.
11460 * doc/invoke.texi: Document maximum allowed option value for
11461 all -falign-* options.
11463 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11466 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11467 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11468 * var-tracking.c (emit_note_insn_var_location): Remove all references
11469 to NOTE_INSN_CALL_ARG_LOCATION.
11470 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11471 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11472 Use copy_rtx_if_shared.
11473 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11474 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11475 (dwarf2out_var_location): Remove handling of
11476 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11478 * final.c (final_scan_insn): Remove all references to
11479 NOTE_INSN_CALL_ARG_LOCATION.
11480 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11481 before dumping final insns.
11482 * except.c (emit_note_eh_region_end): Remove all references to
11483 NOTE_INSN_CALL_ARG_LOCATION.
11484 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11485 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11486 * config/arc/arc.c (hwloop_optimize): Likewise.
11487 * config/arm/arm.c (create_fix_barrier): Likewise.
11488 * config/s390/s390.c (s390_chunkify_start): Likewise.
11489 * config/sh/sh.c (find_barrier): Likewise.
11490 * config/i386/i386.c (rest_of_insert_endbranch,
11491 ix86_seh_fixup_eh_fallthru): Likewise.
11492 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11493 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11494 * config/frv/frv.c (frv_function_prologue): Likewise.
11495 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11497 (note_outside_basic_block_p): Remove all references to
11498 NOTE_INSN_CALL_ARG_LOCATION.
11499 * gengtype.c (adjust_field_rtx_def): Likewise.
11500 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11502 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11503 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11506 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11509 PR tree-optimization/84452
11510 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11511 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11514 2018-02-19 Martin Liska <mliska@suse.cz>
11517 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11519 2018-02-19 Martin Liska <mliska@suse.cz>
11520 Richard Sandiford <richard.sandiford@linaro.org>
11522 PR tree-optimization/82491
11523 * gimple-fold.c (get_base_constructor): Make earlier bail out
11526 2018-02-19 Carl Love <cel@us.ibm.com>
11528 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11529 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11531 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11532 P8V_BUILTIN_VEC_NEG.
11534 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11536 * config/rl78/rl78.md (movdf): New define expand.
11538 2018-02-19 Martin Liska <mliska@suse.cz>
11541 * doc/invoke.texi: Fix typo.
11542 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11544 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11546 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11547 handle rs6000_single_float and rs6000_double_float specially for
11550 2018-02-16 Jeff Law <law@redhat.com>
11552 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11553 (pop_regs): Use it.
11555 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11558 * ipa-inline.c (inline_small_functions): Fix a typo.
11560 2018-02-16 Nathan Sidwell <nathan@acm.org>
11562 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11564 2018-02-16 Carl Love <cel@us.ibm.com>
11566 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11567 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11568 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11569 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11570 expansion to P8V_BUILTIN_VEC_FLOAT2.
11572 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11574 PR rtl-optimization/70023
11575 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11576 src_regno into account.
11578 2018-02-16 Carl Love <cel@us.ibm.com>
11580 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11581 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11582 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11583 * config/rs6000/rs6000.c: Remove case statements for
11584 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11585 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11586 and P9V_BUILTIN_VEC_VINSERT4B.
11587 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11588 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11589 * config/rs6000/vsx.md:
11590 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11593 2018-02-16 Carl Love <cel@us.ibm.com>
11595 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11597 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11599 * config/rs6000/rs6000-c.c: Add the definitions for
11600 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11601 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11602 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11603 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11604 definition for insert4b and define insn *insert3b_internal.
11605 * doc/extend.texi: Add documentation for vec_extract4b.
11607 2018-02-16 Nathan Sidwell <nathan@acm.org>
11609 * doc/extend.texi (Backwards Compatibility): Mention friend
11610 injection. Note for-scope is deprecated.
11611 * doc/invoke.texi (-ffriend-injection): Deprecate.
11613 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11615 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11616 that moved to I2, also allow destinations that are a paradoxical
11617 subreg (instead of a normal reg).
11619 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11622 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11625 2018-02-16 Richard Biener <rguenther@suse.de>
11627 PR tree-optimization/84037
11628 PR tree-optimization/84016
11630 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11631 Adjust vec_construct for the fact we need additional higher latency
11632 128bit inserts for AVX256 and AVX512 vector builds.
11633 (ix86_add_stmt_cost): Scale vector construction cost for
11636 2018-02-16 Richard Biener <rguenther@suse.de>
11638 PR tree-optimization/84417
11639 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11640 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11641 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11643 2018-02-16 Martin Liska <mliska@suse.cz>
11646 * internal-fn.def (ASAN_CHECK): Set proper flags.
11647 (ASAN_MARK): Likewise.
11649 2018-02-16 Julia Koval <julia.koval@intel.com>
11651 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11652 from PTA_CANNONLAKE.
11654 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11657 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11658 Use ++iter rather than iter++ for std::list iterators.
11659 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11660 defer deleting them until all nodes in the forest are processed. Do
11661 free even leaf nodes. Change to_process into auto_vec.
11664 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11665 * vec.h (vec_default_construct): Use memset instead of placement new
11666 if BROKEN_VALUE_INITIALIZATION is defined.
11667 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11668 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11671 PR rtl-optimization/83723
11672 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11673 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11674 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11676 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11678 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11680 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11682 PR rtl-optimization/81443
11683 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11684 from inner REGs to paradoxical SUBREGs.
11686 2018-02-16 Richard Biener <rguenther@suse.de>
11688 PR tree-optimization/84399
11689 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11690 For operands we can analyze at their definition make sure we can
11691 analyze them at each use as well.
11693 2018-02-16 Richard Biener <rguenther@suse.de>
11695 PR tree-optimization/84190
11696 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11697 volatile accesses if the decl isn't volatile.
11699 2018-02-15 Jason Merrill <jason@redhat.com>
11701 PR c++/84314 - ICE with templates and fastcall attribute.
11702 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11703 TYPE_CANONICAL on an existing type.
11705 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11707 PR tree-optimization/84383
11708 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11709 dstoff nor call operand_equal_p if dstbase is NULL.
11711 PR tree-optimization/84334
11712 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11713 also a CONSTANT_CLASS_P, punt.
11715 2018-02-14 Jim Wilson <jimw@sifive.com>
11717 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11718 first SMALL_OPERAND check. New local min_second_step. Move assert
11719 to where locals are set. Add TARGET_RVC support.
11720 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11722 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11724 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11726 2018-02-14 Martin Sebor <msebor@redhat.com>
11728 PR tree-optimization/83698
11729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11730 arrays constrain the offset range to their bounds.
11731 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11732 (builtin_access::overlap): Avoid setting the size of overlap if it's
11734 (maybe_diag_overlap): Also consider arrays when deciding what values
11735 of offsets to include in diagnostics.
11737 2018-02-14 Martin Sebor <msebor@redhat.com>
11740 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11741 that correspond to the kind of a declaration.
11743 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11746 * config/pa/pa.md: Load address of PIC label using the linkage table
11747 if the label is nonlocal.
11749 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11751 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11752 warning message if user requests -maltivec=be.
11753 * doc/invoke.texi: Document deprecation of -maltivec=be.
11755 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11758 * config/rs6000/rs6000-c.c: Update definitions for
11759 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11760 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11762 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11765 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11766 add _get_ssp intrinsics. Remove argument from
11767 __builtin_ia32_rdssp[d|q].
11768 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11769 * config/i386/i386-builtin.def: Remove argument from
11770 __builtin_ia32_rdssp[d|q].
11771 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11772 ix86_expand_special_args_builtin for _rdssp[d|q].
11773 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11774 Clear register before usage.
11775 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11776 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11778 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11780 PR tree-optimization/84357
11781 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11782 operand 1 of an ARRAY_REF too.
11784 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11787 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11788 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11790 (set_of_reg): New struct.
11791 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11792 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11793 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11795 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11796 Split into bitclr, bitset, bitinvert patterns if appropriate.
11797 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11798 use rx_fuse_in_memory_bitop.
11799 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11800 to named insn, correct maximum insn length.
11802 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11805 * machmode.def: Define a complex mode for PARTIAL_INT.
11806 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11808 * doc/rtl.texi: Document CSPImode.
11809 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11811 (msp430_hard_regno_nregs_with_padding): Likewise.
11813 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11816 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11818 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11820 PR rtl-optimization/84169
11821 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11822 we generated a parallel as new i3 and we split that to new i2 and i3
11823 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11824 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11825 those to i2, not i1. Partially rewrite this scan code.
11827 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11830 * stor-layout.c (place_field): For variable length fields, adjust
11831 offset_align afterwards not just based on the field's alignment,
11832 but also on the size.
11834 PR middle-end/84309
11835 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11836 of exps and logs in the use_exp2 case.
11838 2018-02-13 Jeff Law <law@redhat.com>
11840 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11841 entry for "vector".
11843 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11846 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11850 * common.opt (gas-loc-support, gas-locview-support): New.
11851 (ginline-points, ginternal-reset-location-views): New.
11852 * doc/invoke.texi: Document them. Use @itemx where intended.
11853 (gvariable-location-views): Adjust.
11854 * target.def (reset_location_view): New.
11855 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11856 (TARGET_RESET_LOCATION_VIEW): New.
11857 * doc/tm.texi: Rebuilt.
11858 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11859 (dwarf2out_default_as_locview_support): New.
11860 (output_asm_line_debug_info): Use option variables.
11861 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11862 (output_loc_list): Likewise.
11863 (add_high_low_attributes): Check option variables.
11864 Don't output entry view attribute in strict mode.
11865 (gen_inlined_subroutine_die): Check option variables.
11866 (dwarf2out_inline_entry): Likewise.
11867 (init_sections_and_labels): Likewise.
11868 (dwarf2out_early_finish): Likewise.
11869 (maybe_reset_location_view): New, from...
11870 (dwarf2out_var_location): ... here. Call it.
11871 * debug.h (dwarf2out_default_as_loc_support): Declare.
11872 (dwarf2out_default_as_locview_support): Declare.
11873 * hooks.c (hook_int_rtx_insn_0): New.
11874 * hooks.h (hook_int_rtx_insn_0): Declare.
11875 * toplev.c (process_options): Take -gas-loc-support and
11876 -gas-locview-support from dwarf2out. Enable
11877 -gvariable-location-views by default only with locview
11878 assembler support. Enable -ginternal-reset-location-views by
11879 default only if the target defines the corresponding hook.
11880 Enable -ginline-points by default if location views are
11881 enabled; force it disabled if statement frontiers are
11883 * tree-inline.c (expand_call_inline): Check option variables.
11884 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11886 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11888 PR tree-optimization/84321
11889 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11890 handling. Also check whether the anti-range contains any values
11891 that satisfy the mask; switch to a VR_RANGE if not.
11893 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11896 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11898 2018-02-13 Martin Jambor <mjambor@suse.cz>
11901 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11902 of call statements, also set location of a load to a temporary.
11904 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11906 * config/rl78/rl78.c (add_vector_labels): New function.
11907 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11908 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11909 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11910 which checks that no arguments are passed.
11911 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11912 * doc/extend.texi: Documentation for the new attribute.
11914 2018-02-13 Andreas Schwab <schwab@suse.de>
11916 * config/riscv/linux.h (CPP_SPEC): Define.
11918 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11921 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11922 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11923 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11924 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11925 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11926 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11927 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11928 temporarily for AES and PCLMUL builtins.
11930 PR tree-optimization/84339
11931 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11932 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11935 PR middle-end/84309
11936 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11937 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11938 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11940 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11942 * omp-simd-clone.h: New file.
11943 * omp-simd-clone.c: Include omp-simd-clone.h.
11944 (expand_simd_clones): No longer static.
11945 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11946 cgraph.h and omp-simd-clone.h.
11947 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11948 (vect_recog_widen_shift_pattern): Formatting fix.
11949 (vect_pattern_recog_1): Don't check optab for calls.
11952 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11953 operands[2] into a REG before using gen_lowpart on it.
11955 2018-02-12 Jeff Law <law@redhat.com>
11958 * config/sh/sh.c (find_barrier): Consider a sibling call
11961 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11962 successfully back substituting a reg.
11964 2018-02-12 Richard Biener <rguenther@suse.de>
11966 PR tree-optimization/84037
11967 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11968 parameter, move visited init to caller.
11969 (vect_slp_analyze_operations): Separate cost from validity
11970 check, initialize visited once for all instances.
11971 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11973 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11975 (vect_model_promotion_demotion_cost): Likewise.
11976 (vectorizable_bswap): Guard cost modeling with !slp_node
11977 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11979 (vectorizable_call): Likewise.
11980 (vectorizable_conversion): Likewise.
11981 (vectorizable_assignment): Likewise.
11982 (vectorizable_shift): Likewise.
11983 (vectorizable_operation): Likewise.
11984 (vectorizable_store): Likewise.
11985 (vectorizable_load): Likewise.
11986 (vectorizable_condition): Likewise.
11987 (vectorizable_comparison): Likewise.
11989 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11992 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11993 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11994 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11996 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11998 PR middle-end/83665
11999 * params.def (inline-min-speedup): Increase from 8 to 15.
12000 (max-inline-insns-auto): Decrease from 40 to 30.
12001 * ipa-split.c (consider_split): Add some buffer for function to
12002 be considered inlining candidate.
12003 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12006 2018-02-12 Richard Biener <rguenther@suse.de>
12008 PR tree-optimization/84037
12009 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12010 matched stmts if we cannot swap the non-matched ones.
12012 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12014 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12015 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12016 _mm_maskz_scalef_round_ss): New intrinsics.
12017 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12018 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12019 __builtin_ia32_scalefss_round): Remove.
12020 (__builtin_ia32_scalefsd_mask_round,
12021 __builtin_ia32_scalefss_mask_round): New intrinsics.
12022 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12023 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12024 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12025 "<round_constraint>")): Changed to ...
12026 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12027 "<round_scalar_constraint>")): ... this.
12028 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12029 %0, %1, %2<round_op3>}"): Changed to ...
12030 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12031 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12032 %2<round_scalar_mask_op3>}"): ... this.
12033 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12035 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12037 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12038 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12039 (_mm_maskz_sqrt_round_ss): New intrinsics.
12040 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12041 (__builtin_ia32_sqrtsd_mask_round)
12042 (__builtin_ia32_sqrtss_mask_round): New builtins.
12043 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12044 (__builtin_ia32_sqrtss_round): Remove.
12045 (__builtin_ia32_sqrtsd_mask_round)
12046 (__builtin_ia32_sqrtss_mask_round): New builtins.
12047 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12048 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12049 ((match_operand:VF_128 1 "vector_operand"
12050 "xBm,<round_constraint>")): Changed to ...
12051 ((match_operand:VF_128 1 "vector_operand"
12052 "xBm,<round_scalar_constraint>")): ... this.
12053 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12054 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12055 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12056 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12057 %<iptr>1<round_scalar_mask_op3>}): ... this.
12058 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12059 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12061 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12064 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12065 Cast vec_cmpeq result to correct type.
12066 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12067 Cast vec_cmpgt result to correct type.
12069 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12071 * final.c (final_scan_insn_1): Renamed from...
12072 (final_scan_insn): ... this. New wrapper, to recover
12073 seen from the outermost call in recursive ones.
12074 * config/sparc/sparc.c (output_return): Drop seen from call.
12075 (output_sibcall): Likewise.
12076 * config/visium/visium.c (output_branch): Likewise.
12078 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12080 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12083 2018-02-10 Alan Modra <amodra@gmail.com>
12086 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12087 Specify LR as an input.
12089 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12092 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12093 remove_member_access_dummy_vars): New functions.
12094 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12095 lower_omp_1, execute_lower_omp): Use them.
12097 PR rtl-optimization/84308
12098 * shrink-wrap.c (spread_components): Release todo vector.
12100 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12102 PR rtl-optimization/57193
12103 * ira-color.c (struct allocno_color_data): Add member
12104 conflict_allocno_hard_prefs.
12105 (update_conflict_allocno_hard_prefs): New.
12106 (bucket_allocno_compare_func): Add a preference based on
12107 conflict_allocno_hard_prefs.
12108 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12109 (color_allocnos): Remove a dead code. Initiate
12110 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12115 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12116 constraint from =wa to wa. Avoid a subreg on the output operand,
12117 instead use a pseudo and subreg it in a move.
12118 (p9_xxbrd_<mode>): Changed to ...
12119 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12120 (p9_xxbrd_v2df): New expander.
12121 (p9_xxbrw_<mode>): Changed to ...
12122 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12123 (p9_xxbrw_v4sf): New expander.
12125 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12127 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12129 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12132 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12133 multiply in 32-bit mode.
12134 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12135 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12138 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12140 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12141 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12142 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12143 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12145 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12148 * dwarf2out.c (is_trivial_indirect_ref): New function.
12149 (dwarf2out_late_global_decl): Do not generate a location
12150 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12151 and that are not defined in the current unit.
12153 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12155 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12156 instead of a libcall for UNORDERED.
12158 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12161 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12162 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12164 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12167 * config/s390/s390.c (s390_set_current_function): Invoke
12168 s390_indirect_branch_settings also if fndecl didn't change.
12170 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12172 * config/rs6000/rs6000.md (blockage): Set length to zero.
12174 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12176 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12178 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12181 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12182 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12186 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12187 PARALLEL incoming that failed vt_get_decl_and_offset check.
12189 PR middle-end/84237
12190 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12191 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12193 (get_variable_section): For decls in named .bss* sections pass true as
12194 second argument to bss_initializer_p.
12196 2018-02-09 Marek Polacek <polacek@redhat.com>
12197 Jakub Jelinek <jakub@redhat.com>
12200 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12201 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12202 Sync some changes from cxx_fold_indirect_ref.
12204 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12206 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12208 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12209 (BLOCK_INLINE_ENTRY_LABEL): New.
12210 (dwarf2out_var_location): Disregard inline entry markers.
12211 (inline_entry_data): New struct.
12212 (inline_entry_data_hasher): New hashtable type.
12213 (inline_entry_data_hasher::hash): New.
12214 (inline_entry_data_hasher::equal): New.
12215 (inline_entry_data_table): New variable.
12216 (add_high_low_attributes): Add DW_AT_entry_pc and
12217 DW_AT_GNU_entry_view attributes if a pending entry is found
12218 in inline_entry_data_table. Add old entry_pc attribute only
12219 if debug nonbinding markers are disabled.
12220 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12221 markers are enabled.
12222 (block_within_block_p, dwarf2out_inline_entry): New.
12223 (dwarf2out_finish): Check that no entries remained in
12224 inline_entry_data_table.
12225 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12226 (final_scan_insn, notice_source_line): Likewise.
12227 (rest_of_clean_state): Skip inline entry markers.
12228 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12230 * gimple.c (gimple_build_debug_inline_entry): New.
12231 * gimple.h (enum gimple_debug_subcode): Add
12232 GIMPLE_DEBUG_INLINE_ENTRY.
12233 (gimple_build_debug_inline_entry): Declare.
12234 (gimple_debug_inline_entry_p): New.
12235 (gimple_debug_nonbind_marker_p): Adjust.
12236 * insn-notes.def (INLINE_ENTRY): New.
12237 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12238 inline entry marker notes.
12239 (print_insn): Likewise.
12240 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12241 (INSN_DEBUG_MARKER_KIND): Likewise.
12242 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12243 * tree-inline.c (expand_call_inline): Build and insert
12244 debug_inline_entry stmt.
12245 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12246 inline entry blocks early, if nonbind markers are enabled.
12247 (dump_scope_block): Dump fragment info.
12248 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12249 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12250 (gimple_build_debug_inline_entry): New.
12251 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12252 Enable/disable inline entry points too.
12253 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12254 (DEBUG_INSN): Describe inline entry markers.
12256 * common.opt (gvariable-location-views): New.
12257 (gvariable-location-views=incompat5): New.
12258 * config.in: Rebuilt.
12259 * configure: Rebuilt.
12260 * configure.ac: Test assembler for view support.
12261 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12262 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12263 * dwarf2out.c (var_loc_view): New typedef.
12264 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12265 (dwarf2out_locviews_in_attribute): New.
12266 (dwarf2out_locviews_in_loclist): New.
12267 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12268 (enum dw_line_info_opcode): Add LI_adv_address.
12269 (struct dw_line_info_table): Add view.
12270 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12271 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12272 (zero_view_p): New variable.
12273 (ZERO_VIEW_P): New macro.
12274 (output_asm_line_debug_info): New.
12275 (struct var_loc_node): Add view.
12276 (add_AT_view_list, AT_loc_list): New.
12277 (add_var_loc_to_decl): Add view param. Test it against last.
12278 (new_loc_list): Add view params. Record them.
12279 (AT_loc_list_ptr): Handle loc and view lists.
12280 (view_list_to_loc_list_val_node): New.
12281 (print_dw_val): Handle dw_val_class_view_list.
12282 (size_of_die): Likewise.
12283 (value_format): Likewise.
12284 (loc_list_has_views): New.
12285 (gen_llsym): Set vl_symbol too.
12286 (maybe_gen_llsym, skip_loc_list_entry): New.
12287 (dwarf2out_maybe_output_loclist_view_pair): New.
12288 (output_loc_list): Output view list or entries too.
12289 (output_view_list_offset): New.
12290 (output_die): Handle dw_val_class_view_list.
12291 (output_dwarf_version): New.
12292 (output_compilation_unit_header): Use it.
12293 (output_skeleton_debug_sections): Likewise.
12294 (output_rnglists, output_line_info): Likewise.
12295 (output_pubnames, output_aranges): Update version comments.
12296 (output_one_line_info_table): Output view numbers in asm comments.
12297 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12298 Call maybe_gen_llsym.
12299 (loc_list_from_tree_1): Adjust.
12300 (add_AT_location_description): Create view list attribute if
12301 needed, check it's absent otherwise.
12302 (convert_cfa_to_fb_loc_list): Adjust.
12303 (maybe_emit_file): Call output_asm_line_debug_info for test.
12304 (dwarf2out_var_location): Reset views as needed. Precompute
12305 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12306 attribute. Set view.
12307 (new_line_info_table): Reset next view.
12308 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12309 (dwarf2out_source_line): Likewise. Output view resets and labels to
12310 the assembler, or select appropriate line info opcodes.
12311 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12312 (optimize_string_length): Catch it. Adjust.
12313 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12314 dw_val_class_view_list, and remove it if no longer needed.
12315 (hash_loc_list): Hash view numbers.
12316 (loc_list_hasher::equal): Compare them.
12317 (optimize_location_lists): Check whether a view list symbol is
12318 needed, and whether the locview attribute is present, and
12319 whether they match. Remove the locview attribute if no longer
12321 (index_location_lists): Call skip_loc_list_entry for test.
12322 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12323 Use output_dwarf_version.
12324 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12325 (struct dw_val_node): Add val_view_list.
12326 * final.c (SEEN_NEXT_VIEW): New.
12327 (set_next_view_needed): New.
12328 (clear_next_view_needed): New.
12329 (maybe_output_next_view): New.
12330 (final_start_function): Rename to...
12331 (final_start_function_1): ... this. Take pointer to FIRST,
12332 add SEEN parameter. Emit param bindings in the initial view.
12333 (final_start_function): Reintroduce SEEN-less interface.
12334 (final): Rename to...
12335 (final_1): ... this. Take SEEN parameter. Output final pending
12336 next view at the end.
12337 (final): Reintroduce seen-less interface.
12338 (final_scan_insn): Output pending next view before switching
12339 sections or ending a block. Mark the next view as needed when
12340 outputting variable locations. Notify debug backend of section
12341 changes, and of location view changes.
12342 (rest_of_handle_final): Adjust.
12343 * toplev.c (process_options): Autodetect value for debug variable
12344 location views option. Warn on incompat5 without -gdwarf-5.
12345 * doc/invoke.texi (gvariable-location-views): New.
12346 (gvariable-location-views=incompat5): New.
12347 (gno-variable-location-views): New.
12349 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12351 PR tree-optimization/84136
12352 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12353 that the result of find_edge is non-NULL.
12355 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12358 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12359 storing integer register in SImode. Fix cost of 256 and 512
12360 byte aligned SSE register store.
12362 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12364 * config/i386/i386.c (ix86_multiplication_cost): Fix
12365 multiplication cost for TARGET_AVX512DQ.
12367 2018-02-08 Marek Polacek <polacek@redhat.com>
12369 PR tree-optimization/84238
12370 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12373 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12375 PR tree-optimization/84265
12376 * tree-vect-stmts.c (vectorizable_store): Don't treat
12377 VMAT_CONTIGUOUS accesses as grouped.
12378 (vectorizable_load): Likewise.
12380 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12382 PR tree-optimization/81635
12383 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12384 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12385 (test_round_for_mask): New functions.
12386 (wide_int_cc_tests): Call test_round_for_mask.
12387 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12388 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12389 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12390 range returned by get_range_info.
12392 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12395 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12396 * symtab.c: Include builtins.h
12397 (symtab_node::output_to_lto_symbol_table_p): Move here
12398 from lto-streamer-out.c:output_symbol_p.
12399 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12400 (output_symbol_p): Move all logic to symtab.c
12401 (produce_symtab): Update.
12403 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12405 * config/s390/s390-opts.h (enum indirect_branch): Define.
12406 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12407 (s390_indirect_branch_via_thunk)
12408 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12409 (enum s390_indirect_branch_type): Define.
12410 * config/s390/s390.c (struct s390_frame_layout, struct
12411 machine_function): Remove.
12412 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12413 (indirect_branch_table_label_no, indirect_branch_table_name):
12415 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12416 (enum s390_indirect_branch_option): Define.
12417 (s390_return_addr_from_memory): New function.
12418 (s390_handle_string_attribute): New function.
12419 (s390_attribute_table): Add new attribute handler.
12420 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12421 (s390_indirect_branch_via_thunk): New function.
12422 (s390_indirect_branch_via_inline_thunk): New function.
12423 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12424 sibling call optimization for non z10 compiles.
12425 (s390_emit_call): Force indirect branch target to be a single
12426 register. Add r1 clobber for non-z10 compiles.
12427 (s390_emit_epilogue): Emit return jump via return_use expander.
12428 (s390_reorg): Handle JUMP_INSNs as execute targets.
12429 (s390_option_override_internal): Perform validity checks for the
12430 new command line options.
12431 (s390_indirect_branch_attrvalue): New function.
12432 (s390_indirect_branch_settings): New function.
12433 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12434 (s390_output_indirect_thunk_function): New function.
12435 (s390_code_end): Implement target hook.
12436 (s390_case_values_threshold): Implement target hook.
12437 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12439 * config/s390/s390.h (struct s390_frame_layout)
12440 (struct machine_function): Move here from s390.c.
12441 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12442 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12443 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12444 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12445 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12446 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12447 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12448 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12449 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12450 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12451 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12452 (mnemonic attribute): Add values which aren't recognized
12454 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12455 pattern for branch conversion. Fix mnemonic attribute.
12456 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12457 indirect branch via thunk if requested.
12458 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12459 ("*indirect_jump"): Disable for branch conversion using out of
12461 ("indirect_jump_via_thunk<mode>_z10")
12462 ("indirect_jump_via_thunk<mode>")
12463 ("indirect_jump_via_inlinethunk<mode>_z10")
12464 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12465 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12466 ("casesi_jump_via_inlinethunk<mode>_z10")
12467 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12468 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12469 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12470 ("*indirect2_jump"): Disable for branch conversion.
12471 ("casesi_jump"): Turn into expander and expand patterns for branch
12473 ("return_use"): New expander.
12474 ("*return"): Emit return via thunk and rename it to ...
12475 ("*return<mode>"): ... this one.
12476 * config/s390/s390.opt: Add new options and and enum for the
12479 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12481 * lra-constraints.c (match_reload): Unconditionally use
12482 gen_lowpart_SUBREG, rather than selecting between that
12483 and equivalent gen_rtx_SUBREG code.
12485 2018-02-08 Richard Biener <rguenther@suse.de>
12487 PR tree-optimization/84233
12488 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12489 changed flag instead of boguously re-using phi_inserted.
12491 2018-02-08 Martin Jambor <mjambor@suse.cz>
12493 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12494 static local variables.
12496 2018-02-08 Richard Biener <rguenther@suse.de>
12498 PR tree-optimization/84278
12499 * tree-vect-stmts.c (vectorizable_store): When looking for
12500 smaller vector types to perform grouped strided loads/stores
12501 make sure the mode is supported by the target.
12502 (vectorizable_load): Likewise.
12504 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12506 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12507 Increase LDP/STP opportunities by adding adjacent callee-saves.
12509 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12511 PR rtl-optimization/84068
12512 PR rtl-optimization/83459
12513 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12515 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12517 PR tree-optimization/84224
12518 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12519 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12520 non-zero arguments.
12522 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12525 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12526 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12527 count, remove one USE.
12529 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12531 * doc/install.texi (Configuration): Document the
12532 --with-long-double-format={ibm,ieee} PowerPC configuration
12536 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12537 Convert from define_expand to be define_insn_and_split. Rework
12538 float/double/_Float128 conversions to QI/HI/SImode to work with
12539 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12540 conversions to QI/HImode types did a store and then a load to
12541 truncate the value. For conversions to VSX registers, don't split
12542 the insn, instead emit the code directly. Use the code iterator
12543 any_fix to combine signed and unsigned conversions.
12544 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12545 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12546 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12547 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12548 (fix_<mode>di2_hw): Likewise.
12549 (fixuns_<mode>di2_hw): Likewise.
12550 (fix_<mode>si2_hw): Likewise.
12551 (fixuns_<mode>si2_hw): Likewise.
12552 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12553 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12554 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12555 fix<uns>_trunc<SFDF:mode>si2_p8.
12556 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12557 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12558 (fix<uns>_<mode>_mem): Likewise.
12559 (fctiw<u>z_<mode>_mem): Likewise.
12560 (fix<uns>_<mode>_mem): Likewise.
12561 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12562 the register allocator from doing a direct move to the GPRs to do
12563 a store, and instead use the ISA 3.0 store byte/half-word from
12564 vector register instruction. For IEEE 128-bit floating point,
12565 also optimize stores of 32-bit ints.
12566 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12568 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12570 * genextract.c (push_pathstr_operand): New function to support
12572 (walk_rtx): Call push_pathstr_operand.
12573 (print_path): Support [a-zA-Z].
12575 2018-02-07 Richard Biener <rguenther@suse.de>
12577 PR tree-optimization/84037
12578 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12579 (cse_and_gimplify_to_preheader): Declare.
12580 (vect_get_place_in_interleaving_chain): Likewise.
12581 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12583 (_loop_vec_info::~_loop_vec_info): Delete it.
12584 (cse_and_gimplify_to_preheader): New function.
12585 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12586 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12587 (vectorizable_load): Likewise. For grouped stores always base
12588 the IV on the first element.
12589 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12590 condition before gimplifying.
12592 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12594 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12595 *DIV_EXPR and *MOD_EXPR.
12597 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12600 * config/i386/i386.c (ix86_option_override_internal): Mask out
12601 the CF_SET bit when checking -fcf-protection.
12603 2018-02-07 Tom de Vries <tom@codesourcery.com>
12606 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12609 2018-02-07 Richard Biener <rguenther@suse.de>
12611 PR tree-optimization/84204
12612 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12615 PR tree-optimization/84205
12616 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12617 special-case isl_ast_op_zdiv_r.
12619 PR tree-optimization/84223
12620 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12621 Only add conditions from within the region.
12622 (gather_bbs::after_dom_children): Adjust.
12624 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12627 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12628 * config/avr/avr.md: Only post-reload split REG-REG moves if
12629 either register is GENERAL_REG_P.
12631 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12633 PR tree-optimization/84235
12634 * tree-ssa-scopedtables.c
12635 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12636 if the subtraction is performed in floating point type where NaNs are
12637 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12638 build 1. Formatting fix.
12640 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12643 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12644 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12645 and skip it regardless of bb boundaries. Use CALL_P macro,
12646 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12647 unnecessarily, formatting fix.
12649 2018-02-06 Michael Collison <michael.collison@arm.com>
12651 * config/arm/thumb2.md:
12652 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12653 (*thumb_mov_notscc): Ditto.
12655 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12658 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12659 unsigned_fix, not "s".
12661 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12663 * configure.ac (gcc_fn_eh_frame_ro): New function.
12664 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12665 correct .eh_frame permissions.
12666 * configure: Regenerate.
12668 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12670 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12671 irrelevant options.
12673 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12675 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12676 Display warning message for -mno-speculate-indirect-jumps.
12678 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12680 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12682 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12684 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12686 PR tree-optimization/84225
12687 * tree-eh.c (find_trapping_overflow): Only call
12688 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12690 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12693 * config/i386/i386.c: Reimplement the check of possible options
12694 -mibt/-mshstk conbination. Change error messages.
12695 * doc/invoke.texi: Fix a typo: remove extra '='.
12697 2018-02-06 Marek Polacek <polacek@redhat.com>
12699 PR tree-optimization/84228
12700 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12702 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12705 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12706 emitted arch directives.
12707 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12708 __ARM_FEATURE_COPROC before changing architectures.
12710 2018-02-06 Richard Biener <rguenther@suse.de>
12712 * config/i386/i386.c (print_reg): Fix typo.
12713 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12715 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12717 * configure: Regenerate.
12719 2018-02-05 Martin Sebor <msebor@redhat.com>
12721 PR tree-optimization/83369
12722 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12725 2018-02-05 Martin Liska <mliska@suse.cz>
12727 * doc/invoke.texi: Cherry-pick upstream r323995.
12729 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12731 * ira.c (ira_init_register_move_cost): Adjust comment.
12733 2018-02-05 Martin Liska <mliska@suse.cz>
12735 PR gcov-profile/84137
12736 * doc/gcov.texi: Fix typo in documentation.
12738 2018-02-05 Martin Liska <mliska@suse.cz>
12740 PR gcov-profile/83879
12741 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12742 using dlopen functionality.
12744 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12746 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12747 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12748 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12749 _mm_maskz_range_round_ss): New intrinsics.
12750 (__builtin_ia32_rangesd128_round)
12751 (__builtin_ia32_rangess128_round): Remove.
12752 (__builtin_ia32_rangesd128_mask_round,
12753 __builtin_ia32_rangess128_mask_round): New builtins.
12754 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12755 __builtin_ia32_rangess128_round): Remove.
12756 (__builtin_ia32_rangesd128_mask_round,
12757 __builtin_ia32_rangess128_mask_round): New builtins.
12758 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12759 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12760 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12761 "<round_saeonly_constraint>")): Changed to ...
12762 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12763 "<round_saeonly_scalar_constraint>")): ... this.
12764 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12765 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12766 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12767 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12768 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12770 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12772 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12774 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12775 Remove all values except native, 8540 and 8548.
12777 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12779 * config/i386/i386.c (ix86_output_function_return): Pass
12780 INVALID_REGNUM, instead of -1, as invalid register number to
12781 indirect_thunk_name and output_indirect_thunk.
12783 2018-02-02 Julia Koval <julia.koval@intel.com>
12785 * config.gcc: Add -march=icelake.
12786 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12787 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12788 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12789 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12790 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12791 (processor_target_table): Add icelake.
12792 (ix86_option_override_internal): Handle new PTAs.
12793 (get_builtin_code_for_version): Handle icelake.
12794 (M_INTEL_COREI7_ICELAKE): New.
12795 (fold_builtin_cpu): Handle icelake.
12796 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12797 * doc/invoke.texi: Add -march=icelake.
12799 2018-02-02 Julia Koval <julia.koval@intel.com>
12801 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12802 to wide_int_bitmask.
12803 * wide-int-bitmask.h: New.
12805 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12808 * config/i386/i386.md: Replace Pmode with word_mode in
12809 builtin_setjmp_setup and builtin_longjmp to support x32.
12811 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12815 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12817 (rs6000_supported_cpu_names): New static variable.
12818 (linux_cpu_translation_table): Likewise.
12819 (elf_platform) <cpu>: Define new static variable and use it.
12820 Translate kernel AT_PLATFORM name to canonical name if needed.
12821 Error if platform name is unknown.
12823 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12826 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12828 2018-02-01 Jeff Law <law@redhat.com>
12831 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12832 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12833 the scratch if RELEASE_VIA_POP is false.
12834 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12835 If we have to save a temporary register, decrement SIZE appropriately.
12836 Pass new arguments to release_scratch_register_on_entry.
12837 (ix86_adjust_stack_and_probe): Likewise.
12838 (ix86_emit_probe_stack_range): Pass new arguments to
12839 release_scratch_register_on_entry.
12841 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12843 PR rtl-optimization/84157
12844 * combine.c (change_zero_ext): Use REG_P predicate in
12845 front of HARD_REGISTER_P predicate.
12847 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12849 * config/avr/avr.c (avr_option_override): Move disabling of
12850 -fdelete-null-pointer-checks to...
12851 * common/config/avr/avr-common.c (avr_option_optimization_table):
12854 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12856 PR tree-optimization/81635
12857 * tree-data-ref.c (split_constant_offset_1): For types that
12858 wrap on overflow, try to use range info to prove that wrapping
12861 2018-02-01 Renlin Li <renlin.li@arm.com>
12864 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12865 TAILCALL_ADDR_REGS.
12866 (aarch64_register_move_cost): Likewise.
12867 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12868 TAILCALL_ADDR_REGS.
12869 (REG_CLASS_NAMES): Likewise.
12870 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12871 TAILCALL_ADDR_REGS. Remove IP registers.
12872 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12874 2018-02-01 Richard Biener <rguenther@suse.de>
12876 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12877 for specifying RPO order and allow NULL for that.
12878 * domwalk.c (dom_walker::dom_walker): Likewise.
12879 (dom_walker::walk): Handle NULL RPO order.
12880 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12882 (rewrite_update_dom_walker): Likewise.
12883 (mark_def_dom_walker): Likewise.
12885 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12887 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12888 (aarch64_maybe_expand_sve_subreg_move): Declare.
12889 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12890 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12892 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12893 that are semantically a reverse operation.
12894 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12895 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12896 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12898 (aarch64_can_change_mode_class): For big-endian, forbid changes
12899 between two SVE modes if they have different element sizes.
12901 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12903 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12904 the TImode handling for big-endian targets.
12906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12908 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12909 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12911 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12912 Remove BSWAP handing for big-endian targets and use the form of
12913 LD1RQ appropariate for the mode.
12915 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12917 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12918 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12919 duplicated element.
12921 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12924 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12925 check for operands that need to go through aarch64_sve_reload_be.
12927 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12929 PR tree-optimization/81661
12930 PR tree-optimization/84117
12931 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12932 * tree-eh.c: Include gimplify.h.
12933 (find_trapping_overflow, replace_trapping_overflow,
12934 rewrite_to_non_trapping_overflow): New functions.
12935 * tree-vect-loop.c: Include tree-eh.h.
12936 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12937 * tree-data-ref.c: Include tree-eh.h.
12938 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12940 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12942 PR rtl-optimization/84123
12943 * combine.c (change_zero_ext): Check if hard register satisfies
12944 can_change_dest_mode before calling gen_lowpart_SUBREG.
12946 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12949 * ira.c (ira_init_register_move_cost): Remove assert.
12951 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12953 PR rtl-optimization/84071
12954 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12955 * doc/tm.texi: Regenerate.
12957 2018-01-31 Richard Biener <rguenther@suse.de>
12959 PR tree-optimization/84132
12960 * tree-data-ref.c (analyze_miv_subscript): Properly
12961 check whether evolution_function_is_affine_multivariate_p
12962 before calling gcd_of_steps_may_divide_p.
12964 2018-01-31 Julia Koval <julia.koval@intel.com>
12967 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12968 * config/i386/i386.md (rdpid_rex64) New.
12969 (rdpid): Make 32bit only.
12971 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12974 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12975 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12977 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12980 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12982 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12984 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12986 PR rtl-optimization/84071
12987 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12988 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12990 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12992 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12993 (arc_attribute_table): Add 'aux' attribute.
12994 (arc_in_small_data_p): Consider aux like variables.
12995 (arc_is_aux_reg_p): New function.
12996 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12997 (arc_get_aux_arg): New function.
12998 (prepare_move_operands): Handle aux-register access.
12999 (arc_handle_aux_attribute): New function.
13000 * doc/extend.texi (ARC Variable attributes): Add subsection.
13002 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13004 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13005 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13006 (arc_attribute_table): Add 'uncached' attribute.
13007 (arc_print_operand): Print '.di' flag for uncached memory
13009 (arc_in_small_data_p): Do not consider for small data the uncached
13011 (arc_is_uncached_mem_p): New function.
13012 * config/arc/predicates.md (compact_store_memory_operand): Check
13013 for uncached memory accesses.
13014 (nonvol_nonimm_operand): Likewise.
13015 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13017 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13020 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13021 falign-loops=): Add Optimization flag.
13023 2018-01-30 Jeff Law <law@redhat.com>
13026 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13027 INT_REGISTERS_SAVED. Check it prior to calling
13028 get_scratch_register_on_entry.
13029 (ix86_adjust_stack_and_probe): Similarly.
13030 (ix86_emit_probe_stack_range): Similarly.
13031 (ix86_expand_prologue): Corresponding changes.
13033 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13036 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13037 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13039 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13042 * lra-constraints.c (curr_insn_transform): Process AND in the
13045 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13047 PR rtl-optimization/83986
13048 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13049 dependence against last_pending_memory_flush in addition to
13050 pending_jump_insns.
13052 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13054 PR tree-optimization/81611
13055 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13058 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13061 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13064 2018-01-30 Richard Biener <rguenther@suse.de>
13065 Jakub Jelinek <jakub@redhat.com>
13067 PR tree-optimization/84111
13068 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13069 inner loops added during recursion, as they don't have up-to-date
13072 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13075 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13076 (can_inline_edge_by_limits_p): ... here.
13077 (can_early_inline_edge_p, check_callers,
13078 update_caller_keys, update_callee_keys, recursive_inlining,
13079 add_new_edges_to_heap, speculation_useful_p,
13080 inline_small_functions,
13081 inline_small_functions, flatten_function,
13082 inline_to_all_callers_1): Update.
13084 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13086 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13089 2018-01-30 Richard Biener <rguenther@suse.de>
13091 PR tree-optimization/83008
13092 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13093 invariant and constant vector uses in stmts when they need
13094 more than one stmt.
13096 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13099 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13100 * configure: Regenerate.
13102 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13104 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13106 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13107 Use gen_rtx_REG rather than gen_lowpart.
13109 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13111 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13112 rather than 0 when creating partial subregs.
13114 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13116 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13119 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13122 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13123 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13124 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13125 flags. This restores the settings used before the 2017-07-24.
13126 Turning off pre increment/decrement/modify allows IVOPTS to
13127 optimize DF/SF loops where the index is an int.
13129 2018-01-29 Richard Biener <rguenther@suse.de>
13130 Kelvin Nilsen <kelvin@gcc.gnu.org>
13133 * tree-vect-stmts.c (vectorizable_call): Don't call
13134 targetm.vectorize_builtin_md_vectorized_function if callee is
13137 2018-01-22 Carl Love <cel@us.ibm.com>
13139 * doc/extend.tex: Fix typo in second arg in
13140 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13142 2018-01-29 Richard Biener <rguenther@suse.de>
13144 PR tree-optimization/84086
13145 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13146 (flush_ssaname_freelist): When SSA names were released reset
13147 the SCEV hash table.
13149 2018-01-29 Richard Biener <rguenther@suse.de>
13151 PR tree-optimization/84057
13152 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13153 removed paths when removing edges.
13155 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13157 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13158 -mfunction-return=@var{choice}.
13160 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13162 PR diagnostic/84034
13163 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13164 Handle CR like TAB.
13165 (layout::print_source_line): Likewise.
13166 (test_get_line_width_without_trailing_whitespace): Add test cases.
13168 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13170 PR middle-end/84040
13171 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13174 2018-01-26 Jim Wilson <jimw@sifive.com>
13176 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13178 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13181 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13183 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13184 and CMP + SUB-immediate -> SUBS.
13186 2018-01-26 Martin Sebor <msebor@redhat.com>
13188 PR tree-optimization/83896
13189 * tree-ssa-strlen.c (get_string_len): Rename...
13190 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13191 Avoid assuming length is constant.
13192 (handle_char_store): Use HOST_WIDE_INT for string length.
13194 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13197 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13198 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13200 2018-01-26 Richard Biener <rguenther@suse.de>
13202 PR rtl-optimization/84003
13203 * dse.c (record_store): Only record redundant stores when
13204 the earlier store aliases at least all accesses the later one does.
13206 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13208 PR rtl-optimization/83985
13209 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13210 REG_CFA_RESTORE insns.
13211 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13212 REG_CFA_RESTORE insns here.
13215 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13216 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13218 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13220 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13221 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13222 (arc_init): Likewise.
13223 (arc_override_options): Likewise.
13224 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13226 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13228 * config/arc/arc.h (TARGET_DBNZ): Define.
13229 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13230 properly set the tune attribute.
13231 (dbnz): Use TARGET_DBNZ guard.
13232 * config/arc/arc.opt (mtune): Add core3 option.
13234 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13236 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13237 recognize new pic like addresses.
13238 (arc_delegitimize_address): Clean up.
13240 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13242 * config/arc/arc-arches.def: Option mrf16 valid for all
13244 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13245 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13246 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13247 * config/arc/arc-tables.opt: Regenerate.
13248 * config/arc/arc.c (arc_conditional_register_usage): Handle
13249 reduced register file case.
13250 (arc_file_start): Set must have build attributes.
13251 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13252 mrf16 option value.
13253 * config/arc/arc.opt (mrf16): Add new option.
13254 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13255 * config/arc/genmultilib.awk: Handle new mrf16 option.
13256 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13257 * config/arc/t-multilib: Regenerate.
13258 * doc/invoke.texi (ARC Options): Document mrf16 option.
13260 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13262 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13263 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13264 (arc_attribute_table): Add 'secure_call' attribute.
13265 (arc_print_operand): Print secure call operand.
13266 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13268 (arc_is_secure_call_p): New function. * config/arc/arc.md
13269 (call_i): Add support for sjli instruction.
13270 (call_value_i): Likewise.
13271 * config/arc/constraints.md (Csc): New constraint.
13273 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13274 John Eric Martin <John.Martin@emmicro-us.com>
13276 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13277 * config/arc/arc.c (_arc_jli_section): New struct.
13278 (arc_jli_section): New type.
13279 (rc_jli_sections): New static variable.
13280 (arc_handle_jli_attribute): New function.
13281 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13282 (arc_file_end): New function.
13283 (TARGET_ASM_FILE_END): Define.
13284 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13285 (arc_add_jli_section): New function.
13286 (jli_call_scan): Likewise.
13287 (arc_reorg): Call jli_call_scan.
13288 (arc_output_addsi): Remove 'S' from printing asm operand.
13289 (arc_is_jli_call_p): New function.
13290 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13292 (movhi_insn): Likewise.
13293 (movsi_insn): Likewise.
13294 (movsi_set_cc_insn): Likewise.
13295 (loadqi_update): Likewise.
13296 (load_zeroextendqisi_update): Likewise.
13297 (load_signextendqisi_update): Likewise.
13298 (loadhi_update): Likewise.
13299 (load_zeroextendhisi_update): Likewise.
13300 (load_signextendhisi_update): Likewise.
13301 (loadsi_update): Likewise.
13302 (loadsf_update): Likewise.
13303 (movsicc_insn): Likewise.
13304 (bset_insn): Likewise.
13305 (bxor_insn): Likewise.
13306 (bclr_insn): Likewise.
13307 (bmsk_insn): Likewise.
13308 (bicsi3_insn): Likewise.
13309 (cmpsi_cc_c_insn): Likewise.
13310 (movsi_ne): Likewise.
13311 (movsi_cond_exec): Likewise.
13312 (clrsbsi2): Likewise.
13313 (norm_f): Likewise.
13323 (call_i): Remove 'S' asm letter, add jli instruction.
13324 (call_value_i): Likewise.
13325 * config/arc/arc.op (mjli-always): New option.
13326 * config/arc/constraints.md (Cji): New constraint.
13327 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13329 (subsf3_fpx): Likewise.
13330 (mulsf3_fpx): Likewise.
13331 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13333 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13334 function attrbutes.
13335 * doc/invoke.texi (ARC): Document mjli-always option.
13337 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13339 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13340 avoid addition with 0 and use incw and decw where possible.
13342 2018-01-26 Richard Biener <rguenther@suse.de>
13344 PR tree-optimization/81082
13345 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13346 association if it requires casting to unsigned.
13347 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13348 from fold_plusminus_mult_expr to catch important cases late when
13349 range info is available.
13351 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13353 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13354 * configure.ac (hidden_linkonce): New test.
13355 * configure: Regenerate.
13356 * config.in: Regenerate.
13358 2018-01-26 Julia Koval <julia.koval@intel.com>
13360 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13361 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13362 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13363 _mm_mask_bitshuffle_epi64_mask): Fix type.
13364 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13365 USI_FTYPE_V4DI_V4DI_USI): Remove.
13366 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13367 __builtin_ia32_vpshufbitqmb256_mask,
13368 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13369 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13370 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13372 2018-01-26 Alan Modra <amodra@gmail.com>
13375 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13376 UNSPEC_VBPERMQ. Sort other unspecs.
13378 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13380 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13382 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13384 PR middle-end/83055
13385 * predict.c (drop_profile): Do not push/pop cfun; update also
13387 (handle_missing_profiles): Fix logic looking for zero profiles.
13389 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13391 PR middle-end/83977
13392 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13393 on functions with #pragma omp declare simd or functions with simd
13395 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13396 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13397 Remove trailing \n from warning_at calls.
13399 2018-01-25 Tom de Vries <tom@codesourcery.com>
13402 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13403 for neutered workers.
13405 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13408 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13409 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13411 2018-01-24 Jeff Law <law@redhat.com>
13414 * i386.c (get_probe_interval): Move to earlier point.
13415 (ix86_compute_frame_layout): If -fstack-clash-protection and
13416 the frame is larger than the probe interval, then use pushes
13417 to save registers rather than reg->mem moves.
13418 (ix86_expand_prologue): Remove conditional for int_registers_saved
13421 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13424 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13425 min/max for never referenced object.
13427 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13429 PR middle-end/83977
13430 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13432 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13433 attributes from DECL_ATTRIBUTES (decl) without affecting
13434 DECL_ATTRIBUTES (current_function_decl).
13435 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13436 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13438 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13440 PR tree-optimization/83979
13441 * fold-const.c (fold_comparison): Use constant_boolean_node
13442 instead of boolean_{true,false}_node.
13444 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13446 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13449 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13451 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13452 Simplify the clause that sets the length attribute.
13453 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13454 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13455 clause that sets the length attribute.
13456 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13458 2018-01-24 Tom de Vries <tom@codesourcery.com>
13461 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13462 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13463 Add strict parameter.
13464 (prevent_branch_around_nothing): Insert dummy insn between branch to
13465 label and label with no ptx insn inbetween.
13466 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13468 2018-01-24 Tom de Vries <tom@codesourcery.com>
13471 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13472 for neutered threads in warp.
13473 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13475 2018-01-24 Richard Biener <rguenther@suse.de>
13477 PR tree-optimization/83176
13478 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13481 2018-01-24 Richard Biener <rguenther@suse.de>
13483 PR tree-optimization/82819
13484 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13485 code generating pluses that are no-ops in the target precision.
13487 2018-01-24 Richard Biener <rguenther@suse.de>
13489 PR middle-end/84000
13490 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13492 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13494 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13495 to merge probabilities.
13496 * predict.c (probably_never_executed): Also mark as cold functions
13497 with global 0 profile and guessed local profile.
13498 * profile-count.c (profile_probability::combine_with_count): New
13500 * profile-count.h (profile_probability::operator*,
13501 profile_probability::operator*=, profile_probability::operator/,
13502 profile_probability::operator/=): Reduce precision to adjusted
13503 and set value to guessed on contradictory divisions.
13504 (profile_probability::combine_with_freq): Remove.
13505 (profile_probability::combine_wiht_count): Declare.
13506 (profile_count::force_nonzero):: Set to adjusted.
13507 (profile_count::probability_in):: Set quality to adjusted.
13508 * tree-ssa-tail-merge.c (replace_block_by): Use
13509 combine_with_count.
13511 2018-01-23 Andrew Waterman <andrew@sifive.com>
13512 Jim Wilson <jimw@sifive.com>
13514 * config/riscv/riscv.c (riscv_stack_boundary): New.
13515 (riscv_option_override): Set riscv_stack_boundary. Handle
13516 riscv_preferred_stack_boundary_arg.
13517 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13518 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13519 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13520 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13521 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13522 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13524 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13527 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13528 of struct ix86_frame.
13529 (ix86_expand_epilogue): Likewise. Add a local variable for
13530 the reg_save_offset field in struct ix86_frame.
13532 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13534 PR tree-optimization/82604
13535 * tree-loop-distribution.c (enum partition_kind): New enum item
13536 PKIND_PARTIAL_MEMSET.
13537 (partition_builtin_p): Support above new enum item.
13538 (generate_code_for_partition): Ditto.
13539 (compute_access_range): Differentiate cases that equality can be
13540 proven at all loops, the innermost loops or no loops.
13541 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13542 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13543 (finalize_partitions, distribute_loop): Don't fuse partition of
13544 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13545 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13546 parloop is enabled.
13548 2018-01-23 Martin Liska <mliska@suse.cz>
13550 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13551 order to ignore the predictor.
13552 (PRED_POLYMORPHIC_CALL): Likewise.
13553 (PRED_RECURSIVE_CALL): Likewise.
13555 2018-01-23 Martin Liska <mliska@suse.cz>
13557 * tree-profile.c (tree_profiling): Print function header to
13558 aware reader which function we are working on.
13559 * value-prof.c (gimple_find_values_to_profile): Do not print
13560 not interesting value histograms.
13562 2018-01-23 Martin Liska <mliska@suse.cz>
13564 * profile-count.h (enum profile_quality): Add
13565 profile_uninitialized as the first value. Do not number values
13566 as they are zero based.
13567 (profile_count::verify): Update sanity check.
13568 (profile_probability::verify): Likewise.
13570 2018-01-23 Nathan Sidwell <nathan@acm.org>
13572 * doc/invoke.texi (ffor-scope): Deprecate.
13574 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13576 PR tree-optimization/83510
13577 * domwalk.c (set_all_edges_as_executable): New function.
13578 (dom_walker::dom_walker): Convert bool param
13579 "skip_unreachable_blocks" to enum reachability. Move setup of
13580 edge flags to set_all_edges_as_executable and only do it when
13581 reachability is REACHABLE_BLOCKS.
13582 * domwalk.h (enum dom_walker::reachability): New enum.
13583 (dom_walker::dom_walker): Convert bool param
13584 "skip_unreachable_blocks" to enum reachability.
13585 (set_all_edges_as_executable): New decl.
13586 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13587 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13589 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13590 but converting true to REACHABLE_BLOCKS.
13591 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13593 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13594 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13595 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13597 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13598 if check_all_array_refs will be called.
13600 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13602 * tree.c (selftest::test_location_wrappers): Add more test
13605 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13607 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13608 (selftest::test_bit_in_range): Likewise.
13610 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13613 * doc/sourcebuild.texi (vect_float): Say that the selector
13614 only describes the situation when -funsafe-math-optimizations is on.
13615 (vect_float_strict): Document.
13617 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13619 PR tree-optimization/83965
13620 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13621 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13622 instead of checking only for a reduction.
13623 (vect_recog_widen_sum_pattern): Likewise.
13625 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13627 * predict.c (probably_never_executed): Only use precise profile info.
13628 (compute_function_frequency): Skip after inlining hack since we now
13629 have quality checking.
13631 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13633 * profile-count.h (profile_probability::very_unlikely,
13634 profile_probability::unlikely, profile_probability::even): Set
13635 precision to guessed.
13637 2018-01-23 Richard Biener <rguenther@suse.de>
13639 PR tree-optimization/83963
13640 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13641 Properly terminate dominator walk when crossing the exit edge not
13642 when visiting its source block.
13644 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13647 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13648 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13650 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13652 PR tree-optimization/83957
13653 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13654 semicolon after for body surrounded by braces.
13656 PR tree-optimization/83081
13657 * profile-count.h (profile_probability::split): New method.
13658 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13659 Use profile_probability::split.
13660 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13661 when splitting a single conditional jump into 2.
13663 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13665 PR tree-optimization/69452
13666 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13669 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13671 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13672 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13673 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13675 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13677 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13679 * config/rl78/rl78.md (movdi): New define_expand.
13680 * config/rl78/rl78.c (rl78_split_movdi): New function.
13682 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13685 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13687 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13688 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13689 128-bit to produce an UNSPEC move to get the double word with the
13690 signbit and then a shift directly to do signbit.
13691 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13692 implementation with a new version that just does either a direct
13693 move or a regular move. Move memory interface to separate insns.
13694 Move insns so they are next to the expander.
13695 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13696 with signbit move. Split big and little endian case.
13697 (signbit<mode>2_dm_mem_le): Likewise.
13698 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13699 (signbit<mode>2_dm2): Likewise.
13701 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13703 * config/rl78/rl78.md (anddi3): New define_expand.
13705 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13707 * config/rl78/rl78.md (umindi3): New define_expand.
13709 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13711 * config/rl78/rl78.md (smindi3): New define_expand.
13713 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13715 * config/rl78/rl78.md (smaxdi3): New define_expand.
13717 2018-01-22 Carl Love <cel@us.ibm.com>
13719 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13720 LVX_V1TI): Add macro expansion.
13721 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13722 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13723 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13724 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13725 Change check to determine if the instruction is a byte reversing
13726 entry. Fix typo in comment.
13727 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13728 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13729 Add def_builtin calls for new builtins.
13730 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13731 Add define_insn expansion.
13733 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13735 * config/rl78/rl78.md (umaxdi3): New define_expand.
13737 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13739 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13740 for non-QImode registers.
13742 2018-01-22 Richard Biener <rguenther@suse.de>
13744 PR tree-optimization/83963
13745 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13746 including the loop exit block.
13747 (scop_detection::merge_sese): Likewise.
13748 (scop_detection::add_scop): Do it here instead.
13750 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13752 * doc/sourcebuild.texi (arm_softfloat): Document.
13754 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13757 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13758 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13759 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13761 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13762 David Edelsohn <dje.gcc@gmail.com>
13765 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13766 Change "crset eq" to "crset 2".
13767 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13768 (*call_indirect_aix<mode>_nospec): Likewise.
13769 (*call_value_indirect_aix<mode>_nospec): Likewise.
13770 (*call_indirect_elfv2<mode>_nospec): Likewise.
13771 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13772 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13773 change assembly output from . to $.
13774 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13775 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13776 (*tablejump<mode>_internal1_nospec): Likewise.
13778 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13781 * config/sh/sh_optimize_sett_clrt.cc:
13782 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13784 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13786 PR tree-optimization/83940
13787 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13788 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13789 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13790 use it to pass back the definition type.
13791 (vect_check_store_rhs): Likewise rhs_dt_out.
13792 (vect_build_gather_load_calls): Add a mask_dt argument and use
13793 it instead of a call to vect_is_simple_use.
13794 (vectorizable_store): Update calls to vect_check_load_store_mask
13795 and vect_check_store_rhs. Use the dt returned by the latter instead
13796 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13797 instead of calls to vect_is_simple_use. Pass the scalar rather
13798 than the vector operand to vect_is_simple_use when handling
13799 second and subsequent copies of an rhs value.
13800 (vectorizable_load): Update calls to vect_check_load_store_mask
13801 and vect_build_gather_load_calls. Use the cached mask_dt and
13802 gs_info.offset_dt instead of calls to vect_is_simple_use.
13804 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13806 PR middle-end/83945
13807 * tree-emutls.c: Include gimplify.h.
13808 (lower_emutls_2): New function.
13809 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13810 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13811 it before further processing.
13814 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13815 UINTVAL (trueop1) instead of INTVAL (op1).
13817 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13821 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13822 INCOMING_FRAME_SP_OFFSET if not defined.
13823 (scan_trace): Add ENTRY argument. If true and
13824 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13825 emit a note to adjust the CFA offset.
13826 (create_cfi_notes): Adjust scan_trace callers.
13827 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13828 INCOMING_FRAME_SP_OFFSET in the CIE.
13829 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13830 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13832 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13833 * doc/tm.texi: Regenerated.
13835 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13837 PR rtl-optimization/83147
13838 * lra-constraints.c (remove_inheritance_pseudos): Use
13839 lra_substitute_pseudo_within_insn.
13841 2018-01-19 Tom de Vries <tom@codesourcery.com>
13842 Cesar Philippidis <cesar@codesourcery.com>
13845 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13847 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13850 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13851 spaces for function labels.
13853 2018-01-19 Martin Liska <mliska@suse.cz>
13855 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13856 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13857 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13858 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13859 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13860 (PRED_CONST_RETURN): Change from 69 to 65.
13861 (PRED_NULL_RETURN): Change from 91 to 71.
13862 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13863 (PRED_LOOP_GUARD): Change from 66 to 73.
13865 2018-01-19 Martin Liska <mliska@suse.cz>
13867 * predict.c (predict_insn_def): Add new assert.
13868 (struct branch_predictor): Change type to signed integer.
13869 (test_prediction_value_range): Amend test to cover
13870 PROB_UNINITIALIZED.
13871 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13872 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13873 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13874 (PRED_LOOP_IV_COMPARE): Likewise.
13875 * predict.h (PROB_UNINITIALIZED): Define new constant.
13877 2018-01-19 Martin Liska <mliska@suse.cz>
13879 * predict.c (dump_prediction): Add new format for
13880 analyze_brprob.py script which is enabled with -details
13882 * profile-count.h (precise_p): New function.
13884 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13886 PR tree-optimization/83922
13887 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13888 there are no statements that need masking.
13889 (vect_active_double_reduction_p): New function.
13890 (vect_analyze_loop_operations): Use it when handling phis that
13891 are not in the loop header.
13893 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13895 PR tree-optimization/83914
13896 * tree-vect-loop.c (vectorizable_induction): Don't convert
13897 init_expr or apply the peeling adjustment for inductions
13898 that are nested within the vectorized loop.
13900 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13902 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13909 * function.h (gimplify_parameters): Add gimple_seq * argument.
13910 * function.c: Include gimple.h and options.h.
13911 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13912 for the added local temporaries if needed.
13913 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13914 if there are any parameter cleanups, wrap whole body into a
13915 try/finally with the cleanups.
13917 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13920 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13921 Use GET_MODE_CLASS for scalar floating point.
13923 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13927 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13928 Fix call of call_cgraph_insertion_hooks.
13930 2018-01-18 Martin Sebor <msebor@redhat.com>
13932 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13934 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13937 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13940 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13943 * common.opt: (-ffile-prefix-map): New option.
13944 * opts.c (common_handle_option): Defer it.
13945 * opts-global.c (handle_common_deferred_options): Handle it.
13946 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13947 * file-prefix-map.h: New file.
13948 (remap_debug_filename, add_debug_prefix_map): ...here.
13949 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13950 * final.c (debug_prefix_map, add_debug_prefix_map
13951 remap_debug_filename): Move to...
13952 * file-prefix-map.c: New file.
13953 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13954 generalize, get rid of alloca(), use strrchr() instead of strchr().
13955 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13956 Implement in terms of add_prefix_map().
13957 (remap_macro_filename, remap_debug_filename): Implement in term of
13959 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13960 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13961 * dbxout.c: Include file-prefix-map.h.
13962 * varasm.c: Likewise.
13963 * vmsdbgout.c: Likewise.
13964 * xcoffout.c: Likewise.
13965 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13966 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13967 * doc/invoke.texi (-ffile-prefix-map): Document.
13968 (-fdebug-prefix-map): Update description.
13970 2018-01-18 Martin Liska <mliska@suse.cz>
13972 * config/i386/i386.c (indirect_thunk_name): Document that also
13974 (output_indirect_thunk): Document why both instructions
13975 (pause and lfence) are generated.
13977 2018-01-18 Richard Biener <rguenther@suse.de>
13979 PR tree-optimization/83887
13980 * graphite-scop-detection.c
13981 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13982 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13983 (scop_detection::merge_sese): Re-implement with a flood-fill
13984 algorithm that properly finds a SESE region if it exists.
13986 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13989 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13990 pointer_diff optimizations use view_convert instead of convert.
13992 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13994 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13995 Generate different code for -mno-speculate-indirect-jumps.
13996 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13997 (*call_indirect_aix<mode>): Disable for
13998 -mno-speculate-indirect-jumps.
13999 (*call_indirect_aix<mode>_nospec): New define_insn.
14000 (*call_value_indirect_aix<mode>): Disable for
14001 -mno-speculate-indirect-jumps.
14002 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14003 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14004 -mno-speculate-indirect-jumps.
14005 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14007 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14009 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14010 long double type, set the flags for noting the default long double
14011 type, even if we don't pass or return a long double type.
14013 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14016 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14019 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14021 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14022 support for merge[hl].
14023 (fold_mergehl_helper): New helper function.
14024 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14025 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14026 (altivec_vmrglw_direct): Add xxmrglw insn.
14028 2018-01-17 Andrew Waterman <andrew@sifive.com>
14030 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14031 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14033 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14036 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14037 call the lto_location_cache before reading the
14038 DECL_SOURCE_LOCATION of the types.
14040 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14041 Richard Sandiford <richard.sandiford@linaro.org>
14043 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14044 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14045 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14046 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14047 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14049 * config/aarch64/constraints.md (aarch64_movti_operand):
14051 * config/aarch64/predicates.md (Uti): Add new constraint.
14053 2018-01-17 Carl Love <cel@us.ibm.com>
14055 * config/rs6000/vsx.md (define_expand xl_len_r,
14056 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14057 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14059 (define_expand, define_insn): Move the shift left from the
14060 define_insn to the define_expand for lxvl and stxvl instructions.
14061 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14062 and XL_LEN_R definitions to PURE.
14064 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14066 * config/i386/i386.c (indirect_thunk_name): Declare regno
14067 as unsigned int. Compare regno with INVALID_REGNUM.
14068 (output_indirect_thunk): Ditto.
14069 (output_indirect_thunk_function): Ditto.
14070 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14071 in the call to output_indirect_thunk_function.
14073 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14075 PR middle-end/83884
14076 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14077 rather than the size of inner_type to determine the stack slot size
14078 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14080 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14083 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14086 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14088 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14089 endian Linux systems to optionally enable multilibs for selecting
14090 the long double type if the user configured an explicit type.
14091 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14092 have no long double multilibs if not defined.
14093 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14094 warn if the user used -mabi={ieee,ibm}longdouble and we built
14095 multilibs for long double.
14096 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14097 appropriate multilib option.
14098 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14100 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14101 for building long double multilibs.
14102 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14104 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14106 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14109 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14111 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14114 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14117 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14120 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14122 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14123 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14125 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14127 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14128 different rtl trees depending on TARGET_64BIT.
14129 (rs6000_gen_lvx): Likewise.
14131 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14133 * config/visium/visium.md (nop): Tweak comment.
14134 (hazard_nop): Likewise.
14136 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14138 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14139 -mspeculate-indirect-jumps.
14140 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14141 for -mno-speculate-indirect-jumps.
14142 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14143 (*call_value_indirect_elfv2<mode>): Disable for
14144 -mno-speculate-indirect-jumps.
14145 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14146 (indirect_jump): Emit different RTL for
14147 -mno-speculate-indirect-jumps.
14148 (*indirect_jump<mode>): Disable for
14149 -mno-speculate-indirect-jumps.
14150 (*indirect_jump<mode>_nospec): New define_insn.
14151 (tablejump): Emit different RTL for
14152 -mno-speculate-indirect-jumps.
14153 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14154 (tablejumpsi_nospec): New define_expand.
14155 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14156 (tablejumpdi_nospec): New define_expand.
14157 (*tablejump<mode>_internal1): Disable for
14158 -mno-speculate-indirect-jumps.
14159 (*tablejump<mode>_internal1_nospec): New define_insn.
14160 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14163 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14165 * caller-save.c (insert_save): Drop unnecessary parameter. All
14168 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14169 Richard Biener <rguenth@suse.de>
14172 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14173 return early, inline manually is_gimple_sizepos. Make sure if we
14174 call gimplify_expr we don't end up with a gimple constant.
14175 * tree.c (variably_modified_type_p): Don't return true for
14176 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14177 * gimplify.h (is_gimple_sizepos): Remove.
14179 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14181 PR tree-optimization/83857
14182 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14183 vectorizable_live_operation for pure SLP statements.
14184 (vectorizable_live_operation): Handle PHIs.
14186 2018-01-16 Richard Biener <rguenther@suse.de>
14188 PR tree-optimization/83867
14189 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14190 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14192 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14195 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14196 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14197 If off is not INTEGER_CST, issue a may not be aligned warning
14198 rather than isn't aligned. Use isn%'t rather than isn't.
14199 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14201 <case MULT_EXPR>: Improve the case when bottom and one of the
14202 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14203 operand, in that case check if the other operand is multiple of
14204 bottom divided by the INTEGER_CST operand.
14206 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14209 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14210 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14211 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14212 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14213 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14214 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14215 (pa_function_arg_size): New function.
14217 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14219 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14220 in a separate statement.
14222 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14224 PR tree-optimization/83847
14225 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14226 group gathers and scatters.
14228 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14230 PR rtl-optimization/86620
14231 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14233 PR rtl-optimization/83213
14234 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14235 to last if both are JUMP_INSNs.
14237 PR tree-optimization/83843
14238 * gimple-ssa-store-merging.c
14239 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14240 store_immediate_info for bswap/nop orig_stores.
14242 2018-01-15 Andrew Waterman <andrew@sifive.com>
14244 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14246 <UDIV>: Increase cost if !TARGET_DIV.
14248 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14250 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14251 (define_attr "cr_logical_3op"): New.
14252 (cceq_ior_compare): Adjust.
14253 (cceq_ior_compare_complement): Adjust.
14254 (*cceq_rev_compare): Adjust.
14255 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14256 (is_cracked_insn): Adjust.
14257 (insn_must_be_first_in_group): Adjust.
14258 * config/rs6000/40x.md: Adjust.
14259 * config/rs6000/440.md: Adjust.
14260 * config/rs6000/476.md: Adjust.
14261 * config/rs6000/601.md: Adjust.
14262 * config/rs6000/603.md: Adjust.
14263 * config/rs6000/6xx.md: Adjust.
14264 * config/rs6000/7450.md: Adjust.
14265 * config/rs6000/7xx.md: Adjust.
14266 * config/rs6000/8540.md: Adjust.
14267 * config/rs6000/cell.md: Adjust.
14268 * config/rs6000/e300c2c3.md: Adjust.
14269 * config/rs6000/e500mc.md: Adjust.
14270 * config/rs6000/e500mc64.md: Adjust.
14271 * config/rs6000/e5500.md: Adjust.
14272 * config/rs6000/e6500.md: Adjust.
14273 * config/rs6000/mpc.md: Adjust.
14274 * config/rs6000/power4.md: Adjust.
14275 * config/rs6000/power5.md: Adjust.
14276 * config/rs6000/power6.md: Adjust.
14277 * config/rs6000/power7.md: Adjust.
14278 * config/rs6000/power8.md: Adjust.
14279 * config/rs6000/power9.md: Adjust.
14280 * config/rs6000/rs64.md: Adjust.
14281 * config/rs6000/titan.md: Adjust.
14283 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14285 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14286 ix86_indirect_branch_register logic.
14288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14290 * config/i386/constraints.md (Bs): Update
14291 ix86_indirect_branch_register check. Don't check
14292 ix86_indirect_branch_register with GOT_memory_operand.
14294 * config/i386/predicates.md (GOT_memory_operand): Don't check
14295 ix86_indirect_branch_register here.
14296 (GOT32_symbol_operand): Likewise.
14298 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14300 * config/i386/predicates.md (constant_call_address_operand):
14301 Rewrite ix86_indirect_branch_register logic.
14302 (sibcall_insn_operand): Likewise.
14304 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14306 * config/i386/constraints.md (Bs): Replace
14307 ix86_indirect_branch_thunk_register with
14308 ix86_indirect_branch_register.
14310 * config/i386/i386.md (indirect_jump): Likewise.
14311 (tablejump): Likewise.
14312 (*sibcall_memory): Likewise.
14313 (*sibcall_value_memory): Likewise.
14314 Peepholes of indirect call and jump via memory: Likewise.
14315 * config/i386/i386.opt: Likewise.
14316 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14317 (GOT_memory_operand): Likewise.
14318 (call_insn_operand): Likewise.
14319 (sibcall_insn_operand): Likewise.
14320 (GOT32_symbol_operand): Likewise.
14322 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14324 PR middle-end/83837
14325 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14326 type rather than type addr's type points to.
14327 (expand_omp_atomic_mutex): Likewise.
14328 (expand_omp_atomic): Likewise.
14330 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14333 * config/i386/i386.c (output_indirect_thunk_function): Use
14334 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14335 for __x86_return_thunk.
14337 2018-01-15 Richard Biener <rguenther@suse.de>
14339 PR middle-end/83850
14340 * expmed.c (extract_bit_field_1): Fix typo.
14342 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14345 * config/arm/iterators.md (VF): New mode iterator.
14346 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14347 Remove integer-related logic from pattern.
14348 (neon_vabd<mode>_3): Likewise.
14350 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14352 PR middle-end/82694
14353 * common.opt (fstrict-overflow): No longer an alias.
14354 (fwrapv-pointer): New option.
14355 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14356 also for pointer types based on flag_wrapv_pointer.
14357 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14358 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14359 opts->x_flag_wrapv got set.
14360 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14361 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14362 POINTER_TYPE_OVERFLOW_UNDEFINED.
14363 * match.pd: Likewise in address comparison pattern.
14364 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14366 2018-01-15 Richard Biener <rguenther@suse.de>
14369 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14370 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14371 Reset type names to their identifier if their TYPE_DECL doesn't
14372 have linkage (and thus is used for ODR and devirt).
14373 (save_debug_info_for_decl): Remove.
14374 (save_debug_info_for_type): Likewise.
14375 (add_tree_to_fld_list): Adjust.
14376 * tree-pretty-print.c (dump_generic_node): Make dumping of
14377 type names more robust.
14379 2018-01-15 Richard Biener <rguenther@suse.de>
14381 * BASE-VER: Bump to 8.0.1.
14383 2018-01-14 Martin Sebor <msebor@redhat.com>
14386 * builtins.c (check_access): Avoid warning when the no-warning bit
14389 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14391 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14392 * ira-color (allocno_hard_regs_compare): Likewise.
14394 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14397 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14398 Use .pushsection/.popsection.
14400 2018-01-14 Martin Sebor <msebor@redhat.com>
14403 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14405 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14407 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14408 entry from extra_headers.
14409 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14410 extra_headers, make the list bitwise identical to the i?86-*-* one.
14412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14414 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14415 -mcmodel=large with -mindirect-branch=thunk,
14416 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14417 -mfunction-return=thunk-extern.
14418 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14419 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14420 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14422 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14424 * config/i386/i386.c (print_reg): Print the name of the full
14425 integer register without '%'.
14426 (ix86_print_operand): Handle 'V'.
14427 * doc/extend.texi: Document 'V' modifier.
14429 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14431 * config/i386/constraints.md (Bs): Disallow memory operand for
14432 -mindirect-branch-register.
14434 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14435 (GOT_memory_operand): Likewise.
14436 (call_insn_operand): Likewise.
14437 (sibcall_insn_operand): Likewise.
14438 (GOT32_symbol_operand): Likewise.
14439 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14440 for -mindirect-branch-register.
14441 (tablejump): Likewise.
14442 (*sibcall_memory): Likewise.
14443 (*sibcall_value_memory): Likewise.
14444 Disallow peepholes of indirect call and jump via memory for
14445 -mindirect-branch-register.
14446 (*call_pop): Replace m with Bw.
14447 (*call_value_pop): Likewise.
14448 (*sibcall_pop_memory): Replace m with Bs.
14449 * config/i386/i386.opt (mindirect-branch-register): New option.
14450 * doc/invoke.texi: Document -mindirect-branch-register option.
14452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14454 * config/i386/i386-protos.h (ix86_output_function_return): New.
14455 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14456 set function_return_type.
14457 (indirect_thunk_name): Add ret_p to indicate thunk for function
14459 (output_indirect_thunk_function): Pass false to
14460 indirect_thunk_name.
14461 (ix86_output_indirect_branch_via_reg): Likewise.
14462 (ix86_output_indirect_branch_via_push): Likewise.
14463 (output_indirect_thunk_function): Create alias for function
14464 return thunk if regno < 0.
14465 (ix86_output_function_return): New function.
14466 (ix86_handle_fndecl_attribute): Handle function_return.
14467 (ix86_attribute_table): Add function_return.
14468 * config/i386/i386.h (machine_function): Add
14469 function_return_type.
14470 * config/i386/i386.md (simple_return_internal): Use
14471 ix86_output_function_return.
14472 (simple_return_internal_long): Likewise.
14473 * config/i386/i386.opt (mfunction-return=): New option.
14474 (indirect_branch): Mention -mfunction-return=.
14475 * doc/extend.texi: Document function_return function attribute.
14476 * doc/invoke.texi: Document -mfunction-return= option.
14478 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14480 * config/i386/i386-opts.h (indirect_branch): New.
14481 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14482 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14483 with local indirect jump when converting indirect call and jump.
14484 (ix86_set_indirect_branch_type): New.
14485 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14486 (indirectlabelno): New.
14487 (indirect_thunk_needed): Likewise.
14488 (indirect_thunk_bnd_needed): Likewise.
14489 (indirect_thunks_used): Likewise.
14490 (indirect_thunks_bnd_used): Likewise.
14491 (INDIRECT_LABEL): Likewise.
14492 (indirect_thunk_name): Likewise.
14493 (output_indirect_thunk): Likewise.
14494 (output_indirect_thunk_function): Likewise.
14495 (ix86_output_indirect_branch_via_reg): Likewise.
14496 (ix86_output_indirect_branch_via_push): Likewise.
14497 (ix86_output_indirect_branch): Likewise.
14498 (ix86_output_indirect_jmp): Likewise.
14499 (ix86_code_end): Call output_indirect_thunk_function if needed.
14500 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14502 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14503 (ix86_attribute_table): Add indirect_branch.
14504 * config/i386/i386.h (machine_function): Add indirect_branch_type
14505 and has_local_indirect_jump.
14506 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14508 (tablejump): Likewise.
14509 (*indirect_jump): Use ix86_output_indirect_jmp.
14510 (*tablejump_1): Likewise.
14511 (simple_return_indirect_internal): Likewise.
14512 * config/i386/i386.opt (mindirect-branch=): New option.
14513 (indirect_branch): New.
14516 (thunk-inline): Likewise.
14517 (thunk-extern): Likewise.
14518 * doc/extend.texi: Document indirect_branch function attribute.
14519 * doc/invoke.texi: Document -mindirect-branch= option.
14521 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14524 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14526 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14528 * ipa-inline.c (want_inline_small_function_p): Return false if
14529 inlining has already failed with CIF_FINAL_ERROR.
14530 (update_caller_keys): Call want_inline_small_function_p before
14532 (update_callee_keys): Likewise.
14534 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14536 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14538 (rs6000_quadword_masked_address_p): Likewise.
14539 (quad_aligned_load_p): Likewise.
14540 (quad_aligned_store_p): Likewise.
14541 (const_load_sequence_p): Add comment to describe the outer-most loop.
14542 (mimic_memory_attributes_and_flags): New function.
14543 (rs6000_gen_stvx): Likewise.
14544 (replace_swapped_aligned_store): Likewise.
14545 (rs6000_gen_lvx): Likewise.
14546 (replace_swapped_aligned_load): Likewise.
14547 (replace_swapped_load_constant): Capitalize argument name in
14548 comment describing this function.
14549 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14550 and stores that access quad-word aligned addresses and replace
14551 with stvx or lvx instructions when appropriate.
14552 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14553 New function prototype.
14554 (rs6000_quadword_masked_address_p): Likewise.
14555 (rs6000_gen_lvx): Likewise.
14556 (rs6000_gen_stvx): Likewise.
14557 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14558 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14559 when memory address is aligned.
14560 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14561 this split to select lvx instruction when memory address is aligned.
14562 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14563 instruction when memory address is aligned.
14564 (*vsx_le_perm_load_v16qi): Likewise.
14565 (four unnamed splitters): Modify to select the stvx instruction
14566 when memory is aligned.
14568 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14570 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14571 which appears in the queue multiple times.
14573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14574 Alan Hayward <alan.hayward@arm.com>
14575 David Sherwood <david.sherwood@arm.com>
14577 * tree-vectorizer.h (vec_lower_bound): New structure.
14578 (_loop_vec_info): Add check_nonzero and lower_bounds.
14579 (LOOP_VINFO_CHECK_NONZERO): New macro.
14580 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14581 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14582 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14583 fields. Make seg_len the distance travelled, not including the
14585 (dr_direction_indicator): Declare.
14586 (dr_zero_step_indicator): Likewise.
14587 (dr_known_forward_stride_p): Likewise.
14588 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14590 (runtime_alias_check_p): Allow runtime alias checks with
14592 (operator ==): Compare access_size and align.
14593 (prune_runtime_alias_test_list): Rework for new distinction between
14594 the access_size and seg_len.
14595 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14597 (get_segment_min_max): New function.
14598 (create_intersect_range_checks): Use it.
14599 (dr_step_indicator): New function.
14600 (dr_direction_indicator): Likewise.
14601 (dr_zero_step_indicator): Likewise.
14602 (dr_known_forward_stride_p): Likewise.
14603 * tree-loop-distribution.c (data_ref_segment_size): Return
14604 DR_STEP * (niters - 1).
14605 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14607 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14608 (vect_preserves_scalar_order_p): New function, split out from...
14609 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14610 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14611 (vect_vfa_access_size): New function.
14612 (vect_vfa_align): Likewise.
14613 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14614 (dump_lower_bound): New function.
14615 (vect_check_lower_bound): Likewise.
14616 (vect_small_gap_p): Likewise.
14617 (vectorizable_with_step_bound_p): Likewise.
14618 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14619 depencies if the vectorization factor is 1. Convert the checks
14620 for nonzero steps into checks on the bounds of DR_STEP. Try using
14621 a bunds check for variable steps if the minimum required step is
14622 relatively small. Update calls to the dr_with_seg_len
14623 constructor and to vect_compile_time_alias.
14624 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14626 (vect_loop_versioning): Call it.
14627 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14629 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14632 Alan Hayward <alan.hayward@arm.com>
14633 David Sherwood <david.sherwood@arm.com>
14635 * doc/sourcebuild.texi (vect_scatter_store): Document.
14636 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14638 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14640 * genopinit.c (main): Add supports_vec_scatter_store and
14641 supports_vec_scatter_store_cached to target_optabs.
14642 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14643 IFN_MASK_SCATTER_STORE.
14644 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14646 * internal-fn.h (internal_store_fn_p): Declare.
14647 (internal_fn_stored_value_index): Likewise.
14648 * internal-fn.c (scatter_store_direct): New macro.
14649 (expand_scatter_store_optab_fn): New function.
14650 (direct_scatter_store_optab_supported_p): New macro.
14651 (internal_store_fn_p): New function.
14652 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14653 IFN_MASK_SCATTER_STORE.
14654 (internal_fn_mask_index): Likewise.
14655 (internal_fn_stored_value_index): New function.
14656 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14657 for scatter stores.
14658 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14659 * optabs-query.c (supports_vec_scatter_store_p): New function.
14660 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14661 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14662 true for scatter stores.
14663 (vect_gather_scatter_fn_p): Handle scatter stores too.
14664 (vect_check_gather_scatter): Consider using scatter stores if
14665 supports_vec_scatter_store_p.
14666 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14667 scatter stores too.
14668 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14669 internal_fn_stored_value_index.
14670 (check_load_store_masking): Handle scatter stores too.
14671 (vect_get_store_rhs): Make public.
14672 (vectorizable_call): Use internal_store_fn_p.
14673 (vectorizable_store): Handle scatter store internal functions.
14674 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14675 when deciding whether the end of the group has been reached.
14676 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14677 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14678 (mask_scatter_store<mode>): New insns.
14680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14681 Alan Hayward <alan.hayward@arm.com>
14682 David Sherwood <david.sherwood@arm.com>
14684 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14685 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14686 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14688 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14689 Use vect_truncate_gather_scatter_offset if we can't treat the
14690 operation as a normal gather load or scatter store.
14691 (get_group_load_store_type): Take the gather_scatter_info
14692 as argument. Try using a gather load or scatter store for
14693 single-element groups.
14694 (get_load_store_type): Update calls to get_group_load_store_type
14695 and vect_use_strided_gather_scatters_p.
14697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14698 Alan Hayward <alan.hayward@arm.com>
14699 David Sherwood <david.sherwood@arm.com>
14701 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14702 optional tree argument.
14703 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14705 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14706 but continue to use the current value as a fallback.
14707 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14708 to compare the updates.
14709 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14710 (get_load_store_type): Use it when handling a strided access.
14711 (vect_get_strided_load_store_ops): New function.
14712 (vect_get_data_ptr_increment): Likewise.
14713 (vectorizable_load): Handle strided gather loads. Always pass
14714 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14717 Alan Hayward <alan.hayward@arm.com>
14718 David Sherwood <david.sherwood@arm.com>
14720 * doc/md.texi (gather_load@var{m}): Document.
14721 (mask_gather_load@var{m}): Likewise.
14722 * genopinit.c (main): Add supports_vec_gather_load and
14723 supports_vec_gather_load_cached to target_optabs.
14724 * optabs-tree.c (init_tree_optimization_optabs): Use
14725 ggc_cleared_alloc to allocate target_optabs.
14726 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14727 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14729 * internal-fn.h (internal_load_fn_p): Declare.
14730 (internal_gather_scatter_fn_p): Likewise.
14731 (internal_fn_mask_index): Likewise.
14732 (internal_gather_scatter_fn_supported_p): Likewise.
14733 * internal-fn.c (gather_load_direct): New macro.
14734 (expand_gather_load_optab_fn): New function.
14735 (direct_gather_load_optab_supported_p): New macro.
14736 (direct_internal_fn_optab): New function.
14737 (internal_load_fn_p): Likewise.
14738 (internal_gather_scatter_fn_p): Likewise.
14739 (internal_fn_mask_index): Likewise.
14740 (internal_gather_scatter_fn_supported_p): Likewise.
14741 * optabs-query.c (supports_at_least_one_mode_p): New function.
14742 (supports_vec_gather_load_p): Likewise.
14743 * optabs-query.h (supports_vec_gather_load_p): Declare.
14744 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14745 and memory_type field.
14746 (NUM_PATTERNS): Bump to 15.
14747 * tree-vect-data-refs.c: Include internal-fn.h.
14748 (vect_gather_scatter_fn_p): New function.
14749 (vect_describe_gather_scatter_call): Likewise.
14750 (vect_check_gather_scatter): Try using internal functions for
14751 gather loads. Recognize existing calls to a gather load function.
14752 (vect_analyze_data_refs): Consider using gather loads if
14753 supports_vec_gather_load_p.
14754 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14755 (vect_get_gather_scatter_offset_type): Likewise.
14756 (vect_convert_mask_for_vectype): Likewise.
14757 (vect_add_conversion_to_patterm): Likewise.
14758 (vect_try_gather_scatter_pattern): Likewise.
14759 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14760 (vect_vect_recog_func_ptrs): Add it.
14761 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14762 internal_fn_mask_index and internal_gather_scatter_fn_p.
14763 (check_load_store_masking): Take the gather_scatter_info as an
14764 argument and handle gather loads.
14765 (vect_get_gather_scatter_ops): New function.
14766 (vectorizable_call): Check internal_load_fn_p.
14767 (vectorizable_load): Likewise. Handle gather load internal
14769 (vectorizable_store): Update call to check_load_store_masking.
14770 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14771 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14772 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14773 (aarch64_gather_scale_operand_d): New predicates.
14774 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14775 (mask_gather_load<mode>): New insns.
14777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14778 Alan Hayward <alan.hayward@arm.com>
14779 David Sherwood <david.sherwood@arm.com>
14781 * optabs.def (fold_left_plus_optab): New optab.
14782 * doc/md.texi (fold_left_plus_@var{m}): Document.
14783 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14784 * internal-fn.c (fold_left_direct): Define.
14785 (expand_fold_left_optab_fn): Likewise.
14786 (direct_fold_left_optab_supported_p): Likewise.
14787 * fold-const-call.c (fold_const_fold_left): New function.
14788 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14789 * tree-parloops.c (valid_reduction_p): New function.
14790 (gather_scalar_reductions): Use it.
14791 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14792 (vect_finish_replace_stmt): Declare.
14793 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14794 (needs_fold_left_reduction_p): New function, split out from...
14795 (vect_is_simple_reduction): ...here. Accept reductions that
14796 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14797 (vect_force_simple_reduction): Also store the reduction type in
14798 the assignment's STMT_VINFO_REDUC_TYPE.
14799 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14800 (merge_with_identity): New function.
14801 (vect_expand_fold_left): Likewise.
14802 (vectorize_fold_left_reduction): Likewise.
14803 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14804 scalar phi in place for it. Check for target support and reject
14805 cases that would reassociate the operation. Defer the transform
14806 phase to vectorize_fold_left_reduction.
14807 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14808 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14809 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14813 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14814 call to ifc_temp_var.
14816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14817 Alan Hayward <alan.hayward@arm.com>
14818 David Sherwood <david.sherwood@arm.com>
14820 * target.def (legitimize_address_displacement): Take the original
14821 offset as a poly_int.
14822 * targhooks.h (default_legitimize_address_displacement): Update
14824 * targhooks.c (default_legitimize_address_displacement): Likewise.
14825 * doc/tm.texi: Regenerate.
14826 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14827 as an argument, moving assert of ad->disp == ad->disp_term to...
14828 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14829 Try calling targetm.legitimize_address_displacement before expanding
14830 the address rather than afterwards, and adjust for the new interface.
14831 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14832 Match the new hook interface. Handle SVE addresses.
14833 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14834 new hook interface.
14836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14838 * Makefile.in (OBJS): Add early-remat.o.
14839 * target.def (select_early_remat_modes): New hook.
14840 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14841 * doc/tm.texi: Regenerate.
14842 * targhooks.h (default_select_early_remat_modes): Declare.
14843 * targhooks.c (default_select_early_remat_modes): New function.
14844 * timevar.def (TV_EARLY_REMAT): New timevar.
14845 * passes.def (pass_early_remat): New pass.
14846 * tree-pass.h (make_pass_early_remat): Declare.
14847 * early-remat.c: New file.
14848 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14850 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14853 Alan Hayward <alan.hayward@arm.com>
14854 David Sherwood <david.sherwood@arm.com>
14856 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14857 vfm1 with a bound_epilog parameter.
14858 (vect_do_peeling): Update calls accordingly, and move the prologue
14859 call earlier in the function. Treat the base bound_epilog as 0 for
14860 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14861 this base when peeling for gaps.
14862 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14863 with fully-masked loops.
14864 (vect_estimate_min_profitable_iters): Handle the single peeled
14865 iteration in that case.
14867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14868 Alan Hayward <alan.hayward@arm.com>
14869 David Sherwood <david.sherwood@arm.com>
14871 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14872 single-element interleaving even if the size is not a power of 2.
14873 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14874 accesses for single-element interleaving if the group size is
14877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14878 Alan Hayward <alan.hayward@arm.com>
14879 David Sherwood <david.sherwood@arm.com>
14881 * doc/md.texi (fold_extract_last_@var{m}): Document.
14882 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14883 * optabs.def (fold_extract_last_optab): New optab.
14884 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14885 * internal-fn.c (fold_extract_direct): New macro.
14886 (expand_fold_extract_optab_fn): Likewise.
14887 (direct_fold_extract_optab_supported_p): Likewise.
14888 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14889 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14890 EXTRACT_LAST_REDUCTION.
14891 (get_initial_def_for_reduction): Do not create an initial vector
14892 for EXTRACT_LAST_REDUCTION reductions.
14893 (vectorizable_reduction): Leave the scalar phi in place for
14894 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14895 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14896 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14897 transform phase to vectorizable_condition.
14898 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14900 (vect_finish_stmt_generation): ...here.
14901 (vect_finish_replace_stmt): New function.
14902 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14903 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14905 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14908 Alan Hayward <alan.hayward@arm.com>
14909 David Sherwood <david.sherwood@arm.com>
14911 * doc/md.texi (extract_last_@var{m}): Document.
14912 * optabs.def (extract_last_optab): New optab.
14913 * internal-fn.def (EXTRACT_LAST): New internal function.
14914 * internal-fn.c (cond_unary_direct): New macro.
14915 (expand_cond_unary_optab_fn): Likewise.
14916 (direct_cond_unary_optab_supported_p): Likewise.
14917 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14918 loops using EXTRACT_LAST.
14919 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14920 (extract_last_<mode>): ...this optab.
14921 (vec_extract<mode><Vel>): Update accordingly.
14923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14924 Alan Hayward <alan.hayward@arm.com>
14925 David Sherwood <david.sherwood@arm.com>
14927 * target.def (empty_mask_is_expensive): New hook.
14928 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14929 * doc/tm.texi: Regenerate.
14930 * targhooks.h (default_empty_mask_is_expensive): Declare.
14931 * targhooks.c (default_empty_mask_is_expensive): New function.
14932 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14933 if the target says that empty masks are expensive.
14934 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14936 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14939 Alan Hayward <alan.hayward@arm.com>
14940 David Sherwood <david.sherwood@arm.com>
14942 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14943 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14944 (vect_use_loop_mask_for_alignment_p): New function.
14945 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14946 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14947 niters_skip argument. Make sure that the first niters_skip elements
14948 of the first iteration are inactive.
14949 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14950 Update call to vect_set_loop_masks_directly.
14951 (get_misalign_in_elems): New function, split out from...
14952 (vect_gen_prolog_loop_niters): ...here.
14953 (vect_update_init_of_dr): Take a code argument that specifies whether
14954 the adjustment should be added or subtracted.
14955 (vect_update_init_of_drs): Likewise.
14956 (vect_prepare_for_masked_peels): New function.
14957 (vect_do_peeling): Skip prologue peeling if we're using a mask
14958 instead. Update call to vect_update_inits_of_drs.
14959 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14961 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14962 alignment. Do not include the number of peeled iterations in
14963 the minimum threshold in that case.
14964 (vectorizable_induction): Adjust the start value down by
14965 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14966 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14967 Take the number of skipped iterations into account when calculating
14969 * tree-vect-stmts.c (vect_gen_while_not): New function.
14971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14972 Alan Hayward <alan.hayward@arm.com>
14973 David Sherwood <david.sherwood@arm.com>
14975 * doc/sourcebuild.texi (vect_fully_masked): Document.
14976 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14977 default value to 0.
14978 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14980 (vect_analyze_loop_2): ...here. Don't check the vectorization
14981 factor against the number of loop iterations if the loop is
14984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14985 Alan Hayward <alan.hayward@arm.com>
14986 David Sherwood <david.sherwood@arm.com>
14988 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14989 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14990 (dump_groups): Update accordingly.
14991 (iv_use::mem_type): New member variable.
14992 (address_p): New function.
14993 (record_use): Add a mem_type argument and initialize the new
14995 (record_group_use): Add a mem_type argument. Use address_p.
14996 Remove obsolete null checks of base_object. Update call to record_use.
14997 (find_interesting_uses_op): Update call to record_group_use.
14998 (find_interesting_uses_cond): Likewise.
14999 (find_interesting_uses_address): Likewise.
15000 (get_mem_type_for_internal_fn): New function.
15001 (find_address_like_use): Likewise.
15002 (find_interesting_uses_stmt): Try find_address_like_use before
15003 calling find_interesting_uses_op.
15004 (addr_offset_valid_p): Use the iv mem_type field as the type
15005 of the addressed memory.
15006 (add_autoinc_candidates): Likewise.
15007 (get_address_cost): Likewise.
15008 (split_small_address_groups_p): Use address_p.
15009 (split_address_groups): Likewise.
15010 (add_iv_candidate_for_use): Likewise.
15011 (autoinc_possible_for_pair): Likewise.
15012 (rewrite_groups): Likewise.
15013 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15014 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15015 (get_alias_ptr_type_for_ptr_address): New function.
15016 (rewrite_use_address): Rewrite address uses in calls that were
15017 identified by find_address_like_use.
15019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15020 Alan Hayward <alan.hayward@arm.com>
15021 David Sherwood <david.sherwood@arm.com>
15023 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15025 * gimple-expr.h (is_gimple_addressable: Likewise.
15026 * gimple-expr.c (is_gimple_address): Likewise.
15027 * internal-fn.c (expand_call_mem_ref): New function.
15028 (expand_mask_load_optab_fn): Use it.
15029 (expand_mask_store_optab_fn): Likewise.
15031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15032 Alan Hayward <alan.hayward@arm.com>
15033 David Sherwood <david.sherwood@arm.com>
15035 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15036 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15037 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15038 (cond_umax@var{mode}): Document.
15039 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15040 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15041 (cond_umin_optab, cond_umax_optab): New optabs.
15042 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15043 (COND_IOR, COND_XOR): New internal functions.
15044 * internal-fn.h (get_conditional_internal_fn): Declare.
15045 * internal-fn.c (cond_binary_direct): New macro.
15046 (expand_cond_binary_optab_fn): Likewise.
15047 (direct_cond_binary_optab_supported_p): Likewise.
15048 (get_conditional_internal_fn): New function.
15049 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15050 Cope with reduction statements that are vectorized as calls rather
15052 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15053 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15054 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15055 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15056 (UNSPEC_COND_EOR): New unspecs.
15057 (optab): Add mappings for them.
15058 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15059 (sve_int_op, sve_fp_op): New int attributes.
15061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15062 Alan Hayward <alan.hayward@arm.com>
15063 David Sherwood <david.sherwood@arm.com>
15065 * optabs.def (while_ult_optab): New optab.
15066 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15067 * internal-fn.def (WHILE_ULT): New internal function.
15068 * internal-fn.h (direct_internal_fn_supported_p): New override
15069 that takes two types as argument.
15070 * internal-fn.c (while_direct): New macro.
15071 (expand_while_optab_fn): New function.
15072 (convert_optab_supported_p): Likewise.
15073 (direct_while_optab_supported_p): New macro.
15074 * wide-int.h (wi::udiv_ceil): New function.
15075 * tree-vectorizer.h (rgroup_masks): New structure.
15076 (vec_loop_masks): New typedef.
15077 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15078 and fully_masked_p.
15079 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15080 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15081 (vect_max_vf): New function.
15082 (slpeel_make_loop_iterate_ntimes): Delete.
15083 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15084 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15085 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15086 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15087 internal-fn.h, stor-layout.h and optabs-query.h.
15088 (vect_set_loop_mask): New function.
15089 (add_preheader_seq): Likewise.
15090 (add_header_seq): Likewise.
15091 (interleave_supported_p): Likewise.
15092 (vect_maybe_permute_loop_masks): Likewise.
15093 (vect_set_loop_masks_directly): Likewise.
15094 (vect_set_loop_condition_masked): Likewise.
15095 (vect_set_loop_condition_unmasked): New function, split out from
15096 slpeel_make_loop_iterate_ntimes.
15097 (slpeel_make_loop_iterate_ntimes): Rename to..
15098 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15099 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15100 (vect_do_peeling): Update call accordingly.
15101 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15103 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15104 mask_compare_type, can_fully_mask_p and fully_masked_p.
15105 (release_vec_loop_masks): New function.
15106 (_loop_vec_info): Use it to free the loop masks.
15107 (can_produce_all_loop_masks_p): New function.
15108 (vect_get_max_nscalars_per_iter): Likewise.
15109 (vect_verify_full_masking): Likewise.
15110 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15111 retries, and free the mask rgroups before retrying. Check loop-wide
15112 reasons for disallowing fully-masked loops. Make the final decision
15113 about whether use a fully-masked loop or not.
15114 (vect_estimate_min_profitable_iters): Do not assume that peeling
15115 for the number of iterations will be needed for fully-masked loops.
15116 (vectorizable_reduction): Disable fully-masked loops.
15117 (vectorizable_live_operation): Likewise.
15118 (vect_halve_mask_nunits): New function.
15119 (vect_double_mask_nunits): Likewise.
15120 (vect_record_loop_mask): Likewise.
15121 (vect_get_loop_mask): Likewise.
15122 (vect_transform_loop): Handle the case in which the final loop
15123 iteration might handle a partial vector. Call vect_set_loop_condition
15124 instead of slpeel_make_loop_iterate_ntimes.
15125 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15126 (check_load_store_masking): New function.
15127 (prepare_load_store_mask): Likewise.
15128 (vectorizable_store): Handle fully-masked loops.
15129 (vectorizable_load): Likewise.
15130 (supportable_widening_operation): Use vect_halve_mask_nunits for
15132 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15133 (vect_gen_while): New function.
15134 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15135 (aarch64_uqdec<mode>): New insn.
15137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15138 Alan Hayward <alan.hayward@arm.com>
15139 David Sherwood <david.sherwood@arm.com>
15141 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15142 (reduc_xor_scal_optab): New optabs.
15143 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15144 (reduc_xor_scal_@var{m}): Document.
15145 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15146 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15147 internal functions.
15148 * fold-const-call.c (fold_const_call): Handle them.
15149 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15150 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15151 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15152 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15153 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15154 (UNSPEC_XORV): New unspecs.
15155 (optab): Add entries for them.
15156 (BITWISEV): New int iterator.
15157 (bit_reduc_op): New int attributes.
15159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15160 Alan Hayward <alan.hayward@arm.com>
15161 David Sherwood <david.sherwood@arm.com>
15163 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15164 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15165 * optabs.def (vec_shl_insert_optab): New optab.
15166 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15167 (duplicate_and_interleave): Likewise.
15168 * tree-vect-loop.c: Include internal-fn.h.
15169 (neutral_op_for_slp_reduction): New function, split out from
15170 get_initial_defs_for_reduction.
15171 (get_initial_def_for_reduction): Handle option 2 for variable-length
15172 vectors by loading the neutral value into a vector and then shifting
15173 the initial value into element 0.
15174 (get_initial_defs_for_reduction): Replace the code argument with
15175 the neutral value calculated by neutral_op_for_slp_reduction.
15176 Use gimple_build_vector for constant-length vectors.
15177 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15178 but the first group_size elements have a neutral value.
15179 Use duplicate_and_interleave otherwise.
15180 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15181 Update call to get_initial_defs_for_reduction. Handle SLP
15182 reductions for variable-length vectors by creating one vector
15183 result for each scalar result, with the elements associated
15184 with other scalar results stubbed out with the neutral value.
15185 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15186 Require IFN_VEC_SHL_INSERT for double reductions on
15187 variable-length vectors, or SLP reductions that have
15188 a neutral value. Require can_duplicate_and_interleave_p
15189 support for variable-length unchained SLP reductions if there
15190 is no neutral value, such as for MIN/MAX reductions. Also require
15191 the number of vector elements to be a multiple of the number of
15192 SLP statements when doing variable-length unchained SLP reductions.
15193 Update call to vect_create_epilog_for_reduction.
15194 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15195 and remove initial values.
15196 (duplicate_and_interleave): Make public.
15197 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15198 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15201 Alan Hayward <alan.hayward@arm.com>
15202 David Sherwood <david.sherwood@arm.com>
15204 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15205 (can_duplicate_and_interleave_p): New function.
15206 (vect_get_and_check_slp_defs): Take the vector of statements
15207 rather than just the current one. Remove excess parentheses.
15208 Restriction rejectinon of vect_constant_def and vect_external_def
15209 for variable-length vectors to boolean types, or types for which
15210 can_duplicate_and_interleave_p is false.
15211 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15212 (duplicate_and_interleave): New function.
15213 (vect_get_constant_vectors): Use gimple_build_vector for
15214 constant-length vectors and suitable variable-length constant
15215 vectors. Use duplicate_and_interleave for other variable-length
15216 vectors. Don't defer the update when inserting new statements.
15218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15219 Alan Hayward <alan.hayward@arm.com>
15220 David Sherwood <david.sherwood@arm.com>
15222 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15223 min_profitable_iters doesn't go negative.
15225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15226 Alan Hayward <alan.hayward@arm.com>
15227 David Sherwood <david.sherwood@arm.com>
15229 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15230 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15231 * optabs.def (vec_mask_load_lanes_optab): New optab.
15232 (vec_mask_store_lanes_optab): Likewise.
15233 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15234 (MASK_STORE_LANES): Likewise.
15235 * internal-fn.c (mask_load_lanes_direct): New macro.
15236 (mask_store_lanes_direct): Likewise.
15237 (expand_mask_load_optab_fn): Handle masked operations.
15238 (expand_mask_load_lanes_optab_fn): New macro.
15239 (expand_mask_store_optab_fn): Handle masked operations.
15240 (expand_mask_store_lanes_optab_fn): New macro.
15241 (direct_mask_load_lanes_optab_supported_p): Likewise.
15242 (direct_mask_store_lanes_optab_supported_p): Likewise.
15243 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15245 (vect_load_lanes_supported): Likewise.
15246 * tree-vect-data-refs.c (strip_conversion): New function.
15247 (can_group_stmts_p): Likewise.
15248 (vect_analyze_data_ref_accesses): Use it instead of checking
15249 for a pair of assignments.
15250 (vect_store_lanes_supported): Take a masked_p parameter.
15251 (vect_load_lanes_supported): Likewise.
15252 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15253 vect_store_lanes_supported and vect_load_lanes_supported.
15254 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15255 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15256 parameter. Don't allow gaps for masked accesses.
15257 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15258 and vect_load_lanes_supported.
15259 (get_load_store_type): Take a masked_p parameter and update
15260 call to get_group_load_store_type.
15261 (vectorizable_store): Update call to get_load_store_type.
15262 Handle IFN_MASK_STORE_LANES.
15263 (vectorizable_load): Update call to get_load_store_type.
15264 Handle IFN_MASK_LOAD_LANES.
15266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15267 Alan Hayward <alan.hayward@arm.com>
15268 David Sherwood <david.sherwood@arm.com>
15270 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15272 * config/aarch64/aarch64-protos.h
15273 (aarch64_sve_struct_memory_operand_p): Declare.
15274 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15275 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15276 (VPRED, vpred): Handle SVE structure modes.
15277 * config/aarch64/constraints.md (Utx): New constraint.
15278 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15279 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15280 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15281 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15282 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15283 structure modes. Split into pieces after RA.
15284 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15285 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15287 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15288 SVE structure modes.
15289 (aarch64_classify_address): Likewise.
15290 (sizetochar): Move earlier in file.
15291 (aarch64_print_operand): Handle SVE register lists.
15292 (aarch64_array_mode): New function.
15293 (aarch64_sve_struct_memory_operand_p): Likewise.
15294 (TARGET_ARRAY_MODE): Redefine.
15296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15297 Alan Hayward <alan.hayward@arm.com>
15298 David Sherwood <david.sherwood@arm.com>
15300 * target.def (array_mode): New target hook.
15301 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15302 * doc/tm.texi: Regenerate.
15303 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15304 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15305 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15306 targetm.array_mode.
15307 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15311 Alan Hayward <alan.hayward@arm.com>
15312 David Sherwood <david.sherwood@arm.com>
15314 * fold-const.c (fold_binary_loc): Check the argument types
15315 rather than the result type when testing for a vector operation.
15317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15319 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15320 * doc/tm.texi: Regenerate.
15322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15323 Alan Hayward <alan.hayward@arm.com>
15324 David Sherwood <david.sherwood@arm.com>
15326 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15327 (sve): Document new AArch64 extension.
15328 * doc/md.texi (w): Extend the description of the AArch64
15329 constraint to include SVE vectors.
15330 (Upl, Upa): Document new AArch64 predicate constraints.
15331 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15333 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15334 (msve-vector-bits=): New option.
15335 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15336 SVE when these are disabled.
15337 (sve): New extension.
15338 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15339 modes. Adjust their number of units based on aarch64_sve_vg.
15340 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15341 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15342 aarch64_addr_query_type.
15343 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15344 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15345 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15346 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15347 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15348 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15349 (aarch64_simd_imm_zero_p): Delete.
15350 (aarch64_check_zero_based_sve_index_immediate): Declare.
15351 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15352 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15353 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15354 (aarch64_sve_float_mul_immediate_p): Likewise.
15355 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15356 rather than an rtx.
15357 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15358 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15359 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15360 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15361 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15362 (aarch64_regmode_natural_size): Likewise.
15363 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15364 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15366 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15367 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15368 for VG and the SVE predicate registers.
15369 (V_ALIASES): Add a "z"-prefixed alias.
15370 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15371 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15372 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15373 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15374 (REG_CLASS_NAMES): Add entries for them.
15375 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15376 and the predicate registers.
15377 (aarch64_sve_vg): Declare.
15378 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15379 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15380 (REGMODE_NATURAL_SIZE): Define.
15381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15383 * config/aarch64/aarch64.c: Include cfgrtl.h.
15384 (simd_immediate_info): Add a constructor for series vectors,
15385 and an associated step field.
15386 (aarch64_sve_vg): New variable.
15387 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15388 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15389 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15390 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15391 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15392 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15393 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15394 (aarch64_get_mask_mode): New functions.
15395 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15396 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15397 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15398 predicate modes and predicate registers. Explicitly restrict
15399 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15400 to store a vector mode if it is recognized by
15401 aarch64_classify_vector_mode.
15402 (aarch64_regmode_natural_size): New function.
15403 (aarch64_hard_regno_caller_save_mode): Return the original mode
15405 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15406 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15407 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15408 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15410 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15411 does not overlap dest if the function is frame-related. Handle
15413 (aarch64_split_add_offset): New function.
15414 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15415 them aarch64_add_offset.
15416 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15417 and update call to aarch64_sub_sp.
15418 (aarch64_add_cfa_expression): New function.
15419 (aarch64_expand_prologue): Pass extra temporary registers to the
15420 functions above. Handle the case in which we need to emit new
15421 DW_CFA_expressions for registers that were originally saved
15422 relative to the stack pointer, but now have to be expressed
15423 relative to the frame pointer.
15424 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15426 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15427 IP0 and IP1 values for SVE frames.
15428 (aarch64_expand_vec_series): New function.
15429 (aarch64_expand_sve_widened_duplicate): Likewise.
15430 (aarch64_expand_sve_const_vector): Likewise.
15431 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15432 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15433 into the register, rather than emitting a SET directly.
15434 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15435 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15436 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15437 (offset_9bit_signed_scaled_p): New functions.
15438 (aarch64_replicate_bitmask_imm): New function.
15439 (aarch64_bitmask_imm): Use it.
15440 (aarch64_cannot_force_const_mem): Reject expressions involving
15441 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15442 (aarch64_classify_index): Handle SVE indices, by requiring
15443 a plain register index with a scale that matches the element size.
15444 (aarch64_classify_address): Handle SVE addresses. Assert that
15445 the mode of the address is VOIDmode or an integer mode.
15446 Update call to aarch64_classify_symbol.
15447 (aarch64_classify_symbolic_expression): Update call to
15448 aarch64_classify_symbol.
15449 (aarch64_const_vec_all_in_range_p): New function.
15450 (aarch64_print_vector_float_operand): Likewise.
15451 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15452 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15453 and the FP immediates 1.0 and 0.5.
15454 (aarch64_print_address_internal): Handle SVE addresses.
15455 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15456 (aarch64_regno_regclass): Handle predicate registers.
15457 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15459 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15460 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15461 (aarch64_convert_sve_vector_bits): New function.
15462 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15463 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15464 rather than an rtx.
15465 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15466 Handle SVE vector and predicate modes. Accept VL-based constants
15467 that need only one temporary register, and VL offsets that require
15468 no temporary registers.
15469 (aarch64_conditional_register_usage): Mark the predicate registers
15470 as fixed if SVE isn't available.
15471 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15472 Return true for SVE vector and predicate modes.
15473 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15474 rather than an unsigned int. Handle SVE modes.
15475 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15477 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15479 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15480 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15481 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15482 (aarch64_sve_float_mul_immediate_p): New functions.
15483 (aarch64_sve_valid_immediate): New function.
15484 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15485 Explicitly reject structure modes. Check for INDEX constants.
15486 Handle PTRUE and PFALSE constants.
15487 (aarch64_check_zero_based_sve_index_immediate): New function.
15488 (aarch64_simd_imm_zero_p): Delete.
15489 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15490 vector modes. Accept constants in the range of CNT[BHWD].
15491 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15492 ask for an Advanced SIMD mode.
15493 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15494 (aarch64_simd_vector_alignment): Handle SVE predicates.
15495 (aarch64_vectorize_preferred_vector_alignment): New function.
15496 (aarch64_simd_vector_alignment_reachable): Use it instead of
15498 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15499 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15501 (MAX_VECT_LEN): Delete.
15502 (expand_vec_perm_d): Add a vec_flags field.
15503 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15504 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15505 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15507 (aarch64_evpc_rev): Rename to...
15508 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15509 (aarch64_evpc_rev_global): New function.
15510 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15511 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15513 (aarch64_evpc_sve_tbl): New function.
15514 (aarch64_expand_vec_perm_const_1): Update after rename of
15515 aarch64_evpc_rev. Handle SVE permutes too, trying
15516 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15517 than aarch64_evpc_tbl.
15518 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15519 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15520 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15521 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15522 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15523 (aarch64_expand_sve_vcond): New functions.
15524 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15525 of aarch64_vector_mode_p.
15526 (aarch64_dwarf_poly_indeterminate_value): New function.
15527 (aarch64_compute_pressure_classes): Likewise.
15528 (aarch64_can_change_mode_class): Likewise.
15529 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15530 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15531 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15532 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15533 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15534 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15535 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15536 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15538 (Dn, Dl, Dr): Accept const as well as const_vector.
15539 (Dz): Likewise. Compare against CONST0_RTX.
15540 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15541 of "vector" where appropriate.
15542 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15543 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15544 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15545 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15546 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15547 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15548 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15549 (v_int_equiv): Extend to SVE modes.
15550 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15552 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15553 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15554 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15555 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15556 (SVE_COND_FP_CMP): New int iterators.
15557 (perm_hilo): Handle the new unpack unspecs.
15558 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15560 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15561 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15562 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15563 (aarch64_equality_operator, aarch64_constant_vector_operand)
15564 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15565 (aarch64_sve_nonimmediate_operand): Likewise.
15566 (aarch64_sve_general_operand): Likewise.
15567 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15568 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15569 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15570 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15571 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15572 (aarch64_sve_float_arith_immediate): Likewise.
15573 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15574 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15575 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15576 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15577 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15578 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15579 (aarch64_sve_float_arith_operand): Likewise.
15580 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15581 (aarch64_sve_float_mul_operand): Likewise.
15582 (aarch64_sve_vec_perm_operand): Likewise.
15583 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15584 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15585 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15586 as well as const_vector.
15587 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15588 in file. Use CONST0_RTX and CONSTM1_RTX.
15589 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15590 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15591 Use aarch64_simd_imm_zero.
15592 * config/aarch64/aarch64-sve.md: New file.
15593 * config/aarch64/aarch64.md: Include it.
15594 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15595 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15596 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15597 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15598 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15599 (sve): New attribute.
15600 (enabled): Disable instructions with the sve attribute unless
15602 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15603 aarch64_expand_mov_immediate.
15604 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15605 CNT[BHSD] immediates.
15606 (movti): Split CONST_POLY_INT moves into two halves.
15607 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15608 Split additions that need a temporary here if the destination
15609 is the stack pointer.
15610 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15611 (*add<mode>3_poly_1): New instruction.
15612 (set_clobber_cc): New expander.
15614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15616 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15617 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15618 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15619 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15620 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15621 Change innermode from fixed_mode_size to machine_mode.
15622 (simplify_subreg): Update call accordingly. Handle a constant-sized
15623 subreg of a variable-length CONST_VECTOR.
15625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15626 Alan Hayward <alan.hayward@arm.com>
15627 David Sherwood <david.sherwood@arm.com>
15629 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15630 (add_offset_to_base): New function, split out from...
15631 (create_mem_ref): ...here. When handling a scale other than 1,
15632 check first whether the address is valid without the offset.
15633 Add it into the base if so, leaving the index and scale as-is.
15635 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15638 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15639 fold_for_warn before checking if arg2 is INTEGER_CST.
15641 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15643 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15644 (store_multiple_operation): Delete.
15645 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15646 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15647 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15648 guarded by TARGET_STRING.
15649 (rs6000_output_load_multiple): Delete.
15650 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15651 OPTION_MASK_STRING / TARGET_STRING handling.
15652 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15653 (const rs6000_opt_masks) <"string">: Change mask to 0.
15654 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15655 (MASK_STRING): Delete.
15656 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15658 (load_multiple): Delete.
15665 (store_multiple): Delete.
15672 (movmemsi_8reg): Delete.
15673 (corresponding unnamed define_insn): Delete.
15674 (movmemsi_6reg): Delete.
15675 (corresponding unnamed define_insn): Delete.
15676 (movmemsi_4reg): Delete.
15677 (corresponding unnamed define_insn): Delete.
15678 (movmemsi_2reg): Delete.
15679 (corresponding unnamed define_insn): Delete.
15680 (movmemsi_1reg): Delete.
15681 (corresponding unnamed define_insn): Delete.
15682 * config/rs6000/rs6000.opt (mno-string): New.
15683 (mstring): Replace by deprecation warning stub.
15684 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15686 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15688 * regrename.c (regrename_do_replace): If replacing the same
15689 reg multiple times, try to reuse last created gen_raw_REG.
15692 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15693 main to workaround a bug in GDB.
15695 2018-01-12 Tom de Vries <tom@codesourcery.com>
15698 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15700 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15702 PR rtl-optimization/80481
15703 * ira-color.c (get_cap_member): New function.
15704 (allocnos_conflict_by_live_ranges_p): Use it.
15705 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15706 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15708 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15711 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15712 (*saddl_se_1): Ditto.
15713 (*ssubsi_1): Ditto.
15714 (*ssubl_se_1): Ditto.
15716 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15718 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15719 rather than wi::to_widest for DR_INITs.
15720 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15721 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15722 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15724 (vect_analyze_group_access_1): Note that here.
15726 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15728 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15729 polynomial type sizes.
15731 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15733 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15734 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15735 (gimple_add_tmp_var): Likewise.
15737 2018-01-12 Martin Liska <mliska@suse.cz>
15739 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15740 (gimple_alloc_sizes): Likewise.
15741 (dump_gimple_statistics): Use PRIu64 in printf format.
15742 * gimple.h: Change uint64_t to int.
15744 2018-01-12 Martin Liska <mliska@suse.cz>
15746 * tree-core.h: Use uint64_t instead of int.
15747 * tree.c (tree_node_counts): Likewise.
15748 (tree_node_sizes): Likewise.
15749 (dump_tree_statistics): Use PRIu64 in printf format.
15751 2018-01-12 Martin Liska <mliska@suse.cz>
15753 * Makefile.in: As qsort_chk is implemented in vec.c, add
15754 vec.o to linkage of gencfn-macros.
15755 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15756 passing the info to record_node_allocation_statistics.
15757 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15759 * ggc-common.c (struct ggc_usage): Add operator== and use
15760 it in operator< and compare function.
15761 * mem-stats.h (struct mem_usage): Likewise.
15762 * vec.c (struct vec_usage): Remove operator< and compare
15763 function. Can be simply inherited.
15765 2018-01-12 Martin Jambor <mjambor@suse.cz>
15768 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15769 * tree-ssa-math-opts.c: Include domwalk.h.
15770 (convert_mult_to_fma_1): New function.
15771 (fma_transformation_info): New type.
15772 (fma_deferring_state): Likewise.
15773 (cancel_fma_deferring): New function.
15774 (result_of_phi): Likewise.
15775 (last_fma_candidate_feeds_initial_phi): Likewise.
15776 (convert_mult_to_fma): Added deferring logic, split actual
15777 transformation to convert_mult_to_fma_1.
15778 (math_opts_dom_walker): New type.
15779 (math_opts_dom_walker::after_dom_children): New method, body moved
15780 here from pass_optimize_widening_mul::execute, added deferring logic
15782 (pass_optimize_widening_mul::execute): Moved most of code to
15783 math_opts_dom_walker::after_dom_children.
15784 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15785 * config/i386/i386.c (ix86_option_override_internal): Added
15786 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15788 2018-01-12 Richard Biener <rguenther@suse.de>
15791 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15792 inline instance vars.
15794 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15797 * config/rx/rx.c (rx_is_restricted_memory_address):
15798 Handle SUBREG case.
15800 2018-01-12 Richard Biener <rguenther@suse.de>
15802 PR tree-optimization/80846
15803 * target.def (split_reduction): New target hook.
15804 * targhooks.c (default_split_reduction): New function.
15805 * targhooks.h (default_split_reduction): Declare.
15806 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15807 target requests first reduce vectors by combining low and high
15809 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15810 (get_vectype_for_scalar_type_and_size): Export.
15811 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15812 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15813 * doc/tm.texi: Regenerate.
15814 * config/i386/i386.c (ix86_split_reduction): Implement
15815 TARGET_VECTORIZE_SPLIT_REDUCTION.
15817 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15820 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15821 in PIC mode except for TARGET_VXWORKS_RTP.
15822 * config/sparc/sparc.c: Include cfgrtl.h.
15823 (TARGET_INIT_PIC_REG): Define.
15824 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15825 (sparc_pic_register_p): New predicate.
15826 (sparc_legitimate_address_p): Use it.
15827 (sparc_legitimize_pic_address): Likewise.
15828 (sparc_delegitimize_address): Likewise.
15829 (sparc_mode_dependent_address_p): Likewise.
15830 (gen_load_pcrel_sym): Remove 4th parameter.
15831 (load_got_register): Adjust call to above. Remove obsolete stuff.
15832 (sparc_expand_prologue): Do not call load_got_register here.
15833 (sparc_flat_expand_prologue): Likewise.
15834 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15835 (sparc_use_pseudo_pic_reg): New function.
15836 (sparc_init_pic_reg): Likewise.
15837 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15838 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15840 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15842 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15843 Add item for branch_cost.
15845 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15847 PR rtl-optimization/83565
15848 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15849 not extend the result to a larger mode for rotate operations.
15850 (num_sign_bit_copies1): Likewise.
15852 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15855 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15857 Use values-Xc.o for -pedantic.
15858 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15860 2018-01-12 Martin Liska <mliska@suse.cz>
15863 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15865 (possible_polymorphic_call_targets): Use it.
15866 (ipa_devirt): Likewise.
15868 2018-01-12 Martin Liska <mliska@suse.cz>
15870 * profile-count.h (enum profile_quality): Use 0 as invalid
15871 enum value of profile_quality.
15873 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15875 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15876 -mext-string options.
15878 2018-01-12 Richard Biener <rguenther@suse.de>
15880 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15881 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15882 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15884 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15886 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15888 * configure.ac (--with-long-double-format): Add support for the
15889 configuration option to change the default long double format on
15891 * config.gcc (powerpc*-linux*-*): Likewise.
15892 * configure: Regenerate.
15893 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15894 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15895 used without modification.
15897 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15899 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15900 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15901 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15902 MISC_BUILTIN_SPEC_BARRIER.
15903 (rs6000_init_builtins): Likewise.
15904 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15906 (speculation_barrier): New define_insn.
15907 * doc/extend.texi: Document __builtin_speculation_barrier.
15909 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15912 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15913 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15914 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15916 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15917 integral modes instead of "ss" and "sd".
15918 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15919 vectors with 32-bit and 64-bit elements.
15920 (vecdupssescalarmodesuffix): New mode attribute.
15921 (vec_dup<mode>): Use it.
15923 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15926 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15927 frame if argument is passed on stack.
15929 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15932 * ree.c (combine_reaching_defs): Optimize also
15933 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15934 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15936 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15938 PR middle-end/83189
15939 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15941 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15943 PR middle-end/83718
15944 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15945 after they are computed.
15947 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15949 PR tree-optimization/83695
15950 * gimple-loop-linterchange.cc
15951 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15952 reset cached scev information after interchange.
15953 (pass_linterchange::execute): Remove call to scev_reset_htab.
15955 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15957 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15958 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15959 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15960 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15961 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15962 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15963 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15964 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15965 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15966 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15967 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15968 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15969 (V_lane_reg): Likewise.
15970 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15972 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15973 (vfmal_lane_low<mode>_intrinsic,
15974 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15975 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15976 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15977 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15978 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15979 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15981 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15983 * config/arm/arm-cpus.in (fp16fml): New feature.
15984 (ALL_SIMD): Add fp16fml.
15985 (armv8.2-a): Add fp16fml as an option.
15986 (armv8.3-a): Likewise.
15987 (armv8.4-a): Add fp16fml as part of fp16.
15988 * config/arm/arm.h (TARGET_FP16FML): Define.
15989 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15991 * config/arm/arm-modes.def (V2HF): Define.
15992 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15993 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15994 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15995 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15996 vfmsl_low, vfmsl_high): New set of builtins.
15997 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15998 (vfml_op): New code attribute.
15999 (VFMLHALVES): New int iterator.
16000 (VFML, VFMLSEL): New mode attributes.
16001 (V_reg): Define mapping for V2HF.
16002 (V_hi, V_lo): New mode attributes.
16003 (VF_constraint): Likewise.
16004 (vfml_half, vfml_half_selector): New int attributes.
16005 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16007 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16008 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16010 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16011 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16012 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16013 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16015 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16016 Document new effective target and option set.
16018 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16020 * config/arm/arm-cpus.in (armv8_4): New feature.
16021 (ARMv8_4a): New fgroup.
16022 (armv8.4-a): New arch.
16023 * config/arm/arm-tables.opt: Regenerate.
16024 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16025 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16026 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16027 Add matching rules for -march=armv8.4-a and extensions.
16028 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16030 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16033 * config/rx/rx.md (BW): New mode attribute.
16034 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16036 2018-01-11 Richard Biener <rguenther@suse.de>
16038 PR tree-optimization/83435
16039 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16040 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16041 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16044 Alan Hayward <alan.hayward@arm.com>
16045 David Sherwood <david.sherwood@arm.com>
16047 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16049 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16050 (aarch64_print_address_internal): Use it to check for a zero offset.
16052 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16053 Alan Hayward <alan.hayward@arm.com>
16054 David Sherwood <david.sherwood@arm.com>
16056 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16057 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16058 Return a poly_int64 rather than a HOST_WIDE_INT.
16059 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16060 rather than a HOST_WIDE_INT.
16061 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16062 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16063 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16064 final_offset from HOST_WIDE_INT to poly_int64.
16065 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16066 to_constant when getting the number of units in an Advanced SIMD
16068 (aarch64_builtin_vectorized_function): Check for a constant number
16070 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16072 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16073 attribute instead of GET_MODE_NUNITS.
16074 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16075 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16076 GET_MODE_SIZE for fixed-size registers.
16077 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16078 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16079 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16080 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16081 (aarch64_print_operand, aarch64_print_address_internal)
16082 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16083 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16084 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16085 Handle polynomial GET_MODE_SIZE.
16086 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16087 wider than SImode without modification.
16088 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16089 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16090 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16091 passing and returning SVE modes.
16092 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16093 rather than GEN_INT.
16094 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16095 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16096 (aarch64_allocate_and_probe_stack_space): Likewise.
16097 (aarch64_layout_frame): Cope with polynomial offsets.
16098 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16099 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16100 polynomial offsets.
16101 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16102 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16103 poly_int64 rather than a HOST_WIDE_INT.
16104 (aarch64_get_separate_components, aarch64_process_components)
16105 (aarch64_expand_prologue, aarch64_expand_epilogue)
16106 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16107 (aarch64_anchor_offset): New function, split out from...
16108 (aarch64_legitimize_address): ...here.
16109 (aarch64_builtin_vectorization_cost): Handle polynomial
16110 TYPE_VECTOR_SUBPARTS.
16111 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16113 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16114 number of elements from the PARALLEL rather than the mode.
16115 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16116 rather than GET_MODE_BITSIZE.
16117 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16118 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16119 (aarch64_expand_vec_perm_const_1): Handle polynomial
16120 d->perm.length () and d->perm elements.
16121 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16122 Apply to_constant to d->perm elements.
16123 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16124 polynomial CONST_VECTOR_NUNITS.
16125 (aarch64_move_pointer): Take amount as a poly_int64 rather
16127 (aarch64_progress_pointer): Avoid temporary variable.
16128 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16129 the mode attribute instead of GET_MODE.
16131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16132 Alan Hayward <alan.hayward@arm.com>
16133 David Sherwood <david.sherwood@arm.com>
16135 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16136 x exists before using it.
16137 (aarch64_add_constant_internal): Rename to...
16138 (aarch64_add_offset_1): ...this. Replace regnum with separate
16139 src and dest rtxes. Handle the case in which they're different,
16140 including when the offset is zero. Replace scratchreg with an rtx.
16141 Use 2 additions if there is no spare register into which we can
16142 move a 16-bit constant.
16143 (aarch64_add_constant): Delete.
16144 (aarch64_add_offset): Replace reg with separate src and dest
16145 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16146 Use aarch64_add_offset_1.
16147 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16148 an rtx rather than an int. Take the delta as a poly_int64
16149 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16150 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16151 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16152 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16153 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16154 and aarch64_add_sp.
16155 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16156 aarch64_add_constant.
16158 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16160 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16161 Use scalar_float_mode.
16163 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16165 * config/aarch64/aarch64-simd.md
16166 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16167 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16168 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16169 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16170 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16171 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16172 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16173 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16174 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16175 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16177 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16180 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16181 targ_options->x_arm_arch_string is non NULL.
16183 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16185 * config/aarch64/aarch64.h
16186 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16188 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16191 * expmed.c (emit_store_flag_force): Swap if const op0
16192 and change VOIDmode to mode of op0.
16194 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16196 PR rtl-optimization/83761
16197 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16198 than bytes to mode_for_size.
16200 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16202 PR middle-end/83189
16203 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16204 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16207 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16209 PR middle-end/83575
16210 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16211 when in layout mode.
16212 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16213 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16216 2018-01-10 Michael Collison <michael.collison@arm.com>
16218 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16219 * config/aarch64/aarch64-option-extension.def: Add
16220 AARCH64_OPT_EXTENSION of 'fp16fml'.
16221 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16222 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16223 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16224 * config/aarch64/constraints.md (Ui7): New constraint.
16225 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16226 (VFMLA_SEL_W): Ditto.
16229 (VFMLA16_LOW): New int iterator.
16230 (VFMLA16_HIGH): Ditto.
16231 (UNSPEC_FMLAL): New unspec.
16232 (UNSPEC_FMLSL): Ditto.
16233 (UNSPEC_FMLAL2): Ditto.
16234 (UNSPEC_FMLSL2): Ditto.
16235 (f16mac): New code attribute.
16236 * config/aarch64/aarch64-simd-builtins.def
16237 (aarch64_fmlal_lowv2sf): Ditto.
16238 (aarch64_fmlsl_lowv2sf): Ditto.
16239 (aarch64_fmlalq_lowv4sf): Ditto.
16240 (aarch64_fmlslq_lowv4sf): Ditto.
16241 (aarch64_fmlal_highv2sf): Ditto.
16242 (aarch64_fmlsl_highv2sf): Ditto.
16243 (aarch64_fmlalq_highv4sf): Ditto.
16244 (aarch64_fmlslq_highv4sf): Ditto.
16245 (aarch64_fmlal_lane_lowv2sf): Ditto.
16246 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16247 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16248 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16249 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16250 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16251 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16252 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16253 (aarch64_fmlal_lane_highv2sf): Ditto.
16254 (aarch64_fmlsl_lane_highv2sf): Ditto.
16255 (aarch64_fmlal_laneq_highv2sf): Ditto.
16256 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16257 (aarch64_fmlalq_lane_highv4sf): Ditto.
16258 (aarch64_fmlsl_lane_highv4sf): Ditto.
16259 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16260 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16261 * config/aarch64/aarch64-simd.md:
16262 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16263 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16264 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16265 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16266 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16267 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16268 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16269 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16270 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16271 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16272 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16273 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16274 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16275 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16276 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16277 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16278 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16279 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16280 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16281 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16282 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16283 (vfmlsl_low_u32): Ditto.
16284 (vfmlalq_low_u32): Ditto.
16285 (vfmlslq_low_u32): Ditto.
16286 (vfmlal_high_u32): Ditto.
16287 (vfmlsl_high_u32): Ditto.
16288 (vfmlalq_high_u32): Ditto.
16289 (vfmlslq_high_u32): Ditto.
16290 (vfmlal_lane_low_u32): Ditto.
16291 (vfmlsl_lane_low_u32): Ditto.
16292 (vfmlal_laneq_low_u32): Ditto.
16293 (vfmlsl_laneq_low_u32): Ditto.
16294 (vfmlalq_lane_low_u32): Ditto.
16295 (vfmlslq_lane_low_u32): Ditto.
16296 (vfmlalq_laneq_low_u32): Ditto.
16297 (vfmlslq_laneq_low_u32): Ditto.
16298 (vfmlal_lane_high_u32): Ditto.
16299 (vfmlsl_lane_high_u32): Ditto.
16300 (vfmlal_laneq_high_u32): Ditto.
16301 (vfmlsl_laneq_high_u32): Ditto.
16302 (vfmlalq_lane_high_u32): Ditto.
16303 (vfmlslq_lane_high_u32): Ditto.
16304 (vfmlalq_laneq_high_u32): Ditto.
16305 (vfmlslq_laneq_high_u32): Ditto.
16306 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16307 (AARCH64_FL_FOR_ARCH8_4): New.
16308 (AARCH64_ISA_F16FML): New ISA flag.
16309 (TARGET_F16FML): New feature flag for fp16fml.
16310 (doc/invoke.texi): Document new fp16fml option.
16312 2018-01-10 Michael Collison <michael.collison@arm.com>
16314 * config/aarch64/aarch64-builtins.c:
16315 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16316 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16317 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16318 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16319 (AARCH64_ISA_SHA3): New ISA flag.
16320 (TARGET_SHA3): New feature flag for sha3.
16321 * config/aarch64/iterators.md (sha512_op): New int attribute.
16322 (CRYPTO_SHA512): New int iterator.
16323 (UNSPEC_SHA512H): New unspec.
16324 (UNSPEC_SHA512H2): Ditto.
16325 (UNSPEC_SHA512SU0): Ditto.
16326 (UNSPEC_SHA512SU1): Ditto.
16327 * config/aarch64/aarch64-simd-builtins.def
16328 (aarch64_crypto_sha512hqv2di): New builtin.
16329 (aarch64_crypto_sha512h2qv2di): Ditto.
16330 (aarch64_crypto_sha512su0qv2di): Ditto.
16331 (aarch64_crypto_sha512su1qv2di): Ditto.
16332 (aarch64_eor3qv8hi): Ditto.
16333 (aarch64_rax1qv2di): Ditto.
16334 (aarch64_xarqv2di): Ditto.
16335 (aarch64_bcaxqv8hi): Ditto.
16336 * config/aarch64/aarch64-simd.md:
16337 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16338 (aarch64_crypto_sha512su0qv2di): Ditto.
16339 (aarch64_crypto_sha512su1qv2di): Ditto.
16340 (aarch64_eor3qv8hi): Ditto.
16341 (aarch64_rax1qv2di): Ditto.
16342 (aarch64_xarqv2di): Ditto.
16343 (aarch64_bcaxqv8hi): Ditto.
16344 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16345 (vsha512h2q_u64): Ditto.
16346 (vsha512su0q_u64): Ditto.
16347 (vsha512su1q_u64): Ditto.
16348 (veor3q_u16): Ditto.
16349 (vrax1q_u64): Ditto.
16350 (vxarq_u64): Ditto.
16351 (vbcaxq_u16): Ditto.
16352 * config/arm/types.md (crypto_sha512): New type attribute.
16353 (crypto_sha3): Ditto.
16354 (doc/invoke.texi): Document new sha3 option.
16356 2018-01-10 Michael Collison <michael.collison@arm.com>
16358 * config/aarch64/aarch64-builtins.c:
16359 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16360 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16361 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16362 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16363 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16364 (AARCH64_ISA_SM4): New ISA flag.
16365 (TARGET_SM4): New feature flag for sm4.
16366 * config/aarch64/aarch64-simd-builtins.def
16367 (aarch64_sm3ss1qv4si): Ditto.
16368 (aarch64_sm3tt1aq4si): Ditto.
16369 (aarch64_sm3tt1bq4si): Ditto.
16370 (aarch64_sm3tt2aq4si): Ditto.
16371 (aarch64_sm3tt2bq4si): Ditto.
16372 (aarch64_sm3partw1qv4si): Ditto.
16373 (aarch64_sm3partw2qv4si): Ditto.
16374 (aarch64_sm4eqv4si): Ditto.
16375 (aarch64_sm4ekeyqv4si): Ditto.
16376 * config/aarch64/aarch64-simd.md:
16377 (aarch64_sm3ss1qv4si): Ditto.
16378 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16379 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16380 (aarch64_sm4eqv4si): Ditto.
16381 (aarch64_sm4ekeyqv4si): Ditto.
16382 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16383 (sm3part_op): Ditto.
16384 (CRYPTO_SM3TT): Ditto.
16385 (CRYPTO_SM3PART): Ditto.
16386 (UNSPEC_SM3SS1): New unspec.
16387 (UNSPEC_SM3TT1A): Ditto.
16388 (UNSPEC_SM3TT1B): Ditto.
16389 (UNSPEC_SM3TT2A): Ditto.
16390 (UNSPEC_SM3TT2B): Ditto.
16391 (UNSPEC_SM3PARTW1): Ditto.
16392 (UNSPEC_SM3PARTW2): Ditto.
16393 (UNSPEC_SM4E): Ditto.
16394 (UNSPEC_SM4EKEY): Ditto.
16395 * config/aarch64/constraints.md (Ui2): New constraint.
16396 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16397 * config/arm/types.md (crypto_sm3): New type attribute.
16398 (crypto_sm4): Ditto.
16399 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16400 (vsm3tt1aq_u32): Ditto.
16401 (vsm3tt1bq_u32): Ditto.
16402 (vsm3tt2aq_u32): Ditto.
16403 (vsm3tt2bq_u32): Ditto.
16404 (vsm3partw1q_u32): Ditto.
16405 (vsm3partw2q_u32): Ditto.
16406 (vsm4eq_u32): Ditto.
16407 (vsm4ekeyq_u32): Ditto.
16408 (doc/invoke.texi): Document new sm4 option.
16410 2018-01-10 Michael Collison <michael.collison@arm.com>
16412 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16413 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16414 (AARCH64_FL_FOR_ARCH8_4): New.
16415 (AARCH64_FL_V8_4): New flag.
16416 (doc/invoke.texi): Document new armv8.4-a option.
16418 2018-01-10 Michael Collison <michael.collison@arm.com>
16420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16421 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16422 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16423 * config/aarch64/aarch64-option-extension.def: Add
16424 AARCH64_OPT_EXTENSION of 'sha2'.
16425 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16426 (crypto): Disable sha2 and aes if crypto disabled.
16427 (crypto): Enable aes and sha2 if enabled.
16428 (simd): Disable sha2 and aes if simd disabled.
16429 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16431 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16432 (TARGET_SHA2): New feature flag for sha2.
16433 (TARGET_AES): New feature flag for aes.
16434 * config/aarch64/aarch64-simd.md:
16435 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16436 conditional on TARGET_AES.
16437 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16438 (aarch64_crypto_sha1hsi): Make pattern conditional
16440 (aarch64_crypto_sha1hv4si): Ditto.
16441 (aarch64_be_crypto_sha1hv4si): Ditto.
16442 (aarch64_crypto_sha1su1v4si): Ditto.
16443 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16444 (aarch64_crypto_sha1su0v4si): Ditto.
16445 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16446 (aarch64_crypto_sha256su0v4si): Ditto.
16447 (aarch64_crypto_sha256su1v4si): Ditto.
16448 (doc/invoke.texi): Document new aes and sha2 options.
16450 2018-01-10 Martin Sebor <msebor@redhat.com>
16452 PR tree-optimization/83781
16453 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16456 2018-01-11 Martin Sebor <msebor@gmail.com>
16457 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16459 PR tree-optimization/83501
16460 PR tree-optimization/81703
16462 * tree-ssa-strlen.c (get_string_cst): Rename...
16463 (get_string_len): ...to this. Handle global constants.
16464 (handle_char_store): Adjust.
16466 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16467 Jim Wilson <jimw@sifive.com>
16469 * config/riscv/riscv-protos.h (riscv_output_return): New.
16470 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16471 (riscv_attribute_table, riscv_output_return),
16472 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16473 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16474 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16476 (riscv_expand_prologue): Add early return for naked function.
16477 (riscv_expand_epilogue): Likewise.
16478 (riscv_function_ok_for_sibcall): Return false for naked function.
16479 (riscv_set_current_function): New.
16480 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16481 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16482 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16483 * doc/extend.texi (RISC-V Function Attributes): New.
16485 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16487 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16488 check for 128-bit long double before checking TCmode.
16489 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16490 128-bit long doubles before checking TFmode or TCmode.
16491 (FLOAT128_IBM_P): Likewise.
16493 2018-01-10 Martin Sebor <msebor@redhat.com>
16495 PR tree-optimization/83671
16496 * builtins.c (c_strlen): Unconditionally return zero for the empty
16498 Use -Warray-bounds for warnings.
16499 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16500 for non-constant array indices with COMPONENT_REF, arrays of
16501 arrays, and pointers to arrays.
16502 (gimple_fold_builtin_strlen): Determine and set length range for
16503 non-constant character arrays.
16505 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16507 PR middle-end/81897
16508 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16511 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16513 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16515 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16518 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16519 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16520 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16521 indexed_or_indirect_operand predicate.
16522 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16523 (*vsx_le_perm_load_v8hi): Likewise.
16524 (*vsx_le_perm_load_v16qi): Likewise.
16525 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16526 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16527 (*vsx_le_perm_store_v8hi): Likewise.
16528 (*vsx_le_perm_store_v16qi): Likewise.
16529 (eight unnamed splitters): Likewise.
16531 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16533 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16534 * config/rs6000/emmintrin.h: Likewise.
16535 * config/rs6000/mmintrin.h: Likewise.
16536 * config/rs6000/xmmintrin.h: Likewise.
16538 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16541 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16543 * tree.c (tree_nop_conversion): Return true for location wrapper
16545 (maybe_wrap_with_location): New function.
16546 (selftest::check_strip_nops): New function.
16547 (selftest::test_location_wrappers): New function.
16548 (selftest::tree_c_tests): Call it.
16549 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16550 (maybe_wrap_with_location): New decl.
16551 (EXPR_LOCATION_WRAPPER_P): New macro.
16552 (location_wrapper_p): New inline function.
16553 (tree_strip_any_location_wrapper): New inline function.
16555 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16558 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16559 stack_realign_offset for the largest alignment of stack slot
16561 (ix86_find_max_used_stack_alignment): New function.
16562 (ix86_finalize_stack_frame_flags): Use it. Set
16563 max_used_stack_alignment if we don't realign stack.
16564 * config/i386/i386.h (machine_function): Add
16565 max_used_stack_alignment.
16567 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16569 * config/arm/arm.opt (-mbranch-cost): New option.
16570 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16573 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16576 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16577 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16579 2018-01-10 Richard Biener <rguenther@suse.de>
16582 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16583 early out so it also covers the case where we have a non-NULL
16586 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16588 PR tree-optimization/83753
16589 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16590 for non-strided grouped accesses if the number of elements is 1.
16592 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16595 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16596 * i386.h (TARGET_USE_GATHER): Define.
16597 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16599 2018-01-10 Martin Liska <mliska@suse.cz>
16602 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16603 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16605 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16606 CLEANUP_NO_PARTITIONING is not set.
16608 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16610 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16611 for vectors, as a partial revert of r254296.
16612 * rtl.h (const_vec_p): Delete.
16613 (const_vec_duplicate_p): Don't test for vector CONSTs.
16614 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16615 * expmed.c (make_tree): Likewise.
16618 * common.md (E, F): Use CONSTANT_P instead of checking for
16620 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16621 checking for CONST_VECTOR.
16623 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16625 PR middle-end/83575
16626 * predict.c (force_edge_cold): Handle in more sane way edges
16627 with no prediction.
16629 2018-01-09 Carl Love <cel@us.ibm.com>
16631 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16633 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16634 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16635 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16636 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16637 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16638 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16639 * config/rs6000/rs6000-protos.h: Add extern defition for
16640 rs6000_generate_float2_double_code.
16641 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16643 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16644 (float2_v2df): Add define_expand.
16646 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16649 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16650 op_mode in the force_to_mode call.
16652 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16654 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16655 instead of checking each element individually.
16656 (aarch64_evpc_uzp): Likewise.
16657 (aarch64_evpc_zip): Likewise.
16658 (aarch64_evpc_ext): Likewise.
16659 (aarch64_evpc_rev): Likewise.
16660 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16661 instead of checking each element individually. Return true without
16663 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16664 whether all selected elements come from the same input, instead of
16665 checking each element individually. Remove calls to gen_rtx_REG,
16666 start_sequence and end_sequence and instead assert that no rtl is
16669 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16671 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16672 order of HIGH and CONST checks.
16674 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16676 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16677 if the destination isn't an SSA_NAME.
16679 2018-01-09 Richard Biener <rguenther@suse.de>
16681 PR tree-optimization/83668
16682 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16684 (canonicalize_loop_form): ... here, renamed from ...
16685 (canonicalize_loop_closed_ssa_form): ... this and amended to
16686 swap successor edges for loop exit blocks to make us use
16687 the RPO order we need for initial schedule generation.
16689 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16691 PR tree-optimization/64811
16692 * match.pd: When optimizing comparisons with Inf, avoid
16693 introducing or losing exceptions from comparisons with NaN.
16695 2018-01-09 Martin Liska <mliska@suse.cz>
16698 * asan.c (shadow_mem_size): Add gcc_assert.
16700 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16702 Don't save registers in main().
16705 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16706 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16707 * config/avr/avr.c (avr_set_current_function): Don't error if
16708 naked, OS_task or OS_main are specified at the same time.
16709 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16711 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16713 * common/config/avr/avr-common.c (avr_option_optimization_table):
16714 Switch on -mmain-is-OS_task for optimizing compilations.
16716 2018-01-09 Richard Biener <rguenther@suse.de>
16718 PR tree-optimization/83572
16719 * graphite.c: Include cfganal.h.
16720 (graphite_transform_loops): Connect infinite loops to exit
16721 and remove fake edges at the end.
16723 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16725 * ipa-inline.c (edge_badness): Revert accidental checkin.
16727 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16730 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16731 symbols; not inline clones.
16733 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16736 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16737 hard registers. Formatting fixes.
16739 PR preprocessor/83722
16740 * gcc.c (try_generate_repro): Pass
16741 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16742 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16745 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16746 Kito Cheng <kito.cheng@gmail.com>
16748 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16749 (riscv_leaf_function_p): Delete.
16750 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16752 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16754 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16756 (do_ifelse): New function.
16757 (do_isel): New function.
16758 (do_sub3): New function.
16759 (do_add3): New function.
16760 (do_load_mask_compare): New function.
16761 (do_overlap_load_compare): New function.
16762 (expand_compare_loop): New function.
16763 (expand_block_compare): Call expand_compare_loop() when appropriate.
16764 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16765 option description.
16766 (-mblock-compare-inline-loop-limit): New option.
16768 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16771 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16772 Reverse order of second and third operands in first alternative.
16773 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16774 of first and second elements in UNSPEC_VPERMR vector.
16775 (altivec_expand_vec_perm_le): Likewise.
16777 2018-01-08 Jeff Law <law@redhat.com>
16779 PR rtl-optimizatin/81308
16780 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16781 (process_switch): If group_case_labels makes a change, then set
16783 (pass_convert_switch::execute): If a switch is converted, then
16784 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16786 PR rtl-optimization/81308
16787 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16790 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16792 PR target/83663 - Revert r255946
16793 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16794 generation for cases where splatting a value is not useful.
16795 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16796 across a vec_duplicate and a paradoxical subreg forming a vector
16797 mode to a vec_concat.
16799 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16801 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16802 -march=armv8.3-a variants.
16803 * config/arm/t-multilib: Likewise.
16804 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16806 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16808 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16810 (cceq_ior_compare_complement): Give it a name so I can use it, and
16811 change boolean_or_operator predicate to boolean_operator so it can
16812 be used to generate a crand.
16813 (eqne): New code iterator.
16814 (bd/bd_neg): New code_attrs.
16815 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16816 a single define_insn.
16817 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16818 decrement (bdnzt/bdnzf/bdzt/bdzf).
16819 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16820 with the new names of the branch decrement patterns, and added the
16821 names of the branch decrement conditional patterns.
16823 2018-01-08 Richard Biener <rguenther@suse.de>
16825 PR tree-optimization/83563
16826 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16829 2018-01-08 Richard Biener <rguenther@suse.de>
16831 PR middle-end/83713
16832 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16834 2018-01-08 Richard Biener <rguenther@suse.de>
16836 PR tree-optimization/83685
16837 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16838 references to abnormals.
16840 2018-01-08 Richard Biener <rguenther@suse.de>
16843 * dwarf2out.c (output_indirect_strings): Handle empty
16844 skeleton_debug_str_hash.
16845 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16847 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16849 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16850 (emit_store_direct): Likewise.
16851 (arc_trampoline_adjust_address): Likewise.
16852 (arc_asm_trampoline_template): New function.
16853 (arc_initialize_trampoline): Use asm_trampoline_template.
16854 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16855 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16856 * config/arc/arc.md (flush_icache): Delete pattern.
16858 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16860 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16861 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16864 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16867 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16868 by not USED_FOR_TARGET.
16869 (make_pass_resolve_sw_modes): Likewise.
16871 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16873 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16876 2018-01-08 Richard Biener <rguenther@suse.de>
16878 PR middle-end/83580
16879 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16881 2018-01-08 Richard Biener <rguenther@suse.de>
16883 PR middle-end/83517
16884 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16886 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16888 PR middle-end/81897
16889 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16890 basic blocks with a small number of successors.
16891 (convert_control_dep_chain_into_preds): Improve handling of
16893 (dump_predicates): Split apart into...
16894 (dump_pred_chain): ...here...
16895 (dump_pred_info): ...and here.
16896 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16897 (can_chain_union_be_invalidated_p): Improve check for invalidation
16899 (uninit_uses_cannot_happen): Avoid unnecessary if
16900 convert_control_dep_chain_into_preds yielded nothing.
16902 2018-01-06 Martin Sebor <msebor@redhat.com>
16904 PR tree-optimization/83640
16905 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16906 subtracting negative offset from size.
16907 (builtin_access::overlap): Adjust offset bounds of the access to fall
16908 within the size of the object if possible.
16910 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16912 PR rtl-optimization/83699
16913 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16914 extract_bit_field_as_subreg to cases in which the extracted
16915 value is also a vector.
16917 * lra-constraints.c (process_alt_operands): Test for the equivalence
16918 substitutions when detecting a possible reload cycle.
16920 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16923 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16924 by default if flag_selective_schedling{,2}. Formatting fixes.
16926 PR rtl-optimization/83682
16927 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16928 if it has non-VECTOR_MODE element mode.
16929 (vec_duplicate_p): Likewise.
16931 PR middle-end/83694
16932 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16933 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16935 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16938 * config/i386/i386-builtin.def
16939 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16940 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16941 Require also OPTION_MASK_ISA_AVX512F in addition to
16942 OPTION_MASK_ISA_GFNI.
16943 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16944 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16945 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16946 to OPTION_MASK_ISA_GFNI.
16947 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16948 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16949 OPTION_MASK_ISA_AVX512BW.
16950 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16951 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16952 addition to OPTION_MASK_ISA_GFNI.
16953 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16954 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16955 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16956 to OPTION_MASK_ISA_GFNI.
16957 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16958 a requirement for all ISAs rather than any of them with a few
16960 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16962 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16963 bitmasks to be enabled with 3 exceptions, instead of requiring any
16964 enabled ISA with lots of exceptions.
16965 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16966 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16967 Change avx512bw in isa attribute to avx512f.
16968 * config/i386/sgxintrin.h: Add license boilerplate.
16969 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16970 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16971 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16972 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16974 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16975 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16976 temporarily sse2 rather than sse if not enabled already.
16979 * config/i386/sse.md (VI248_VLBW): Rename to ...
16980 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16981 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16982 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16983 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16984 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16985 mode iterator instead of VI248_VLBW.
16987 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16989 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16990 (record_modified): Skip clobbers; add debug output.
16991 (param_change_prob): Use sreal frequencies.
16993 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16995 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16996 punt for user-aligned variables.
16998 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17000 * tree-chrec.c (chrec_contains_symbols): Return true for
17003 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17006 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17007 of (x|y) == x for BICS pattern.
17009 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17011 PR tree-optimization/83605
17012 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17013 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17016 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17018 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17019 * config/epiphany/rtems.h: New file.
17021 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17022 Uros Bizjak <ubizjak@gmail.com>
17025 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17026 QIreg_operand instead of register_operand predicate.
17027 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17028 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17029 comments instead of -fmitigate[-_]rop.
17031 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17034 * cgraphunit.c (symbol_table::compile): Switch to text_section
17035 before calling assembly_start debug hook.
17036 * run-rtl-passes.c (run_rtl_passes): Likewise.
17039 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17041 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17042 range_int_cst_p rather than !symbolic_range_p before calling
17043 extract_range_from_multiplicative_op_1.
17045 2018-01-04 Jeff Law <law@redhat.com>
17047 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17048 redundant test in assertion.
17050 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17052 * doc/rtl.texi: Document machine_mode wrapper classes.
17054 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17056 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17057 using tree_to_uhwi.
17059 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17061 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17062 the VEC_PERM_EXPR fold to fail.
17064 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17067 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17068 to switched_sections.
17070 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17073 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17074 test for d.testing.
17076 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17079 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17080 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17082 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17085 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17086 is BLKmode and bitpos not zero or mode change is needed.
17088 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17091 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17094 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17097 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17098 instead of MULT rtx. Update all corresponding splitters.
17099 (*saddl_se): Ditto.
17100 (*ssub<modesuffix>): Ditto.
17101 (*ssubl_se): Ditto.
17102 (*cmp_sadd_di): Update split patterns.
17103 (*cmp_sadd_si): Ditto.
17104 (*cmp_sadd_sidi): Ditto.
17105 (*cmp_ssub_di): Ditto.
17106 (*cmp_ssub_si): Ditto.
17107 (*cmp_ssub_sidi): Ditto.
17108 * config/alpha/predicates.md (const23_operand): New predicate.
17109 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17110 Look for ASHIFT, not MULT inner operand.
17111 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17113 2018-01-04 Martin Liska <mliska@suse.cz>
17115 PR gcov-profile/83669
17116 * gcov.c (output_intermediate_file): Add version to intermediate
17118 * doc/gcov.texi: Document new field 'version' in intermediate
17119 file format. Fix location of '-k' option of gcov command.
17121 2018-01-04 Martin Liska <mliska@suse.cz>
17124 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17126 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17128 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17130 2018-01-03 Martin Sebor <msebor@redhat.com>
17132 PR tree-optimization/83655
17133 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17134 checking calls with invalid arguments.
17136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17138 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17139 (vectorizable_mask_load_store): Delete.
17140 (vectorizable_call): Return false for masked loads and stores.
17141 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17142 instead of gimple_assign_rhs1.
17143 (vectorizable_load): Handle IFN_MASK_LOAD.
17144 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17148 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17150 (vectorizable_mask_load_store): ...here.
17151 (vectorizable_load): ...and here.
17153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17155 * tree-vect-stmts.c (vect_build_all_ones_mask)
17156 (vect_build_zero_merge_argument): New functions, split out from...
17157 (vectorizable_load): ...here.
17159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17161 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17163 (vectorizable_mask_load_store): ...here.
17164 (vectorizable_store): ...and here.
17166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17168 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17170 (vectorizable_mask_load_store): ...here.
17172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17174 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17175 (vect_model_store_cost): Take a vec_load_store_type instead of a
17177 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17178 (vect_model_store_cost): Take a vec_load_store_type instead of a
17180 (vectorizable_mask_load_store): Update accordingly.
17181 (vectorizable_store): Likewise.
17182 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17186 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17187 IFN_MASK_LOAD calls here rather than...
17188 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17191 Alan Hayward <alan.hayward@arm.com>
17192 David Sherwood <david.sherwood@arm.com>
17194 * expmed.c (extract_bit_field_1): For vector extracts,
17195 fall back to extract_bit_field_as_subreg if vec_extract
17198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17199 Alan Hayward <alan.hayward@arm.com>
17200 David Sherwood <david.sherwood@arm.com>
17202 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17203 they are variable or constant sized.
17204 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17205 slots for constant-sized data.
17207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17208 Alan Hayward <alan.hayward@arm.com>
17209 David Sherwood <david.sherwood@arm.com>
17211 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17212 handling COND_EXPRs with boolean comparisons, try to find a better
17213 basis for the mask type than the boolean itself.
17215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17217 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17218 is calculated and how it can be overridden.
17219 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17220 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17222 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17226 Alan Hayward <alan.hayward@arm.com>
17227 David Sherwood <david.sherwood@arm.com>
17229 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17230 Remove the mode argument.
17231 (aarch64_simd_valid_immediate): Remove the mode and inverse
17233 * config/aarch64/iterators.md (bitsize): New iterator.
17234 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17235 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17236 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17237 aarch64_simd_valid_immediate.
17238 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17239 (aarch64_reg_or_bic_imm): Likewise.
17240 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17241 with an insn_type enum and msl with a modifier_type enum.
17242 Replace element_width with a scalar_mode. Change the shift
17243 to unsigned int. Add constructors for scalar_float_mode and
17244 scalar_int_mode elements.
17245 (aarch64_vect_float_const_representable_p): Delete.
17246 (aarch64_can_const_movi_rtx_p)
17247 (aarch64_simd_scalar_immediate_valid_for_move)
17248 (aarch64_simd_make_constant): Update call to
17249 aarch64_simd_valid_immediate.
17250 (aarch64_advsimd_valid_immediate_hs): New function.
17251 (aarch64_advsimd_valid_immediate): Likewise.
17252 (aarch64_simd_valid_immediate): Remove mode and inverse
17253 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17254 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17255 and aarch64_float_const_representable_p on the result.
17256 (aarch64_output_simd_mov_immediate): Remove mode argument.
17257 Update call to aarch64_simd_valid_immediate and use of
17258 simd_immediate_info.
17259 (aarch64_output_scalar_simd_mov_immediate): Update call
17262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17263 Alan Hayward <alan.hayward@arm.com>
17264 David Sherwood <david.sherwood@arm.com>
17266 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17267 (mode_nunits): Likewise CONST_MODE_NUNITS.
17268 * machmode.def (ADJUST_NUNITS): Document.
17269 * genmodes.c (mode_data::need_nunits_adj): New field.
17270 (blank_mode): Update accordingly.
17271 (adj_nunits): New variable.
17272 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17274 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17275 listed in adj_nunits.
17276 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17277 listed in adj_nunits. Don't emit case statements for such modes.
17278 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17279 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17280 nothing if adj_nunits is nonnull.
17281 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17282 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17283 (emit_mode_fbit): Update use of print_maybe_const_decl.
17284 (emit_move_size): Likewise. Treat the array as non-const
17286 (emit_mode_adjustments): Handle adj_nunits.
17288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17290 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17291 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17292 (VECTOR_MODES): Use it.
17293 (make_vector_modes): Take the prefix as an argument.
17295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17296 Alan Hayward <alan.hayward@arm.com>
17297 David Sherwood <david.sherwood@arm.com>
17299 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17300 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17301 for MODE_VECTOR_BOOL.
17302 * machmode.def (VECTOR_BOOL_MODE): Document.
17303 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17304 (make_vector_bool_mode): New function.
17305 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17307 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17308 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17310 * stor-layout.c (int_mode_for_mode): Likewise.
17311 * tree.c (build_vector_type_for_mode): Likewise.
17312 * varasm.c (output_constant_pool_2): Likewise.
17313 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17314 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17315 for MODE_VECTOR_BOOL.
17316 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17317 of mode class checks.
17318 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17319 instead of a list of mode class checks.
17320 (expand_vector_scalar_condition): Likewise.
17321 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17324 Alan Hayward <alan.hayward@arm.com>
17325 David Sherwood <david.sherwood@arm.com>
17327 * machmode.h (mode_size): Change from unsigned short to
17329 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17330 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17331 or if measurement_type is not polynomial.
17332 (fixed_size_mode::includes_p): Check for constant-sized modes.
17333 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17334 return a poly_uint16 rather than an unsigned short.
17335 (emit_mode_size): Change the type of mode_size from unsigned short
17336 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17337 (emit_mode_adjustments): Cope with polynomial vector sizes.
17338 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17340 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17342 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17343 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17344 * caller-save.c (setup_save_areas): Likewise.
17345 (replace_reg_with_saved_mem): Likewise.
17346 * calls.c (emit_library_call_value_1): Likewise.
17347 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17348 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17349 (gen_lowpart_for_combine): Likewise.
17350 * convert.c (convert_to_integer_1): Likewise.
17351 * cse.c (equiv_constant, cse_insn): Likewise.
17352 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17353 (cselib_subst_to_values): Likewise.
17354 * dce.c (word_dce_process_block): Likewise.
17355 * df-problems.c (df_word_lr_mark_ref): Likewise.
17356 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17357 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17358 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17359 (rtl_for_decl_location): Likewise.
17360 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17361 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17362 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17363 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17364 (expand_expr_real_1): Likewise.
17365 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17366 (pad_below): Likewise.
17367 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17368 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17369 * ira.c (get_subreg_tracking_sizes): Likewise.
17370 * ira-build.c (ira_create_allocno_objects): Likewise.
17371 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17372 (ira_sort_regnos_for_alter_reg): Likewise.
17373 * ira-costs.c (record_operand_costs): Likewise.
17374 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17375 (resolve_simple_move): Likewise.
17376 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17377 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17378 (lra_constraints): Likewise.
17379 (CONST_POOL_OK_P): Reject variable-sized modes.
17380 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17381 (add_pseudo_to_slot, lra_spill): Likewise.
17382 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17383 * optabs-query.c (get_best_extraction_insn): Likewise.
17384 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17385 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17386 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17387 * recog.c (offsettable_address_addr_space_p): Likewise.
17388 * regcprop.c (maybe_mode_change): Likewise.
17389 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17390 * regrename.c (build_def_use): Likewise.
17391 * regstat.c (dump_reg_info): Likewise.
17392 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17393 (find_reloads, find_reloads_subreg_address): Likewise.
17394 * reload1.c (eliminate_regs_1): Likewise.
17395 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17396 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17397 (simplify_binary_operation_1, simplify_subreg): Likewise.
17398 * targhooks.c (default_function_arg_padding): Likewise.
17399 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17400 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17401 (verify_gimple_assign_ternary): Likewise.
17402 * tree-inline.c (estimate_move_cost): Likewise.
17403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17404 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17405 (get_address_cost_ainc): Likewise.
17406 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17407 (vect_supportable_dr_alignment): Likewise.
17408 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17409 (vectorizable_reduction): Likewise.
17410 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17411 (vectorizable_operation, vectorizable_load): Likewise.
17412 * tree.c (build_same_sized_truth_vector_type): Likewise.
17413 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17414 * var-tracking.c (emit_note_insn_var_location): Likewise.
17415 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17416 (ADDR_VEC_ALIGN): Likewise.
17418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17419 Alan Hayward <alan.hayward@arm.com>
17420 David Sherwood <david.sherwood@arm.com>
17422 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17424 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17425 or if measurement_type is polynomial.
17426 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17427 * combine.c (make_extraction): Likewise.
17428 * dse.c (find_shift_sequence): Likewise.
17429 * dwarf2out.c (mem_loc_descriptor): Likewise.
17430 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17431 (extract_bit_field, extract_low_bits): Likewise.
17432 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17433 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17434 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17435 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17436 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17437 * reload.c (find_reloads): Likewise.
17438 * reload1.c (alter_reg): Likewise.
17439 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17440 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17441 * tree-if-conv.c (predicate_mem_writes): Likewise.
17442 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17443 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17444 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17445 * valtrack.c (dead_debug_insert_temp): Likewise.
17446 * varasm.c (mergeable_constant_section): Likewise.
17447 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17450 Alan Hayward <alan.hayward@arm.com>
17451 David Sherwood <david.sherwood@arm.com>
17453 * expr.c (expand_assignment): Cope with polynomial mode sizes
17454 when assigning to a CONCAT.
17456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17457 Alan Hayward <alan.hayward@arm.com>
17458 David Sherwood <david.sherwood@arm.com>
17460 * machmode.h (mode_precision): Change from unsigned short to
17462 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17464 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17465 or if measurement_type is not polynomial.
17466 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17467 in which the mode is already known to be a scalar_int_mode.
17468 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17469 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17471 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17472 for GET_MODE_PRECISION.
17473 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17474 for GET_MODE_PRECISION.
17475 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17477 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17478 (expand_field_assignment, make_extraction): Likewise.
17479 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17480 (get_last_value): Likewise.
17481 * convert.c (convert_to_integer_1): Likewise.
17482 * cse.c (cse_insn): Likewise.
17483 * expr.c (expand_expr_real_1): Likewise.
17484 * lra-constraints.c (simplify_operand_subreg): Likewise.
17485 * optabs-query.c (can_atomic_load_p): Likewise.
17486 * optabs.c (expand_atomic_load): Likewise.
17487 (expand_atomic_store): Likewise.
17488 * ree.c (combine_reaching_defs): Likewise.
17489 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17490 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17491 * tree.h (type_has_mode_precision_p): Likewise.
17492 * ubsan.c (instrument_si_overflow): Likewise.
17494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17495 Alan Hayward <alan.hayward@arm.com>
17496 David Sherwood <david.sherwood@arm.com>
17498 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17499 polynomial numbers of units.
17500 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17501 (valid_vector_subparts_p): New function.
17502 (build_vector_type): Remove temporary shim and take the number
17503 of units as a poly_uint64 rather than an int.
17504 (build_opaque_vector_type): Take the number of units as a
17505 poly_uint64 rather than an int.
17506 * tree.c (build_vector_from_ctor): Handle polynomial
17507 TYPE_VECTOR_SUBPARTS.
17508 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17509 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17510 (build_vector_from_val): If the number of units is variable,
17511 use build_vec_duplicate_cst for constant operands and
17512 VEC_DUPLICATE_EXPR otherwise.
17513 (make_vector_type): Remove temporary is_constant ().
17514 (build_vector_type, build_opaque_vector_type): Take the number of
17515 units as a poly_uint64 rather than an int.
17516 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17518 * cfgexpand.c (expand_debug_expr): Likewise.
17519 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17520 (store_constructor, expand_expr_real_1): Likewise.
17521 (const_scalar_mask_from_tree): Likewise.
17522 * fold-const-call.c (fold_const_reduction): Likewise.
17523 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17524 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17525 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17526 (fold_relational_const): Likewise.
17527 (native_interpret_vector): Likewise. Change the size from an
17528 int to an unsigned int.
17529 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17530 TYPE_VECTOR_SUBPARTS.
17531 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17532 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17533 duplicating a non-constant operand into a variable-length vector.
17534 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17535 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17536 * ipa-icf.c (sem_variable::equals): Likewise.
17537 * match.pd: Likewise.
17538 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17539 * print-tree.c (print_node): Likewise.
17540 * stor-layout.c (layout_type): Likewise.
17541 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17542 * tree-cfg.c (verify_gimple_comparison): Likewise.
17543 (verify_gimple_assign_binary): Likewise.
17544 (verify_gimple_assign_ternary): Likewise.
17545 (verify_gimple_assign_single): Likewise.
17546 * tree-pretty-print.c (dump_generic_node): Likewise.
17547 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17548 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17549 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17550 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17551 (vect_shift_permute_load_chain): Likewise.
17552 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17553 (expand_vector_condition, optimize_vector_constructor): Likewise.
17554 (lower_vec_perm, get_compute_type): Likewise.
17555 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17556 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17557 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17558 (vect_recog_mask_conversion_pattern): Likewise.
17559 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17560 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17561 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17562 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17563 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17564 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17565 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17566 (supportable_widening_operation): Likewise.
17567 (supportable_narrowing_operation): Likewise.
17568 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17570 * varasm.c (output_constant): Likewise.
17572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17573 Alan Hayward <alan.hayward@arm.com>
17574 David Sherwood <david.sherwood@arm.com>
17576 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17577 so that both the length == 3 and length != 3 cases set up their
17578 own permute vectors. Add comments explaining why we know the
17579 number of elements is constant.
17580 (vect_permute_load_chain): Likewise.
17582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17583 Alan Hayward <alan.hayward@arm.com>
17584 David Sherwood <david.sherwood@arm.com>
17586 * machmode.h (mode_nunits): Change from unsigned char to
17588 (ONLY_FIXED_SIZE_MODES): New macro.
17589 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17590 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17591 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17593 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17594 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17595 or if measurement_type is not polynomial.
17596 * genmodes.c (ZERO_COEFFS): New macro.
17597 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17599 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17600 Use ZERO_COEFFS when emitting initializers.
17601 * data-streamer.h (bp_pack_poly_value): New function.
17602 (bp_unpack_poly_value): Likewise.
17603 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17604 for GET_MODE_NUNITS.
17605 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17606 for GET_MODE_NUNITS.
17607 * tree.c (make_vector_type): Remove temporary shim and make
17608 the real function take the number of units as a poly_uint64
17609 rather than an int.
17610 (build_vector_type_for_mode): Handle polynomial nunits.
17611 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17612 * emit-rtl.c (const_vec_series_p_1): Likewise.
17613 (gen_rtx_CONST_VECTOR): Likewise.
17614 * fold-const.c (test_vec_duplicate_folding): Likewise.
17615 * genrecog.c (validate_pattern): Likewise.
17616 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17617 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17618 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17619 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17620 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17621 * rtlanal.c (subreg_get_info): Likewise.
17622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17623 (vect_grouped_load_supported): Likewise.
17624 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17625 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17626 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17627 (simplify_const_unary_operation, simplify_binary_operation_1)
17628 (simplify_const_binary_operation, simplify_ternary_operation)
17629 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17630 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17631 instead of CONST_VECTOR_NUNITS.
17632 * varasm.c (output_constant_pool_2): Likewise.
17633 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17634 explicit-encoded elements in the XVEC for variable-length vectors.
17636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17638 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17641 Alan Hayward <alan.hayward@arm.com>
17642 David Sherwood <david.sherwood@arm.com>
17644 * coretypes.h (fixed_size_mode): Declare.
17645 (fixed_size_mode_pod): New typedef.
17646 * builtins.h (target_builtins::x_apply_args_mode)
17647 (target_builtins::x_apply_result_mode): Change type to
17648 fixed_size_mode_pod.
17649 * builtins.c (apply_args_size, apply_result_size, result_vector)
17650 (expand_builtin_apply_args_1, expand_builtin_apply)
17651 (expand_builtin_return): Update accordingly.
17653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17655 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17656 * cselib.c (cselib_hash_rtx): Likewise.
17657 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17658 CONST_VECTOR encoding.
17660 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17661 Jeff Law <law@redhat.com>
17664 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17665 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17666 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17667 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17670 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17671 explicitly probe *sp in a noreturn function if there were any callee
17672 register saves or frame pointer is needed.
17674 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17677 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17678 BLKmode for ternary, binary or unary expressions.
17681 * var-tracking.c (delete_vta_debug_insn): New inline function.
17682 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17683 insns from get_insns () to NULL instead of each bb separately.
17684 Use delete_vta_debug_insn. No longer static.
17685 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17686 delete_vta_debug_insns callers.
17687 * rtl.h (delete_vta_debug_insns): Declare.
17688 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17689 instead of variable_tracking_main.
17691 2018-01-03 Martin Sebor <msebor@redhat.com>
17693 PR tree-optimization/83603
17694 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17695 arguments past the endof the argument list in functions declared
17696 without a prototype.
17697 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17698 Avoid checking when arguments are null.
17700 2018-01-03 Martin Sebor <msebor@redhat.com>
17703 * doc/extend.texi (attribute const): Fix a typo.
17704 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17705 issuing -Wsuggest-attribute for void functions.
17707 2018-01-03 Martin Sebor <msebor@redhat.com>
17709 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17710 offset_int::from instead of wide_int::to_shwi.
17711 (maybe_diag_overlap): Remove assertion.
17712 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17713 * gimple-ssa-sprintf.c (format_directive): Same.
17714 (parse_directive): Same.
17715 (sprintf_dom_walker::compute_format_length): Same.
17716 (try_substitute_return_value): Same.
17718 2018-01-03 Jeff Law <law@redhat.com>
17720 PR middle-end/83654
17721 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17722 non-constant residual for zero at runtime and avoid probing in
17723 that case. Reorganize code for trailing problem to mirror handling
17726 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17728 PR tree-optimization/83501
17729 * tree-ssa-strlen.c (get_string_cst): New.
17730 (handle_char_store): Call get_string_cst.
17732 2018-01-03 Martin Liska <mliska@suse.cz>
17734 PR tree-optimization/83593
17735 * tree-ssa-strlen.c: Include tree-cfg.h.
17736 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17737 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17738 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17740 (strlen_dom_walker::before_dom_children): Call
17741 gimple_purge_dead_eh_edges. Dump tranformation with details
17743 (strlen_dom_walker::before_dom_children): Update call by adding
17744 new argument cleanup_eh.
17745 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17747 2018-01-03 Martin Liska <mliska@suse.cz>
17750 * cif-code.def (VARIADIC_THUNK): New enum value.
17751 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17754 2018-01-03 Jan Beulich <jbeulich@suse.com>
17756 * sse.md (mov<mode>_internal): Tighten condition for when to use
17757 vmovdqu<ssescalarsize> for TI and OI modes.
17759 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17761 Update copyright years.
17763 2018-01-03 Martin Liska <mliska@suse.cz>
17766 * ipa-visibility.c (function_and_variable_visibility): Skip
17767 functions with noipa attribure.
17769 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17771 * gcc.c (process_command): Update copyright notice dates.
17772 * gcov-dump.c (print_version): Ditto.
17773 * gcov.c (print_version): Ditto.
17774 * gcov-tool.c (print_version): Ditto.
17775 * gengtype.c (create_file): Ditto.
17776 * doc/cpp.texi: Bump @copying's copyright year.
17777 * doc/cppinternals.texi: Ditto.
17778 * doc/gcc.texi: Ditto.
17779 * doc/gccint.texi: Ditto.
17780 * doc/gcov.texi: Ditto.
17781 * doc/install.texi: Ditto.
17782 * doc/invoke.texi: Ditto.
17784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17786 * vector-builder.h (vector_builder::m_full_nelts): Change from
17787 unsigned int to poly_uint64.
17788 (vector_builder::full_nelts): Update prototype accordingly.
17789 (vector_builder::new_vector): Likewise.
17790 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17791 (vector_builder::operator ==): Likewise.
17792 (vector_builder::finalize): Likewise.
17793 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17794 Take the number of elements as a poly_uint64 rather than an
17796 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17797 from unsigned int to poly_uint64.
17798 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17799 (vec_perm_indices::new_vector): Likewise.
17800 (vec_perm_indices::length): Likewise.
17801 (vec_perm_indices::nelts_per_input): Likewise.
17802 (vec_perm_indices::input_nelts): Likewise.
17803 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17804 number of elements per input as a poly_uint64 rather than an
17805 unsigned int. Use the original encoding for variable-length
17806 vectors, rather than clamping each individual element.
17807 For the second and subsequent elements in each pattern,
17808 clamp the step and base before clamping their sum.
17809 (vec_perm_indices::series_p): Handle polynomial element counts.
17810 (vec_perm_indices::all_in_range_p): Likewise.
17811 (vec_perm_indices_to_tree): Likewise.
17812 (vec_perm_indices_to_rtx): Likewise.
17813 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17814 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17815 (tree_vector_builder::new_binary_operation): Handle polynomial
17816 element counts. Return false if we need to know the number
17817 of elements at compile time.
17818 * fold-const.c (fold_vec_perm): Punt if the number of elements
17819 isn't known at compile time.
17821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17823 * vec-perm-indices.h (vec_perm_builder): Change element type
17824 from HOST_WIDE_INT to poly_int64.
17825 (vec_perm_indices::element_type): Update accordingly.
17826 (vec_perm_indices::clamp): Handle polynomial element_types.
17827 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17828 (vec_perm_indices::all_in_range_p): Likewise.
17829 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17831 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17832 polynomial vec_perm_indices element types.
17833 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17834 * fold-const.c (fold_vec_perm): Likewise.
17835 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17836 * tree-vect-generic.c (lower_vec_perm): Likewise.
17837 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17838 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17839 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17846 rather than an int. Use plus_constant.
17847 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17848 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17851 Alan Hayward <alan.hayward@arm.com>
17852 David Sherwood <david.sherwood@arm.com>
17854 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17855 a HOST_WIDE_INT to a poly_int64.
17857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17858 Alan Hayward <alan.hayward@arm.com>
17859 David Sherwood <david.sherwood@arm.com>
17861 * calls.c (load_register_parameters): Cope with polynomial
17862 mode sizes. Require a constant size for BLKmode parameters
17863 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17864 forces a parameter to be padded at the lsb end in order to
17865 fill a complete number of words, require the parameter size
17866 to be ordered wrt UNITS_PER_WORD.
17868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17869 Alan Hayward <alan.hayward@arm.com>
17870 David Sherwood <david.sherwood@arm.com>
17872 * reload1.c (spill_stack_slot_width): Change element type
17873 from unsigned int to poly_uint64_pod.
17874 (alter_reg): Treat mode sizes as polynomial.
17876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17877 Alan Hayward <alan.hayward@arm.com>
17878 David Sherwood <david.sherwood@arm.com>
17880 * reload.c (complex_word_subreg_p): New function.
17881 (reload_inner_reg_of_subreg, push_reload): Use it.
17883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17884 Alan Hayward <alan.hayward@arm.com>
17885 David Sherwood <david.sherwood@arm.com>
17887 * lra-constraints.c (process_alt_operands): Reject matched
17888 operands whose sizes aren't ordered.
17889 (match_reload): Refer to this check here.
17891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17892 Alan Hayward <alan.hayward@arm.com>
17893 David Sherwood <david.sherwood@arm.com>
17895 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17896 that the mode size is in the set {1, 2, 4, 8, 16}.
17898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17899 Alan Hayward <alan.hayward@arm.com>
17900 David Sherwood <david.sherwood@arm.com>
17902 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17903 Use plus_constant instead of gen_rtx_PLUS.
17905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17906 Alan Hayward <alan.hayward@arm.com>
17907 David Sherwood <david.sherwood@arm.com>
17909 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17910 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17911 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17912 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17913 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17914 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17915 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17916 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17917 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17918 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17920 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17921 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17922 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17923 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17924 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17925 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17926 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17927 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17928 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17929 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17931 * expr.c (emit_move_resolve_push): Treat the input and result
17932 of PUSH_ROUNDING as a poly_int64.
17933 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17934 (emit_push_insn): Likewise.
17935 * lra-eliminations.c (mark_not_eliminable): Likewise.
17936 * recog.c (push_operand): Likewise.
17937 * reload1.c (elimination_effects): Likewise.
17938 * rtlanal.c (nonzero_bits1): Likewise.
17939 * calls.c (store_one_arg): Likewise. Require the padding to be
17940 known at compile time.
17942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17943 Alan Hayward <alan.hayward@arm.com>
17944 David Sherwood <david.sherwood@arm.com>
17946 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17947 Use plus_constant instead of gen_rtx_PLUS.
17949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17950 Alan Hayward <alan.hayward@arm.com>
17951 David Sherwood <david.sherwood@arm.com>
17953 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17954 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17961 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17962 via stack temporaries. Treat the mode size as polynomial too.
17964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17965 Alan Hayward <alan.hayward@arm.com>
17966 David Sherwood <david.sherwood@arm.com>
17968 * expr.c (expand_expr_real_2): When handling conversions involving
17969 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17970 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17971 as a poly_uint64 too.
17973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17974 Alan Hayward <alan.hayward@arm.com>
17975 David Sherwood <david.sherwood@arm.com>
17977 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17980 Alan Hayward <alan.hayward@arm.com>
17981 David Sherwood <david.sherwood@arm.com>
17983 * combine.c (can_change_dest_mode): Handle polynomial
17984 REGMODE_NATURAL_SIZE.
17985 * expmed.c (store_bit_field_1): Likewise.
17986 * expr.c (store_constructor): Likewise.
17987 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17988 and polynomial REGMODE_NATURAL_SIZE.
17989 (gen_lowpart_common): Likewise.
17990 * reginfo.c (record_subregs_of_mode): Likewise.
17991 * rtlanal.c (read_modify_subreg_p): Likewise.
17993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17994 Alan Hayward <alan.hayward@arm.com>
17995 David Sherwood <david.sherwood@arm.com>
17997 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17998 numbers of elements.
18000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18001 Alan Hayward <alan.hayward@arm.com>
18002 David Sherwood <david.sherwood@arm.com>
18004 * match.pd: Cope with polynomial numbers of vector elements.
18006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18007 Alan Hayward <alan.hayward@arm.com>
18008 David Sherwood <david.sherwood@arm.com>
18010 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18011 in a POINTER_PLUS_EXPR.
18013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18014 Alan Hayward <alan.hayward@arm.com>
18015 David Sherwood <david.sherwood@arm.com>
18017 * omp-simd-clone.c (simd_clone_subparts): New function.
18018 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18019 (ipa_simd_modify_function_body): Likewise.
18021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18022 Alan Hayward <alan.hayward@arm.com>
18023 David Sherwood <david.sherwood@arm.com>
18025 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18026 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18027 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18028 (expand_vector_condition, vector_element): Likewise.
18029 (subparts_gt): New function.
18030 (get_compute_type): Use subparts_gt.
18031 (count_type_subparts): Delete.
18032 (expand_vector_operations_1): Use subparts_gt instead of
18033 count_type_subparts.
18035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18036 Alan Hayward <alan.hayward@arm.com>
18037 David Sherwood <david.sherwood@arm.com>
18039 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18040 (vect_compile_time_alias): ...this new function. Do the calculation
18041 on poly_ints rather than trees.
18042 (vect_prune_runtime_alias_test_list): Update call accordingly.
18044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18045 Alan Hayward <alan.hayward@arm.com>
18046 David Sherwood <david.sherwood@arm.com>
18048 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18050 (vect_schedule_slp_instance): Likewise.
18052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18053 Alan Hayward <alan.hayward@arm.com>
18054 David Sherwood <david.sherwood@arm.com>
18056 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18057 constant and extern definitions for variable-length vectors.
18058 (vect_get_constant_vectors): Note that the number of units
18059 is known to be constant.
18061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18062 Alan Hayward <alan.hayward@arm.com>
18063 David Sherwood <david.sherwood@arm.com>
18065 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18066 of units as polynomial. Choose between WIDE and NARROW based
18069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18070 Alan Hayward <alan.hayward@arm.com>
18071 David Sherwood <david.sherwood@arm.com>
18073 * tree-vect-stmts.c (simd_clone_subparts): New function.
18074 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18077 Alan Hayward <alan.hayward@arm.com>
18078 David Sherwood <david.sherwood@arm.com>
18080 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18081 vectors as polynomial. Use build_index_vector for
18082 IFN_GOMP_SIMD_LANE.
18084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18085 Alan Hayward <alan.hayward@arm.com>
18086 David Sherwood <david.sherwood@arm.com>
18088 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18089 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18090 for variable-length vectors.
18091 (vectorizable_mask_load_store): Treat the number of units as
18092 polynomial, asserting that it is constant if the condition has
18093 already been enforced.
18094 (vectorizable_store, vectorizable_load): Likewise.
18096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18097 Alan Hayward <alan.hayward@arm.com>
18098 David Sherwood <david.sherwood@arm.com>
18100 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18101 of units as polynomial. Punt if we can't tell at compile time
18102 which vector contains the final result.
18104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18105 Alan Hayward <alan.hayward@arm.com>
18106 David Sherwood <david.sherwood@arm.com>
18108 * tree-vect-loop.c (vectorizable_induction): Treat the number
18109 of units as polynomial. Punt on SLP inductions. Use an integer
18110 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18111 cast of such a series for variable-length floating-point
18114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18115 Alan Hayward <alan.hayward@arm.com>
18116 David Sherwood <david.sherwood@arm.com>
18118 * tree.h (build_index_vector): Declare.
18119 * tree.c (build_index_vector): New function.
18120 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18121 of units as polynomial, forcibly converting it to a constant if
18122 vectorizable_reduction has already enforced the condition.
18123 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18124 to create a {1,2,3,...} vector.
18125 (vectorizable_reduction): Treat the number of units as polynomial.
18126 Choose vectype_in based on the largest scalar element size rather
18127 than the smallest number of units. Enforce the restrictions
18130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18131 Alan Hayward <alan.hayward@arm.com>
18132 David Sherwood <david.sherwood@arm.com>
18134 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18135 number of units as polynomial.
18137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18138 Alan Hayward <alan.hayward@arm.com>
18139 David Sherwood <david.sherwood@arm.com>
18141 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18142 * target.def (autovectorize_vector_sizes): Return the vector sizes
18143 by pointer, using vector_sizes rather than a bitmask.
18144 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18145 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18146 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18148 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18149 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18150 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18151 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18152 * omp-general.c (omp_max_vf): Likewise.
18153 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18154 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18155 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18156 * tree-vect-slp.c (vect_slp_bb): Likewise.
18157 * doc/tm.texi: Regenerate.
18158 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18160 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18161 the vector size as a poly_uint64 rather than an unsigned int.
18162 (current_vector_size): Change from an unsigned int to a poly_uint64.
18163 (get_vectype_for_scalar_type): Update accordingly.
18164 * tree.h (build_truth_vector_type): Take the size and number of
18165 units as a poly_uint64 rather than an unsigned int.
18166 (build_vector_type): Add a temporary overload that takes
18167 the number of units as a poly_uint64 rather than an unsigned int.
18168 * tree.c (make_vector_type): Likewise.
18169 (build_truth_vector_type): Take the number of units as a poly_uint64
18170 rather than an unsigned int.
18172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18173 Alan Hayward <alan.hayward@arm.com>
18174 David Sherwood <david.sherwood@arm.com>
18176 * target.def (get_mask_mode): Take the number of units and length
18177 as poly_uint64s rather than unsigned ints.
18178 * targhooks.h (default_get_mask_mode): Update accordingly.
18179 * targhooks.c (default_get_mask_mode): Likewise.
18180 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18181 * doc/tm.texi: Regenerate.
18183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18184 Alan Hayward <alan.hayward@arm.com>
18185 David Sherwood <david.sherwood@arm.com>
18187 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18188 * omp-general.c (omp_max_vf): Likewise.
18189 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18190 (expand_omp_simd): Handle polynomial safelen.
18191 * omp-low.c (omplow_simd_context): Add a default constructor.
18192 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18193 (lower_rec_simd_input_clauses): Update accordingly.
18194 (lower_rec_input_clauses): Likewise.
18196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18197 Alan Hayward <alan.hayward@arm.com>
18198 David Sherwood <david.sherwood@arm.com>
18200 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18201 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18202 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18203 (vect_analyze_slp_cost): Likewise.
18204 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18205 (vect_model_load_cost): Likewise.
18207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18208 Alan Hayward <alan.hayward@arm.com>
18209 David Sherwood <david.sherwood@arm.com>
18211 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18212 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18213 from an unsigned int * to a poly_uint64_pod *.
18214 (calculate_unrolling_factor): New function.
18215 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18218 Alan Hayward <alan.hayward@arm.com>
18219 David Sherwood <david.sherwood@arm.com>
18221 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18222 from an unsigned int to a poly_uint64.
18223 (_loop_vec_info::slp_unrolling_factor): Likewise.
18224 (_loop_vec_info::vectorization_factor): Change from an int
18226 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18227 (vect_get_num_vectors): New function.
18228 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18229 (vect_get_num_copies): Use vect_get_num_vectors.
18230 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18231 to an unsigned int *.
18232 (vect_analyze_data_refs): Change min_vf from an int * to a
18234 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18235 than an unsigned HOST_WIDE_INT.
18236 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18237 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18238 to an unsigned int *.
18239 (vect_analyze_data_ref_dependences): Likewise.
18240 (vect_compute_data_ref_alignment): Handle polynomial vf.
18241 (vect_enhance_data_refs_alignment): Likewise.
18242 (vect_prune_runtime_alias_test_list): Likewise.
18243 (vect_shift_permute_load_chain): Likewise.
18244 (vect_supportable_dr_alignment): Likewise.
18245 (dependence_distance_ge_vf): Take the vectorization factor as a
18246 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18247 (vect_analyze_data_refs): Change min_vf from an int * to a
18249 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18250 vfm1 as a poly_uint64 rather than an int. Make the same change
18251 for the returned bound_scalar.
18252 (vect_gen_vector_loop_niters): Handle polynomial vf.
18253 (vect_do_peeling): Likewise. Update call to
18254 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18255 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18257 * tree-vect-loop.c (vect_determine_vectorization_factor)
18258 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18259 (vect_get_known_peeling_cost): Likewise.
18260 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18261 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18262 (vect_transform_loop): Likewise. Use the lowest possible VF when
18263 updating the upper bounds of the loop.
18264 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18265 rather than an int.
18266 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18267 polynomial unroll factors.
18268 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18269 (vect_make_slp_decision): Likewise.
18270 (vect_supported_load_permutation_p): Likewise, and polynomial
18272 (vect_analyze_slp_cost): Handle polynomial vf.
18273 (vect_slp_analyze_node_operations): Likewise.
18274 (vect_slp_analyze_bb_1): Likewise.
18275 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18276 than an unsigned HOST_WIDE_INT.
18277 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18278 (vectorizable_load): Handle polynomial vf.
18279 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18281 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18284 Alan Hayward <alan.hayward@arm.com>
18285 David Sherwood <david.sherwood@arm.com>
18287 * match.pd: Handle bit operations involving three constants
18288 and try to fold one pair.
18290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18292 * tree-vect-loop-manip.c: Include gimple-fold.h.
18293 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18294 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18295 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18296 Add a path that uses a step of VF instead of 1, but disable it
18298 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18299 and niters_no_overflow parameters. Update calls to
18300 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18301 Create a new SSA name if the latter choses to use a ste other
18302 than zero, and return it via niters_vector_mult_vf_var.
18303 * tree-vect-loop.c (vect_transform_loop): Update calls to
18304 vect_do_peeling, vect_gen_vector_loop_niters and
18305 slpeel_make_loop_iterate_ntimes.
18306 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18307 (vect_gen_vector_loop_niters): Update declarations after above changes.
18309 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18311 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18312 128-bit round to integer instructions.
18313 (ceil<mode>2): Likewise.
18314 (btrunc<mode>2): Likewise.
18315 (round<mode>2): Likewise.
18317 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18319 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18320 unaligned VSX load/store on P8/P9.
18321 (expand_block_clear): Allow the use of unaligned VSX
18322 load/store on P8/P9.
18324 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18326 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18328 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18329 swap associated with both a load and a store.
18331 2018-01-02 Andrew Waterman <andrew@sifive.com>
18333 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18334 * config/riscv/riscv.md (clear_cache): Use it.
18336 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18338 * web.c: Remove out-of-date comment.
18340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18342 * expr.c (fixup_args_size_notes): Check that any existing
18343 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18344 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18345 (emit_single_push_insn): ...here.
18347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18349 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18350 (const_vector_encoded_nelts): New function.
18351 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18352 (const_vector_int_elt, const_vector_elt): Declare.
18353 * emit-rtl.c (const_vector_int_elt_1): New function.
18354 (const_vector_elt): Likewise.
18355 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18356 of CONST_VECTOR_ELT.
18358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18360 * expr.c: Include rtx-vector-builder.h.
18361 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18362 directly on the tree encoding.
18363 (const_vector_from_tree): Likewise.
18364 * optabs.c: Include rtx-vector-builder.h.
18365 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18366 sequence of "u" values.
18367 * vec-perm-indices.c: Include rtx-vector-builder.h.
18368 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18369 directly on the vec_perm_indices encoding.
18371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18373 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18374 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18375 * rtx-vector-builder.h: New file.
18376 * rtx-vector-builder.c: Likewise.
18377 * rtl.h (rtx_def::u2): Add a const_vector field.
18378 (CONST_VECTOR_NPATTERNS): New macro.
18379 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18380 (CONST_VECTOR_DUPLICATE_P): Likewise.
18381 (CONST_VECTOR_STEPPED_P): Likewise.
18382 (CONST_VECTOR_ENCODED_ELT): Likewise.
18383 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18384 (unwrap_const_vec_duplicate): Likewise.
18385 (const_vec_series_p): Check for a non-duplicated vector encoding.
18386 Say that the function only returns true for integer vectors.
18387 * emit-rtl.c: Include rtx-vector-builder.h.
18388 (gen_const_vec_duplicate_1): Delete.
18389 (gen_const_vector): Call gen_const_vec_duplicate instead of
18390 gen_const_vec_duplicate_1.
18391 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18392 (gen_const_vec_duplicate): Use rtx_vector_builder.
18393 (gen_const_vec_series): Likewise.
18394 (gen_rtx_CONST_VECTOR): Likewise.
18395 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18396 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18397 Build a new vector rather than modifying a CONST_VECTOR in-place.
18398 (handle_special_swappables): Update call accordingly.
18399 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18400 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18401 Build a new vector rather than modifying a CONST_VECTOR in-place.
18402 (handle_special_swappables): Update call accordingly.
18404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18406 * simplify-rtx.c (simplify_const_binary_operation): Use
18407 CONST_VECTOR_ELT instead of XVECEXP.
18409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18411 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18412 the selector elements to be different from the data elements
18413 if the selector is a VECTOR_CST.
18414 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18415 ssizetype for the selector.
18417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18419 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18420 before testing each element individually.
18421 * tree-vect-generic.c (lower_vec_perm): Likewise.
18423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18425 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18426 * selftest-run-tests.c (selftest::run_tests): Call it.
18427 * vector-builder.h (vector_builder::operator ==): New function.
18428 (vector_builder::operator !=): Likewise.
18429 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18430 (vec_perm_indices::all_from_input_p): New function.
18431 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18432 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18433 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18434 instead of reading the VECTOR_CST directly. Detect whether both
18435 vector inputs are the same before constructing the vec_perm_indices,
18436 and update the number of inputs argument accordingly. Use the
18437 utility functions added above. Only construct sel2 if we need to.
18439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18441 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18442 the broadcast of the low byte.
18443 (expand_mult_highpart): Use an explicit encoding for the permutes.
18444 * optabs-query.c (can_mult_highpart_p): Likewise.
18445 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18446 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18447 (vectorizable_bswap): Likewise.
18448 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18449 explicit encoding for the power-of-2 permutes.
18450 (vect_permute_store_chain): Likewise.
18451 (vect_grouped_load_supported): Likewise.
18452 (vect_permute_load_chain): Likewise.
18454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18456 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18457 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18458 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18459 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18460 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18461 (vect_gen_perm_mask_any): Likewise.
18463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18465 * int-vector-builder.h: New file.
18466 * vec-perm-indices.h: Include int-vector-builder.h.
18467 (vec_perm_indices): Redefine as an int_vector_builder.
18468 (auto_vec_perm_indices): Delete.
18469 (vec_perm_builder): Redefine as a stand-alone class.
18470 (vec_perm_indices::vec_perm_indices): New function.
18471 (vec_perm_indices::clamp): Likewise.
18472 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18473 (vec_perm_indices::new_vector): New function.
18474 (vec_perm_indices::new_expanded_vector): Update for new
18475 vec_perm_indices class.
18476 (vec_perm_indices::rotate_inputs): New function.
18477 (vec_perm_indices::all_in_range_p): Operate directly on the
18478 encoded form, without computing elided elements.
18479 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18480 encoding. Update for new vec_perm_indices class.
18481 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18482 the given vec_perm_builder.
18483 (expand_vec_perm_var): Update vec_perm_builder constructor.
18484 (expand_mult_highpart): Use vec_perm_builder instead of
18485 auto_vec_perm_indices.
18486 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18487 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18488 or double series encoding as appropriate.
18489 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18490 vec_perm_indices instead of auto_vec_perm_indices.
18491 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18492 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18493 (vect_permute_store_chain): Likewise.
18494 (vect_grouped_load_supported): Likewise.
18495 (vect_permute_load_chain): Likewise.
18496 (vect_shift_permute_load_chain): Likewise.
18497 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18498 (vect_transform_slp_perm_load): Likewise.
18499 (vect_schedule_slp_instance): Likewise.
18500 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18501 (vectorizable_mask_load_store): Likewise.
18502 (vectorizable_bswap): Likewise.
18503 (vectorizable_store): Likewise.
18504 (vectorizable_load): Likewise.
18505 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18506 vec_perm_indices instead of auto_vec_perm_indices. Use
18507 tree_to_vec_perm_builder to read the vector from a tree.
18508 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18509 vec_perm_builder instead of a vec_perm_indices.
18510 (have_whole_vector_shift): Use vec_perm_builder and
18511 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18512 truncation to calc_vec_perm_mask_for_shift.
18513 (vect_create_epilog_for_reduction): Likewise.
18514 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18515 from auto_vec_perm_indices to vec_perm_indices.
18516 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18517 instead of changing individual elements.
18518 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18519 the vector in d.perm.
18520 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18521 from auto_vec_perm_indices to vec_perm_indices.
18522 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18523 instead of changing individual elements.
18524 (arm_vectorize_vec_perm_const): Use new_vector to install
18525 the vector in d.perm.
18526 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18527 Update vec_perm_builder constructor.
18528 (rs6000_expand_interleave): Likewise.
18529 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18530 (rs6000_expand_interleave): Likewise.
18532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18534 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18535 to qimode could truncate the indices.
18536 * optabs.c (expand_vec_perm_var): Likewise.
18538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18540 * Makefile.in (OBJS): Add vec-perm-indices.o.
18541 * vec-perm-indices.h: New file.
18542 * vec-perm-indices.c: Likewise.
18543 * target.h (vec_perm_indices): Replace with a forward class
18545 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18546 * optabs.h: Include vec-perm-indices.h.
18547 (expand_vec_perm): Delete.
18548 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18549 (expand_vec_perm_const): Declare.
18550 * target.def (vec_perm_const_ok): Replace with...
18551 (vec_perm_const): ...this new hook.
18552 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18553 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18554 * doc/tm.texi: Regenerate.
18555 * optabs.def (vec_perm_const): Delete.
18556 * doc/md.texi (vec_perm_const): Likewise.
18557 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18558 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18559 expand_vec_perm for constant permutation vectors. Assert that
18560 the mode of variable permutation vectors is the integer equivalent
18561 of the mode that is being permuted.
18562 * optabs-query.h (selector_fits_mode_p): Declare.
18563 * optabs-query.c: Include vec-perm-indices.h.
18564 (selector_fits_mode_p): New function.
18565 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18566 is defined, instead of checking whether the vec_perm_const_optab
18567 exists. Use targetm.vectorize.vec_perm_const instead of
18568 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18569 fit in the vector mode before using a variable permute.
18570 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18571 vec_perm_indices instead of an rtx.
18572 (expand_vec_perm): Replace with...
18573 (expand_vec_perm_const): ...this new function. Take the selector
18574 as a vec_perm_indices rather than an rtx. Also take the mode of
18575 the selector. Update call to shift_amt_for_vec_perm_mask.
18576 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18577 Use vec_perm_indices::new_expanded_vector to expand the original
18578 selector into bytes. Check whether the indices fit in the vector
18579 mode before using a variable permute.
18580 (expand_vec_perm_var): Make global.
18581 (expand_mult_highpart): Use expand_vec_perm_const.
18582 * fold-const.c: Includes vec-perm-indices.h.
18583 * tree-ssa-forwprop.c: Likewise.
18584 * tree-vect-data-refs.c: Likewise.
18585 * tree-vect-generic.c: Likewise.
18586 * tree-vect-loop.c: Likewise.
18587 * tree-vect-slp.c: Likewise.
18588 * tree-vect-stmts.c: Likewise.
18589 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18591 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18592 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18593 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18594 (aarch64_vectorize_vec_perm_const): ...this new function.
18595 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18596 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18597 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18598 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18599 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18601 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18603 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18604 check for NEON modes.
18605 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18606 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18607 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18608 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18610 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18611 the old VEC_PERM_CONST conditions.
18612 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18613 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18614 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18615 (ia64_vectorize_vec_perm_const_ok): Merge into...
18616 (ia64_vectorize_vec_perm_const): ...this new function.
18617 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18618 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18619 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18620 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18621 * config/mips/mips.c (mips_expand_vec_perm_const)
18622 (mips_vectorize_vec_perm_const_ok): Merge into...
18623 (mips_vectorize_vec_perm_const): ...this new function.
18624 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18625 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18626 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18627 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18628 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18629 (rs6000_expand_vec_perm_const): Delete.
18630 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18632 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18633 (altivec_expand_vec_perm_const_le): Take each operand individually.
18634 Operate on constant selectors rather than rtxes.
18635 (altivec_expand_vec_perm_const): Likewise. Update call to
18636 altivec_expand_vec_perm_const_le.
18637 (rs6000_expand_vec_perm_const): Delete.
18638 (rs6000_vectorize_vec_perm_const_ok): Delete.
18639 (rs6000_vectorize_vec_perm_const): New function.
18640 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18641 an element count and rtx array.
18642 (rs6000_expand_extract_even): Update call accordingly.
18643 (rs6000_expand_interleave): Likewise.
18644 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18645 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18646 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18647 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18648 (rs6000_expand_vec_perm_const): Delete.
18649 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18650 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18651 (altivec_expand_vec_perm_const_le): Take each operand individually.
18652 Operate on constant selectors rather than rtxes.
18653 (altivec_expand_vec_perm_const): Likewise. Update call to
18654 altivec_expand_vec_perm_const_le.
18655 (rs6000_expand_vec_perm_const): Delete.
18656 (rs6000_vectorize_vec_perm_const_ok): Delete.
18657 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18658 reference to the SPE evmerge intructions.
18659 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18660 an element count and rtx array.
18661 (rs6000_expand_extract_even): Update call accordingly.
18662 (rs6000_expand_interleave): Likewise.
18663 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18664 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18666 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18668 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18670 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18671 vector mode and that that mode matches the mode of the data
18673 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18674 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18675 directly using expand_vec_perm_1 when forcing selectors into
18677 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18681 * optabs-query.h (can_vec_perm_p): Delete.
18682 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18683 * optabs-query.c (can_vec_perm_p): Split into...
18684 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18685 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18686 particular selector is valid.
18687 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18688 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18689 (vect_grouped_load_supported): Likewise.
18690 (vect_shift_permute_load_chain): Likewise.
18691 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18692 (vect_transform_slp_perm_load): Likewise.
18693 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18694 (vectorizable_bswap): Likewise.
18695 (vect_gen_perm_mask_checked): Likewise.
18696 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18697 implementations of variable permutation vectors into account
18698 when deciding which selector to use.
18699 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18700 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18701 with a false third argument.
18702 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18703 to test whether the constant selector is valid and can_vec_perm_var_p
18704 to test whether a variable selector is valid.
18706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18708 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18709 * optabs-query.c (can_vec_perm_p): Likewise.
18710 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18711 instead of vec_perm_indices.
18712 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18713 (vect_gen_perm_mask_checked): Likewise,
18714 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18715 (vect_gen_perm_mask_checked): Likewise,
18717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18719 * optabs-query.h (qimode_for_vec_perm): Declare.
18720 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18721 (qimode_for_vec_perm): ...this new function.
18722 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18724 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18726 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18727 does not have a conditional at the top.
18729 2018-01-02 Richard Biener <rguenther@suse.de>
18731 * ipa-inline.c (big_speedup_p): Fix expression.
18733 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18736 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18739 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18743 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18744 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18745 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18746 cond_taken_branch_cost 3->4.
18748 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18750 PR tree-optimization/83581
18751 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18752 TODO_cleanup_cfg if any changes have been made.
18754 PR middle-end/83608
18755 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18756 convert_modes if target mode has the right side, but different mode
18759 PR middle-end/83609
18760 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18761 last argument when extracting from CONCAT. If either from_real or
18762 from_imag is NULL, use expansion through memory. If result is not
18763 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18764 the parts directly to inner mode, if even that fails, use expansion
18767 PR middle-end/83623
18768 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18769 check for bswap in mode rather than HImode and use that in expand_unop
18772 Copyright (C) 2018 Free Software Foundation, Inc.
18774 Copying and distribution of this file, with or without modification,
18775 are permitted in any medium without royalty provided the copyright
18776 notice and this notice are preserved.