1 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
3 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4 Use a simpler align directive also if MAXSKIP = ALIGN-1.
5 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
12 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17 2018-07-04 Martin Liska <mliska@suse.cz>
18 Jonathan Wakely <jwakely@redhat.com>
20 * coverage.c: Use correct type.
21 * doc/invoke.texi: Language correction.
23 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
26 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
27 ENDBRANCH for non-tail call which may return via indirect branch.
28 * doc/extend.texi: Document indirect_return attribute.
30 2018-07-03 Martin Sebor <msebor@redhat.com>
32 PR tree-optimization/86274
33 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
35 (format_floating): Correct handling of infinities and NaNs.
37 2018-07-03 Martin Sebor <msebor@redhat.com>
39 * print-tree.c (print_real_cst): New function.
40 (print_node_brief): Call it.
43 2018-07-03 Jeff Law <law@redhat.com>
45 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
46 into a single pattern.
48 * config/h8300/h8300.md (ors code_iterator): New.
49 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
50 a single pattern and single splitter.
51 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
52 (iorqi3_1, xorqi3_1): Likewise.
53 (iorqi3, xorqi3 expanders): Similarly.
55 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
56 (movmd_internal) into a single pattern using the P mode iterator.
57 (movmd splitters): Similarly.
58 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
59 (movsd splitters): Similarly.
61 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
62 ADDB, ADDW and ADDL into a single ADD attribute which selects the
63 right table based on the size of the operand.
64 * config/h8300/h8300.md (length_table): Corresponding changes. All
65 references to "addb", "addw" and "addl" changed to "add".
66 (btst patterns): Merge two variants into a single pattern.
67 (tstqi, tsthi): Likewise.
68 (addhi3_incdec, addsi3_incdec): Likewise.
69 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
70 (mulhi3, mulsi3): Likewise.
71 (udivhi3, udivsi3): Likewise.
72 (divhi3, divsi3): Likewise.
73 (andorqi3, andorhi3, andorsi3): Likewise.
75 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
78 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
79 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
81 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
83 PR tree-optimization/85694
84 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
85 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
87 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
88 (<u>avg<mode>3_ceil): New patterns.
90 2018-07-03 David Malcolm <dmalcolm@redhat.com>
92 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
93 scan-tree-dump directive.
94 * gcc.dg/vect/slp-perm-2.c: Likewise.
95 * gcc.dg/vect/slp-perm-3.c: Likewise.
96 * gcc.dg/vect/slp-perm-5.c: Likewise.
97 * gcc.dg/vect/slp-perm-6.c: Likewise.
98 * gcc.dg/vect/slp-perm-7.c: Likewise.
99 * gcc.dg/vect/slp-perm-8.c: Likewise.
101 2018-07-03 Marek Polacek <polacek@redhat.com>
104 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
106 2018-07-03 Richard Biener <rguenther@suse.de>
109 * tree-ssa-structalias.c (find_func_clobbers): Properly
110 handle indirect calls.
112 2018-07-03 Jeff Law <law@redhat.com>
114 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
115 (shifts): New code iterator.
116 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
117 expander. Fix HImode handling on H8/SX.
118 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
119 (subqi3, subhi3, subsi3 expanders): Likewise.
120 (andqi3, andhi3, andsi3 expanders): Likewise.
121 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
122 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
123 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
124 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
125 (zero_extendqihi2, zero_extendqisi2): Likewise.
126 (extendqihi2, extendqisi2): Likewise.
127 (rotlqi3, rotlhi3, rotlsi3): Likewise.
128 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
129 (rotlqi3_1, rotlhi3_1): Likewise.
130 (logicalhi3_sn, logicalsi3_sn): Likewise.
131 (logicalhi3, logicalsi3): Likewise.
133 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
135 * tree-vect-patterns.c (vect_recog_rotate_pattern)
136 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
137 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
138 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
139 type to append_pattern_def_seq instead of creating a stmt_vec_info
141 (build_mask_conversion): Likewise. Remove vinfo argument.
142 (vect_add_conversion_to_patterm): Likewise, renaming to...
143 (vect_add_conversion_to_pattern): ...this.
144 (vect_recog_mask_conversion_pattern): Update call to
145 build_mask_conversion. Pass the vector type to
146 append_pattern_def_seq here too.
147 (vect_recog_gather_scatter_pattern): Update call to
148 vect_add_conversion_to_pattern.
150 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
152 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
153 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
154 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
155 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
156 STMT_VINFO_PATTERN_DEF_SEQ to null here.
157 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
158 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
159 append_pattern_def_seq instead of new_pattern_def_seq.
160 (vect_recog_divmod_pattern): Do both of the above.
161 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
164 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
166 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
167 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
168 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
169 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
170 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
171 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
172 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
173 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
174 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
175 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
176 parameter with a single stmt_vec_info.
177 (vect_recog_func_ptr): Likewise.
178 (vect_recog_gather_scatter_pattern): Likewise, folding in...
179 (vect_try_gather_scatter_pattern): ...this.
180 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
181 the stmt_vec_info of the statement to be matched. Don't clear
182 STMT_VINFO_RELATED_STMT.
183 (vect_pattern_recog): Update call accordingly.
185 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
187 PR tree-optimization/85694
188 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
189 (uavgM3_ceil): Document new optabs.
190 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
191 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
193 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
194 (savg_ceil_optab): New optabs.
195 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
196 (vect_vect_recog_func_ptrs): Add it.
197 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
198 constant directly from the associated lhs.
200 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
202 * tree-vect-patterns.c (vect_split_statement): New function.
203 (vect_convert_input): Use it to try to split an existing cast.
205 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
207 * poly-int.h (print_hex): New function.
208 * dumpfile.h (dump_dec, dump_hex): Declare.
209 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
210 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
211 min_input_precision, operation_precision and operation_sign.
212 * tree-vect-patterns.c (vect_get_range_info): New function.
213 (vect_same_loop_or_bb_p, vect_single_imm_use)
214 (vect_operation_fits_smaller_type): Delete.
215 (vect_look_through_possible_promotion): Add an optional
216 single_use_p parameter.
217 (vect_recog_over_widening_pattern): Rewrite to use new
218 stmt_vec_info infomration. Handle one operation at a time.
219 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
220 (vect_truncatable_operation_p, vect_set_operation_type)
221 (vect_set_min_input_precision): New functions.
222 (vect_determine_min_output_precision_1): Likewise.
223 (vect_determine_min_output_precision): Likewise.
224 (vect_determine_precisions_from_range): Likewise.
225 (vect_determine_precisions_from_users): Likewise.
226 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
227 (vect_vect_recog_func_ptrs): Put over_widening first.
229 (vect_pattern_recog): Call vect_determine_precisions.
231 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
233 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
234 statements that have been replaced by further pattern statements.
235 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
237 2018-07-03 Richard Biener <rguenther@suse.de>
239 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
240 always set *dt. Dump vectype in vectype overload.
241 * dumpfile.h (dump_gimple_expr): New function.
242 (dump_gimple_expr_loc): Likewise.
243 * dumpfile.c (dump_gimple_expr): New function.
244 (dump_gimple_expr_loc): Likewise.
246 2018-07-02 Jeff Law <law@redhat.com>
248 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
249 the H8/300, H8/300H and H8/S variants into a single pattern.
250 (movhi_h8300, movqi_h8300hs): Similarly.
251 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
252 (QHI mode iterator): New.
254 * config/h8300/h8300.md: Remove trailing whitespace.
256 2018-07-02 Jim Wilson <jimw@sifive.com>
258 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
259 instead of emit_insn for interrupt returns.
260 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
261 (riscv_sret, riscv_uret): Likewise.
263 2018-07-02 David Malcolm <dmalcolm@redhat.com>
265 * pretty-print.c (selftest::test_pp_format): Move save and restore
266 of quotes to class auto_fix_quotes, and add an instance.
267 * selftest.c: Include "intl.h".
268 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
269 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
270 * selftest.h (selftest::auto_fix_quotes): New class.
272 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
274 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
275 (aarch64_sve_prepare_conditional_op): Remove.
276 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
277 Allow aarch64_simd_reg_or_zero as select operand; remove
278 the aarch64_sve_prepare_conditional_op call.
279 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
280 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
281 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
282 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
283 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
284 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
285 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
286 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
287 and a splitters to match all of the *_any patterns.
288 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
290 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
291 (SVE_COND_FP_BINARY_REV): Remove.
292 (sve_int_op_rev, sve_fp_op_rev): New.
293 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
294 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
295 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
296 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
297 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
298 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
299 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
300 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
301 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
303 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
304 Remove match_dup 1 from the inner unspec.
305 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
307 * config/aarch64/aarch64.md (movprfx): New attr.
308 (length): Default movprfx to 8.
309 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
310 (*madd<SVE_I>, *msub<SVE_I): Likewise.
311 (*<su>mul<SVE_I>3_highpart): Likewise.
312 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
313 (*v<ASHIFT><SVE_I>3): Likewise.
314 (*<su><MAXMIN><SVE_I>3): Likewise.
315 (*<su><MAXMIN><SVE_F>3): Likewise.
316 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
317 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
318 (*div<SVE_F>4): Likewise.
320 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
322 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
325 2018-07-02 Richard Biener <rguenther@suse.de>
327 PR tree-optimization/86363
328 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
329 memset argument refers to a non-variable address.
331 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
333 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
334 {PLUS,MINUS}_EXPR code to...
335 (adjust_symbolic_bound): ...here,
336 (combine_bound): ...here,
337 (set_value_range_with_overflow): ...and here.
339 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
341 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
343 (extract_range_from_abs_expr): ...here.
345 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
347 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
348 -fno-omit-frame-pointer when not optimizing.
350 2018-07-02 Martin Liska <mliska@suse.cz>
353 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
354 (propagate_nothrow): Likewise.
356 2018-07-02 Martin Liska <mliska@suse.cz>
359 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
361 2018-07-02 David Malcolm <dmalcolm@redhat.com>
363 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
364 function in r262149, changing "loc" param from source_location to
365 const dump_location_t &.
366 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
367 declaration, as above.
369 2018-07-01 Paul Koning <ni1d@arrl.net>
371 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
372 -munit-asm, -mgnu-asm, -mdec-asm.
373 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
374 (pdp11_output_labelref): New.
375 (pdp11_output_def): New.
376 (pdp11_output_addr_vec_elt): New.
377 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
378 %# and %@ format codes.
379 (pdp11_option_override): New.
380 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
381 (pdp11_output_ident): New.
382 (pdp11_asm_named_section): New.
383 (pdp11_asm_init_sections): New.
384 (pdp11_file_start): New.
385 (pdp11_file_end): New.
386 (output_ascii): Use .ascii/.asciz for -mdec-asm.
387 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
388 %o, like %c but octal.
389 (pdp11_option_override): New.
390 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
392 (DATA_SECTION_ASM_OP): Ditto.
393 (READONLY_DATA_SECTION_ASM_OP): New.
394 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
395 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
396 (ASM_OUTPUT_LABELREF): Ditto.
397 (ASM_OUTPUT_DEF): Ditto.
398 (ASM_OUTPUT_EXTERNAL): New.
399 (ASM_OUTPUT_SOURCE_FILENAME): New.
400 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
401 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
402 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
403 %# and %@ format codes.
404 * config/pdp11/pdp11.opt (mgnu-asm): New.
405 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
406 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
407 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
409 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
411 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
412 dereferencing path[] beyond its length.
414 (debug_all_paths): New.
415 (rewire_first_differing_edge): New.
416 (adjust_paths_after_duplication): New.
417 (duplicate_thread_path): Call adjust_paths_after_duplication.
419 (thread_through_all_blocks): Add new argument to
420 duplicate_thread_path.
422 2018-06-30 Jim Wilson <jimw@sifive.com>
424 * config/riscv/predicates.md (p2m1_shift_operand): New.
425 (high_mask_shift_operand): New.
426 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
427 pattern using p2m1_shift_operand.
428 (lshsi3_zero_extend_3+2): New combiner pattern using
429 high_mask_shift_operand.
431 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
433 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
435 (vect_recog_rotate_pattern): ...here.
436 (vect_convert_input): Try to insert casts of invariants in the
438 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
439 preheader to be empty.
441 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
443 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
444 vector type. If given, install it in the new statement's
446 (vect_element_precision): New function.
447 (vect_unpromoted_value): New struct.
448 (vect_unpromoted_value::vect_unpromoted_value): New function.
449 (vect_unpromoted_value::set_op): Likewise.
450 (vect_look_through_possible_promotion): Likewise.
451 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
452 (vect_widened_op_tree, vect_convert_input): Likewise.
453 (vect_convert_inputs, vect_convert_output): Likewise.
454 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
455 to handle the optional cast of the multiplication result and
456 vect_widened_op_tree to detect the widened multiplication itself.
457 Do not require the input and output of promotion casts to have
458 the same sign, but base the signedness of the operation on the
459 input rather than the result. If the pattern includes two
460 promotions, check that those promotions have the same sign.
461 Do not restrict the MULT_EXPR handling to a double-width result;
462 handle quadruple-width results and wider. Use vect_convert_inputs
463 to convert the inputs to the common type.
464 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
465 to handle the optional cast of the ABS result. Also allow a sign
466 change or a sign extension between the ABS and MINUS.
467 Use vect_widened_op_tree to detect the widened subtraction and use
468 vect_convert_inputs to convert the inputs to the common type.
469 (vect_handle_widen_op_by_const): Delete.
470 (vect_recog_widen_op_pattern): New function.
471 (vect_recog_widen_mult_pattern): Use it.
472 (vect_recog_widen_shift_pattern): Likewise.
473 (vect_recog_widen_sum_pattern): Use
474 vect_look_through_possible_promotion to handle the promoted
477 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
479 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
480 the containing gimple_seq *.
481 * gimple-iterator.h (gsi_for_stmt): Declare it.
482 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
483 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
484 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
485 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
486 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
488 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
490 (vect_mark_pattern_stmts): ...here. Handle cases in which the
491 statement being replaced is part of an existing pattern
492 definition sequence, inserting the new pattern statements before
494 (vect_pattern_recog_1): Don't return a bool. If the statement
495 is already part of a pattern, instead apply pattern matching
496 to the pattern definition statements. Don't clear the
497 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
498 (vect_pattern_recog): Don't break after the first match;
499 continue processing the pattern definition statements instead.
500 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
502 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
504 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
505 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
506 (vect_recog_widen_sum_pattern): Use it.
508 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
510 * tree-vect-loop.c (vectorizable_reduction): Assert that the
511 phi is not a pattern statement and has not been replaced by
513 * tree-vect-patterns.c (type_conversion_p): Don't check
514 STMT_VINFO_IN_PATTERN_P.
515 (vect_recog_vector_vector_shift_pattern): Likewise.
516 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
517 the pattern statement rather than the original statement; check
518 directly for a WIDEN_MULT_EXPR here.
519 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
520 vect_is_simple_use to return the pattern statement rather
521 than the original statement; use is_pattern_stmt_p to check
522 for such a pattern statement.
523 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
524 to return the pattern statement rather than the original statement;
525 don't do the same transformation here.
526 (vect_is_simple_use): If the defining statement has been replaced
527 by a pattern statement, return the pattern statement instead.
528 Remove the corresponding (local) transformation from the vectype
531 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
533 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
534 end and default to null.
535 * tree-vect-loop.c (vect_create_epilog_for_reduction)
536 (vectorizable_reduction): Update calls accordingly, dropping the
537 gimple ** argument if the passed-back statement isn't needed.
538 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
539 (vect_recog_rotate_pattern): Likewise.
540 (vect_recog_mask_conversion_pattern): Likewise.
541 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
542 (vect_mask_constant_operand_p): Likewise.
543 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
544 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
545 (get_group_load_store_type, get_load_store_type): Likewise.
546 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
547 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
548 (vectorizable_conversion, vectorizable_assignment): Likewise.
549 (vectorizable_shift, vectorizable_operation): Likewise.
550 (vectorizable_store, vect_is_simple_cond): Likewise.
551 (vectorizable_condition, vectorizable_comparison): Likewise.
552 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
553 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
554 and move it to the end. Cope with null def_stmt_outs.
556 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
558 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
560 2018-06-29 Jeff Law <law@redhat.com>
562 * config/v850/v850.c (v850_legitimate_address_p): Handle large
563 displacements for TARGET_V850E2V3 and newer.
564 (TARGET_LRA_P): Remove. Defaults to LRA now.
565 * config/v850/v850.md (sign23byte_load): Remove.
566 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
567 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
569 2018-06-29 Martin Liska <mliska@suse.cz>
572 * coverage.c (coverage_init): Mangle full path name.
573 * doc/invoke.texi: Document the change.
574 * gcov-io.c (mangle_path): New.
575 * gcov-io.h (mangle_path): Likewise.
576 * gcov.c (mangle_name): Use mangle_path for path mangling.
578 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
580 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
581 if starting source register is not even.
583 2018-06-29 Martin Liska <mliska@suse.cz>
585 PR tree-optimization/86263
586 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
587 Make edge redirection.
589 2018-06-29 David Malcolm <dmalcolm@redhat.com>
591 * dumpfile.c (dump_loc): Add indentation based on scope depth.
592 (dump_scope_depth): New variable.
593 (get_dump_scope_depth): New function.
594 (dump_begin_scope): New function.
595 (dump_end_scope): New function.
596 * dumpfile.h (get_dump_scope_depth): New declaration.
597 (dump_begin_scope): New declaration.
598 (dump_end_scope): New declaration.
599 (class auto_dump_scope): New class.
600 (AUTO_DUMP_SCOPE): New macro.
601 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
604 2018-06-29 Richard Biener <rguenther@suse.de>
606 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
607 compute_all_dependences succeeds.
608 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
609 exceed --param loop-max-datarefs-for-datadeps.
611 2018-06-29 Jakub Jelinek <jakub@redhat.com>
613 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
615 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
618 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
619 alternative 0 in preferred_for_speed attribute.
621 2018-06-28 Paul Koning <ni1d@arrl.net>
623 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
624 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
625 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
626 * config/pdp11/pdp11.md: Correct "length" attribute calculation
627 for shift insn patterns.
629 2018-06-28 David Malcolm <dmalcolm@redhat.com>
631 * cgraph.c (cgraph_node::get_body): Replace assignments to
632 "dump_file" with calls to set_dump_file.
633 * dumpfile.c (alt_dump_file): Make static, and group with...
634 (alt_flags): ...this definition.
635 (dumps_are_enabled): New variable.
636 (refresh_dumps_are_enabled): New function.
637 (set_dump_file): New function.
638 (set_alt_dump_file): New function.
639 (gcc::dump_manager::dump_start): Replace assignments to
640 "dump_file" and "alt_dump_file" with calls to set_dump_file and
642 (gcc::dump_manager::dump_finish): Likewise.
643 * dumpfile.h (alt_dump_file): Delete decl.
644 (dumps_are_enabled): New variable decl.
645 (set_dump_file): New function decl.
646 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
648 * tree-nested.c (lower_nested_functions): Replace assignments to
649 "dump_file" with calls to set_dump_file.
651 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
653 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
654 goto_locus of each outgoing edge of each basic block.
656 2018-06-28 Richard Biener <rguenther@suse.de>
658 * dwarf2out.c (decl_scope_table): Remove.
659 (push_decl_scope): Likewise.
660 (pop_decl_scope): Likewise.
661 (gen_type_die_for_member): Do not call push/pop_decl_scope.
662 (gen_struct_or_union_type_die): Likewise.
663 (gen_tagged_type_die): Likewise.
664 (dwarf2out_init): Do not initialize decl_scope_table.
665 (dwarf2out_c_finalize): Do not free it.
667 2018-06-28 Richard Biener <rguenther@suse.de>
669 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
670 deciding whether to not re-use a DIE.
672 2018-06-28 Richard Biener <rguenther@suse.de>
674 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
675 DW_AT_abstract_origin attribute.
677 2018-06-28 Martin Liska <mliska@suse.cz>
679 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
680 Use newly introduced constants.
681 * tree-switch-conversion.h (struct jump_table_cluster):
682 Define max_ratio_for_size and max_ratio_for_speed.
684 2018-06-28 Martin Liska <mliska@suse.cz>
686 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
687 Add new checking assert to catch invalid state.
688 (jump_table_cluster::can_be_handled): Handle single case
690 (jump_table_cluster::is_beneficial): Bail out for such case.
691 (bit_test_cluster::find_bit_tests):
692 Add new checking assert to catch invalid state.
693 (bit_test_cluster::can_be_handled): Handle single case
695 (bit_test_cluster::is_beneficial): Bail out for such case.
696 (switch_decision_tree::analyze_switch_statement):
699 2018-06-28 Martin Liska <mliska@suse.cz>
701 * common.opt: Introduce -completion option.
702 * gcc.c (driver_handle_option): Handle it.
703 (driver::main): Print completions if completion
705 * opt-suggestions.c (option_proposer::get_completions):
707 (option_proposer::suggest_completion): Likewise.
708 (option_proposer::find_param_completions): Likewise.
709 (verify_autocompletions): Likewise.
710 (test_completion_valid_options): Likewise.
711 (test_completion_valid_params): Likewise.
712 (in_completion_p): Likewise.
713 (empty_completion_p): Likewise.
714 (test_completion_partial_match): Likewise.
715 (test_completion_garbage): Likewise.
716 (opt_proposer_c_tests): Likewise.
717 * opt-suggestions.h: Declare new functions.
718 * opts.c (common_handle_option): Handle OPT__completion_.
719 * selftest-run-tests.c (selftest::run_tests): Add
720 opt_proposer_c_tests.
721 * selftest.c (assert_str_startswith): New.
722 * selftest.h (assert_str_startswith): Likewise.
723 (opt_proposer_c_tests): New.
724 (ASSERT_STR_STARTSWITH): Likewise.
726 2018-06-28 Martin Liska <mliska@suse.cz>
728 * Makefile.in: Add opt-suggestions.o.
729 * gcc-main.c: Include opt-suggestions.h.
730 * gcc.c (driver::driver): Likewise.
731 (driver::~driver): Remove m_option_suggestions.
732 (driver::build_option_suggestions): Moved to option_proposer.
733 (driver::suggest_option): Likewise.
734 (driver::handle_unrecognized_options): Use option_proposer.
735 * gcc.h (class driver): Add new memver m_option_proposer.
736 * opt-suggestions.c: New file.
737 * opt-suggestions.h: New file.
739 2018-06-28 Martin Liska <mliska@suse.cz>
741 * vec.h (class auto_string_vec): New (moved from auto_argvec).
742 (auto_string_vec::~auto_string_vec): Likewise.
744 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
746 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
747 prevent_decl_creation_for_types fields up and add reset_location field.
748 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
749 statement if id->reset_location is true.
750 (copy_edges_for_bb): Do not set goto_locus on the new edges if
751 id->reset_location is true.
752 (copy_phis_for_bb): Force input_location on the arguments if
753 id->reset_location is true.
754 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
755 is set on the function to be inlined.
757 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
759 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
761 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
763 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
765 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
766 hard registers for the clobbered pseudo.
768 2018-06-27 Paul Koning <ni1d@arrl.net>
770 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
771 mutually exclusive options.
772 * config/pdp11/constraints.md (h): New constraint.
773 (O): Update definition to match shift code generation.
775 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
777 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
779 (output_jump): Change arguments.
780 (pdp11_fixed_cc_regs): New function.
781 (pdp11_cc_mode): Ditto.
782 (pdp11_expand_shift): Ditto.
783 (pdp11_assemble_shift): Ditto.
784 (pdp11_small_shift): Ditto.
785 (pdp11_branch_cost): Remove.
786 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
788 (pdp11_register_move_cost): Update for CC registers.
789 (pdp11_rtx_costs): Add case for LSHIFTRT.
790 (pdp11_output_jump): Add CCNZ mode conditional branches.
791 (notice_update_cc_on_set): Remove.
792 (pdp11_cc_mode): New function.
793 (simple_memory_operand): Correct pre/post decrement case.
794 (no_side_effect_operand): New function.
795 (pdp11_regno_reg_class): Add CC_REGS class.
796 (pdp11_fixed_cc_regs): New function.
797 (pdp11_small_shift): New function.
798 (pdp11_expand_shift): New function to expand shift insns.
799 (pdp11_assemble_shift): New function to output shifts.
800 (pdp11_branch_cost): Remove.
801 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
802 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
804 (PTRDIFF_TYPE): Ditto.
805 (ADJUST_INSN_LENGTH): New macro.
806 (FIXED_REGISTERS): Add CC registers.
807 (CALL_USED_REGISTERS): Ditto.
809 (REG_CLASS_NAMES): Ditto.
810 (REG_CLASS_CONTENTS): Ditto.
811 (SELECT_CC_MODE): Use new function.
812 (TARGET_FLAGS_REGNUM): New macro.
813 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
814 (cc0_reg_rtx): Remove.
815 (CC_STATUS_MDEP): Remove.
816 (CC_STATUS_MDEFP_INIT): Remove.
818 (NOTICE_UPDATE_CC): Remove.
819 (REGISTER_NAMES): Add CC registers.
820 (BRANCH_COST): Change to constant 1.
821 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
823 * config/pdp11/pdp11.opt (mbcopy): Remove.
824 (mbcopy-builtin): Remove.
825 (mbranch-cheap): Remove.
826 (mbranch-expensive): Remove.
827 * config/pdp11/predicates.md (expand_shift_operand): Update to
828 match shift code generation.
829 (ccnz_operator): New predicate.
830 * doc/invoke.texi (PDP-11 Options): Remove deleted options
831 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
832 Remove non-existent option -mabshi, -mno-abshi. Document mutually
834 * doc/md.texi (PDP-11): Document new D and h constraints. Update
835 description of O constraint.
837 2018-06-27 Jeff Law <law@redhat.com>
838 Austin Law <austinklaw@gmail.com>
840 * config/v850/v850.md (addsi3_set_flags): New pattern.
841 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
842 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
843 (zero_extendhisi2_v850_set_flags): Likewise.
844 (zero_extendqisi2_v850_set_flags): Likewise.
845 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
846 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
847 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
849 * config/v850/v850-protos.h (notice_update_cc): Remove.
850 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
851 (v850_print_operand): Handle 'D' and "d".
852 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
853 Add handling of arithmetic/logical operations compared against zero.
854 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
855 Do not look at v850_compare_op, instead get mode from last argument.
856 (v850_gen_compare): Remove
857 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
858 after reload for prologue insns.
859 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
861 (construct_save_jarl): Likewise.
862 (TARGET_FLAGS_REGNUM): Define.
863 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
864 (NOTICE_UPDATE_CC): Remove.
865 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
866 than cc0. Conditionalize on reload_completed.
867 (cmpsi_insn, setfcc_insn): Likewise.
868 (tst1 splitter): Turn into define_and_split which sets the flags
870 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
871 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
872 (cstoresf4, cstoredf4): Clobber the flags.
873 (cmpsi, cmpsf, cmpdf): Remove expanders.
874 (setf_insn): Remove pattern.
875 (addsi3): Turn into define_and_split which clobbers the flags after
876 reload and a suitable pattern (addsi3_clobber_flags) for use after
878 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
879 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
880 (ashrsi3, ashrsi3_v850e2): Likewise.
881 (bins): Clobber the flags.
882 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
883 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
884 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
885 (call_value_internal_short, call_value_internal_long): Likewise.
886 (callt_save_interrupt, callt_return_interrupt): Likewise.
887 (save_interrupt, return_interrupt): Likewise.
888 (callt_save_all_interrupt, save_all_interrupt): Likewise.
889 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
890 (restore_all_interrupt, _restore_all_interrupt): Likewise.
891 (All FP comparisons): Only allow after reload has completed.
893 (divh, divhu): Tweak output template.
894 (branch_z_normal, branch_z_invert): Remove
895 (branch_nz_normal, branch_nz_invert): Likewise.
896 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
898 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
899 * config/v850/v850.c (notice_update_cc): Remove.
900 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
901 (CC_NO_CARRY): Likewise.
902 (NOTICE_UPDATE_CC): Define to nothing.
903 * config/v850/v850.md: Remove block comment on cc0 handling
904 Remove "cc" attribute from all patterns. Remove cc_status handling
905 from all patterns. Minor formatting fixes.
907 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
909 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
910 (cortex-a76.cortex-a55): Likewise.
911 * config/aarch64/aarch64-tune.md: Regenerate.
912 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
913 cortex-a76.cortex-a55.
915 2018-06-27 Jeff Law <law@redhat.com>
917 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
918 (MULTILIB_DIRNAMES): Similarly.
920 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
922 * gimple.h (gimple_return_retbnd): Delete.
923 (gimple_return_set_retbnd): Likewise.
924 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
925 gimple_return_set_retbnd.
926 * gimple-pretty-print.c (dump_gimple_return): Remove call to
927 gimple_return_retbnd and adjust.
928 * tree-inline.h (struct copy_body_data): Remove retbnd field.
929 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
930 Explicitly return NULL in a couple more cases. Move assertion
931 on debug statements and remove unreachable code.
932 (reset_debug_binding): Do not test id->retbnd.
933 (expand_call_inline): Do not set it.
935 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
937 * configure.ac: Add --disable-gcov option.
938 * configure: Regenerate.
939 * Makefile.in: Honour @enable_gcov@.
940 * doc/install.texi: Document --disable-gcov.
942 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
944 * config/arm/arm-cpus.in (cortex-a76): New entry.
945 (cortex-a76.cortex-a55): Likewise.
946 * config/arm/arm-tables.opt: Regenerate.
947 * config/arm/arm-tune.md: Likewise.
948 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
949 * doc/invoke.texi (ARM Options): Document cortex-a76 and
950 cortex-a76.cortex-a55.
952 2018-06-27 Tamar Christina <tamar.christina@arm.com>
955 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
957 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
959 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
961 (EPILOGUE_USES): Likewise.
963 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
965 * tree-inline.c (remap_location): New function extracted from...
966 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
967 (copy_phis_for_bb): ...here. Call remap_location.
968 (copy_cfg_body): Adjust call to copy_edges_for_bb.
970 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
972 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
973 unaligned vsx for 16B memset.
975 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
978 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
979 ieee128_float_type_node to long_double_type_node unless
980 TARGET_LONG_DOUBLE_128 is set.
982 2018-06-26 David Malcolm <dmalcolm@redhat.com>
984 * cfgloop.c (get_loop_location): Convert return type from
985 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
986 by implicit construction from rtx_insn *, and using
987 dump_user_location_t::from_function_decl for the fallback case.
988 * cfgloop.h (get_loop_location): Convert return type from
989 location_t to dump_user_location_t.
990 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
991 dump_printf_loc to pass in a dump_location_t rather than a
992 location_t, via the gimple stmt.
993 * coverage.c (get_coverage_counts): Update calls to
994 dump_printf_loc to pass in dump_location_t rather than a
996 * doc/optinfo.texi (Dump types): Convert example of
997 dump_printf_loc from taking "locus" to taking "insn". Update
998 description of the "_loc" calls to cover dump_location_t.
999 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1001 (dump_user_location_t::dump_user_location_t): New constructors,
1002 from gimple *stmt and rtx_insn *.
1003 (dump_user_location_t::from_function_decl): New function.
1004 (dump_loc): Make static.
1005 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1006 const dump_location_t &.
1007 (dump_generic_expr_loc): Delete.
1008 (dump_printf_loc): Convert param "loc" from location_t to
1009 const dump_location_t &.
1010 (selftest::test_impl_location): New function.
1011 (selftest::dumpfile_c_tests): New function.
1012 * dumpfile.h: Include "profile-count.h".
1013 (class dump_user_location_t): New class.
1014 (struct dump_impl_location_t): New struct.
1015 (class dump_location_t): New class.
1016 (dump_printf_loc): Convert 2nd param from source_location to
1017 const dump_location_t &.
1018 (dump_generic_expr_loc): Delete.
1019 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1020 const dump_location_t &.
1021 * gimple-fold.c (fold_gimple_assign): Update call to
1022 dump_printf_loc to pass in a dump_location_t rather than a
1023 location_t, via the gimple stmt.
1024 (gimple_fold_call): Likewise.
1025 * gimple-loop-interchange.cc
1026 (loop_cand::analyze_iloop_reduction_var): Update for change to
1027 check_reduction_path.
1028 (tree_loop_interchange::interchange): Update for change to
1030 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1031 change in return-type of find_loop_location.
1032 (graphite_regenerate_ast_isl): Likewise.
1033 * graphite-optimize-isl.c (optimize_isl): Likewise.
1034 * graphite.c (graphite_transform_loops): Likewise.
1035 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1036 pass in a dump_location_t rather than a location_t, via the
1038 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1039 * ipa.c (walk_polymorphic_call_targets): Likewise.
1040 * loop-unroll.c (report_unroll): Convert "locus" param from
1041 location_t to dump_location_t.
1042 (decide_unrolling): Update for change to get_loop_location's
1044 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1045 location_t to dump_user_location_t.
1046 (grid_find_single_omp_among_assignments_1): Updates calls to
1047 dump_printf_loc to pass in a dump_location_t rather than a
1048 location_t, via the gimple stmt.
1049 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1050 location_t to dump_location_t. Updates calls to dump_printf_loc
1051 to pass in a dump_location_t rather than a location_t, via the
1053 (grid_inner_loop_gridifiable_p): Likewise.
1054 (grid_dist_follows_simple_pattern): Likewise.
1055 (grid_gfor_follows_tiling_pattern): Likewise.
1056 (grid_target_follows_gridifiable_pattern): Likewise.
1057 (grid_attempt_target_gridification): Convert initialization
1058 of local "grid" from memset to zero-initialization; FIXME: does
1059 this require C++11? Update call to dump_printf_loc to pass in a
1060 optinfo_location rather than a location_t, via the gimple stmt.
1061 * profile.c (read_profile_edge_counts): Updates call to
1062 dump_printf_loc to pass in a dump_location_t rather than a
1064 (compute_branch_probabilities): Likewise.
1065 * selftest-run-tests.c (selftest::run_tests): Call
1067 * selftest.h (dumpfile_c_tests): New decl.
1068 * tree-loop-distribution.c (pass_loop_distribution::execute):
1069 Update for change in return type of find_loop_location.
1070 * tree-parloops.c (parallelize_loops): Likewise.
1071 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1072 "locus" from location_t to dump_user_location_t.
1073 (canonicalize_loop_induction_variables): Likewise.
1074 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1075 for change in return type of find_loop_location.
1076 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1077 to dump_printf_loc to pass in a dump_location_t rather than a
1078 location_t, via the stmt.
1079 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1081 * tree-vect-loop-manip.c (find_loop_location): Convert return
1082 type from source_location to dump_user_location_t.
1083 (vect_do_peeling): Update for above change.
1084 (vect_loop_versioning): Update for change in type of
1086 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1087 from location_t to dump_user_location_t.
1088 (vect_estimate_min_profitable_iters): Update for change in type
1090 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1091 location_t to dump_location_t.
1092 (vect_slp_bb): Update for change in type of vect_location.
1093 * tree-vectorizer.c (vect_location): Convert from source_location
1094 to dump_user_location_t.
1095 (try_vectorize_loop_1): Update for change in vect_location's type.
1096 (vectorize_loops): Likewise.
1097 (increase_alignment): Likewise.
1098 * tree-vectorizer.h (vect_location): Convert from source_location
1099 to dump_user_location_t.
1100 (find_loop_location): Convert return type from source_location to
1101 dump_user_location_t.
1102 (check_reduction_path): Convert 1st param from location_t to
1103 dump_user_location_t.
1104 * value-prof.c (check_counter): Update call to dump_printf_loc to
1105 pass in a dump_user_location_t rather than a location_t; update
1106 call to error_at for change in type of "locus".
1107 (check_ic_target): Update call to dump_printf_loc to
1108 pass in a dump_user_location_t rather than a location_t, via the
1111 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1113 * config/s390/s390.h (enum processor_flags): Do not use
1114 default tune parameter when -march was specified.
1116 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1119 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1120 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1123 2018-06-26 Richard Biener <rguenther@suse.de>
1125 PR tree-optimization/86287
1127 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1128 (vect_analyze_loop): Initialize n_stmts.
1130 2018-06-26 Richard Biener <rguenther@suse.de>
1133 * fold-const.c (fold_convertible_p): Pointer extension
1136 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1139 * dwarf2out.c (loc_list_has_views): Adjust comments.
1140 (dw_loc_list): Split single cross-partition range with
1143 2018-06-25 Jeff Law <law@redhat.com>
1145 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1146 on -mbig-switch by default.
1148 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1150 (const_float_0_operand): Remove unused predicate.
1151 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1152 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1153 (recipsf2): New expander. Original pattern now called
1155 (recipdf2, recipdf2_insn): Similarly.
1156 (rsqrtsf2, rsqrtsf2_insn): Similarly
1157 (rsqrtdf2, rsqrtdf2_insn): Similarly
1159 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1161 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1162 Simplify logic for FreeBSD (twice).
1164 2018-06-25 Martin Sebor <msebor@redhat.com>
1166 PR tree-optimization/86204
1167 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1168 a strnlen result if it's less than the length of the string.
1170 2018-06-25 Martin Sebor <msebor@redhat.com>
1172 PR tree-optimization/85700
1173 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1174 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1175 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1177 2018-06-25 Martin Sebor <msebor@redhat.com>
1179 * doc/extend.texi (Zero-length arrays): Update and clarify.
1181 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1183 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1184 added IEEE/IBM long double multilib support on PowerPC little
1185 endian Linux systems.
1186 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1187 (MULTILIB_DEFAULTS): Likewise.
1188 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1190 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1191 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1192 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1194 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1197 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1198 (REORDER_45): Likewise.
1200 2018-06-25 Jeff Law <law@redhat.com>
1202 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1203 dividend to 32 bits. Adjust length.
1204 (udivmodhi4): Cleanup output template. Fix length.
1206 2018-06-25 Carl Love <cel@us.ibm.com>
1208 * config/rs6000/vsx.md: Change word selector to prefered location.
1210 2018-06-25 Richard Biener <rguenther@suse.de>
1212 PR tree-optimization/86304
1213 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1214 epilogue-if-converted loops as well.
1216 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1218 * lto-section-out.c (lto_begin_section): Do not print section
1219 name for noaddr and unnumbered dumps.
1221 2018-06-25 Richard Biener <rguenther@suse.de>
1223 * tree-vectorizer.h (struct vec_info_shared): New structure
1224 with parts split out from struct vec_info and loop_nest from
1225 struct _loop_vec_info.
1226 (struct vec_info): Adjust accordingly.
1227 (struct _loop_vec_info): Likewise.
1228 (LOOP_VINFO_LOOP_NEST): Adjust.
1229 (LOOP_VINFO_DATAREFS): Likewise.
1230 (LOOP_VINFO_DDRS): Likewise.
1231 (struct _bb_vec_info): Likewise.
1232 (BB_VINFO_DATAREFS): Likewise.
1233 (BB_VINFO_DDRS): Likewise.
1234 (struct _stmt_vec_info): Add dr_aux member.
1235 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1236 (DR_MISALIGNMENT_UNINITIALIZED): New.
1237 (set_dr_misalignment): Adjust.
1238 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1239 (vect_analyze_loop): Adjust prototype.
1240 (vect_analyze_loop_form): Likewise.
1241 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1242 Compute dependences lazily.
1243 (vect_record_base_alignments): Use shared datarefs/ddrs.
1244 (vect_verify_datarefs_alignment): Likewise.
1245 (vect_analyze_data_refs_alignment): Likewise.
1246 (vect_analyze_data_ref_accesses): Likewise.
1247 (vect_analyze_data_refs): Likewise.
1248 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1249 constructor parameter for shared part.
1250 (vect_analyze_loop_form): Pass in shared part and adjust.
1251 (vect_analyze_loop_2): Pass in storage for the number of
1252 stmts. Move loop nest finding to the caller. Compute
1254 (vect_analyze_loop): Pass in shared part.
1255 (vect_transform_loop): Verify shared datarefs are unchanged.
1256 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1257 constructor parameter for shared part.
1258 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1259 (vect_slp_bb): Verify shared datarefs are unchanged before
1261 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1263 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1264 DR_MISALIGNMENT_UNINITIALIZED.
1265 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1266 parameter for shared part.
1267 (vec_info::~vec_info): Adjust.
1268 (vec_info_shared::vec_info_shared): New.
1269 (vec_info_shared::~vec_info_shared): Likewise.
1270 (vec_info_shared::save_datarefs): Likewise.
1271 (vec_info_shared::check_datarefs): Likewise.
1272 (try_vectorize_loop_1): Construct shared part live for analyses
1273 of a single loop for multiple vector sizes.
1274 * tree-parloops.c (gather_scalar_reductions): Adjust.
1276 2018-06-25 Richard Biener <rguenther@suse.de>
1278 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1279 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1280 (vect_analyze_data_refs): Remove similar code from here and
1281 simplify accordingly.
1283 2018-06-25 Richard Biener <rguenther@suse.de>
1285 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1286 for reverse storage order accesses rather than asserting
1287 they cannot happen here.
1289 2018-06-25 Tom de Vries <tdevries@suse.de>
1292 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1293 Use data16 instead of .byte for insn prefix.
1295 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1298 * parser.c (make_char_string_pack): Pass this literal chars
1299 through cpp_interpret_string.
1300 (cp_parser_userdef_numeric_literal): Check the result of
1301 make_char_string_pack.
1303 2018-06-24 Maya Rashish <coypu@sdf.org>
1305 * ginclude/stddef.h: Simplify conditions around avoiding
1306 re-definition of __size_t.
1308 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1310 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1311 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1313 2018-06-22 Maya Rashish <coypu@sdf.org>
1315 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1316 made to OpenBSD and NetBSD through the years.
1318 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1320 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1321 behavior of vec_pack (vector double, vector double) to match
1322 behavior of vec_float2 (vector double, vector double).
1324 2018-06-22 Olivier Hainque <hainque@adacore.com>
1326 * gimplify.c (gimplify_function_tree): Prevent creation
1327 of a trampoline for the address of the current function
1328 passed to entry/exit instrumentation hooks.
1330 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1333 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1336 2018-06-22 Martin Liska <mliska@suse.cz>
1338 PR tree-optimization/86263
1339 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1340 Bail out if is_enabled is false.
1341 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1343 (jump_table_cluster::is_enabled): New function.
1345 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1347 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1348 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1349 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1350 (lto_input_ts_binfo_tree_pointers): Likewise.
1351 * tree-streamer-out.c (streamer_write_tree_bitfields,
1352 write_ts_binfo_tree_pointers): Likewise.
1353 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1355 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1357 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1359 2018-06-22 Martin Liska <mliska@suse.cz>
1361 * symbol-summary.h (get): Make it pure and inline move
1362 functionality from ::get function.
1363 (get): Remove and inline into ::get and ::get_create.
1364 (get_create): Move code from ::get function.
1366 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1369 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1370 -x assembler-with-cpp.
1372 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1374 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1375 _FILE_OFFSET_BITS=64 for C++.
1377 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1379 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1380 conversion insn that shows up when pr85657-3.c is compiled using
1381 IEEE 128-bit long double.
1382 (neg<mode>2_internal): Use the correct mode to check whether the
1383 mode is IBM extended.
1384 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1385 multiply and divide external functions from being created more
1388 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1390 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1392 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1393 the edge can be forwarded.
1394 (cfg_layout_merge_blocks): Likewise.
1396 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1398 * except.c (finish_eh_generation): Commit edge insertions only after
1399 the EH edges have been redirected from post-landing to landing pads.
1401 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1403 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1404 create_tmp_var_for to create the FRAME decl.
1405 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1407 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1409 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1410 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1411 debug statement when resetting its value.
1412 (expand_call_inline): Copy the locus of the call onto the assignment
1413 of the return value, if any. Use local variable in more cases.
1415 2018-06-21 Martin Liska <mliska@suse.cz>
1417 * ipa-pure-const.c (propagate_nothrow): Use
1418 funct_state_summaries->get.
1419 (dump_malloc_lattice): Likewise.
1420 (propagate_malloc): Likewise.
1422 2018-06-21 Richard Biener <rguenther@suse.de>
1424 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1425 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1426 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1428 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1429 BLOCK_ABSTRACT_ORIGIN unconditionally.
1431 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1433 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1435 * ipa-reference.c (ipa_reference_c_finalize): Delete
1436 ipa_ref_opt_sum_summaries and set it to NULL.
1438 2018-06-21 Tom de Vries <tdevries@suse.de>
1440 PR tree-optimization/85859
1441 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1442 test with comment from bb_no_side_effects_p.
1444 2018-06-21 Richard Biener <rguenther@suse.de>
1446 PR tree-optimization/86232
1447 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1448 max for constant niter.
1450 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1452 * config/aarch64/aarch64-simd.md
1453 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1455 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1457 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1458 Make opernads of the unspec commutative.
1460 2018-06-21 Richard Biener <rguenther@suse.de>
1462 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1463 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1464 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1465 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1467 (vect_record_base_alignments): Do not record base alignment
1468 for gathers or scatters.
1469 (vect_compute_data_ref_alignment): Drop return value that is always
1470 true. Bail out early for gathers or scatters.
1471 (vect_enhance_data_refs_alignment): Bail out early for gathers
1473 (vect_find_same_alignment_drs): Likewise.
1474 (vect_analyze_data_refs_alignment): Remove dead code.
1475 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1476 (vect_analyze_data_refs): For possible gathers or scatters do
1477 not create an alternate DR, just check their possible validity
1478 and mark them. Adjust DECL_NONALIASED handling to not rely
1480 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1481 update inits of gathers or scatters.
1482 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1483 Also copy gather/scatter flag to pattern vinfo.
1485 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1487 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1488 behavior of vec_packsu (vector unsigned long long, vector unsigned
1489 long long) to match behavior of vec_packs with same signature.
1491 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1492 Thomas Schwinge <thomas@codesourcery.com>
1493 Cesar Philippidis <cesar@codesourcery.com>
1495 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1496 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1497 (gimplify_adjust_omp_clauses): Likewise.
1498 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1499 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1500 (gimplify_omp_target_update): Update handling of acc update and
1502 * omp-low.c (install_var_field): Remove unused parameter
1503 base_pointers_restrict.
1504 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1505 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1507 (omp_target_base_pointers_restrict_p): Delete.
1508 (scan_omp_target): Update call to scan_sharing_clauses.
1509 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1511 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1512 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1513 (convert_local_omp_clauses): Likewise.
1514 * tree-pretty-print.c (dump_omp_clause): Likewise.
1515 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1517 (omp_clause_code_name): Likewise.
1519 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1522 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1525 PR tree-optimization/86231
1526 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1527 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1529 2018-06-20 Tom de Vries <tdevries@suse.de>
1531 PR tree-optimization/86097
1532 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1533 iv type if signedness of iv type is not the same as that of *nit.
1535 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1537 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1538 EDGE_EH edges, verify they are all EDGE_EH.
1540 2018-06-20 Maya Rashish <coypu@sdf.org>
1542 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1544 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1546 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1547 * config/aarch64/aarch64.c (xgene1_tunings): Add
1548 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1549 (aarch64_mode_valid_for_sched_fusion_p):
1550 Allow 16-byte modes.
1551 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1552 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1554 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1556 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1557 * config/aarch64/iterators.md (VQ2): New mode iterator.
1559 2018-06-20 Martin Liska <mliska@suse.cz>
1561 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1562 Change default ratio from 10 to 8.
1564 2018-06-20 Martin Liska <mliska@suse.cz>
1566 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1568 (bit_test_cluster::find_bit_tests): Likewise.
1569 (switch_decision_tree::analyze_switch_statement): Find clusters.
1570 * tree-switch-conversion.h (struct jump_table_cluster): Document
1573 2018-06-20 Martin Liska <mliska@suse.cz>
1575 * tree-switch-conversion.c (switch_conversion::collect):
1576 Record m_uniq property.
1577 (switch_conversion::expand): Bail out for special conditions.
1578 (group_cluster::~group_cluster): New.
1579 (group_cluster::group_cluster): Likewise.
1580 (group_cluster::dump): Likewise.
1581 (jump_table_cluster::emit): New.
1582 (switch_decision_tree::fix_phi_operands_for_edges): New.
1583 (struct case_node): Remove struct.
1584 (jump_table_cluster::can_be_handled): New.
1585 (case_values_threshold): Moved to header.
1586 (reset_out_edges_aux): Likewise.
1587 (jump_table_cluster::is_beneficial): New.
1588 (bit_test_cluster::can_be_handled): Likewise.
1589 (add_case_node): Remove.
1590 (bit_test_cluster::is_beneficial): New.
1591 (case_bit_test::cmp): New.
1592 (bit_test_cluster::emit): New.
1593 (expand_switch_as_decision_tree_p): Remove.
1594 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1595 (fix_phi_operands_for_edge): Likewise.
1596 (switch_decision_tree::analyze_switch_statement): New.
1597 (compute_cases_per_edge): Move ...
1598 (switch_decision_tree::compute_cases_per_edge): ... here.
1599 (try_switch_expansion): Likewise.
1600 (switch_decision_tree::try_switch_expansion): Likewise.
1601 (record_phi_operand_mapping): Likewise.
1602 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1603 (emit_case_decision_tree): Likewise.
1604 (switch_decision_tree::emit): Likewise.
1605 (balance_case_nodes): Likewise.
1606 (switch_decision_tree::balance_case_nodes): Likewise.
1607 (dump_case_nodes): Likewise.
1608 (switch_decision_tree::dump_case_nodes): Likewise.
1609 (emit_jump): Likewise.
1610 (switch_decision_tree::emit_jump): Likewise.
1611 (emit_cmp_and_jump_insns): Likewise.
1612 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1613 (emit_case_nodes): Likewise.
1614 (switch_decision_tree::emit_case_nodes): Likewise.
1615 (conditional_probability): Remove.
1616 * tree-switch-conversion.h (enum cluster_type): New.
1618 (struct cluster): Likewise.
1619 (cluster::cluster): Likewise.
1620 (struct simple_cluster): Likewise.
1621 (simple_cluster::simple_cluster): Likewise.
1622 (struct group_cluster): Likewise.
1623 (struct jump_table_cluster): Likewise.
1624 (struct bit_test_cluster): Likewise.
1625 (struct min_cluster_item): Likewise.
1626 (struct case_tree_node): Likewise.
1627 (case_tree_node::case_tree_node): Likewise.
1628 (jump_table_cluster::case_values_threshold): Likewise.
1629 (struct case_bit_test): Likewise.
1630 (struct switch_decision_tree): Likewise.
1631 (struct switch_conversion): Likewise.
1632 (switch_decision_tree::reset_out_edges_aux): Likewise.
1634 2018-06-20 Martin Liska <mliska@suse.cz>
1636 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1637 (hoist_edge_and_branch_if_true): Likewise.
1638 (expand_switch_using_bit_tests_p): Likewise.
1639 (struct case_bit_test): Likewise.
1640 (case_bit_test_cmp): Likewise.
1641 (emit_case_bit_tests): Likewise.
1642 (switch_conversion::switch_conversion): New class.
1643 (struct switch_conv_info): Remove old struct.
1644 (collect_switch_conv_info): More to ...
1645 (switch_conversion::collect): ... this.
1646 (check_range): Likewise.
1647 (switch_conversion::check_range): Likewise.
1648 (check_all_empty_except_final): Likewise.
1649 (switch_conversion::check_all_empty_except_final): Likewise.
1650 (check_final_bb): Likewise.
1651 (switch_conversion::check_final_bb): Likewise.
1652 (create_temp_arrays): Likewise.
1653 (switch_conversion::create_temp_arrays): Likewise.
1654 (free_temp_arrays): Likewise.
1655 (gather_default_values): Likewise.
1656 (switch_conversion::gather_default_values): Likewise.
1657 (build_constructors): Likewise.
1658 (switch_conversion::build_constructors): Likewise.
1659 (constructor_contains_same_values_p): Likewise.
1660 (switch_conversion::contains_same_values_p): Likewise.
1661 (array_value_type): Likewise.
1662 (switch_conversion::array_value_type): Likewise.
1663 (build_one_array): Likewise.
1664 (switch_conversion::build_one_array): Likewise.
1665 (build_arrays): Likewise.
1666 (switch_conversion::build_arrays): Likewise.
1667 (gen_def_assigns): Likewise.
1668 (switch_conversion::gen_def_assigns): Likewise.
1669 (prune_bbs): Likewise.
1670 (switch_conversion::prune_bbs): Likewise.
1671 (fix_phi_nodes): Likewise.
1672 (switch_conversion::fix_phi_nodes): Likewise.
1673 (gen_inbound_check): Likewise.
1674 (switch_conversion::gen_inbound_check): Likewise.
1675 (process_switch): Use the newly created class.
1676 (switch_conversion::expand): New.
1677 (switch_conversion::~switch_conversion): New.
1678 * tree-switch-conversion.h: New file.
1680 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1682 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1683 tree-vect-patterns.c.
1684 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1685 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1686 (vect_recog_sad_pattern): Likewise.
1687 (vect_recog_widen_sum_pattern): Likewise.
1688 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1689 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1690 (vect_recog_rotate_pattern): Likewise.
1691 (vect_recog_mult_pattern): Likewise.
1692 (vect_recog_vector_vector_shift_pattern): Likewise.
1693 (vect_recog_divmod_pattern): Likewise.
1694 (vect_recog_mixed_size_cond_pattern): Likewise.
1695 (vect_recog_bool_pattern): Likewise.
1696 (vect_recog_mask_conversion_pattern): Likewise.
1697 (vect_try_gather_scatter_pattern): Likewise.
1698 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1699 (vect_recog_over_widening_pattern): Likewise.
1700 (vect_recog_gather_scatter_pattern): Likewise.
1701 (vect_recog_func_ptr): Move from tree-vectorizer.h
1702 (vect_vect_recog_func_ptrs): Move further down the file.
1703 (vect_recog_func): Likewise. Remove the third argument.
1704 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1705 (vect_pattern_recog_1): Expect the pattern function to do any
1706 necessary target tests. Also expect it to provide a vector type.
1707 Remove the type_in handling.
1709 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1711 * tree-vect-patterns.c (vect_pattern_detected): New function.
1712 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1713 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1714 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1715 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1716 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1717 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1718 (vect_recog_mask_conversion_pattern)
1719 (vect_try_gather_scatter_pattern): Likewise.
1721 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1723 * tree-vect-patterns.c (vect_get_internal_def): New function.
1724 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1725 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1726 (search_type_for_mask_1): Use it.
1728 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1730 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1731 redundant WIDEN_SUM_EXPR handling.
1732 (vect_recog_sad_pattern): Likewise.
1734 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1736 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1737 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1738 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1739 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1740 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1742 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1744 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1745 use the stmt_vec_info of the original bb statement for the
1746 new zero assignment, even if the call is part of a pattern.
1748 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1750 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1751 that the sequence is attached to the original statement rather
1752 than the pattern statement.
1753 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1754 PATTERN_DEF_SEQ from the original statement rather than
1755 the main pattern statement.
1756 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1757 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1758 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1760 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1762 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1763 definition statements before the early exit for statements that aren't
1765 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1767 (vect_transform_loop): ...here. Process pattern definition
1768 statements without first checking whether the main pattern
1769 statement is live or relevant.
1771 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1773 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1774 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1776 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1778 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1779 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1780 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1781 (expand_block_compare): Change select_block_compare_mode call.
1782 (expand_strncmp_align_check): Use new functions, fix comment.
1783 (emit_final_str_compare_gpr): New function.
1784 (expand_strn_compare): Refactor and clean up code.
1785 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1787 2018-06-19 Tony Reix <tony.reix@atos.com>
1788 Damien Bergamini <damien.bergamini@atos.com>
1789 David Edelsohn <dje.gcc@gmail.com>
1791 * collect2.c (static_obj): New variable.
1792 (static_libs): New variable.
1793 (is_in_list): Uncomment declaration.
1794 (main): Track AIX libraries linked statically.
1795 (is_in_list): Uncomment definition.
1796 (scan_prog_file): Don't add AIX shared libraries initializer
1797 to constructor list if linking statically.
1799 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1801 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1803 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1805 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1807 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1810 2018-06-19 Martin Liska <mliska@suse.cz>
1812 * config/i386/i386.c (ix86_can_inline_p): Do not use
1813 ipa_fn_summaries::get_create.
1814 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1816 (devirtualization_time_bonus): Likewise.
1817 (ipcp_propagate_stage): Likewise.
1818 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1819 (edge_set_predicate): Likewise.
1820 (evaluate_conditions_for_known_args): Likewise.
1821 (evaluate_properties_for_edge): Likewise.
1822 (ipa_call_summary::reset): Tranform to ...
1823 (ipa_call_summary::~ipa_call_summary): ... this.
1824 (ipa_fn_summary::reset): Transform to ...
1825 (ipa_fn_summary::~ipa_fn_summary): ... this.
1826 (ipa_fn_summary_t::remove): Rename to ...
1827 (ipa_fn_summary_t::remove_callees): ... this.
1828 (ipa_fn_summary_t::duplicate): Use placement new
1829 instead of memory copy.
1830 (ipa_call_summary_t::duplicate): Likewise.
1831 (ipa_call_summary_t::remove): Remove.
1832 (dump_ipa_call_summary): Change get_create to get.
1833 (ipa_dump_fn_summary): Dump only when summary exists.
1834 (analyze_function_body): Use symbol_summary::get instead
1836 (compute_fn_summary): Likewise.
1837 (estimate_edge_devirt_benefit): Likewise.
1838 (estimate_edge_size_and_time): Likewise.
1839 (inline_update_callee_summaries): Likewise.
1840 (remap_edge_change_prob): Likewise.
1841 (remap_edge_summaries): Likewise.
1842 (ipa_merge_fn_summary_after_inlining): Likewise.
1843 (write_ipa_call_summary): Likewise.
1844 (ipa_fn_summary_write): Likewise.
1845 (ipa_free_fn_summary): Likewise.
1846 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1847 (struct ipa_call_summary): Likewise.
1848 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1850 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1851 (estimate_size_after_inlining): Likewise.
1852 (estimate_growth): Likewise.
1853 (growth_likely_positive): Likewise.
1854 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1855 (inline_call): Likewise.
1856 * ipa-inline.c (caller_growth_limits): Likewise.
1857 (can_inline_edge_p): Likewise.
1858 (can_inline_edge_by_limits_p): Likewise.
1859 (compute_uninlined_call_time): Likewise.
1860 (compute_inlined_call_time): Likewise.
1861 (want_inline_small_function_p): Likewise.
1862 (edge_badness): Likewise.
1863 (update_caller_keys): Likewise.
1864 (update_callee_keys): Likewise.
1865 (inline_small_functions): Likewise.
1866 (inline_to_all_callers_1): Likewise.
1867 (dump_overall_stats): Likewise.
1868 (early_inline_small_functions): Likewise.
1869 (early_inliner): Likewise.
1870 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1871 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1872 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1873 * ipa-split.c (execute_split_functions): Likewise.
1874 * symbol-summary.h: Likewise.
1875 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1877 2018-06-19 Richard Biener <rguenther@suse.de>
1879 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1880 (vectorize_loops): ... here. Fix dbgcnt handling.
1881 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1883 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1886 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1887 ieee128 argument takes up only one (vector) register, not two (floating
1890 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1892 * gimplify.c (gimplify_init_constructor): Really never clear for an
1893 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1895 2018-06-19 Richard Biener <rguenther@suse.de>
1897 PR tree-optimization/86179
1898 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1899 after failed recognition.
1901 2018-06-18 Martin Sebor <msebor@redhat.com>
1904 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1905 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1906 Handle integer subtraction.
1907 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1908 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1910 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1912 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1913 param from rtx to rtx_insn *.
1914 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1916 (frv_ifcvt_modify_insn): Likwise.
1917 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1918 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1919 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1920 as_a <rtx_insn *> cast to local "unprotected_region" once
1921 it's been established that it's not NULL or pc_rtx.
1922 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1923 param "sethi" from rtx to rtx_insn *.
1924 (nds32_group_float_insns): Likewise for param "insn".
1925 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1927 (vax_output_int_subtract): Likewise.
1928 * config/vax/vax.c (vax_output_int_add): Likewise for param
1930 (vax_output_int_subtract): Likewise.
1931 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1932 (emit_pattern_after): Likewise for param "after".
1933 (emit_insn_after): Likewise.
1934 (emit_jump_insn_after): Likewise.
1935 (emit_call_insn_after): Likewise.
1936 (emit_debug_insn_after): Likewise.
1937 (emit_pattern_before): Likewise for param "before".
1938 (emit_insn_before): Likewise.
1939 (emit_jump_insn_before): Likewise.
1940 * final.c (get_insn_template): Likewise for param "insn", removing
1942 * output.h (get_insn_template): Likewise for 2nd param.
1943 * rtl.h (emit_insn_before): Likewise.
1944 (emit_jump_insn_before): Likewise.
1945 (emit_debug_insn_before_noloc): Likewise.
1946 (emit_insn_after): Likewise.
1947 (emit_jump_insn_after): Likewise.
1948 (emit_call_insn_after): Likewise.
1949 (emit_debug_insn_after): Likewise.
1950 (set_insn_deleted): Likewise for param.
1952 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1955 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1956 floating point modes, so that IFmode is numerically greater than
1957 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1958 to declare the ordering. This prevents IFmode from being
1959 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1960 machine. Include rs6000-modes.h to share the fractional values
1961 between genmodes* and the rest of the compiler.
1965 * config/rs6000/rs6000-modes.h: New file.
1966 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1967 meaning of rs6000_long_double_size so that 126..128 selects an
1968 appropriate 128-bit floating point type.
1969 (rs6000_option_override_internal): Likewise.
1970 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1971 (TARGET_LONG_DOUBLE_128): Change the meaning of
1972 rs6000_long_double_size so that 126..128 selects an appropriate
1973 128-bit floating point type.
1974 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1975 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1976 source and destination to match the standard usage.
1977 (truncifkf2): Likewise.
1978 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1979 ISA 2.07 to use an explicit clobber, instead of passing in a
1981 (copysign<mode>3_soft): Likewise.
1983 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1985 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1986 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1987 (vect_slp_analyze_instance_dependence): Likewise.
1988 (vect_enhance_data_refs_alignment): Likewise.
1989 (vect_analyze_data_refs_alignment): Likewise.
1990 (vect_slp_analyze_and_verify_instance_alignment
1991 (vect_analyze_data_ref_accesses): Likewise.
1992 (vect_prune_runtime_alias_test_list): Likewise.
1993 (vect_analyze_data_refs): Likewise.
1994 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1995 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1996 (vect_analyze_scalar_cycles_1): Likewise.
1997 (vect_get_loop_niters): Likewise.
1998 (vect_analyze_loop_form_1): Likewise.
1999 (vect_update_vf_for_slp): Likewise.
2000 (vect_analyze_loop_operations): Likewise.
2001 (vect_analyze_loop): Likewise.
2002 (vectorizable_induction): Likewise.
2003 (vect_transform_loop): Likewise.
2004 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2005 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2006 (vect_make_slp_decision): Likewise.
2007 (vect_detect_hybrid_slp): Likewise.
2008 (vect_slp_analyze_operations): Likewise.
2009 (vect_slp_bb): Likewise.
2010 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2011 (vectorizable_bswap): Likewise.
2012 (vectorizable_call): Likewise.
2013 (vectorizable_simd_clone_call): Likewise.
2014 (vectorizable_conversion): Likewise.
2015 (vectorizable_assignment): Likewise.
2016 (vectorizable_shift): Likewise.
2017 (vectorizable_operation): Likewise.
2018 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2020 2018-06-18 Martin Sebor <msebor@redhat.com>
2022 PR tree-optimization/81384
2023 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2024 * builtins.c (expand_builtin_strnlen): New function.
2025 (expand_builtin): Call it.
2026 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2027 * builtins.def (BUILT_IN_STRNLEN): New.
2028 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2029 Warn for bounds in excess of maximum object size.
2030 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2031 single-value ranges. Handle strnlen.
2032 (handle_builtin_strlen): Handle strnlen.
2033 (strlen_check_and_optimize_stmt): Same.
2034 * doc/extend.texi (Other Builtins): Document strnlen.
2036 2018-06-18 Maya Rashish <coypu@sdf.org>
2038 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2039 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2040 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2042 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2044 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2046 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2048 * tree.c (escaped_string::escape): Replace cast to char * by
2049 const_cast<char *> (unescaped).
2051 2018-06-18 Nick Clifton <nickc@redhat.com>
2054 * tree.c (escaped_string): New class. Converts an unescaped
2055 string into its escaped equivalent.
2056 (warn_deprecated_use): Use the new class to convert the
2057 deprecation message, if present.
2058 (test_escaped_strings): New self test.
2059 (test_c_tests): Add test_escaped_strings.
2060 * doc/extend.texi (deprecated): Add a note that the
2061 deprecation message is affected by the -fmessage-length
2062 option, and that control characters will be escaped.
2063 (#pragma GCC error): Document this pragma.
2064 (#pragma GCC warning): Likewise.
2065 * doc/invoke.texi (-fmessage-length): Document this option's
2066 effect on the #warning and #error preprocessor directives and
2067 the deprecated attribute.
2069 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2071 * tree.c (decl_value_expr_lookup): Revert latest change.
2072 (decl_value_expr_insert): Likewise.
2074 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2076 * gimplify.c (nonlocal_vlas): Delete.
2077 (nonlocal_vla_vars): Likewise.
2078 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2080 (gimplify_body): Do not create and destroy nonlocal_vlas.
2081 * tree-nested.c: Include diagnostic.h.
2082 (use_pointer_in_frame): Tweak.
2083 (lookup_field_for_decl): Add assertion and declare the transformation.
2084 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2085 internal error when the reference is in a wrong context. Do not
2086 create a debug decl by default.
2087 (note_nonlocal_block_vlas): Delete.
2088 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2089 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2090 create a debug decl by default.
2091 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2092 call is in a wrong context.
2093 (fixup_vla_decls): New function.
2094 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2095 debug variables were created.
2096 * tree.c (decl_value_expr_lookup): Add checking assertion.
2097 (decl_value_expr_insert): Likewise.
2099 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2102 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2103 * tree-scalar-evolution.c (interpret_expr): Likewise.
2104 (expression_expensive_p): Likewise.
2105 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2106 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2107 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2108 (ssa_defined_by_minus_one_stmt_p): New.
2110 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2113 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2114 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2115 * dojump.c (do_jump): Likewise.
2116 * expr.c (expand_expr_real_2): Check operand type's sign.
2117 * fold-const.c (const_unop): Handle ABSU_EXPR.
2118 (fold_abs_const): Likewise.
2119 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2120 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2121 (strip_sign_op_1): Likesise.
2122 * match.pd: Add new pattern to generate ABSU_EXPR.
2123 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2124 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2125 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2126 * tree-inline.c (estimate_operator_cost): Likewise.
2127 * tree-pretty-print.c (dump_generic_node): Likewise.
2128 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2129 * tree.def (ABSU_EXPR): New.
2131 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2134 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2135 documented as preserved for backward compatibility only.
2136 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2138 PR rtl-optimization/86108
2139 * bb-reorder.c (create_forwarder_block): Renamed to ...
2140 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2141 jump from new landing pad to the second part.
2142 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2145 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2148 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2149 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2150 Only call store_expr for halves if the mode is the same.
2153 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2154 Fix up comment formatting.
2156 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2158 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2159 (typed_splay_tree::closure,
2160 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2161 (typed_splay_tree::typed_splay_tree,
2162 typed_splay_tree::operator =): Declared private.
2163 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2164 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2165 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2166 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2167 typed_splay_tree::splay_tree_splay,
2168 typed_splay_tree::splay_tree_foreach_helper,
2169 typed_splay_tree::splay_tree_insert,
2170 typed_splay_tree::splay_tree_remove,
2171 typed_splay_tree::splay_tree_lookup,
2172 typed_splay_tree::splay_tree_predecessor,
2173 typed_splay_tree::splay_tree_successor,
2174 typed_splay_tree::splay_tree_min,
2175 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2176 (typed_splay_tree::root, typed_splay_tree::comp,
2177 typed_splay_tree::delete_key,
2178 typed_splay_tree::delete_value): New data members.
2179 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2180 typed_splay_tree::remove.
2182 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2184 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2185 -mginv and -mno-ginv to the assembler.
2186 * config/mips/mips.opt (-mcrc): New option.
2188 * doc/invoke.text (-mcrc): Document.
2191 2018-06-15 Nick Clifton <nickc@redhat.com>
2194 * tree.c (escaped_string): New class. Converts an unescaped
2195 string into its escaped equivalent.
2196 (warn_deprecated_use): Use the new class to convert the
2197 deprecation message, if present.
2198 (test_escaped_strings): New self test.
2199 (test_c_tests): Add test_escaped_strings.
2200 * doc/extend.texi (deprecated): Add a note that the
2201 deprecation message is affected by the -fmessage-length
2202 option, and that control characters will be escaped.
2203 (#pragma GCC error): Document this pragma.
2204 (#pragma GCC warning): Likewise.
2205 * doc/invoke.texi (-fmessage-length): Document this option's
2206 effect on the #warning and #error preprocessor directives and
2207 the deprecated attribute.
2209 2018-06-15 Richard Biener <rguenther@suse.de>
2211 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2212 here, also noting vector size used.
2213 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2214 size used in MSG_OPTIMIZED_LOCATIONS dump.
2215 (pass_slp_vectorize::execute): Adjust.
2217 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2220 * config/arc/arc.c (arc_return_address_register): Fix
2223 2018-06-15 Richard Biener <rguenther@suse.de>
2226 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2227 leave useless conversion stripping to force_gimple_operand_gsi.
2228 (gimplify_build2): Likewise.
2229 (gimplify_build1): Likewise.
2231 2018-06-15 Richard Biener <rguenther@suse.de>
2234 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2235 before adjusting their block.
2237 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2239 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2240 multilibs for *-*-rtems*.
2241 * config/riscv/t-rtems: New file.
2243 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2246 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2247 unsigned_type_for returns NULL.
2250 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2251 subregs of multi-word pseudos unless the float mode has word size.
2253 2018-06-14 Richard Biener <rguenther@suse.de>
2256 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2257 duplicate and harmful type_hash_canon.
2258 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2260 2018-06-14 Richard Biener <rguenther@suse.de>
2263 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2266 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2268 * config/rtems.h (STDINT_LONG32): Define.
2270 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2271 Prachi Godbole <prachi.godbole@imgtec.com>
2273 * config/mips/mips-cpus.def: Define P6600.
2274 * config/mips/mips-tables.opt: Regenerate.
2275 * config/mips/mips.c (mips_ucbranch_type): New enum.
2276 (mips_rtx_cost_data): Add support for P6600.
2277 (mips_issue_rate): Likewise.
2278 (mips_multipass_dfa_lookahead): Likewise.
2279 (mips_avoid_hazard): Likewise.
2280 (mips_reorg_process_insns): Likewise.
2281 (mips_classify_branch_p6600): New function.
2282 * config/mips/mips.h (TUNE_P6600): New define.
2283 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2284 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2285 * config/mips/mips.md: Include p6600.md.
2286 (processor): Add p6600.
2287 * config/mips/p6600.md: New file.
2288 * doc/invoke.texi: Add p6600 to supported architectures.
2290 2018-06-13 Martin Sebor <msebor@redhat.com>
2292 PR tree-optimization/86114
2293 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2295 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2297 2018-06-13 Richard Biener <rguenther@suse.de>
2299 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2300 Properly set vector type of the intermediate stmt.
2301 * tree-vect-stmts.c (vectorizable_operation): The destination
2302 var always has vectype_out type.
2304 2018-06-13 Jeff Law <law@redhat.com>
2306 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2307 integer 0 for argument to print_rtl_with_bb.
2308 (rl78_reorg): Likewise.
2310 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2312 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2313 from rtx to rtx_insn *.
2314 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2316 (add_sched_insns_for_speculation): Likewise for local "target",
2317 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2318 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2319 from rtx_insn ** to rtx_code_label **.
2320 (reorg_emit_nops): Likewise.
2321 (c6x_reorg): Likewise for local "call_labels".
2322 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2324 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2325 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2326 the loops over LABEL_REFS.
2327 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2329 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2330 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2331 (split_branches): Strengthen local "olabel" from rtx to
2332 rtx_insn *, adding a safe_as_a cast.
2333 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2335 (add_insn_after): Likewise for first two params.
2336 (add_insn_before): Likewise.
2337 (remove_insn): Likewise for param.
2338 (emit_pattern_before_noloc): Likewise for second and third params.
2339 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2340 (emit_call_insn_before_noloc): Likewise.
2341 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2343 (emit_barrier_before): Likewise.
2344 (emit_label_before): Strengthen "label" param from "rtx" to
2345 "rtx_code_label *". Strengthen "before" param from "rtx" to
2347 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2349 (emit_pattern_after_noloc): Likewise.
2350 (emit_insn_after_noloc): Likewise.
2351 (emit_jump_insn_after_noloc): Likewise.
2352 (emit_call_insn_after_noloc): Likewise.
2353 (emit_debug_insn_after_noloc): Likewise.
2354 (emit_barrier_after): Likewise.
2355 (emit_label_after): Likewise for both params.
2356 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2357 "loc" param from "int" to "location_t".
2358 (emit_insn_after_setloc): Likewise.
2359 (emit_jump_insn_after_setloc): Likewise.
2360 (emit_call_insn_after_setloc): Likewise.
2361 (emit_debug_insn_after_setloc): Likewise.
2362 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2363 "loc" param from "int" to "location_t".
2364 (emit_pattern_before): Convert NULL_RTX to NULL.
2365 (emit_insn_before_setloc): Convert "loc" param from "int" to
2367 (emit_jump_insn_before_setloc): Likewise.
2368 (emit_call_insn_before_setloc): Likewise.
2369 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2370 rtx_insn *. Convert "loc" param from "int" to "location_t".
2371 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2372 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2373 Convert 3rd param from "int" to "location_t".
2374 (emit_barrier_before, emit_barrier_after, next_real_insn):
2375 Strengthen param from rtx to rtx_insn *.
2376 (emit_label_before): Strengthen 1st param from "rtx" to
2377 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2379 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2380 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2381 Strengthen 2nd param from "rtx" to "rtx_insn *".
2382 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2383 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2384 Likewise. Convert 3rd param from "int" to "location_t".
2385 (emit_label_after): Strengthen 1st param from "rtx" to
2387 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2389 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2390 from "rtx" to "rtx_insn *".
2392 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2394 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2395 bodies streamed in with -Q.
2396 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2397 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2398 * gimple-streamer-out.c: Include gimple-pretty-print.h
2399 (output_bb): Dump stmts streamed.
2400 * lto-section-out.c: Include print-tree.h
2401 (lto_begin_section): Dump sections created.
2402 (lto_output_decl_index): Dump decl encoded.
2403 * lto-streamer-out.c: Include print-tree.h
2404 (create_output_block): Dump output block created.
2405 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2406 (output_function): Dump function output.
2407 (output_constructor): Dump constructor streamed.
2408 (write_global_stream): Output indexes encoded.
2409 (produce_asm_for_decls): Dump streams encoded.
2410 * lto-streamer.c (streamer_dump_file): New global var.
2411 * lto-streamer.h (streamer_dump_file): Declare.
2412 * passes.c (ipa_write_summaries): Initialize streamer dump.
2413 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2416 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2419 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2420 offsets for register save directives. Emit a second batch of save
2421 directives, if need be, when the function accesses prior frames.
2423 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2425 * config/arc/fpu.md (fmasf4): Force operand to register.
2426 (fnmasf4): Likewise.
2428 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2430 * config/arc/arc-protos.h (arc_pad_return): Remove.
2431 * config/arc/arc.c (machine_function): Remove force_short_suffix
2433 (arc_print_operand): Adjust printing of '&'.
2434 (arc_verify_short): Remove conditional printing of short suffix.
2435 (arc_final_prescan_insn): Remove reference to size_reason.
2436 (pad_return): New function.
2437 (arc_reorg): Call pad_return.
2438 (arc_pad_return): Remove.
2439 (arc_init_machine_status): Remove reference to force_short_suffix.
2440 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2441 (attr length): When attribute iscompact is true force to 2
2442 regardless; in the case of maybe check if we want to force the
2443 instruction to have 4 bytes length.
2444 (nopv): Change it to generate 4 byte long nop as well.
2445 (blockage): New pattern.
2446 (simple_return): Remove call to arc_pad_return.
2447 (p_return_i): Likewise.
2449 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2451 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2453 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2455 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2458 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2460 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2461 for ARC700 and ARCv2.
2463 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2466 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2467 operands[2] instead of operands[1].
2470 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2472 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2473 case, check whether the outer register overlaps an unallocatable
2474 register, not just whether it fits the required class.
2476 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2478 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2480 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2481 (memrefs_conflict_p): Likewise.
2482 (init_alias_analysis): Likewise.
2483 * cfgexpand.c (expand_debug_expr): Likewise.
2484 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2485 * cse.c (fold_rtx): Likewise.
2486 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2487 * expr.c (emit_block_move_hints): Likewise.
2488 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2489 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2490 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2491 (emit_group_store): Likewise.
2492 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2493 to read the PRE/POST_MODIFY increment.
2494 * calls.c (store_one_arg): Use strip_offset.
2495 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2497 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2499 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2500 (simplify_binary_operation_1): Extend CONST_INT handling to
2502 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2503 than a HOST_WIDE_INT.
2504 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2506 (adjust_mems, add_stores): Update accodingly.
2507 (vt_canonicalize_addr): Track polynomial offsets.
2508 (emit_note_insn_var_location): Likewise.
2509 (vt_add_function_parameter): Likewise.
2510 (vt_initialize): Likewise.
2512 2018-06-12 Jeff Law <law@redhat.com>
2514 * config.gcc (alpha*-*-freebsd*): Remove.
2515 * config/alpha/freebsd.h: Remove.
2517 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2520 * spellcheck-tree.c (levenshtein_distance): Rename to...
2521 (get_edit_distance): ...this, and update for underlying renaming.
2522 * spellcheck-tree.h (levenshtein_distance): Rename to...
2523 (get_edit_distance): ...this.
2524 * spellcheck.c (levenshtein_distance): Rename to...
2525 (get_edit_distance): ...this. Convert from Levenshtein distance
2526 to Damerau-Levenshtein distance by supporting transpositions of
2527 adjacent characters. Rename "v1" to "v_next" and "v0" to
2529 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2530 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2531 update for underlying renaming.
2532 (selftest::levenshtein_distance_unit_test): Rename to...
2533 (selftest::test_get_edit_distance_unit): ...this, and update for
2534 underlying renaming.
2535 (selftest::test_find_closest_string): Add example from PR 69968
2536 where transposition helps
2537 (selftest::test_metric_conditions): Update for renaming.
2538 (selftest::test_metric_conditions): Likewise.
2539 (selftest::spellcheck_c_tests): Likewise.
2540 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2541 (get_edit_distance): ...this.
2542 (best_match::consider): Update for renaming.
2544 2018-06-12 Martin Sebor <msebor@redhat.com>
2546 PR tree-optimization/85259
2547 * builtins.c (compute_objsize): Handle constant offsets.
2548 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2549 true iff a warning has been issued.
2550 * gimple.h (gimple_nonartificial_location): New function.
2551 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2552 gimple_nonartificial_location and handle -Wno-system-headers.
2553 (handle_builtin_stxncpy): Same.
2555 2018-06-12 Martin Sebor <msebor@redhat.com>
2558 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2560 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2562 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2563 BUILTIN_VEC_XST entries for pointer to double and long long.
2565 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2568 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2570 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2573 2018-06-12 Martin Liska <mliska@suse.cz>
2575 * doc/options.texi: Document IntegerRange.
2577 2018-06-12 Martin Liska <mliska@suse.cz>
2579 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2580 * opt-functions.awk: Handle Deprecated flag.
2581 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2583 (read_cmdline_option): Report warning for a deprecated option.
2584 * opts.h (struct cl_option): Add new field cl_deprecated.
2585 (CL_ERR_DEPRECATED): New.
2587 2018-06-12 Martin Liska <mliska@suse.cz>
2589 * doc/options.texi: Document Deprecated option flag.
2591 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2593 * config/arc/arc-arch.h (arc_extras): New enum.
2594 (arc_cpu_t):Add field extra.
2595 (arc_cpu_types): Consider the extras.
2596 * config/arc/arc-cpus.def: Add extras info.
2597 * config/arc/arc-opts.h (processor_type): Consider extra field.
2598 * config/arc/arc.c (arc_override_options): Handle extra field.
2600 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2602 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2603 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2605 * config/arc/arc.c (arc_init): Update pic warning.
2606 (irq_range): Update irq range parsing warnings.
2607 (arc_override_options): Update various warning messages.
2608 (arc_handle_aux_attribute): Likewise.
2610 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2612 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2614 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2616 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2617 compiler messages directives.
2619 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2621 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2622 * config/mips/mips-tables.opt: Regenerate.
2623 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2625 * doc/invoke.texi: Document -march=i6500.
2627 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2629 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2630 (i6400_gpmul): Add cpu_unit.
2631 (i6400_gpdiv): Likewise.
2632 (i6400_msa_add_d): Update reservations.
2633 (i6400_msa_int_add) Likewise.
2634 (i6400_msa_short_logic3) Likewise.
2635 (i6400_msa_short_logic2) Likewise.
2636 (i6400_msa_short_logic) Likewise.
2637 (i6400_msa_move) Likewise.
2638 (i6400_msa_cmp) Likewise.
2639 (i6400_msa_short_float2) Likewise.
2640 (i6400_msa_div_d) Likewise.
2641 (i6400_msa_long_logic1) Likewise.
2642 (i6400_msa_long_logic2) Likewise.
2643 (i6400_msa_mult) Likewise.
2644 (i6400_msa_long_float2) Likewise.
2645 (i6400_msa_long_float4) Likewise.
2646 (i6400_msa_long_float5) Likewise.
2647 (i6400_msa_long_float8) Likewise.
2648 (i6400_fpu_fadd): Include frint type.
2649 (i6400_fpu_store): New define_insn_reservation.
2650 (i6400_fpu_load): Likewise.
2651 (i6400_fpu_move): Likewise.
2652 (i6400_fpu_fcmp): Likewise.
2653 (i6400_fpu_fmadd): Likewise.
2654 (i6400_int_mult): Include imul3nc type and update reservation.
2655 (i6400_int_div): Include idiv3 type and update reservation.
2656 (i6400_int_load): Update to check type not move_type.
2657 (i6400_int_store): Likewise.
2658 (i6400_int_prefetch): Set zero latency.
2660 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2662 * gcc.c: Document new %@{...} sequence.
2663 (LINK_COMMAND_SPEC): Use it for the -L switches.
2664 (cpp_unique_options): Use it for the -I switches.
2665 (at_file_argbuf): New global variable.
2666 (in_at_file): Likewise.
2667 (alloc_args): Create at_file_argbuf.
2668 (clear_args): Truncate at_file_argbuf.
2669 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2670 (open_at_file): New function.
2671 (close_at_file): Likewise.
2672 (create_at_file): Delete.
2673 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2676 (validate_switches_from_spec): Deal with %@{...} sequence.
2677 (validate_switches): Likewise.
2678 (driver::finalize): Call clear_args.
2680 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2682 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2684 2018-06-11 Martin Sebor <msebor@redhat.com>
2686 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2687 (Wno-class-memaccess): Add @opindex.
2688 (Wno-templates, Wno-multiple-inheritance): Same.
2689 (Wno-virtual-inheritance, Wno-namespaces): Same.
2690 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2691 (Wno-format-overflow, Wno-format-truncation): Same.
2692 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2693 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2694 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2695 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2696 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2697 (Wno-misspelled-isr): Same.
2699 2018-06-11 Martin Sebor <msebor@redhat.com>
2701 * PR tree-optimization/86083
2702 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2704 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2706 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2708 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2711 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2712 on the correct operand.
2713 (*movdi_internal64): Ditto.
2715 2018-06-11 Martin Liska <mliska@suse.cz>
2717 PR tree-optimization/86089
2718 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2720 2018-06-11 Julia Koval <julia.koval@intel.com>
2722 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2723 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2724 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2726 2018-06-11 Olivier Hainque <hainque@adacore.com>
2728 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2729 for Ada with strict dwarf2.
2731 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2734 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2737 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2739 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2741 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2743 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2746 2018-06-08 Martin Liska <mliska@suse.cz>
2748 * tree-cfg.h (debug_function): Fix argument type to match
2751 2018-06-08 Martin Liska <mliska@suse.cz>
2753 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2754 Remove usage of MPX-related (and removed) fields.
2755 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2757 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2759 * cfg.c (debug): Use TDF_NONE rather than 0.
2760 * cfghooks.c (debug): Likewise.
2761 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2762 (struct dump_option_value_info): Convert to...
2763 (struct kv_pair): ...this template type.
2764 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2766 (optinfo_verbosity_options): Likewise.
2767 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2769 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2770 than int for "optgroup_flags" param.
2771 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2773 (dump_dec): Likewise.
2774 (dump_finish): Use TDF_NONE rather than 0.
2775 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2776 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2777 than 0. Update for change to option_ptr.
2778 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2779 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2780 0. Update for changes to optinfo_verbosity_options and
2782 (opt_info_switch_p): Convert optgroup_flags from int to
2784 (dump_basic_block): Use dump_flags_t rather than int
2785 for "dump_kind" param.
2786 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2787 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2788 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2789 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2790 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2791 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2792 TDF_NONE): Convert from macros to...
2793 (enum dump_flag): ...this new enum.
2794 (dump_flags_t): Update to use enum.
2795 (operator|, operator&, operator~, operator|=, operator&=):
2796 Implement for dump_flags_t.
2797 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2798 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2799 Convert from macros to...
2800 (enum optgroup_flag): ...this new enum.
2801 (optgroup_flags_t): New typedef.
2802 (operator|, operator|=): Implement for optgroup_flags_t.
2803 (struct dump_file_info): Convert field "alt_flags" to
2804 dump_flags_t. Convert field "optgroup_flags" to
2806 (dump_basic_block): Use dump_flags_t rather than int for param.
2807 (dump_generic_expr_loc): Likewise.
2808 (dump_dec): Likewise.
2809 (dump_register): Convert param "optgroup_flags" to
2811 (opt_info_enable_passes): Likewise.
2812 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2814 * gimple-pretty-print.c (debug): Likewise.
2815 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2816 (merged_store_group::apply_stores): Likewise.
2817 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2818 * gimple.c (verify_gimple_pp): Likewise.
2819 * graphite-poly.c (print_pbb_body): Likewise.
2820 * passes.c (pass_manager::register_one_dump_file): Convert
2821 local "optgroup_flags" to optgroup_flags_t.
2822 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2824 (debug_body): Likewise.
2825 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2826 to optgroup_flags_t.
2827 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2829 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2830 (convert_mult_to_fma): Likewise.
2831 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2832 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2833 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2834 "dump_kind" to dump_flags_t.
2836 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2838 * config/rs6000/rs6000.c (min, max): Delete.
2840 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2842 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2843 -mabi=spe and -mabi=no-spe.
2845 2018-06-08 Martin Liska <mliska@suse.cz>
2847 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2848 where we expect an existing summary.
2850 2018-06-08 Martin Liska <mliska@suse.cz>
2852 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2853 * ipa-inline.h (estimate_edge_growth): Likewise.
2855 2018-06-08 Martin Liska <mliska@suse.cz>
2857 * cgraph.c (function_version_hasher::hash): Use
2858 cgraph_node::get_uid ().
2859 (function_version_hasher::equal):
2860 * cgraph.h (cgraph_node::get_uid): New method.
2861 * ipa-inline.c (update_caller_keys): Use
2862 cgraph_node::get_uid ().
2863 (update_callee_keys): Likewise.
2864 * ipa-utils.c (searchc): Likewise.
2865 (ipa_reduced_postorder): Likewise.
2866 * lto-cgraph.c (input_node): Likewise.
2867 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2868 * symbol-summary.h (symtab_insertion): Likewise.
2869 (symtab_removal): Likewise.
2870 (symtab_duplication): Likewise.
2871 * tree-pretty-print.c (dump_function_header): Likewise.
2872 * tree-sra.c (convert_callers_for_node): Likewise.
2874 2018-06-08 Martin Liska <mliska@suse.cz>
2876 * cgraph.c (symbol_table::create_edge): Always assign a new
2878 (symbol_table::free_edge): Do not recycle numbers.
2879 * cgraph.h (cgraph_edge::get): New method.
2880 * symbol-summary.h (symtab_removal): Use it.
2881 (symtab_duplication): Likewise.
2882 (call_summary::hashable_uid): Remove.
2884 2018-06-08 Martin Liska <mliska@suse.cz>
2886 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2887 (initialize_growth_caches): Remove.
2888 (free_growth_caches): Likewise.
2889 (do_estimate_edge_time): Use edge_growth_cache.
2890 (do_estimate_edge_size): Likewise.
2891 (do_estimate_edge_hints): Likewise.
2892 * ipa-inline.c (reset_edge_caches): Likewise.
2893 (recursive_inlining): Likewise.
2894 (inline_small_functions): Likewise.
2895 * ipa-inline.h (initialize_growth_caches): Remove.
2896 (estimate_edge_size): Likewise.
2897 (estimate_edge_time): Likewise.
2898 (estimate_edge_hints): Likewise.
2899 (reset_edge_growth_cache): Likewise.
2900 * symbol-summary.h (call_summary::remove): New method.
2902 2018-06-08 Martin Liska <mliska@suse.cz>
2904 * ipa-cp.c (class edge_clone_summary): New summary.
2905 (grow_edge_clone_vectors): Remove.
2906 (ipcp_edge_duplication_hook): Remove.
2907 (class edge_clone_summary_t): New call_summary class.
2908 (ipcp_edge_removal_hook): Remove.
2909 (edge_clone_summary_t::duplicate): New function.
2910 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2911 (create_specialized_node): Likewise.
2912 (ipcp_driver): Initialize edge_clone_summaries and do not
2915 2018-06-08 Martin Liska <mliska@suse.cz>
2917 * symbol-summary.h (get): New function.
2918 (call_summary::m_initialize_when_cloning): New class member.
2920 2018-06-08 Martin Liska <mliska@suse.cz>
2922 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2923 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2924 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2925 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2926 (remove_cgraph_node_from_order): Use the removed_nodes set.
2927 (do_per_function_toporder): Likwise.
2928 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2929 instead of summary_uid.
2930 (symtab_removal): Likewise.
2931 (symtab_duplication): Likewise.
2933 2018-06-08 Martin Liska <mliska@suse.cz>
2935 * ipa-cp.c (ipcp_store_bits_results): Use
2936 ipcp_transformation_sum.
2937 (ipcp_store_vr_results): Likewise.
2938 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2940 (ipcp_transformation_initialize): ... this.
2941 (ipa_set_node_agg_value_chain):
2942 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2943 (write_ipcp_transformation_info): Likewise.
2944 (read_ipcp_transformation_info): Likewise.
2945 (ipcp_update_bits): Likewise.
2946 (ipcp_update_vr): Likewise.
2947 (ipcp_transform_function): Likewise.
2948 * ipa-prop.h: Rename ipcp_transformation_summary to
2949 ipcp_transformation.
2950 (class ipcp_transformation_t): New function summary.
2951 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2952 (ipa_get_agg_replacements_for_node): Likewise.
2954 2018-06-08 Martin Liska <mliska@suse.cz>
2956 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2958 (class funct_state_summary_t): New function_summary class.
2959 (has_function_state): Remove.
2960 (get_function_state): Likewise.
2961 (set_function_state): Likewise.
2962 (add_new_function): Likewise.
2963 (funct_state_summary_t::insert): New function.
2964 (duplicate_node_data): Remove.
2965 (remove_node_data): Remove.
2966 (funct_state_summary_t::duplicate): New function.
2967 (register_hooks): Create new funct_state_summaries.
2968 (pure_const_generate_summary): Use it.
2969 (pure_const_write_summary): Likewise.
2970 (pure_const_read_summary): Likewise.
2971 (propagate_pure_const): Likewise.
2972 (propagate_nothrow): Likewise.
2973 (dump_malloc_lattice): Likewise.
2974 (propagate_malloc): Likewise.
2975 (execute): Do not register hooks, just remove summary
2977 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2980 2018-06-08 Martin Liska <mliska@suse.cz>
2982 * ipa-reference.c (remove_node_data): Remove.
2983 (duplicate_node_data): Likewise.
2984 (class ipa_ref_var_info_summary_t): New class.
2985 (class ipa_ref_opt_summary_t): Likewise.
2986 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2987 (get_reference_optimization_summary): Use
2988 ipa_ref_opt_sum_summaries.
2989 (set_reference_vars_info): Remove.
2990 (set_reference_optimization_summary): Likewise.
2991 (ipa_init): Create summaries.
2992 (init_function_info): Use function summary.
2993 (ipa_ref_opt_summary_t::duplicate): New function.
2994 (ipa_ref_opt_summary_t::remove): New function.
2995 (get_read_write_all_from_node): Fix GNU coding style.
2996 (propagate): Use function summary.
2997 (write_node_summary_p): Fix GNU coding style.
2998 (stream_out_bitmap): Likewise.
2999 (ipa_reference_read_optimization_summary): Use function summary.
3000 (ipa_reference_c_finalize): Do not release hooks.
3002 2018-06-08 Martin Liska <mliska@suse.cz>
3004 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3005 (analyze_function_body): Extract multiple calls of get_create.
3006 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3007 * ipa-inline.c (recursive_inlining): Use ::get method.
3008 * ipa-inline.h (estimate_edge_growth): Likewise.
3010 2018-06-08 Martin Liska <mliska@suse.cz>
3012 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3014 (hsa_function_summary::hsa_function_summary): Use the new enum
3016 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3017 * hsa-gen.c (hsa_get_host_function): Likewise.
3018 (get_brig_function_name): Likewise.
3019 * ipa-hsa.c (process_hsa_functions): Likewise.
3020 (ipa_hsa_write_summary): Likewise.
3021 * symbol-summary.h (symtab_duplication): Use ::get function/
3022 (get): New function.
3024 2018-06-08 Martin Liska <mliska@suse.cz>
3026 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3028 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3029 (hsa_register_kernel): Likewise.
3030 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3031 * hsa-gen.c (hsa_get_host_function): Likewise.
3032 (get_brig_function_name): Likewise.
3033 (generate_hsa): Likewise.
3034 (pass_gen_hsail::execute): Likewise.
3035 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3036 (devirtualization_time_bonus): Likewise.
3037 (ipcp_propagate_stage): Likewise.
3038 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3039 (edge_set_predicate): Likewise.
3040 (evaluate_conditions_for_known_args): Likewise.
3041 (evaluate_properties_for_edge): Likewise.
3042 (ipa_fn_summary::reset): Likewise.
3043 (ipa_fn_summary_t::duplicate): Likewise.
3044 (dump_ipa_call_summary): Likewise.
3045 (ipa_dump_fn_summary): Likewise.
3046 (analyze_function_body): Likewise.
3047 (compute_fn_summary): Likewise.
3048 (estimate_edge_devirt_benefit): Likewise.
3049 (estimate_edge_size_and_time): Likewise.
3050 (estimate_calls_size_and_time): Likewise.
3051 (estimate_node_size_and_time): Likewise.
3052 (inline_update_callee_summaries): Likewise.
3053 (remap_edge_change_prob): Likewise.
3054 (remap_edge_summaries): Likewise.
3055 (ipa_merge_fn_summary_after_inlining): Likewise.
3056 (ipa_update_overall_fn_summary): Likewise.
3057 (read_ipa_call_summary): Likewise.
3058 (inline_read_section): Likewise.
3059 (write_ipa_call_summary): Likewise.
3060 (ipa_fn_summary_write): Likewise.
3061 (ipa_free_fn_summary): Likewise.
3062 * ipa-hsa.c (process_hsa_functions): Likewise.
3063 (ipa_hsa_write_summary): Likewise.
3064 (ipa_hsa_read_section): Likewise.
3065 * ipa-icf.c (sem_function::merge): Likewise.
3066 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3067 (do_estimate_edge_time): Likewise.
3068 (estimate_size_after_inlining): Likewise.
3069 (estimate_growth): Likewise.
3070 (growth_likely_positive): Likewise.
3071 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3072 (inline_call): Likewise.
3073 * ipa-inline.c (caller_growth_limits): Likewise.
3074 (can_inline_edge_p): Likewise.
3075 (can_inline_edge_by_limits_p): Likewise.
3076 (compute_uninlined_call_time): Likewise.
3077 (compute_inlined_call_time): Likewise.
3078 (want_inline_small_function_p): Likewise.
3079 (edge_badness): Likewise.
3080 (update_caller_keys): Likewise.
3081 (update_callee_keys): Likewise.
3082 (recursive_inlining): Likewise.
3083 (inline_small_functions): Likewise.
3084 (inline_to_all_callers_1): Likewise.
3085 (dump_overall_stats): Likewise.
3086 (early_inline_small_functions): Likewise.
3087 (early_inliner): Likewise.
3088 * ipa-inline.h (estimate_edge_growth): Likewise.
3089 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3090 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3091 * ipa-prop.h (IPA_NODE_REF): Likewise.
3092 (IPA_EDGE_REF): Likewise.
3093 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3094 (propagate_malloc): Likewise.
3095 * ipa-split.c (execute_split_functions): Likewise.
3096 * symbol-summary.h: Rename get to get_create.
3098 (get_create): Likewise.
3099 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3101 2018-06-08 Martin Liska <mliska@suse.cz>
3103 * symbol-summary.h (release): Move definition out of class
3105 (symtab_removal): Likewise.
3106 (symtab_duplication): Likewise.
3108 2018-06-08 Martin Liska <mliska@suse.cz>
3110 * symbol-summary.h (function_summary): Move constructor
3111 implementation out of class declaration.
3112 (release): Likewise.
3113 (symtab_insertion): Likewise.
3114 (symtab_removal): Likewise.
3115 (symtab_duplication): Likewise.
3118 2018-06-08 Martin Liska <mliska@suse.cz>
3120 * Makefile.in: Remove support for MPX (macros, related functions,
3121 fields in cgraph_node, ...).
3122 * builtin-types.def (BT_BND): Likewise.
3123 (BT_FN_BND_CONST_PTR): Likewise.
3124 (BT_FN_CONST_PTR_BND): Likewise.
3125 (BT_FN_VOID_PTR_BND): Likewise.
3126 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3127 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3128 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3129 (expand_builtin_mempcpy_with_bounds): Likewise.
3130 (expand_builtin_memset_with_bounds): Likewise.
3131 (expand_builtin_memset_args): Likewise.
3132 (std_expand_builtin_va_start): Likewise.
3133 (expand_builtin): Likewise.
3134 (expand_builtin_with_bounds): Likewise.
3135 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3136 (DEF_LIB_BUILTIN_CHKP): Likewise.
3137 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3138 (DEF_CHKP_BUILTIN): Likewise.
3139 (BUILT_IN_MEMCPY): Likewise.
3140 (BUILT_IN_MEMMOVE): Likewise.
3141 (BUILT_IN_MEMPCPY): Likewise.
3142 (BUILT_IN_MEMSET): Likewise.
3143 (BUILT_IN_STPCPY): Likewise.
3144 (BUILT_IN_STRCAT): Likewise.
3145 (BUILT_IN_STRCHR): Likewise.
3146 (BUILT_IN_STRCPY): Likewise.
3147 (BUILT_IN_STRLEN): Likewise.
3148 (BUILT_IN_MEMCPY_CHK): Likewise.
3149 (BUILT_IN_MEMMOVE_CHK): Likewise.
3150 (BUILT_IN_MEMPCPY_CHK): Likewise.
3151 (BUILT_IN_MEMSET_CHK): Likewise.
3152 (BUILT_IN_STPCPY_CHK): Likewise.
3153 (BUILT_IN_STRCAT_CHK): Likewise.
3154 (BUILT_IN_STRCPY_CHK): Likewise.
3155 * calls.c (store_bounds): Likewise.
3156 (emit_call_1): Likewise.
3157 (special_function_p): Likewise.
3158 (maybe_warn_nonstring_arg): Likewise.
3159 (initialize_argument_information): Likewise.
3160 (finalize_must_preallocate): Likewise.
3161 (compute_argument_addresses): Likewise.
3162 (expand_call): Likewise.
3163 * cfgexpand.c (expand_call_stmt): Likewise.
3164 (expand_return): Likewise.
3165 (expand_gimple_stmt_1): Likewise.
3166 (pass_expand::execute): Likewise.
3167 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3168 (cgraph_node::remove): Likewise.
3169 (cgraph_node::dump): Likewise.
3170 (cgraph_node::verify_node): Likewise.
3171 * cgraph.h (chkp_function_instrumented_p): Likewise.
3172 (symtab_node::get_alias_target): Likewise.
3173 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3174 (cgraph_local_p): Likewise.
3175 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3176 (cgraph_edge::rebuild_references): Likewise.
3177 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3178 (walk_polymorphic_call_targets): Likewise.
3179 (cgraph_node::expand_thunk): Likewise.
3180 (symbol_table::output_weakrefs): Likewise.
3181 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3182 (ix86_handle_option): Likewise.
3183 * config/i386/constraints.md: Likewise.
3184 * config/i386/i386-builtin-types.def (BND): Likewise.
3188 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3189 (BDESC_FIRST): Likewise.
3191 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3192 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3193 * config/i386/i386.c (enum reg_class): Likewise.
3194 (ix86_target_string): Likewise.
3195 (ix86_option_override_internal): Likewise.
3196 (ix86_conditional_register_usage): Likewise.
3197 (ix86_valid_target_attribute_inner_p): Likewise.
3198 (ix86_set_indirect_branch_type): Likewise.
3199 (ix86_set_current_function): Likewise.
3200 (ix86_function_arg_regno_p): Likewise.
3201 (init_cumulative_args): Likewise.
3202 (ix86_function_arg_advance): Likewise.
3203 (ix86_function_arg): Likewise.
3204 (ix86_pass_by_reference): Likewise.
3205 (ix86_function_value_regno_p): Likewise.
3206 (ix86_function_value_1): Likewise.
3207 (ix86_function_value_bounds): Likewise.
3208 (ix86_return_in_memory): Likewise.
3209 (ix86_setup_incoming_vararg_bounds): Likewise.
3210 (ix86_va_start): Likewise.
3211 (indirect_thunk_need_prefix): Likewise.
3212 (print_reg): Likewise.
3213 (ix86_print_operand): Likewise.
3214 (ix86_expand_call): Likewise.
3215 (ix86_output_function_return): Likewise.
3216 (reg_encoded_number): Likewise.
3217 (BDESC_VERIFYS): Likewise.
3218 (ix86_init_mpx_builtins): Likewise.
3219 (ix86_init_builtins): Likewise.
3220 (ix86_emit_cmove): Likewise.
3221 (ix86_emit_move_max): Likewise.
3222 (ix86_expand_builtin): Likewise.
3223 (ix86_builtin_mpx_function): Likewise.
3224 (ix86_get_arg_address_for_bt): Likewise.
3225 (ix86_load_bounds): Likewise.
3226 (ix86_store_bounds): Likewise.
3227 (ix86_load_returned_bounds): Likewise.
3228 (ix86_store_returned_bounds): Likewise.
3229 (ix86_class_likely_spilled_p): Likewise.
3230 (ix86_hard_regno_mode_ok): Likewise.
3231 (x86_order_regs_for_local_alloc): Likewise.
3232 (ix86_mitigate_rop): Likewise.
3233 (ix86_bnd_prefixed_insn_p): Likewise.
3234 (ix86_mpx_bound_mode): Likewise.
3235 (ix86_make_bounds_constant): Likewise.
3236 (ix86_initialize_bounds): Likewise.
3237 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3238 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3239 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3240 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3241 (TARGET_CHKP_BOUND_MODE): Likewise.
3242 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3243 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3244 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3245 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3246 * config/i386/i386.h (TARGET_MPX): Likewise.
3247 (TARGET_MPX_P): Likewise.
3248 (VALID_BND_REG_MODE): Likewise.
3249 (FIRST_BND_REG): Likewise.
3250 (LAST_BND_REG): Likewise.
3251 (enum reg_class): Likewise.
3252 (BND_REG_P): Likewise.
3253 (BND_REGNO_P): Likewise.
3254 (BNDmode): Likewise.
3255 (ADJUST_INSN_LENGTH): Likewise.
3256 * config/i386/i386.md: Likewise.
3257 * config/i386/i386.opt: Likewise.
3258 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3259 (defined): Likewise.
3260 (LINK_MPX): Likewise.
3261 (MPX_SPEC): Likewise.
3262 (LIBMPX_SPEC): Likewise.
3263 (LIBMPXWRAPPERS_SPEC): Likewise.
3264 (CHKP_SPEC): Likewise.
3265 * config/i386/predicates.md: Likewise.
3266 * dbxout.c (dbxout_type): Likewise.
3267 * doc/extend.texi: Likewise.
3268 * doc/invoke.texi: Likewise.
3269 * doc/md.texi: Likewise.
3270 * doc/tm.texi: Likewise.
3271 * doc/tm.texi.in: Likewise.
3272 * dwarf2out.c (is_base_type): Likewise.
3273 (gen_formal_types_die): Likewise.
3274 (gen_subprogram_die): Likewise.
3275 (gen_type_die_with_usage): Likewise.
3276 (gen_decl_die): Likewise.
3277 (dwarf2out_late_global_decl): Likewise.
3278 * expr.c (expand_assignment): Likewise.
3279 (emit_storent_insn): Likewise.
3280 (store_expr_with_bounds): Likewise.
3281 (store_expr): Likewise.
3282 (expand_expr_real_1): Likewise.
3283 * expr.h (store_expr_with_bounds): Likewise.
3284 * function.c (use_register_for_decl): Likewise.
3285 (struct bounds_parm_data): Likewise.
3286 (assign_parms_augmented_arg_list): Likewise.
3287 (assign_parm_find_entry_rtl): Likewise.
3288 (assign_parm_is_stack_parm): Likewise.
3289 (assign_parm_load_bounds): Likewise.
3290 (assign_bounds): Likewise.
3291 (assign_parms): Likewise.
3292 (expand_function_start): Likewise.
3293 * gcc.c (CHKP_SPEC): Likewise.
3294 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3295 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3296 (wrestrict_dom_walker::check_call): Likewise.
3297 * gimple.c (gimple_build_call_from_tree): Likewise.
3298 * gimple.h (enum gf_mask): Likewise.
3299 (gimple_call_with_bounds_p): Likewise.
3300 (gimple_call_set_with_bounds): Likewise.
3301 * gimplify.c (gimplify_init_constructor): Likewise.
3302 * ipa-cp.c (initialize_node_lattices): Likewise.
3303 (propagate_constants_across_call): Likewise.
3304 (find_more_scalar_values_for_callers_subset): Likewise.
3305 * ipa-hsa.c (process_hsa_functions): Likewise.
3306 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3307 * ipa-icf.c (sem_function::merge): Likewise.
3308 * ipa-inline.c (early_inliner): Likewise.
3309 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3310 (warn_function_cold): Likewise.
3311 (propagate_pure_const): Likewise.
3312 * ipa-ref.h (enum GTY): Likewise.
3313 * ipa-split.c (find_retbnd): Likewise.
3314 (consider_split): Likewise.
3315 (split_function): Likewise.
3316 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3317 * ipa.c (walk_polymorphic_call_targets): Likewise.
3318 (symbol_table::remove_unreachable_nodes): Likewise.
3319 (process_references): Likewise.
3320 (cgraph_build_static_cdtor_1): Likewise.
3321 * lto-cgraph.c (lto_output_node): Likewise.
3322 (output_refs): Likewise.
3323 (compute_ltrans_boundary): Likewise.
3324 (input_overwrite_node): Likewise.
3325 (input_node): Likewise.
3326 (input_cgraph_1): Likewise.
3327 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3328 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3329 (class pass_chkp_instrumentation_passes): Likewise.
3330 (make_pass_chkp_instrumentation_passes): Likewise.
3331 * passes.def: Likewise.
3332 * rtl.h (struct GTY): Likewise.
3333 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3334 * stor-layout.c (layout_type): Likewise.
3335 * symtab.c: Likewise.
3336 * target.def: Likewise.
3337 * targhooks.c (default_chkp_bound_type): Likewise.
3338 (default_chkp_bound_mode): Likewise.
3339 (default_builtin_chkp_function): Likewise.
3340 (default_chkp_function_value_bounds): Likewise.
3341 (default_chkp_make_bounds_constant): Likewise.
3342 (default_chkp_initialize_bounds): Likewise.
3343 * targhooks.h (default_chkp_bound_type): Likewise.
3344 (default_chkp_bound_mode): Likewise.
3345 (default_builtin_chkp_function): Likewise.
3346 (default_chkp_function_value_bounds): Likewise.
3347 (default_chkp_make_bounds_constant): Likewise.
3348 (default_chkp_initialize_bounds): Likewise.
3349 * toplev.c (compile_file): Likewise.
3350 (process_options): Likewise.
3351 * tree-core.h (DEF_BUILTIN): Likewise.
3352 (DEF_BUILTIN_CHKP): Likewise.
3353 * tree-inline.c (declare_return_variable): Likewise.
3354 (remap_gimple_stmt): Likewise.
3355 (copy_bb): Likewise.
3356 (initialize_inlined_parameters): Likewise.
3357 (expand_call_inline): Likewise.
3358 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3359 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3360 (make_pass_ipa_chkp_produce_thunks): Likewise.
3361 (make_pass_chkp): Likewise.
3362 (make_pass_chkp_opt): Likewise.
3363 (make_pass_chkp_instrumentation_passes): Likewise.
3364 * tree-pretty-print.c (dump_generic_node): Likewise.
3365 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3366 * tree-ssa-dce.c (propagate_necessity): Likewise.
3367 (eliminate_unnecessary_stmts): Likewise.
3368 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3369 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3370 * tree-ssa-sccvn.h: Likewise.
3371 * tree-ssa-strlen.c (get_string_length): Likewise.
3372 (valid_builtin_call): Likewise.
3373 (adjust_last_stmt): Likewise.
3374 (handle_builtin_strchr): Likewise.
3375 (handle_builtin_strcpy): Likewise.
3376 (handle_builtin_stxncpy): Likewise.
3377 (handle_builtin_memcpy): Likewise.
3378 (handle_builtin_strcat): Likewise.
3379 (strlen_check_and_optimize_stmt): Likewise.
3380 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3381 * tree-streamer-in.c: Likewise.
3382 * tree-streamer.c (record_common_node): Likewise.
3383 * tree.c (tree_code_size): Likewise.
3384 (wide_int_to_tree_1): Likewise.
3385 (type_contains_placeholder_1): Likewise.
3386 (build_common_tree_nodes): Likewise.
3387 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3388 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3389 (POINTER_BOUNDS_P): Likewise.
3390 (BOUNDED_TYPE_P): Likewise.
3391 (BOUNDED_P): Likewise.
3392 (CALL_WITH_BOUNDS_P): Likewise.
3393 (pointer_bounds_type_node): Likewise.
3394 * value-prof.c (gimple_ic): Likewise.
3395 * var-tracking.c (vt_add_function_parameters): Likewise.
3396 * varasm.c (make_decl_rtl): Likewise.
3397 (assemble_start_function): Likewise.
3398 (output_constant): Likewise.
3399 (maybe_assemble_visibility): Likewise.
3400 * varpool.c (ctor_for_folding): Likewise.
3401 * chkp-builtins.def: Remove.
3402 * ipa-chkp.c: Remove.
3403 * ipa-chkp.h: Remove.
3404 * rtl-chkp.c: Remove.
3405 * rtl-chkp.h: Remove.
3406 * tree-chkp-opt.c: Remove.
3407 * tree-chkp.c: Remove.
3408 * tree-chkp.h: Remove.
3410 2018-06-07 Carl Love <cel@us.ibm.com>
3412 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3413 vextract_fp_from_shortl): Add BE support.
3415 2018-06-07 Paul Koning <ni1d@arrl.net>
3417 * compare-elim.c (try_merge_compare): Don't merge compare if
3418 address contains a side effect.
3419 (try_eliminate_compare): Likewise.
3421 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3423 * config.gcc: Support "tremont".
3424 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3425 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3427 * config/i386/i386.c (m_TREMONT): Define.
3428 (processor_target_table): Add "tremont".
3429 (PTA_TREMONT): Define.
3430 (ix86_lea_outperforms): Add TARGET_TREMONT.
3431 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3432 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3433 and M_INTEL_GOLDMONT_PLUS.
3434 (fold_builtin_cpu): Add "tremont".
3435 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3436 (ix86_option_override_internal): Add "tremont".
3437 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3438 (processor_type): Add PROCESSOR_TREMONT.
3439 * config/i386/x86-tune.def: Add m_TREMONT.
3440 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3442 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3444 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3445 symbol defined for msp430i* devices to be lower case.
3447 gcc/testsuite/gcc.target/msp430/
3448 * msp430i-device-symbol.c: New test.
3449 * msp430f-device-symbol.c: New test.
3450 * msp430.h: New test header file.
3452 2018-06-07 Richard Biener <rguenther@suse.de>
3454 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3455 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3456 Properly wrap signed arithmetic if overflow wraps.
3458 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3460 PR tree-optimization/69615
3461 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3462 of a cast from a same precision integral SSA_NAME in a bb dominated
3463 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3464 cast to utype if rhs2 has already a compatible type.
3466 2018-06-07 Richard Biener <rguenther@suse.de>
3468 PR tree-optimization/85935
3469 * graphite-scop-detection.c (find_params_in_bb): Analyze
3470 condition operands with respect to the correct loop. Assert
3471 the analysis doesn't fail.
3473 2018-06-04 Carl Love <cel@us.ibm.com>
3475 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3476 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3477 as it is slightly cheaper.
3478 (first_match_or_eos_index_<mode>):
3479 Calculate index using natural element order.
3480 (first_match_index_<mode>):
3481 Calculate index using natural element order.
3482 (first_match_or_eos_index_<mode>):
3483 Calculate index using natural order.
3484 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3485 for BE and LE modes.
3486 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3487 P9V_BUILTIN_VCLZLSBB_V16QI.
3488 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3491 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3493 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3494 indentation and line wrap for many prototypes. Add missing
3495 @smallexample directives around block of prototypes for vec_xl and
3498 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3500 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3501 track if we pass or return IEEE 128-bit floating point.
3502 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3503 C++ mangling that is compatible with GCC 8.1.
3504 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3505 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3506 floating point types.
3507 (rs6000_function_arg_advance_1): Likewise.
3508 (rs6000_mangle_type): Optionally generate mangled names that match
3509 what GCC 8.1 generated for IEEE 128-bit floating point types.
3510 (rs6000_globalize_decl_name): If we have an external function that
3511 passes or returns IEEE 128-bit types, generate a weak reference
3512 from the mangled name used in GCC 8.1 to the current mangled
3514 (rs6000_init_builtins): Make __ibm128 use the long double type if
3515 long double is IBM extended double. Make __float128 use the long
3516 double type if long double is IEEE 128-bit.
3519 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3520 macro for __ibm128 built-in functions.
3521 (PACK_IF): Add __ibm128 pack/unpack functions.
3522 (UNPACK_IF): Likewise.
3523 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3524 enable long double built-in functions if long double is IEEE
3525 128-bit floating point.
3526 (rs6000_invalid_builtin): Update long double built-in function
3528 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3529 functions, adjust the built-in function to use the long double
3530 built-in function if __ibm128 and long double are the same type.
3531 * doc/extend.texi (PowerPC builtins): Update documention for
3532 __builtin_{,un}pack_longdouble. Add documentation for
3533 __builtin_{,un}pack_ibm128.
3535 2018-06-06 Jim Wilson <jimw@sifive.com>
3537 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3538 (struct machine_function): New field interrupt_mode.
3539 (riscv_handle_type_attribute): New function. Add forward declaration.
3540 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3541 (riscv_expand_epilogue): Check interrupt_mode field.
3542 (riscv_set_current_function): Check interrupt attribute args and
3543 set interrupt_mode field.
3544 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3545 (riscv_sret, riscv_uret): New.
3546 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3547 new arguments to interrupt attribute.
3549 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3552 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3553 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3555 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3557 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3558 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3559 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3560 several redundant entries.
3562 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3564 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3565 type from "rtx" to "rtx_insn *".
3566 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3567 for local "call_insn", removing cast.
3568 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3570 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3572 PR tree-optimization/86066
3573 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3574 for BIT_INSERT_EXPR stores.
3576 2018-06-06 Richard Biener <rguenther@suse.de>
3578 PR tree-optimization/86062
3579 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3580 component refs ontop
3581 of to be offsetted base.
3583 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3585 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3586 to be static and remove check on interrupt attribute name.
3588 gcc/testsuite/gcc.target/msp430/
3589 * function-attributes-4.c: New test.
3590 * static-interrupts.c: New test.
3592 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3594 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3595 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3597 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3600 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3602 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3603 Remove second argument, change how error is called.
3604 (aarch64_layout_arg): Remove second argument from
3605 aarch64_err_no_fpadvsimd call.
3606 (aarch64_init_cumulative_args): Ditto.
3607 (aarch64_gimplify_va_arg_expr): Ditto.
3608 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3610 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3612 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3613 (*simple_return_indirect_internal<mode>): Rename from
3614 simple_return_indirect_internal. Use W mode iterator.
3615 (rstorssp): New expander.
3616 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3617 (clrssbsy): New expander.
3618 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3620 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3622 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3624 (cmse_check_pointed_object): Likewise.
3626 2018-06-05 Martin Liska <mliska@suse.cz>
3628 PR gcov-profile/47618
3629 * doc/invoke.texi: Document how -fprofile-dir format
3632 2018-06-05 Richard Biener <rguenther@suse.de>
3634 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3635 removal pretend DOM info isn't available so we do not update
3636 it and only remove edges, not dominated blocks. Actually free
3637 DOM info in case we removed something. Remove unreachable blocks.
3638 (mfb_keep_latches): Work with either DOM info or marked backedges.
3639 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3640 first. Mark backedges if DOM info isn't available.
3641 (Re-)compute DOM info after cleanup_control_flow_pre.
3643 2018-06-05 Richard Biener <rguenther@suse.de>
3645 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3646 (locus_discrim_hasher::hash): Adjust.
3647 (locus_discrim_hasher::equal): Likewise.
3648 (next_discriminator_for_locus): Work on line directly.
3649 (same_line_p): Pass in expanded locus1 as well.
3650 (assign_discriminators): Avoid redundant location expansions.
3652 2018-06-05 Richard Biener <rguenther@suse.de>
3654 PR tree-optimization/86046
3655 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3656 if required after clearing TREE_ADDRESSABLE.
3658 2018-06-05 Richard Biener <rguenther@suse.de>
3660 PR tree-optimization/86047
3661 * tree-ssa-loop.c (for_each_index): Glob handling of all
3662 decls and constants and really handle all of them.
3664 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3667 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3668 qualifier_void_pointer and qualifier_const_void_pointer.
3669 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3670 (arm_init_builtins): Handle the above.
3671 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3672 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3673 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3676 2018-06-05 Martin Liska <mliska@suse.cz>
3678 * auto-profile.c (read_autofdo_file): Do not use
3679 gcov_ctr_summary struct.
3680 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3681 * coverage.c (struct counts_entry): Likewise.
3682 (read_counts_file): Read just single summary entry.
3683 (get_coverage_counts): Use gcov_summary struct.
3684 * coverage.h (get_coverage_counts): Likewise.
3685 * gcov-dump.c (dump_working_sets): Likewise.
3686 (tag_summary): Dump just single summary.
3687 * gcov-io.c (gcov_write_summary): Write just histogram
3689 (gcov_read_summary): Read just single summary.
3690 (compute_working_sets): Use gcov_summary struct.
3691 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3692 of GCOV_COUNTERS_SUMMABLE.
3693 (GCOV_COUNTERS_SUMMABLE): Remove.
3694 (GCOV_FIRST_VALUE_COUNTER): Replace with
3695 GCOV_COUNTER_V_INTERVAL.
3696 (struct gcov_ctr_summary): Remove.
3697 (struct gcov_summary): Directly use fields of former
3699 (compute_working_sets): Use gcov_summary struct.
3700 * gcov.c (read_count_file): Do not use ctrs fields.
3701 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3703 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3705 * profile.c: Likewise.
3706 * profile.h: Likewise.
3708 2018-06-05 Martin Liska <mliska@suse.cz>
3710 PR gcov-profile/84846
3711 * gcov.c (output_lines): Print working directory only
3712 in intermediate format.
3714 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3716 * config/s390/s390-builtin-types.def: Add void function type.
3717 * config/s390/s390-builtins.def: Use the function type for the
3720 2018-06-04 Jim Wilson <jimw@sifive.com>
3722 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3724 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3725 and maybe_eh_return. Change regno to unsigned int. Use new args to
3726 handle EH_RETURN_DATA_REGNO registers properly.
3727 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3728 (riscv_expand_epilogue): Update comment. Change argument name and
3729 type. Update code to use new name and type. Pass new args to
3730 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3732 * config/riscv/riscv.md (NORMAL_RETURN): New.
3733 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3734 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3735 (eh_return): Call gen_eh_return_internal and emit barrier.
3736 (eh_return_internal): Call riscv_expand_epilogue.
3738 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3740 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3741 bit_insertion field and declare can_be_merged_into method.
3742 (merged_store_group::can_be_merged_into): New method.
3743 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3744 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3745 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3747 2018-06-04 Richard Biener <rguenther@suse.de>
3749 PR tree-optimization/85955
3750 * builtins.c (fold_builtin_sincos): Convert pointers to
3751 destination to appropriate type before dereferencing.
3753 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3755 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3757 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3759 * expr.c (expand_expr_real_1): Force the operand into memory if
3760 its TYPE_MODE is BLKmode and if there is no integer mode for
3761 the number of bits being extracted.
3763 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3767 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3768 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3770 2018-06-04 Richard Biener <rguenther@suse.de>
3772 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3773 (cleanup_tree_cfg_noloop): ... single caller. Do
3774 start_recording_case_labels later.
3776 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3778 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3779 to _IMMINTRIN_H_INCLUDED.
3780 * config/i386/pconfigintrin.h: Ditto.
3781 * config/i386/waitpkgintrin.h: Ditto.
3782 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3783 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3784 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3785 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3786 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3787 waitpkgintrin.h and cldemoteintrin.h.
3789 2018-06-04 Richard Biener <rguenther@suse.de>
3791 PR tree-optimization/86038
3792 * tracer.c (find_best_successor): Check probability for
3793 being initialized, bail out if not.
3795 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3798 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3799 of bits to ignore when comparing architectures.
3801 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3803 PR tree-optimization/69615
3804 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3805 maximum or minimum of the type, try to merge it also as if
3806 range1 is + [-, x - 1] or + [x + 1, -].
3809 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3811 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3813 PR tree-optimization/86034
3814 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3815 the unsigned bitfield type in a bit insertion sequence if it does not
3816 have a larger precision than the bitfield size.
3817 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3819 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3821 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3823 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3825 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3826 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3827 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3828 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3830 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3832 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3833 Disable -fdelete-null-pointer-checks for ELF toolchain.
3835 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3836 Kito Cheng <kito.cheng@gmail.com>
3838 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3839 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3840 (nds32*-*-*): Add glibc and uclibc conditions.
3841 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3842 (TARGET_EXCEPT_UNWIND_INFO): Define.
3843 * config/nds32/elf.h: New file.
3844 * config/nds32/linux.h: New file.
3845 * config/nds32/nds32-elf.opt: New file.
3846 * config/nds32/nds32-linux.opt: New file.
3847 * config/nds32/nds32-fp-as-gp.c
3848 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3849 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3851 (nds32_asm_file_end): Ditto.
3852 (nds32_print_operand): Ditto.
3853 (nds32_insert_attributes): Ditto.
3854 (nds32_init_libfuncs): New function.
3855 (TARGET_HAVE_TLS): Define.
3856 (TARGET_INIT_LIBFUNCS): Define.
3857 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3859 (TARGET_ELF): Apply different mcmodel setting.
3860 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3861 been migrated into elf.h and linux.h files.
3862 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3863 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3864 (mcmodel): The content has been migrated into nds32-elf.opt and
3865 nds32-linux.opt files.
3866 * config/nds32/t-elf: New file.
3867 * config/nds32/t-linux: New file.
3869 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3870 Shiva Chen <shiva0217@gmail.com>
3872 * config/nds32/constants.md (unspec_volatile_element): Add
3873 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3874 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3876 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3877 (make_pass_nds32_fp_as_gp): Declare.
3878 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3880 (nds32_asm_function_end_prologue): Remove unused asm output.
3881 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3882 (nds32_asm_file_start): Output necessary fp_as_gp information.
3883 (nds32_option_override): Adjust register usage.
3884 (nds32_expand_prologue): Consider fp_as_gp situation.
3885 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3886 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3889 (simple_return): Ditto.
3890 (omit_fp_begin): Output special directive for fp_as_gp.
3891 (omit_fp_end): Output special directive for fp_as_gp.
3892 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3893 mforbid-fp-as-gp): New options.
3895 2018-06-01 Mark Wielaard <mark@klomp.org>
3897 * dwarf2out.c (dwarf2out_finish): Remove generation of
3898 DW_AT_loclists_base.
3900 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3902 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3903 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3904 (struct merged_store_group): Add bit_insertion field.
3905 (dump_char_array): Use standard hexadecimal format.
3906 (merged_store_group::merged_store_group): Set bit_insertion to false.
3907 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3908 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3909 also print the mask in the dump file.
3910 (pass_store_merging::gate): Minor tweak.
3911 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3912 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3913 stores with INTEGER_CST stores.
3914 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3915 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3916 and use it throughout. Generate bit insertion sequences if need be.
3917 (pass_store_merging::process_store): Remove redundant condition.
3918 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3920 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3922 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3923 the 128-bit floating point types. Fix function comment.
3925 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3927 * config/aarch64/aarch64-simd.md
3928 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3930 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3933 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3935 PR tree-optimization/85989
3936 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3938 (backprop::intersect_uses): Check it when deciding whether this
3939 is a backedge reference.
3940 (backprop::process_block): Add each phi to m_visited_phis
3941 after visiting it, then clear it at the end.
3943 2018-06-01 Richard Biener <rguenther@suse.de>
3945 * tree-vectorizer.h (vect_dr_stmt): New function.
3946 (vect_get_load_cost): Adjust.
3947 (vect_get_store_cost): Likewise.
3948 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3949 Use vect_dr_stmt instead of DR_SMTT.
3950 (vect_record_base_alignments): Likewise.
3951 (vect_calculate_target_alignment): Likewise.
3952 (vect_compute_data_ref_alignment): Likewise and make static.
3953 (vect_update_misalignment_for_peel): Likewise.
3954 (vect_verify_datarefs_alignment): Likewise.
3955 (vector_alignment_reachable_p): Likewise.
3956 (vect_get_data_access_cost): Likewise. Pass down
3957 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3958 (vect_get_peeling_costs_all_drs): Likewise.
3959 (vect_peeling_hash_get_lowest_cost): Likewise.
3960 (vect_enhance_data_refs_alignment): Likewise.
3961 (vect_find_same_alignment_drs): Likewise.
3962 (vect_analyze_data_refs_alignment): Likewise.
3963 (vect_analyze_group_access_1): Likewise.
3964 (vect_analyze_group_access): Likewise.
3965 (vect_analyze_data_ref_access): Likewise.
3966 (vect_analyze_data_ref_accesses): Likewise.
3967 (vect_vfa_segment_size): Likewise.
3968 (vect_small_gap_p): Likewise.
3969 (vectorizable_with_step_bound_p): Likewise.
3970 (vect_prune_runtime_alias_test_list): Likewise.
3971 (vect_analyze_data_refs): Likewise.
3972 (vect_supportable_dr_alignment): Likewise.
3973 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3974 (vect_gen_prolog_loop_niters): Likewise.
3975 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3976 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3978 (vect_recog_mask_conversion_pattern): Likewise.
3979 (vect_try_gather_scatter_pattern): Likewise.
3980 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3981 to vect_get_store_cost.
3982 (vect_get_store_cost): Get stmt_info instead of DR.
3983 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3984 (vect_get_load_cost): Get stmt_info instead of DR.
3986 2018-06-01 Richard Biener <rguenther@suse.de>
3989 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3990 vars wrapped in MEM_REFs.
3992 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3994 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3995 Fix subreg tests so that we only return a choice between
3996 GENERAL_REGS and FP_REGS if the original classes included both.
3998 2018-06-01 Richard Biener <rguenther@suse.de>
4001 * tree-ssa-structalias.c (get_function_part_constraint):
4002 Handle NULL fi->decl.
4003 (find_func_aliases_for_call): Properly handle indirect
4004 fi from direct call.
4005 (find_func_clobbers): Likewise.
4006 (ipa_pta_execute): Likewise.
4007 (create_variable_info_for): For functions that are ifunc_resolver
4008 resolve to a varinfo that contains the result of the resolver call.
4009 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4012 2018-05-31 Michael Collison <michael.collison@arm.com>
4014 * config/aarch64/aarch64.md:
4015 (*fix_to_zero_extenddfdi2): New pattern.
4016 * gcc.target/aarch64/fix_extend1.c: New testcase.
4018 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4022 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4023 and BUILT_IN_STRNCMP_EQ.
4024 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4025 BUILT_IN_STRNCMP_EQ.
4026 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4027 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4028 (gimple_fold_builtin): Likewise.
4029 * tree-ssa-strlen.c (compute_string_length): New function.
4030 (determine_min_obsize): New function.
4031 (handle_builtin_string_cmp): New function to handle calls to
4032 string compare functions.
4033 (strlen_optimize_stmt): Add handling to builtin string compare
4035 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4036 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4037 * tree.c (build_common_builtin_nodes): Add new defines of
4038 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4040 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4043 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4044 functions with naked attribute.
4046 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4048 * config/i386/sse.md (avx_vec_concat<mode>):
4049 Substitute concat_tg_mode mode attribute with xtg_mode.
4050 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4051 (concat_tg_mode): Remove mode attribute.
4053 2018-05-31 Martin Sebor <msebor@redhat.com>
4056 * calls.c (alloc_max_size): Correct a logic error/typo.
4057 Treat excessive arguments as infinite. Warn for invalid arguments.
4058 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4060 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4063 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4064 and movx for Haswell.
4066 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4067 Cesar Philippidis <cesar@codesourcery.com>
4070 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4071 when emitting error on private/firstprivate reductions.
4072 * omp-low.c (lower_omp_target): Avoid reference-type processing
4073 on pointers for firstprivate clause.
4075 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4077 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4080 * config/aarch64/aarch64-simd.md
4081 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4082 (aarch64_ld1_x3_<mode>): Likewise
4083 (aarch64_st1x2<VALLDIF:mode>): Likewise
4084 (aarch64_st1_x2_<mode>): Likewise
4085 (aarch64_st1x3<VALLDIF:mode>): Likewise
4086 (aarch64_st1_x3_<mode>): Likewise
4087 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4088 (vld1_s8_x3): Likewise.
4089 (vld1_u16_x3): Likewise.
4090 (vld1_s16_x3): Likewise.
4091 (vld1_u32_x3): Likewise.
4092 (vld1_s32_x3): Likewise.
4093 (vld1_u64_x3): Likewise.
4094 (vld1_s64_x3): Likewise.
4095 (vld1_f16_x3): Likewise.
4096 (vld1_f32_x3): Likewise.
4097 (vld1_f64_x3): Likewise.
4098 (vld1_p8_x3): Likewise.
4099 (vld1_p16_x3): Likewise.
4100 (vld1_p64_x3): Likewise.
4101 (vld1q_u8_x3): Likewise.
4102 (vld1q_s8_x3): Likewise.
4103 (vld1q_u16_x3): Likewise.
4104 (vld1q_s16_x3): Likewise.
4105 (vld1q_u32_x3): Likewise.
4106 (vld1q_s32_x3): Likewise.
4107 (vld1q_u64_x3): Likewise.
4108 (vld1q_s64_x3): Likewise.
4109 (vld1q_f16_x3): Likewise.
4110 (vld1q_f32_x3): Likewise.
4111 (vld1q_f64_x3): Likewise.
4112 (vld1q_p8_x3): Likewise.
4113 (vld1q_p16_x3): Likewise.
4114 (vld1q_p64_x3): Likewise.
4115 (vst1_s64_x2): Likewise.
4116 (vst1_u64_x2): Likewise.
4117 (vst1_f64_x2): Likewise.
4118 (vst1_s8_x2): Likewise.
4119 (vst1_p8_x2): Likewise.
4120 (vst1_s16_x2): Likewise.
4121 (vst1_p16_x2): Likewise.
4122 (vst1_s32_x2): Likewise.
4123 (vst1_u8_x2): Likewise.
4124 (vst1_u16_x2): Likewise.
4125 (vst1_u32_x2): Likewise.
4126 (vst1_f16_x2): Likewise.
4127 (vst1_f32_x2): Likewise.
4128 (vst1_p64_x2): Likewise.
4129 (vst1q_s8_x2): Likewise.
4130 (vst1q_p8_x2): Likewise.
4131 (vst1q_s16_x2): Likewise.
4132 (vst1q_p16_x2): Likewise.
4133 (vst1q_s32_x2): Likewise.
4134 (vst1q_s64_x2): Likewise.
4135 (vst1q_u8_x2): Likewise.
4136 (vst1q_u16_x2): Likewise.
4137 (vst1q_u32_x2): Likewise.
4138 (vst1q_u64_x2): Likewise.
4139 (vst1q_f16_x2): Likewise.
4140 (vst1q_f32_x2): Likewise.
4141 (vst1q_f64_x2): Likewise.
4142 (vst1q_p64_x2): Likewise.
4143 (vst1_s64_x3): Likewise.
4144 (vst1_u64_x3): Likewise.
4145 (vst1_f64_x3): Likewise.
4146 (vst1_s8_x3): Likewise.
4147 (vst1_p8_x3): Likewise.
4148 (vst1_s16_x3): Likewise.
4149 (vst1_p16_x3): Likewise.
4150 (vst1_s32_x3): Likewise.
4151 (vst1_u8_x3): Likewise.
4152 (vst1_u16_x3): Likewise.
4153 (vst1_u32_x3): Likewise.
4154 (vst1_f16_x3): Likewise.
4155 (vst1_f32_x3): Likewise.
4156 (vst1_p64_x3): Likewise.
4157 (vst1q_s8_x3): Likewise.
4158 (vst1q_p8_x3): Likewise.
4159 (vst1q_s16_x3): Likewise.
4160 (vst1q_p16_x3): Likewise.
4161 (vst1q_s32_x3): Likewise.
4162 (vst1q_s64_x3): Likewise.
4163 (vst1q_u8_x3): Likewise.
4164 (vst1q_u16_x3): Likewise.
4165 (vst1q_u32_x3): Likewise.
4166 (vst1q_u64_x3): Likewise.
4167 (vst1q_f16_x3): Likewise.
4168 (vst1q_f32_x3): Likewise.
4169 (vst1q_f64_x3): Likewise.
4170 (vst1q_p64_x3): Likewise.
4172 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4174 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4175 user_label_prefix to name.
4177 * tree-core.h: Update comment about the format of NAME string
4178 passed to handler in attribute_spec.
4180 * config/msp430/msp430.md: Remove erroneous subreg expression from
4181 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4182 zero_extend{q,h}isi2.
4184 2018-05-30 Borislav Petkov <bp@suse.de>
4186 * doc/extend.texi: Document some architecture specific
4187 constraints and sort entries.
4189 2018-05-30 Martin Sebor <msebor@redhat.com>
4192 * builtins.c (expand_builtin_stpcpy_1): New function.
4193 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4194 only if the former succeeds.
4196 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4198 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4201 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4203 * doc/invoke.texi (-flinker-output): Document
4205 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4207 * passes.c (ipa_write_summaries): Only modify statements if body
4209 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4210 incrementally linking.
4211 (ipa_passes): Likewise.
4212 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4213 pass down resolution info.
4214 * common.opt (flag_incremental_link): Update info.
4215 * gcc.c (plugin specs): Turn flinker-output=* to
4216 -plugin-opt=-linker-output-known
4217 * toplev.c (compile_file): Also cut compilation when doing incremental
4219 * flag-types. (enum lto_partition_model): Add
4220 LTO_LINKER_OUTPUT_NOLTOREL.
4221 (invoke.texi): Add -flinker-output docs.
4222 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4223 link same way as WPA; do not stream in dead initializers.
4225 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4226 darf2out_register_external_decl): Support incremental link.
4228 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4230 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4232 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4234 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4235 it down to simple_object_copy_lto_debug_sections.
4236 (run_gcc): Determine incremental LTO link time and configure
4237 lto1 into non-wpa mode, disable renaming of debug sections.
4239 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4241 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4242 descriptions of various incorrectly documented functions.
4244 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4247 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4248 address check not strict.
4250 2018-05-30 Richard Biener <rguenther@suse.de>
4252 PR tree-optimization/85964
4253 * tracer.c (better_p): Drop initialized count check, we only
4254 call the function with initialized counts now.
4255 (find_best_successor): Do find a best edge if one
4256 has uninitialized count.
4257 (find_best_predecessor): Likewise. Do BB frequency check only
4258 if count is initialized.
4260 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4262 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4263 (aarch64_ldrstr_offset_compare): New.
4264 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4265 load/store orderings.
4266 (aarch64_gen_adjusted_ldpstp): Likewise.
4268 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4270 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4271 Check for subset of GENERAL_REGS and FP_REGS.
4272 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4275 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4277 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4278 and wi::to_poly_offset. Add the current offset and then check
4279 whether the sum fits, rather than using an unchecked addition of
4280 a checked term. Check for a shwi rather than a uhwi.
4281 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4282 (store_constructor): Use poly_int_tree_p.
4283 (expand_expr_real_1): Likewise.
4284 * function.c (assign_temp): Likewise.
4285 * fold-const.c (const_binop): Use poly_int_tree_p and
4287 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4289 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4290 to_poly_offset for MEM offsets.
4291 * ipa-icf.c (sem_variable::equals): Likewise.
4292 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4293 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4294 wi::to_poly_offset for BIT_FIELD_REF offsets.
4295 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4297 * var-tracking.c (emit_note_insn_var_location): Use
4298 tree_to_poly_uint64.
4300 2018-05-29 Jim Wilson <jimw@sifive.com>
4302 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4304 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4307 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4308 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4310 (sse4_1_round<mode>2): Use nonimmediate_operand
4311 for operand 1 predicate.
4313 2018-05-29 Martin Sebor <msebor@redhat.com>
4314 Richard Biener <rguenther@suse.de>
4317 * calls.c (get_size_range): Call determine_value_range instead
4318 of get_value_range..
4319 * tree-vrp.h (determine_value_range): Declared new function.
4320 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4322 2018-05-29 Richard Biener <rguenther@suse.de>
4324 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4325 sure to use non-pattern stmts for get_earlier_stmt arguments.
4326 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4327 called on pattern stmts.
4328 (get_later_stmt): Likewise.
4330 2018-05-29 Martin Liska <mliska@suse.cz>
4332 PR gcov-profile/85759
4333 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4336 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4338 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4340 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4344 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4345 VEC_PACK_FLOAT_EXPR): New tree codes.
4346 * tree-pretty-print.c (op_code_prio): Handle
4347 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4348 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4349 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4350 * tree-inline.c (estimate_operator_cost): Likewise.
4351 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4352 * fold-const.c (const_binop): Likewise.
4353 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4354 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4355 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4356 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4357 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4358 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4359 * expr.c (expand_expr_real_2): Likewise.
4360 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4361 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4362 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4364 * optabs.c (expand_widen_pattern_expr): For
4365 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4366 sign from result type rather than operand's type.
4367 (expand_binop_directly): For vec_packu_float_optab and
4368 vec_packs_float_optab allow result type to be different from operand's
4370 * optabs-tree.c (optab_for_tree_code): Handle
4371 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4372 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4373 * tree-vect-generic.c (expand_vector_operations_1): Handle
4374 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4375 VEC_PACK_FLOAT_EXPR.
4376 * tree-vect-stmts.c (supportable_widening_operation): Handle
4378 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4379 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4380 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4381 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4382 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4384 (vec_pack<floatprefix>_float_<mode>): New expander.
4385 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4387 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4388 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4389 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4390 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4391 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4393 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4394 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4395 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4396 VEC_PACK_FLOAT_EXPR): Document.
4398 2018-05-29 Richard Biener <rguenther@suse.de>
4400 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4402 (stmt_vec_info_vec): Make pointer.
4403 (init_stmt_vec_info_vec): Remove.
4404 (free_stmt_vec_info_vec): Likewise.
4405 (set_stmt_vec_info_vec): New function.
4406 (free_stmt_vec_infos): Likewise.
4407 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4408 (set_vinfo_for_stmt): Likewise.
4409 (get_earlier_stmt): Likewise.
4410 (get_later_stmt): Likewise.
4411 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4412 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4413 (vec_info::~vec_info): Free stmt_vec_infos.
4414 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4415 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4416 (pass_slp_vectorize::execute): Likewise.
4417 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4418 (free_stmt_vec_info_vec): Likewise.
4419 (set_stmt_vec_info_vec): New function.
4420 (free_stmt_vec_infos): Likewise.
4421 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4422 the global stmt_vec_info_vec.
4423 * tree-parloops.c (gather_scalar_reductions): Use
4424 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4427 2018-05-29 Richard Biener <rguenther@suse.de>
4429 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4431 2018-05-29 Martin Liska <mliska@suse.cz>
4432 David Malcolm <dmalcolm@redhat.com>
4434 * vec.c (test_reverse): New.
4435 (vec_c_tests): Add new test.
4436 * vec.h (vl_ptr>::reverse): New function.
4438 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4440 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4442 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4445 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4447 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4449 2018-05-28 Richard Biener <rguenther@suse.de>
4451 PR tree-optimization/85933
4452 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4453 look at stmts marked as vectorizable.
4455 2018-05-28 Richard Biener <rguenther@suse.de>
4457 PR tree-optimization/85934
4458 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4459 vector boolean check before scalar optimization.
4461 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4463 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4466 2018-05-28 Mark Wielaard <mark@klomp.org>
4468 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4469 if it is an expression containing a minus sign.
4471 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4473 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4475 2018-05-27 Paul Koning <ni1d@arrl.net>
4477 * config/pdp11/pdp11.md (truncsihi2): Remove.
4479 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4480 Chung-Ju Wu <jasonwucj@gmail.com>
4482 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4484 (unaligned_store_dw): Ditto.
4485 * config/nds32/nds32-memory-manipulation.c
4486 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4487 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4488 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4489 (emit_setmem_word_loop): Rename to ...
4490 (emit_setmem_doubleword_loop): ... this.
4491 (nds32_gen_dup_4_byte_to_word_value): New function.
4492 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4493 (nds32_expand_setmem_loop): Refine implementation.
4494 (nds32_expand_setmem_loop_v3m): Ditto.
4495 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4498 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4500 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4502 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4504 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4505 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4506 machine->attr_no_prologue_p.
4507 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4508 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4509 (nds32_expand_epilogue): Consider attr_naked_p.
4510 (nds32_expand_epilogue_v3pop): Likewise.
4511 (nds32_can_use_return_insn): Likewise.
4512 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4513 attr_no_prologue_p fields.
4514 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4516 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4519 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4521 * config/i386/sse.md
4522 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4524 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4526 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4528 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4530 (*<floatsuffix>floatv2div2sf2): Rename to ...
4531 (*float<floatunssuffix>v2div2sf2): ... this.
4532 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4533 (float<floatunssuffix>v2div2sf2_mask): ... this.
4534 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4535 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4536 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4538 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4540 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4542 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4544 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4546 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4548 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4549 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4550 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4551 gen_ufix_truncv8dfv8si2.
4552 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4553 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4554 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4555 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4556 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4557 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4558 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4559 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4561 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4565 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4567 2018-05-25 Jim Wilson <jimw@sifive.com>
4569 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4570 * config/riscv/riscv.c (struct machine_function): Add
4571 interrupt_handler_p and attribute_checked_p fields.
4572 (riscv_attribute_table): Add interrupt.
4573 (riscv_interrupt_type_p): New.
4574 (riscv_save_reg_p): Save extra regs for interrupt handler.
4575 (riscv_use_save_libcall): Return false for interrupt handler.
4576 (riscv_first_stack_step): Add forward declaration.
4577 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4578 for interrupt handler with large frame. Use it for saved reg list.
4579 (riscv_expand_prologue): Move flag_stack_usage_info support to
4580 eliminate duplication.
4581 (riscv_expand_epilogue): Generate mret for interrupt handler.
4582 (riscv_epilogue_uses): New.
4583 (riscv_can_use_return_insn): Return false for interrupt handler.
4584 (riscv_function_ok_for_sibcall): Likewise.
4585 (riscv_set_current_function): Add interrupt handler support.
4586 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4587 * config/riscv/riscv.md (UNSPECV_MRET): New.
4589 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4591 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4593 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4595 PR tree-optimization/85712
4596 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4597 this candidate has already been replaced in-situ by a copy.
4599 2018-05-25 Jason Merrill <jason@redhat.com>
4601 PR c++/80485 - inline function non-zero address.
4602 * symtab.c (nonzero_address): Check DECL_COMDAT.
4604 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4607 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4608 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4609 extension of SImode operation. Use const123_operand predicate.
4610 (*saddsi_1): Remove.
4611 (*saddl_se_1): Ditto.
4613 (*ssubl_se_1): Ditto.
4614 * config/alpha/predicates.md (const123_operand): New predicate.
4615 * config/alpha/constraints.md (P): Use IN_RANGE.
4617 2018-05-25 Richard Biener <rguenther@suse.de>
4619 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4621 (ref_maybe_used_by_stmt_p): Likewise.
4622 (stmt_may_clobber_ref_p): Likewise.
4623 (stmt_may_clobber_ref_p_1): Likewise.
4624 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4626 (ref_maybe_used_by_stmt_p): Likewise.
4627 (stmt_may_clobber_ref_p): Likewise.
4628 (stmt_may_clobber_ref_p_1): Likewise.
4629 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4630 the alias oracle to disambiguate DRs with stmts DR analysis
4632 (vect_analyze_data_refs): Do not give up on not analyzable
4633 DRs for BB vectorization. Remove code truncating the dataref
4636 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4639 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4640 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4641 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4643 2018-05-25 Richard Biener <rguenther@suse.de>
4645 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4646 function, combining stmt data ref gathering and fatal analysis
4648 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4649 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4650 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4651 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4652 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4654 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4656 PR tree-optimization/85720
4657 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4658 SCC if all partitions are builtins.
4659 (version_loop_by_alias_check): New parameter. Generate cancelable
4660 runtime alias check if all partitions are builtins.
4661 (distribute_loop): Update call to above function.
4663 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4665 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4666 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4667 (parm_default_def_partition_arg): Ditto.
4668 (set_parm_default_def_partition): Ditto.
4669 (get_parm_default_def_partitions): Ditto and make it static.
4670 (get_undefined_value_partitions): Ditto and make it static.
4671 (remove_ssa_form): Refactor call to init_var_map here.
4672 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4673 computation for loop region.
4674 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4675 (register_default_def): Delete.
4676 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4677 (parm_default_def_partition_arg): Ditto.
4678 (set_parm_default_def_partition): Ditto.
4679 (get_parm_default_def_partitions): Ditto and make it static.
4680 (get_undefined_value_partitions): Ditto and make it static.
4681 (coalesce_with_default, coalesce_with_default): Update comment.
4682 (create_coalesce_list_for_region): New func factored out from
4683 create_outofssa_var_map.
4684 (populate_coalesce_list_for_outofssa): New func factored out from
4685 create_outofssa_var_map and coalesce_ssa_name.
4686 (create_outofssa_var_map): Delete.
4687 (coalesce_ssa_name): Refactor to support live range computation.
4688 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4689 (get_parm_default_def_partitions): Delete.
4690 (get_undefined_value_partitions): Ditto.
4691 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4692 computation for loop region.
4693 (new_tree_live_info, loe_visit_block): Ditto.
4694 (live_worklist, set_var_live_on_entry): Ditto.
4695 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4696 * tree-ssa-live.h (struct _var_map): New fields.
4697 (init_var_map): Change decl.
4698 (region_contains_p): New.
4700 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4702 * tree-ssa-live.h (live_merge_and_clear): Delete.
4704 2018-05-25 Richard Biener <rguenther@suse.de>
4707 * tree-dump.c (dequeue_and_dump): Remove access to removed
4708 operand 2 of a SWITCH_EXPR.
4710 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4712 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4713 multiplication and division.
4714 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4715 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4716 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4717 (cond_udiv_optab, cond_umod_optab): New optabs.
4718 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4719 (IFN_COND_RDIV): New internal functions.
4720 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4721 TRUNC_MOD_EXPR and RDIV_EXPR.
4722 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4723 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4725 (SVE_INT_BINARY): Include mult.
4726 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4727 (optab, sve_int_op): Handle mult.
4728 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4730 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4731 for SVE_INT_BINARY_SD.
4733 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4735 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4736 (optab, sve_int_op): Handle div and udiv.
4737 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4738 for SVE_INT_BINARY_SD.
4739 (*<optab><mode>3): New insn for the same.
4741 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4743 * tree-vect-patterns.c: Include predict.h.
4744 (vect_recog_divmod_pattern): Restrict check for division support
4745 to when optimizing for size.
4747 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4749 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4750 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4751 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4752 (gimple_match_op::set_op): Likewise.
4753 (gimple_resimplify4): Declare.
4754 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4755 (expr::gen_transform): Likewise.
4756 (decision_tree::gen): Generate a simplification routine for 4 operands.
4757 * gimple-match-head.c (gimple_simplify): Add an overload for
4758 4 operands. In the top-level function, handle up to 4 call
4759 arguments and call gimple_resimplify4.
4760 (gimple_resimplify4): New function.
4761 (build_call_internal): Pass a fourth operand.
4762 (maybe_push_to_seq): Likewise.
4763 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4764 Fold VEC_COND_EXPRs of an operation and a default value into
4765 an IFN_COND_* function if possible.
4766 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4768 (SVE_COND_FP_BINARY): Include them.
4769 (optab, sve_fp_op): Handle them.
4770 (SVE_INT_BINARY_REV): New code iterator.
4771 (SVE_COND_FP_BINARY_REV): New int iterator.
4772 (commutative): New int attribute.
4773 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4775 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4777 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4778 (*cond_<optab><mode>): New patterns for reversed operands.
4780 2018-05-25 Richard Biener <rguenther@suse.de>
4782 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4783 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4784 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4785 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4786 * tree-vect-data-refs.c (everywhere): Adjust users.
4787 * tree-vect-loop.c (everywhere): Likewise.
4788 * tree-vect-slp.c (everywhere): Likewise.
4789 * tree-vect-stmts.c (everywhere): Likewise.
4790 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4792 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4794 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4796 (gcc_cv_as_section_exclude): ... this.
4797 Try Solaris as #exclude syntax.
4798 * configure: Regenerate.
4799 * config.in: Regenerate.
4800 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4802 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4803 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4805 * varasm.c (default_elf_asm_named_section): Don't check if
4806 HAVE_GAS_SECTION_EXCLUDE is defined.
4808 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4810 * doc/md.texi: Update the documentation of the cond_* optabs
4811 to mention the new final operand. Fix GET_MODE_NUNITS call.
4812 Describe the scalar case too.
4813 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4814 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4816 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4817 (get_conditional_internal_fn): Update comment.
4818 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4819 accumulator value as a final argument to conditional functions.
4820 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4821 a define_expand and add an "else" operand. Assert for now that
4822 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4823 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4824 (*cond_<optab><mode>): New patterns.
4825 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4826 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4827 (UNSPEC_COND_EOR): Delete.
4828 (optab): Remove associated mappings.
4829 (SVE_INT_BINARY): New code iterator.
4830 (sve_int_op): Remove int attribute and add "minus" to the code
4832 (SVE_COND_INT_OP): Delete.
4833 (SVE_COND_FP_OP): Rename to...
4834 (SVE_COND_FP_BINARY): ...this.
4836 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4838 * optabs.c (can_reuse_operands_p): New function.
4839 (maybe_legitimize_operands): Try to reuse the results for
4842 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4844 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4845 Add {q} suffix to insn mnemonic.
4847 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4849 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4850 (msp430_warn_func_return): New.
4852 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4854 * fold-const.c (tree_nonzero_bits): New function.
4855 * fold-const.h (tree_nonzero_bits): Likewise.
4856 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4857 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4859 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4863 * varasm.c (assemble_alias): Check ifunc_resolver only on
4866 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4869 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4870 when memory input operand is handled.
4872 2018-05-24 Luis Machado <luis.machado@linaro.org>
4874 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4876 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4878 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4880 * match.pd: Delay FMA folds until after vectorization.
4882 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4885 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4886 address check not strict.
4888 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4890 * gimple-match.h (gimple_match_op): New class.
4891 (mprts_hook): Replace parameters with a gimple_match_op *.
4892 (maybe_build_generic_op): Likewise.
4893 (gimple_simplified_result_is_gimple_val): Replace parameters with
4894 a const gimple_match_op *.
4895 (gimple_simplify): Replace code_helper * and tree * parameters with
4896 a gimple_match_op * parameter.
4897 (gimple_resimplify1): Replace code_helper *, tree and tree *
4898 parameters with a gimple_match_op * parameter.
4899 (gimple_resimplify2): Likewise.
4900 (gimple_resimplify3): Likewise.
4901 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4902 parameters with a gimple_match_op * parameter.
4903 * gimple-match-head.c (gimple_simplify): Change prototypes of
4904 auto-generated functions to take a gimple_match_op * instead of
4905 separate code_helper * and tree * parameters. Make the same
4906 change in the top-level overload and update calls to the
4907 gimple_resimplify routines. Update calls to the auto-generated
4908 functions and to maybe_push_res_to_seq in the publicly-facing
4909 operation-specific gimple_simplify overloads.
4910 (gimple_match_op::MAX_NUM_OPS): Define.
4911 (gimple_resimplify1): Replace rcode and ops with a single res_op
4912 parameter. Update call to gimple_simplify.
4913 (gimple_resimplify2): Likewise.
4914 (gimple_resimplify3): Likewise.
4915 (mprts_hook): Replace parameters with a gimple_match_op *.
4916 (maybe_build_generic_op): Likewise.
4917 (build_call_internal): Replace type, nargs and ops with
4918 a gimple_match_op *.
4919 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4920 with a single gimple_match_op *. Update calls to mprts_hook,
4921 build_call_internal and gimple_simplified_result_is_gimple_val.
4922 Factor out code that is common to the tree_code and combined_fn cases.
4923 * genmatch.c (expr::gen_transform): Replace tem_code and
4924 tem_ops with a gimple_match_op called tem_op. Update calls
4925 to the gimple_resimplify functions and maybe_push_res_to_seq.
4926 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4927 res_ops. Update call to the gimple_resimplify functions.
4928 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4929 (decision_tree::gen): Make the functions take a gimple_match_op *
4930 called res_op instead of separate res_code and res_ops parameters.
4931 Update call accordingly.
4932 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4933 and ops with a single res_op parameter. Update calls to
4934 maybe_build_generic_op and maybe_push_res_to_seq.
4935 (fold_stmt_1): Update calls to gimple_simplify and
4936 replace_stmt_with_simplification.
4937 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4938 and gimple_simplified_result_is_gimple_val.
4939 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4941 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4942 with a gimple_match_op *.
4943 (vn_nary_build_or_lookup): Likewise. Update call to
4944 vn_nary_build_or_lookup_1.
4945 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4946 gimple_match_op *. Update calls to the gimple_resimplify routines
4947 and to gimple_simplified_result_is_gimple_val.
4948 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4949 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4950 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4951 (visit_nary_op): Likewise.
4952 (visit_reference_op_load): Likewise.
4954 2018-05-23 Luis Machado <luis.machado@linaro.org>
4956 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4957 modifier for printing the step amount.
4959 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4962 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4965 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4967 * doc/sourcebuild.texi (Endianness): New subsubsection.
4969 2018-05-23 Luis Machado <luis.machado@linaro.org>
4971 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4972 <prefetch_dynamic_strides>: New const bool field.
4973 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4974 prefetch_dynamic_strides.
4975 (exynosm1_prefetch_tune): Likewise.
4976 (thunderxt88_prefetch_tune): Likewise.
4977 (thunderx_prefetch_tune): Likewise.
4978 (thunderx2t99_prefetch_tune): Likewise.
4979 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4981 (aarch64_override_options_internal): Update to set
4982 PARAM_PREFETCH_DYNAMIC_STRIDES.
4983 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4984 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4985 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4986 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4987 prefetch-dynamic-strides setting.
4989 2018-05-23 Luis Machado <luis.machado@linaro.org>
4991 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4992 <minimum_stride>: New const int field.
4993 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4994 minimum_stride field defaulting to -1.
4995 (exynosm1_prefetch_tune): Likewise.
4996 (thunderxt88_prefetch_tune): Likewise.
4997 (thunderx_prefetch_tune): Likewise.
4998 (thunderx2t99_prefetch_tune): Likewise.
4999 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5000 <default_opt_level>: Set to 3.
5001 (aarch64_override_options_internal): Update to set
5002 PARAM_PREFETCH_MINIMUM_STRIDE.
5003 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5004 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5005 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5006 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5007 stride is constant and is below the minimum stride threshold.
5009 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5011 * config/arm/arm-cpus.in (mode26): Delete.
5012 (armv4): Delete mode26 reference.
5013 * config/arm/arm.c (arm_configure_build_target): Delete use of
5016 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5018 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5020 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5021 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5023 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5024 (floatunsdidf2): Ditto.
5026 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5028 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5029 (fixuns_trunc<mode>si2_avx512f): Ditto.
5030 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5031 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5032 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5034 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5036 PR rtl-optimization/79985
5037 * df-scan.c (df_insn_refs_collect): Remove special case for
5038 global registers and asm statements.
5040 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5042 * extend.texi (Global Register Variables): Rewrite the bullet list.
5043 Note that the register is available for allocation. Note that access
5044 via inline asm must use constraints. Add note about async-signal
5045 handlers. Remove paragraph about automagic register selection.
5047 2018-05-23 Richard Biener <rguenther@suse.de>
5049 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5050 of fixed offset from memset VN.
5052 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5054 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5056 (alloc_cand_and_find_basis): Initialize first_interp field.
5057 (slsr_process_mul): Modify first_interp field.
5058 (slsr_process_add): Likewise.
5059 (slsr_process_cast): Modify first_interp field for each new
5061 (slsr_process_copy): Likewise.
5062 (dump_candidate): Dump first_interp field.
5063 (replace_mult_candidate): Process all interpretations, not just
5065 (replace_rhs_if_not_dup): Likewise.
5066 (replace_one_candidate): Likewise.
5068 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5070 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5072 (aarch64_needs_frame_chain): New function.
5073 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5075 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5078 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5079 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5080 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5081 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5082 as true for strict-align.
5083 (aarch64_can_inline_p): Perform checks even when callee has no
5084 attributes to check for strict alignment.
5085 * doc/extend.texi (AArch64 Function Attributes): Document
5087 * doc/invoke.texi: (AArch64 Options): Likewise.
5089 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5091 PR tree-optimization/85853
5092 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5093 the handling of the root of the node to...
5094 (vect_slp_analyze_node_operations_1): ...this new function,
5095 and run the whole thing with the child nodes' def types
5096 set according to their SLP node's def type.
5098 2018-05-23 Richard Biener <rguenther@suse.de>
5101 * tree-data-ref.c (create_runtime_alias_checks): Defer
5102 and ignore overflow warnings.
5104 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5106 PR tree-optimization/85822
5107 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5110 2018-05-23 Richard Biener <rguenther@suse.de>
5112 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5113 memset constants via native_interpret_expr.
5115 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5118 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5120 (cgraph_node::create_alias): Likewise.
5121 (cgraph_node::get_availability): Check ifunc_resolver instead
5122 of looking up ifunc attribute.
5123 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5124 * varasm.c (do_assemble_alias): Likewise.
5125 (assemble_alias): Likewise.
5126 (default_binds_local_p_3): Likewise.
5127 * cgraph.h (cgraph_node): Add ifunc_resolver.
5128 (cgraph_node::only_called_directly_or_aliased_p): Return false
5130 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5132 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5133 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5134 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5135 instead of looking up ifunc attribute.
5137 2018-05-22 Luis Machado <luis.machado@linaro.org>
5139 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5141 2018-05-22 Martin Sebor <msebor@redhat.com>
5144 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5145 only when expasion succeeds.
5146 (expand_builtin_strcmp): Same.
5147 (expand_builtin_strncmp): Same.
5149 2018-05-22 Martin Sebor <msebor@redhat.com>
5151 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5153 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5154 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5156 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5157 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5159 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5160 Add check for legitimate_address.
5161 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5162 (aarch64_swap_ldrstr_operands): New.
5163 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5166 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5167 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5169 * config/aarch64/aarch64.md: New patterns to generate stp
5171 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5172 single words and double words.
5173 (load_pair_sw, load_pair_dw): Likewise.
5174 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5176 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5178 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5179 for different mode ldpstp and add peephole for merged zero stores.
5181 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5183 (aarch64_gen_store_pair): Rename calls to match new patterns.
5184 (aarch64_gen_load_pair): Rename calls to match new patterns.
5185 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5186 (load_pair<DREG:mode><DREG2:mode>): ... This.
5187 (store_pair<mode>): Rename to...
5188 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5189 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5191 (V_INT_EQUIV): Handle SImode.
5192 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5195 2018-05-22 Martin Sebor <msebor@redhat.com>
5198 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5199 or ajust the presumed bound on an operation to avoid unnecessary
5202 2018-05-22 Martin Sebor <msebor@redhat.com>
5204 PR tree-optimization/85826
5205 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5206 assuming that a DECL necesarily has a constant size.
5208 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5211 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5213 2018-05-22 Richard Biener <rguenther@suse.de>
5215 PR tree-optimization/85834
5216 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5217 non-constant and non-zero memset arguments.
5219 2018-05-22 Martin Liska <mliska@suse.cz>
5222 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5224 2018-05-22 Richard Biener <rguenther@suse.de>
5226 PR tree-optimization/85863
5227 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5228 comparisons when vectype is specified.
5229 (vectorizable_condition): Do not specify vectype for
5230 vect_is_simple_cond when SLP vectorizing.
5232 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5235 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5236 define __ibm128 as long double.
5237 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5238 as a distinct type when IEEE 128-bit support is enabled.
5239 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5240 128-bit types to use the correct functions.
5241 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5242 convert between 128-bit floating point types that have different
5243 modes but the same representation, instead of using gen_lowpart to
5245 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5247 (IFKF_reg): New attributes to give the register constraints for
5249 (extend<mode>tf2_internal): New insns to mark an explicit
5250 conversion between 128-bit floating point types that have a
5251 different mode but share the same representation.
5253 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5255 PR tree-optimization/85814
5256 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5257 a null return from get_strinfo when unsharing the next
5258 strinfo in the chain.
5260 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5263 * varasm.c (weak_finish): Clean up weak_decls.
5265 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5267 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5268 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5269 UNSPEC_UADALP values.
5270 * config/aarch64/iterators.md (ABAL): New int iterator.
5273 (sur): Add mappings for the above.
5274 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5276 (aarch64_<sur>abal<mode>_4): Likewise.
5277 (aarch64_<sur>adalp<mode>_3): Likewise.
5278 (<sur>sadv16qi): New define_expand.
5280 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5282 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5283 (*movdf_internal): Ditto.
5284 (*rcpsf2_sse): Ditto.
5285 (*rsqrtsf2_sse): Ditto.
5286 (*sqrt<mode>2_sse): Ditto.
5288 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5290 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5292 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5293 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5294 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5295 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5297 * config/aarch64/arm_neon.h: Likewise.
5298 * config/aarch64/iterators.md (VQ_I): New.
5300 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5302 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5304 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5307 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5309 * config/nds32/constraints.md (S): New constraint.
5310 * config/nds32/nds32.md (call_internal): Use constraint S.
5311 (call_value_internal): Likewise.
5312 (sibcall_internal): Likewise.
5313 (sibcall_value_internal): Likewise.
5315 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5316 Chung-Ju Wu <jasonwucj@gmail.com>
5318 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5321 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5322 Chung-Ju Wu <jasonwucj@gmail.com>
5324 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5325 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5326 (nds32_rtx_costs_impl): Simplify.
5327 (nds32_address_cost_impl): Simplify.
5328 (nds32_init_rtx_costs): New function.
5329 (nds32_rtx_costs_speed_prefer): Likewise.
5330 (nds32_rtx_costs_size_prefer): Likewise.
5331 (nds32_address_cost_speed_prefer): Likewise.
5332 (nds32_address_cost_speed_fwprop): Likewise.
5333 (nds32_address_cost_size_prefer): Likewise.
5334 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5335 * config/nds32/nds32.c (nds32_option_override): Use
5336 nds32_init_rtx_costs function.
5338 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5340 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5341 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5342 (TARGET_PIPELINE_N8): Likewise.
5343 (TARGET_PIPELINE_N10): Likewise.
5344 (TARGET_PIPELINE_N13): Likewise.
5345 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5347 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5349 * config/nds32/nds32-fpu.md: Update copyright year.
5351 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5353 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5355 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5357 * config/nds32/nds32.c
5358 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5359 * config/nds32/nds32.opt (minline-asm-r15): New option.
5361 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5363 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5365 * config/nds32/nds32.md (abssi2): New pattern.
5367 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5369 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5370 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5371 Merge insn pattern from sse_cvtsi2ss<round_name> and
5372 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5373 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5374 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5375 using SWI48 mode iterator.
5376 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5377 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5378 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5379 pattern from sse_cvttss2si<round_saeonly_name>
5380 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5381 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5382 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5383 using SWI48 mode iterator.
5384 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5385 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5386 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5387 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5388 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5389 using SWI48 mode iterator.
5390 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5391 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5392 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5393 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5394 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5395 SWI48 mode iterator.
5396 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5397 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5398 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5399 pattern from sse_cvttsd2si<round_saeonly_name>
5400 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5402 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5404 * config/nds32/nds32-md-auxiliary.c
5405 (nds32_valid_smw_lwm_base_p): Refine.
5406 (nds32_output_smw_single_word): Refine.
5407 (nds32_output_smw_double_word): New.
5408 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5410 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5412 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5413 (nds32_output_stack_pop): Refine.
5414 (nds32_expand_unaligned_load): Refine.
5415 (nds32_expand_unaligned_store): Refine.
5417 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5418 Chung-Ju Wu <jasonwucj@gmail.com>
5420 * config/nds32/constants.md: Add TP_REGNUM constant.
5421 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5422 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5424 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5425 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5426 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5427 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5428 and PIC code generation.
5429 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5431 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5433 * config/nds32/nds32.md: Support TLS and PIC.
5434 * config/nds32/nds32.c: Support TLS and PIC.
5435 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5436 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5439 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5441 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5442 mode with E_ prefix.
5444 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5445 Chung-Ju Wu <jasonwucj@gmail.com>
5447 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5448 * config/nds32/nds32-md-auxiliary.c
5449 (symbolic_reference_mentioned_p): New.
5450 (nds32_legitimize_ict_address): New.
5451 (nds32_expand_ict_move): New.
5452 (nds32_indirect_call_referenced_p): New.
5453 (nds32_symbol_binds_local_p): Delete.
5454 (nds32_long_call_p): Modify.
5455 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5456 * config/nds32/nds32-protos.h
5457 (symbolic_reference_mentioned_p): Declare.
5458 (nds32_legitimize_ict_address): Declare.
5459 (nds32_expand_ict_move): Declare.
5460 (nds32_indirect_call_referenced_p): Declare.
5461 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5462 (nds32_relax_group): Use nds32_ict_const_p as condition.
5463 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5464 (nds32_asm_file_start): Output ict_model directive in asm code.
5465 (nds32_legitimate_address_p): Consider indirect call.
5466 (nds32_print_operand): Consider indirect call.
5467 (nds32_print_operand_address): Consider indirect call.
5468 (nds32_insert_attributes): Handle "indirect_call" attribute.
5469 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5470 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5471 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5472 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5473 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5474 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5475 (TARGET_ICT_MODEL_SMALL): Define.
5476 (TARGET_ICT_MODEL_LARGE): Define.
5477 * config/nds32/nds32.md (movsi): Consider ict model.
5478 (call, call_value): Consider ict model.
5479 (sibcall, sibcall_value): Consider ict model.
5480 * config/nds32/nds32.opt (mict-model): New option.
5481 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5484 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5485 Monk Chiang <sh.chiang04@gmail.com>
5486 Jim Wilson <jimw@sifive.com>
5488 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5489 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5490 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5491 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5492 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5493 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5494 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5495 compute save_libcall_adjustment properly.
5496 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5497 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5498 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5499 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5500 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5501 (ABI_SPEC): Handle mabi=ilp32e.
5502 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5503 (RVE): Add RVE mask.
5504 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5505 <-march>: Add rv32e as an example.
5507 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5510 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5511 (intra_create_variable_infos): Handle C++ constructors.
5513 2018-05-18 Martin Liska <mliska@suse.cz>
5515 * passes.def: Remove a redundant pass.
5517 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5520 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5522 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5524 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5526 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5528 (armv2, armv2a, armv3, armv3m): Delete architectures.
5529 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5530 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5531 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5533 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5534 (*mulsidi3adddi): Likewise.
5535 (mulsidi3): Likewise.
5536 (*mulsidi3_nov6): Likewise.
5537 (umulsidi3): Likewise.
5538 (umulsidi3_nov6): Likewise.
5539 (umaddsidi4): Likewise.
5540 (*umulsidi3adddi): Likewise.
5541 (smulsi3_highpart): Likewise.
5542 (*smulsi3_highpart_nov6): Likewise.
5543 (umulsi3_highpart): Likewise.
5544 (*umulsi3_highpart_nov6): Likewise.
5545 * config/arm/arm.h (arm_arch3m): Delete.
5546 * config/arm/arm.c (arm_arch3m): Delete.
5547 (arm_option_override_internal): Update armv3-related comment.
5548 (arm_configure_build_target): Delete use of isa_bit_mode32.
5549 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5550 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5551 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5553 (mulusa3): Likewise.
5554 * config/arm/arm-protos.h (arm_arch3m): Delete.
5555 * config/arm/arm-tables.opt: Regenerate.
5556 * config/arm/arm-tune.md: Likewise.
5557 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5558 deleted architectures.
5560 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5562 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5563 (armv5t, armv5te): New features.
5564 (ARMv5, ARMv5e): Delete fgroups.
5565 (ARMv5t, ARMv5te): Adjust for above changes.
5567 (armv5, armv5e): Delete arches.
5568 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5570 (*call_reg_arm): Likewise.
5571 (*call_value_reg_armv5): Likewise.
5572 (*call_value_reg_arm): Likewise.
5573 (*call_symbol): Likewise.
5574 (*call_value_symbol): Likewise.
5575 (*sibcall_insn): Likewise.
5576 (*sibcall_value_insn): Likewise.
5578 (prefetch): Likewise.
5579 (define_split and define_peephole2 dependent on arm_arch5):
5581 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5583 (TARGET_ARM_QBIT): Likewise.
5584 (TARGET_DSP_MULTIPLY): Likewise.
5585 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5586 (arm_arch5, arm_arch5e): Delete.
5587 (arm_arch5t, arm_arch5te): Declare.
5588 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5589 (arm_arch5t): Declare.
5590 (arm_option_reconfigure_globals): Update for the above.
5591 (arm_options_perform_arch_sanity_checks): Update comment, replace
5592 use of arm_arch5 with arm_arch5t.
5593 (use_return_insn): Likewise.
5594 (arm_emit_call_insn): Likewise.
5595 (output_return_instruction): Likewise.
5596 (arm_final_prescan_insn): Likewise.
5597 (arm_coproc_builtin_available): Likewise.
5598 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5599 arm_arch5e with arm_arch5t and arm_arch5te.
5600 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5601 (arm_arch5t, arm_arch5te): Declare.
5602 * config/arm/arm-tables.opt: Regenerate.
5603 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5604 * config/arm/t-multilib: Likewise.
5605 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5606 instead of arm_arch5.
5607 (*call_reg_thumb1): Likewise.
5608 (*call_value_reg_thumb1_v5): Likewise.
5609 (*call_value_reg_thumb1): Likewise.
5610 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5612 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5614 2018-05-18 Martin Liska <mliska@suse.cz>
5616 PR gcov-profile/84846
5617 * doc/gcov.texi: Document -t option of gcov tool.
5619 2018-05-18 Martin Liska <mliska@suse.cz>
5621 PR gcov-profile/84846
5622 * gcov.c (print_usage): Add new -t option.
5623 (process_args): Handle the option.
5624 (generate_results): Use stdout as output when requested by
5627 2018-05-18 Martin Liska <mliska@suse.cz>
5629 PR gcov-profile/84846
5630 * coverage.c (coverage_init): Write PWD to .gcno file.
5631 * doc/gcov.texi: Document how working directory is printed.
5632 * gcov-dump.c (dump_gcov_file): Print PWD.
5633 * gcov.c (output_intermediate_file): Likewise.
5634 (read_graph_file): Read PWD string.
5635 (output_lines): Print PWD.
5637 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5640 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5641 for retval and return false if all args to phi are zero.
5643 2018-05-18 Richard Biener <rguenther@suse.de>
5645 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5647 (evrp_dom_walker::before_dom_children): Call it.
5649 2018-05-18 Richard Biener <rguenther@suse.de>
5651 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5652 results when processing array refs with variable index.
5654 2018-05-18 Toon Moene <toon@moene.org>
5656 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5657 directly after that of -floop-interchange. Indicate that both
5658 options are enabled by default when specifying -O3.
5660 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5662 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5663 iterator. Delete separate integer-mode vec_set<mode> expander.
5664 (aarch64_simd_vec_setv2di): Delete.
5665 (vec_setv2di): Delete.
5666 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5667 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5668 the "w, r" alternative.
5670 2018-05-18 Martin Liska <mliska@suse.cz>
5672 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5673 * tree-pass.h (make_pass_lower_switch_O0): New function.
5674 * tree-switch-conversion.c (node_has_low_bound): Remove.
5675 (node_has_high_bound): Likewise.
5676 (node_is_bounded): Likewise.
5677 (class pass_lower_switch): Make it a template type and create
5679 (pass_lower_switch::execute): Add template argument.
5680 (make_pass_lower_switch): New function.
5681 (make_pass_lower_switch_O0): New function.
5682 (do_jump_if_equal): Remove.
5683 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5684 all the hard work to tree optimization passes.
5686 2018-05-18 Martin Liska <mliska@suse.cz>
5688 * dbgcnt.c (limit_low): Renamed from limit.
5689 (limit_high): New variable.
5690 (dbg_cnt_is_enabled): Check for upper limit.
5691 (dbg_cnt): Adjust dumping.
5692 (dbg_cnt_set_limit_by_index): Add new argument for high
5694 (dbg_cnt_set_limit_by_name): Likewise.
5695 (dbg_cnt_process_single_pair): Parse new format.
5696 (dbg_cnt_process_opt): Use strtok.
5697 (dbg_cnt_list_all_counters): Remove 'value' and add
5699 * doc/invoke.texi: Document changes.
5701 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5703 * doc/sourcebuild.texi (scalar_all_fma): Document.
5704 * tree.def (FMA_EXPR): Delete.
5705 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5706 * internal-fn.c (ternary_direct): New macro.
5707 (expand_ternary_optab_fn): Likewise.
5708 (direct_ternary_optab_supported_p): Likewise.
5709 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5710 * builtins.c (fold_builtin_fma): Delete.
5711 (fold_builtin_3): Don't call it.
5712 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5713 * expr.c (expand_expr_real_2): Likewise.
5714 * fold-const.c (operand_equal_p): Likewise.
5715 (fold_ternary_loc): Likewise.
5716 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5717 * gimple.c (DEFTREECODE): Likewise.
5718 * gimplify.c (gimplify_expr): Likewise.
5719 * optabs-tree.c (optab_for_tree_code): Likewise.
5720 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5721 * tree-eh.c (operation_could_trap_p): Likewise.
5722 (stmt_could_throw_1_p): Likewise.
5723 * tree-inline.c (estimate_operator_cost): Likewise.
5724 * tree-pretty-print.c (dump_generic_node): Likewise.
5725 (op_code_prio): Likewise.
5726 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5727 * tree-ssa-operands.c (get_expr_operands): Likewise.
5728 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5729 * fold-const-call.h (fold_fma): Delete.
5730 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5731 CFN_FNMA and CFN_FNMS.
5733 * genmatch.c (combined_fn): New enum.
5734 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5735 (commutative_op): New function.
5736 (commutate): Use it. Handle more than 2 operands.
5737 (dt_operand::gen_gimple_expr): Use commutative_op.
5738 (parser::parse_expr): Allow :c to be used with non-binary
5739 operators if the commutative operand is known.
5740 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5741 CFN_FMS, CFN_FNMA and CFN_FNMS.
5742 (backprop::process_assign_use): Remove FMA_EXPR handling.
5743 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5744 (gen_hsa_fma): New function.
5745 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5746 IFN_FNMA and IFN_FNMS.
5747 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5748 * gimple-fold.h (follow_all_ssa_edges): Declare.
5749 * gimple-fold.c (follow_all_ssa_edges): New function.
5750 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5751 gimple_build interface and use follow_all_ssa_edges to fold the result.
5752 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5753 instead of checking for optabs directly.
5754 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5755 rather than FMA_EXPRs.
5756 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5757 call to IFN_FMA instead of an FMA_EXPR.
5759 2018-05-17 Jim Wilson <jimw@sifive.com>
5761 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5762 sign extended promoted subreg, and the range does not have the sign bit
5763 set, then do a sign extend.
5765 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5766 test, check for sign extended subreg and/or constant operands, and
5767 do a sign extend in that case.
5769 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5771 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5772 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5774 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5775 Change logics_shift_reg to logics_shift_imm.
5776 (thunderx2t99_fp_loadpair_basic): Delete.
5777 (thunderx2t99_fp_storepair_basic): Delete.
5778 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5779 (thunderx2t99_asimd_polynomial): Delete.
5780 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5781 and neon_fp_mul_d_scalar_q.
5782 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5783 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5784 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5785 (thunderx2t99_asimd_lut): Add missing tbl types.
5786 (thunderx2t99_asimd_ext): Delete.
5787 (thunderx2t99_asimd_load1_1_mult): Delete.
5788 (thunderx2t99_asimd_load1_2_mult): Delete.
5789 (thunderx2t99_asimd_load1_ldp): New.
5790 (thunderx2t99_asimd_load1): New.
5791 (thunderx2t99_asimd_load2): Add missing *load2* types.
5792 (thunderx2t99_asimd_load3): New.
5793 (thunderx2t99_asimd_load4): New.
5794 (thunderx2t99_asimd_store1_1_mult): Delete.
5795 (thunderx2t99_asimd_store1_2_mult): Delete.
5796 (thunderx2t99_asimd_store2_mult): Delete.
5797 (thunderx2t99_asimd_store2_onelane): Delete.
5798 (thunderx2t99_asimd_store_stp): New.
5799 (thunderx2t99_asimd_store1): New.
5800 (thunderx2t99_asimd_store2): New.
5801 (thunderx2t99_asimd_store3): New.
5802 (thunderx2t99_asimd_store4): New.
5804 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5806 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5807 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5809 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5810 Segher Boessenkool <segher@kernel.crashing.org>
5813 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5816 2018-05-17 Richard Biener <rguenther@suse.de>
5818 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5819 for pruning loop and prune defs feeding only already visited PHIs.
5821 2018-05-17 Richard Biener <rguenther@suse.de>
5823 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5825 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5826 Richard Biener <rguenther@suse.de>
5828 PR tree-optimization/85793
5829 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5830 for VMAT_ELEMENTWISE.
5832 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5834 * internal-fn.h (lookup_internal_fn): Declare
5835 * internal-fn.c (lookup_internal_fn): New function.
5836 * gimple.c (gimple_build_call_from_tree): Handle calls to
5838 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5839 internal function names.
5840 * tree-pretty-print.c (dump_generic_node): Likewise.
5841 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5843 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5845 * gimple-fold.h (gimple_build): Make the function forms take
5846 combined_fn rather than built_in_function.
5847 (gimple_simplify): Likewise.
5848 * gimple-match-head.c (gimple_simplify): Likewise.
5849 * gimple-fold.c (gimple_build): Likewise.
5850 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5851 rather than gimple_build_call_internal.
5852 (get_initial_defs_for_reduction): Likewise.
5853 (vect_create_epilog_for_reduction): Likewise.
5854 (vectorizable_live_operation): Likewise.
5856 2018-05-17 Martin Liska <mliska@suse.cz>
5858 * gimple-ssa-sprintf.c (format_directive): Do not use
5859 space in between 'G_' and '('.
5861 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5864 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5865 even if the mask is not all ones.
5868 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5870 (ix86_gimple_fold_builtin): Likewise.
5873 * config/i386/i386.c: Include tree-vector-builder.h.
5874 (ix86_vector_shift_count): New function.
5875 (ix86_fold_builtin): Fold shift builtins by scalar count.
5876 (ix86_gimple_fold_builtin): Likewise.
5878 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5879 _mm512_setzero): New intrinsics.
5881 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5882 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5884 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5885 code generation for cases where splatting a value is not useful.
5886 * simplify-rtx.c (simplify_ternary_operation): Simplify
5887 vec_merge across a vec_duplicate and a paradoxical subreg forming
5888 a vector mode to a vec_concat.
5890 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5892 * config.gcc: Support "goldmont-plus".
5893 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5895 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5896 PROCESSOR_GOLDMONT_PLUS.
5897 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5898 (processor_target_table): Add "goldmont-plus".
5899 (PTA_GOLDMONT_PLUS): Define.
5900 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5901 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5902 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5903 (fold_builtin_cpu): Add "goldmont-plus".
5904 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5905 (ix86_option_override_internal): Add "goldmont-plus".
5906 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5907 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5908 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5909 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5911 2018-05-17 Richard Biener <rguenther@suse.de>
5913 PR tree-optimization/85757
5914 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5915 remove defs that only feed that PHI from further processing.
5917 2018-05-16 Jim Wilson <jimw@sifive.com>
5919 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5921 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5923 2018-05-16 Mark Wielaard <mark@klomp.org>
5925 * dwarf2out.c (count_index_strings): New function.
5926 (output_indirect_strings): Call count_index_strings and generate
5927 header for dwarf_version >= 5.
5929 2018-05-16 Mark Wielaard <mark@klomp.org>
5931 * dwarf2out.c (dwarf_FORM): New function.
5932 (set_indirect_string): Use dwarf_FORM.
5933 (reset_indirect_string): Likewise.
5934 (size_of_die): Likewise.
5935 (value_format): Likewise.
5936 (output_die): Likewise.
5937 (add_skeleton_AT_string): Likewise.
5938 (output_macinfo_op): Likewise.
5939 (index_string): Likewise.
5940 (output_index_string_offset): Likewise.
5941 (output_index_string): Likewise.
5942 (count_index_strings): Likewise.
5944 2018-05-16 Carl Love <cel@us.ibm.com>
5946 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5947 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5949 2018-05-16 Martin Jambor <mjambor@suse.cz>
5951 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5952 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5954 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5956 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5957 (fnma<mode>4): Likewise.
5958 (fms<mode>4): Likewise.
5959 (fnms<mode>4): Likewise.
5960 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5961 (aarch64_fnma<mode>4): Likewise.
5962 (aarch64_fms<mode>4): Likewise.
5963 (aarch64_fnms<mode>4): Likewise.
5964 (aarch64_fnmadd<mode>4): Likewise.
5966 2018-05-16 Jason Merrill <jason@redhat.com>
5968 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5970 2018-05-16 Richard Biener <rguenther@suse.de>
5972 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5973 (dump_stmt_cost): Declare.
5974 (add_stmt_cost): Dump cost we add.
5975 (add_stmt_costs): New function.
5976 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5978 (vect_analyze_stmt): Adjust prototype.
5979 (vectorizable_condition): Likewise.
5980 (vectorizable_live_operation): Likewise.
5981 (vectorizable_reduction): Likewise.
5982 (vectorizable_induction): Likewise.
5983 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5984 cost vector to pass to vectorizable_ and record afterwards.
5985 (vect_model_reduction_cost): Take cost vector argument and adjust.
5986 (vect_model_induction_cost): Likewise.
5987 (vectorizable_reduction): Likewise.
5988 (vectorizable_induction): Likewise.
5989 (vectorizable_live_operation): Likewise.
5990 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5991 SLP_TREE_NUMBER_OF_VEC_STMTS.
5992 (vect_analyze_slp_cost_1): Remove.
5993 (vect_analyze_slp_cost): Likewise.
5994 (vect_slp_analyze_node_operations): Take visited args and
5995 a target cost vector. Avoid processing already visited stmt sets.
5996 (vect_slp_analyze_operations): Use a local cost vector to gather
5997 costs and register those of non-discarded instances.
5998 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5999 (vect_schedule_slp_instance): Remove copying of
6000 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6002 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6003 adding cost. Record cost entry location.
6004 (vect_prologue_cost_for_slp_op): Function to compute cost of
6005 a constant or invariant generated for SLP vect in the prologue,
6006 split out from vect_analyze_slp_cost_1.
6007 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6008 (vect_model_promotion_demotion_cost): Likewise.
6009 (vect_model_store_cost): Likewise, make static.
6010 (vect_model_load_cost): Likewise.
6011 (vectorizable_bswap): Add cost vector arg and adjust.
6012 (vectorizable_call): Likewise.
6013 (vectorizable_simd_clone_call): Likewise.
6014 (vectorizable_conversion): Likewise.
6015 (vectorizable_assignment): Likewise.
6016 (vectorizable_shift): Likewise.
6017 (vectorizable_operation): Likewise.
6018 (vectorizable_store): Likewise.
6019 (vectorizable_load): Likewise.
6020 (vectorizable_condition): Likewise.
6021 (vectorizable_comparison): Likewise.
6022 (can_vectorize_live_stmts): Likewise.
6023 (vect_analyze_stmt): Likewise.
6024 (vect_transform_stmt): Adjust calls to vectorizable_*.
6025 * tree-vectorizer.c: Include gimple-pretty-print.h.
6026 (dump_stmt_cost): New function.
6028 2018-05-16 Richard Biener <rguenther@suse.de>
6030 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6031 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6032 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6033 (check_name): New callback.
6034 (dse_classify_store): Track cycles via a visited bitmap of PHI
6035 defs and simplify handling of in-loop and across loop dead stores
6036 and properly fail for loop-variant refs. Handle byte-tracking with
6037 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6040 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6042 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6043 (vect_get_mask_type_for_stmt): Likewise.
6044 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6046 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6047 to determine the statement's vector type and the vector type that
6048 should be used for calculating nunits. Deal with cases in which
6049 the type has to be deferred.
6050 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6051 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6052 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6053 (vect_determine_vf_for_stmt): New functions, split out from...
6054 (vect_determine_vectorization_factor): ...here.
6055 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6056 (vect_get_mask_type_for_stmt): New functions, split out from
6057 vect_determine_vectorization_factor.
6059 2018-05-16 Richard Biener <rguenther@suse.de>
6061 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6062 verify the [VEC_]COND_EXPR embedded comparison.
6064 2018-05-15 Martin Sebor <msebor@redhat.com>
6066 PR tree-optimization/85753
6067 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6068 RECORD_TYPE in addition to ARRAY_TYPE.
6070 2018-05-15 Martin Sebor <msebor@redhat.com>
6073 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6075 2018-05-15 Richard Biener <rguenther@suse.de>
6077 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6078 add by_clobber_p one. Change algorithm to collect all defs
6079 representing uses we need to walk and try reducing them to
6080 a single one before failing.
6081 (dse_dom_walker::dse_optimize_stmt): Adjust.
6083 2018-05-13 Mark Wielaard <mark@klomp.org>
6085 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6086 (size_of_loc_descr): Likewise.
6087 (output_loc_operands): Likewise.
6088 (output_loc_operands_raw): Likewise.
6089 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6090 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6091 (hash_loc_operands): Likewise.
6092 (compare_loc_operands): Likewise.
6094 2018-05-14 Mark Wielaard <mark@klomp.org>
6096 * dwarf2out.c (count_index_addrs): New function.
6097 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6099 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6101 PR tree-optimization/83648
6102 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6103 return value as malloc candidate.
6105 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6108 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6109 param as true in call to suggest_attribute.
6111 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6113 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6114 -mreadonly-in-sdata.
6116 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6118 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6120 (aarch64_crypto_aesd_fused): Likewise.
6122 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6124 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6125 (movsi_aarch64): Likewise.
6126 (load_pairsi): Likewise.
6127 (load_pairdi): Likewise.
6128 (store_pairsi): Likewise.
6129 (store_pairdi): Likewise.
6130 (load_pairsf): Likewise.
6131 (load_pairdf): Likewise.
6132 (store_pairsf): Likewise.
6133 (store_pairdf): Likewise.
6134 (zero_extend): Likewise.
6135 (trunc): Swap alternatives.
6136 (fcvt_target): Add '?' to prefer w over r.
6138 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6141 * config/i386/i386.md: Disallow non-commutative arithmetics in
6142 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6143 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6144 in the peephole2 before it.
6146 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6148 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6149 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6150 (ix86_handle_option): Handle -mcldemote.
6151 * config.gcc: New header.
6152 * config/i386/cldemoteintrin.h: New file.
6153 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6154 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6156 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6157 OPTION_MASK_ISA_CLDEMOTE.
6158 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6159 (ix86_valid_target_attribute_inner_p): Ditto.
6160 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6161 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6162 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6163 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6164 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6166 * config/i386/i386.opt: Add -mcldemote.
6167 * config/i386/x86intrin.h: New header.
6168 * doc/invoke.texi: Add -mcldemote.
6170 2018-05-14 Richard Biener <rguenther@suse.de>
6172 * doc/match-and-simplify.texi: Adjust :s documentation.
6174 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6176 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6177 intended memcpy size.
6178 (REORDER_45): Likewise.
6180 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6182 * sort.cc: New file.
6183 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6184 * vec.c (qsort_chk): Use gcc_qsort.
6185 * Makefile.in (OBJS-libcommon): Add sort.o.
6186 (build/sort.o): New target. Use it...
6187 (BUILD_RTL): ... here, and...
6188 (build/gencfn-macros): ... here, and...
6189 (build/genmatch): ... here.
6191 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6192 Chung-Ju Wu <jasonwucj@gmail.com>
6194 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6195 * config/nds32/nds32-graywolf.md: New file.
6196 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6197 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6199 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6200 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6201 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6202 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6203 * config/nds32/pipelines.md: Include n15 settings.
6205 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6206 Chung-Ju Wu <jasonwucj@gmail.com>
6208 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6209 * config/nds32/nds32-n13.md: New file.
6210 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6211 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6213 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6214 * config/nds32/nds32.md (pipeline_model): Add n13.
6215 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6216 * config/nds32/pipelines.md: Include n13 settings.
6218 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6219 Chung-Ju Wu <jasonwucj@gmail.com>
6221 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6222 * config/nds32/nds32-n10.md: New file.
6223 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6224 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6226 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6227 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6228 * config/nds32/nds32.md (pipeline_model): Add n10.
6229 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6230 * config/nds32/pipelines.md: Include n10 settings.
6232 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6233 Kito Cheng <kito.cheng@gmail.com>
6234 Chung-Ju Wu <jasonwucj@gmail.com>
6236 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6237 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6238 Add enum values for DSP extension instructions.
6239 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6241 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6242 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6244 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6245 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6246 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6247 * config/nds32/nds32-intrinsic.md: Likewise.
6248 * config/nds32/nds32_intrinsic.h: Likewise.
6249 * config/nds32/nds32-md-auxiliary.c: Likewise.
6250 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6251 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6252 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6253 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6254 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6255 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6256 TYPE_DMAC in switch statement.
6257 * config/nds32/nds32.c: New checking and implementation for DSP
6258 extension instructions.
6259 * config/nds32/nds32.h: Likewise.
6260 * config/nds32/nds32.md: Likewise.
6261 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6262 * config/nds32/predicates.md: Implement new predicates for DSP
6265 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6267 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6268 Reformat alternatives and attributes so it is easier to identify
6269 which constraints/attributes go with which instruction.
6270 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6271 (mov<mode>_softfloat32, FMOVE64): Likewise.
6272 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6273 (mov<mode>_softfloat64, FMOVE64): Likewise.
6275 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6277 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6279 (Basic PowerPC Built-in Functions): The new name of the
6280 subsection previously known as "PowerPC Built-in Functions".
6281 (Basic PowerPC Built-in Functions Available on all Configurations):
6283 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6284 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6285 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6286 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6288 2018-05-11 Martin Jambor <mjambor@suse.cz>
6291 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6294 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6297 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6299 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6301 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6302 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6303 (ix86_handle_option): Handle -mwaitpkg.
6304 * config.gcc: New header.
6305 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6306 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6307 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6309 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6310 OPTION_MASK_ISA_WAITPKG.
6311 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6312 (ix86_option_override_internal): Add PTA_WAITPKG.
6313 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6314 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6315 IX86_BUILTIN_TPAUSE.
6316 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6317 __builtin_ia32_umwait and __builtin_ia32_tpause.
6318 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6319 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6320 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6321 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6322 UNSPECV_TPAUSE): New.
6323 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6324 * config/i386/i386.opt: Add -mwaitpkg.
6325 * config/i386/waitpkgintrin.h: New file.
6326 * config/i386/x86intrin.h: New header.
6327 * doc/invoke.texi: Add -mwaitpkg.
6329 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6332 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6334 (cortex-m0): Use armv6s-m isa.
6335 (cortex-m0plus): Likewise.
6336 (cortex-m1): Likewise.
6337 (cortex-m0.small-multiply): Likewise.
6338 (cortex-m0plus.small-multiply): Likewise.
6339 (cortex-m1.small-multiply): Likewise.
6341 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6342 Jakub Jelinek <jakub@redhat.com>
6344 PR tree-optimization/85692
6345 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6346 source permute as well.
6348 2018-05-11 Martin Liska <mliska@suse.cz>
6351 * doc/extend.texi: Document LLVM style format for no_sanitize
6354 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6356 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6357 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6358 (mode_supports_vsx_dform_quad): Likewise.
6359 (mode_supports_vmx_dform): Move these functions to be next to the
6360 other mode_supports functions.
6361 (mode_supports_dq_form): Likewise.
6362 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6363 mode_supports_dq_form.
6364 (reg_offset_addressing_ok_p): Likewise.
6365 (offsettable_ok_by_alignment): Likewise.
6366 (rs6000_legitimate_offset_address_p): Likewise.
6367 (legitimate_lo_sum_address_p): Likewise.
6368 (rs6000_legitimize_address): Likewise.
6369 (rs6000_legitimize_reload_address): Likewise.
6370 (rs6000_secondary_reload_inner): Likewise.
6371 (rs6000_preferred_reload_class): Likewise.
6372 (rs6000_output_move_128bit): Likewise.
6374 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6376 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6377 Generate SImode target register for null target.
6378 <case IX86_BUILTIN_XGETBV>: Ditto.
6379 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6380 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6382 2018-05-10 Carl Love <cel@us.ibm.com>
6384 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6385 dcbtt and dcbtstt if operands[2] is 0.
6387 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6390 * config/i386/sse.md (usadv64qi): New expander.
6392 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6394 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6395 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6396 -maltivec=be support.
6397 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6398 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6399 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6400 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6401 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6402 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6403 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6404 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6405 altivec_vsumsws): Adjust.
6406 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6407 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6408 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6410 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6411 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6412 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6413 (altivec_lve<VI_char>x): Delete expand.
6414 (*altivec_lve<VI_char>x_internal): Rename to...
6415 (altivec_lve<VI_char>x): ... this.
6416 (altivec_lvxl_<mode>): Delete expand.
6417 (*altivec_lvxl_<mode>_internal): Rename to ...
6418 (altivec_lvxl_<mode>): ... this.
6419 (altivec_stvxl_<mode>): Delete expand.
6420 (*altivec_stvxl_<mode>_internal): Rename to ...
6421 (altivec_stvxl_<mode>): ... this.
6422 (altivec_stve<VI_char>x): Delete expand.
6423 (*altivec_stve<VI_char>x_internal): Rename to ...
6424 (altivec_stve<VI_char>x): ... this.
6425 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6426 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6427 reduc_plus_scal_<mode>): Adjust.
6428 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6430 (rs6000_cpu_cpp_builtins): Adjust.
6431 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6432 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6433 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6434 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6435 -maltivec=be support.
6436 (rs6000_split_vec_extract_var): Adjust.
6437 (rs6000_split_v4si_init): Adjust.
6438 (swap_selector_for_mode): Delete.
6439 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6440 altivec_expand_stvex_be): Delete.
6441 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6442 -maltivec=be support.
6443 (rs6000_gimple_fold_builtin): Ditto.
6444 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6446 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6447 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6448 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6449 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6450 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6451 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6452 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6453 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6454 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6455 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6456 anonymous split): Adjust.
6457 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6458 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6460 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6462 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6463 when --with-gxx-include-dir is also specified.
6464 * configure: Regenerate.
6466 2018-05-09 Jim Wilson <jimw@sifive.com>
6469 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6470 * config/riscv/t-withmultilib: New.
6471 * config/riscv/withmultilib.h: New.
6472 * doc/install.texi: Document RISC-V --with-multilib-list support.
6474 2018-05-09 Richard Biener <rguenther@suse.de>
6476 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6478 (vect_bb_vectorization_profitable_p): Adjust. Compute
6479 actual scalar cost using the cost vector and the add_stmt_cost
6482 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6484 PR rtl-optimization/85645
6485 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6486 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6488 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6490 PR rtl-optimization/85645
6491 * shrink-wrap.c (spread_components): Return a boolean saying if
6492 anything was changed.
6493 (try_shrink_wrapping_separate): Iterate spread_components until
6494 nothing changes anymore.
6496 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6498 PR rtl-optimization/85645
6499 * regrename.c (build_def_use): Also kill the chains that include the
6500 destination of a REG_CFA_REGISTER note.
6502 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6504 PR rtl-optimization/85645
6505 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6506 insn that has a REG_CFA_REGISTER note.
6508 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6510 * cfgexpand.c (expand_clobber): New function.
6511 (expand_gimple_stmt_1): Use it.
6512 * tree-vect-stmts.c (vect_clobber_variable): New function,
6514 (vectorizable_simd_clone_call): ...here.
6515 (vectorizable_store): Emit a clobber either side of an
6516 IFN_STORE_LANES sequence.
6517 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6519 2018-05-09 Tom de Vries <tom@codesourcery.com>
6522 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6523 (define_insn "trap_if_false"): Add exit after trap.
6525 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6527 PR rtl-optimization/85638
6528 * bb-reorder.c: Include common/common-target.h.
6529 (create_forwarder_block): New function extracted from...
6530 (fix_up_crossing_landing_pad): ...here. Rename into...
6531 (dw2_fix_up_crossing_landing_pad): ...this.
6532 (sjlj_fix_up_crossing_landing_pad): New function.
6533 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6534 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6535 from both partitions and exit the loop after one iteration.
6537 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6540 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6542 (Basic PowerPC Built-in Functions): The new name of the
6543 subsection previously known as "PowerPC Built-in Functions".
6544 (Basic PowerPC Built-in Functions Available on all Configurations):
6546 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6548 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6550 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6552 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6555 2018-05-08 Jim Wilson <jimw@sifive.com>
6557 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6558 (LD_EMUL_SUFFIX): New.
6559 (LINK_SPEC): Use it.
6561 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6563 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6565 (Basic PowerPC Built-in Functions): The new name of the
6566 subsection previously known as "PowerPC Built-in Functions".
6567 (Basic PowerPC Built-in Functions Available on all Configurations):
6569 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6571 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6573 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6575 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6578 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6581 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6582 after cmpelim optimization.
6584 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6586 * config.gcc: Support "goldmont".
6587 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6588 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6590 * config/i386/i386.c (m_GOLDMONT): Define.
6591 (processor_target_table): Add "goldmont".
6592 (PTA_GOLDMONT): Define.
6593 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6594 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6595 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6596 (fold_builtin_cpu): Add "goldmont".
6597 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6598 (ix86_option_override_internal): Add "goldmont".
6599 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6600 (processor_type): Add PROCESSOR_GOLDMONT.
6601 * config/i386/i386.md: Add CPU "glm".
6602 * config/i386/glm.md: New file.
6603 * config/i386/x86-tune.def: Add m_GOLDMONT.
6604 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6606 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6609 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6611 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6612 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6613 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6616 * config/i386/i386.c (ix86_fold_builtin): Handle
6617 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6620 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6621 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6623 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6626 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6627 (check_arch): Likewise.
6628 (check_fpu): Return the result rather than printing it.
6629 (end arch): Fix operator precedence.
6630 (end cpu): Likewise.
6631 (END): Print the result from check_fpu.
6633 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6634 Alan Hayward <alan.hayward@arm.com>
6635 David Sherwood <david.sherwood@arm.com>
6637 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6638 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6639 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6640 (*fcmuo<mode>_and): New patterns.
6642 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6644 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6645 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6646 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6647 (cmp_op, sve_imm_con): New code attributes.
6648 (SVE_COND_INT_CMP, imm_con): Delete.
6649 (cmp_op): Remove above unspecs from int attribute.
6650 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6652 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6653 comparison-specific unspecs.
6654 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6655 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6656 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6657 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6658 (*vec_fcm<cmp_op><mode>): Rename to...
6659 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6660 (*vec_fcmuo<mode>): Rename to...
6661 (*fcmuo<mode>): ...this and adjust likewise.
6662 (*pred_fcm<cmp_op><mode>): New pattern.
6663 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6664 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6666 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6668 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6669 (aarch64_emit_sve_predicated_cond): New function.
6670 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6671 (aarch64_emit_unspec_cond_or): Replace with...
6672 (aarch64_emit_sve_or_conds): ...this new function. Use
6673 aarch64_emit_sve_ptrue_op for the individual comparisons and
6674 aarch64_emit_binop to OR them together.
6675 (aarch64_emit_inverted_unspec_cond): Replace with...
6676 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6677 aarch64_emit_sve_ptrue_op for the comparison and
6678 aarch64_emit_unop to invert the result.
6679 (aarch64_expand_sve_vec_cmp_float): Update after the above
6680 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6682 2018-05-07 Nathan Sidwell <nathan@acm.org>
6684 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6685 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6686 (Backwards Compatibility): Likewise.
6688 2018-05-07 Luis Machado <luis.machado@linaro.org>
6692 2018-05-07 Luis Machado <luis.machado@linaro.org>
6694 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6695 <prefetch_dynamic_strides>: New const bool field.
6696 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6697 prefetch_dynamic_strides.
6698 (exynosm1_prefetch_tune): Likewise.
6699 (thunderxt88_prefetch_tune): Likewise.
6700 (thunderx_prefetch_tune): Likewise.
6701 (thunderx2t99_prefetch_tune): Likewise.
6702 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6704 (aarch64_override_options_internal): Update to set
6705 PARAM_PREFETCH_DYNAMIC_STRIDES.
6706 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6707 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6708 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6709 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6710 prefetch-dynamic-strides setting.
6712 2018-05-07 Luis Machado <luis.machado@linaro.org>
6714 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6715 <minimum_stride>: New const int field.
6716 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6717 minimum_stride field.
6718 (exynosm1_prefetch_tune): Likewise.
6719 (thunderxt88_prefetch_tune): Likewise.
6720 (thunderx_prefetch_tune): Likewise.
6721 (thunderx2t99_prefetch_tune): Likewise.
6722 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6723 (aarch64_override_options_internal): Update to set
6724 PARAM_PREFETCH_MINIMUM_STRIDE.
6725 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6726 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6727 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6728 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6729 stride is constant and is below the minimum stride threshold.
6731 2018-05-07 Luis Machado <luis.machado@linaro.org>
6733 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6736 2018-05-07 Luis Machado <luis.machado@linaro.org>
6738 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6739 <prefetch_dynamic_strides>: New const bool field.
6740 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6741 prefetch_dynamic_strides.
6742 (exynosm1_prefetch_tune): Likewise.
6743 (thunderxt88_prefetch_tune): Likewise.
6744 (thunderx_prefetch_tune): Likewise.
6745 (thunderx2t99_prefetch_tune): Likewise.
6746 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6748 (aarch64_override_options_internal): Update to set
6749 PARAM_PREFETCH_DYNAMIC_STRIDES.
6750 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6751 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6752 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6753 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6754 prefetch-dynamic-strides setting.
6756 2018-05-07 Luis Machado <luis.machado@linaro.org>
6758 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6759 <minimum_stride>: New const int field.
6760 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6761 minimum_stride field.
6762 (exynosm1_prefetch_tune): Likewise.
6763 (thunderxt88_prefetch_tune): Likewise.
6764 (thunderx_prefetch_tune): Likewise.
6765 (thunderx2t99_prefetch_tune): Likewise.
6766 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6767 (aarch64_override_options_internal): Update to set
6768 PARAM_PREFETCH_MINIMUM_STRIDE.
6769 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6770 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6771 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6772 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6773 stride is constant and is below the minimum stride threshold.
6775 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6778 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6779 the type is addressable. Don't force op into register if it has
6782 2018-05-05 Roland McGrath <mcgrathr@google.com>
6785 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6786 any section for which we don't know a specific type it should have,
6787 regardless of name. Previously this was done only for the exact
6788 names ".init_array", ".fini_array", and ".preinit_array".
6789 (default_elf_asm_named_section): Add comment about
6790 relationship with default_section_type_flags and SECTION_NOTYPE.
6791 (get_section): Don't consider it a type conflict if one side has
6792 SECTION_NOTYPE and the other doesn't, as long as neither has the
6793 SECTION_BSS et al used in the default_section_type_flags logic.
6795 2018-05-05 Tom de Vries <tom@codesourcery.com>
6798 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6799 (workaround_barsyncs): New function.
6800 (nvptx_reorg): Use workaround_barsyncs.
6801 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6802 (define_expand "nvptx_membar_cta"): New define_expand.
6803 (define_insn "*nvptx_membar_cta"): New insn.
6805 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6807 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6808 To improve optimization opportunities.
6809 * builtin-types.def: The new needed builtin types for the above.
6811 2018-05-04 Richard Biener <rguenther@suse.de>
6813 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6814 * gimple-ssa-store-merging.c
6815 (imm_store_chain_info::output_merged_store): Remove redundant create,
6816 release split_store vector contents on failure.
6817 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6818 scalar stmt vector on cache hit.
6820 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6822 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6824 * config.gcc (powerpc-xilinx-eabi*): Remove.
6825 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6827 (fusion_addis_mem_combo_load): Ditto.
6828 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6830 (rs6000_cpu_cpp_builtins): Ditto.
6831 * config/rs6000/rs6000-linux.c
6832 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6833 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6834 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6836 (rs6000_setup_reg_addr_masks): Ditto.
6837 (rs6000_init_hard_regno_mode_ok): Ditto.
6838 (rs6000_option_override_internal): Ditto.
6839 (legitimate_lo_sum_address_p): Ditto.
6840 (rs6000_legitimize_address): Ditto.
6841 (rs6000_legitimize_reload_address): Ditto.
6842 (rs6000_legitimate_address_p): Ditto.
6843 (abi_v4_pass_in_fpr): Ditto.
6844 (setup_incoming_varargs): Ditto.
6845 (rs6000_gimplify_va_arg): Ditto.
6846 (rs6000_split_multireg_move): Ditto.
6847 (rs6000_savres_strategy): Ditto.
6848 (rs6000_emit_prologue_components): Ditto.
6849 (rs6000_emit_epilogue_components): Ditto.
6850 (rs6000_emit_prologue): Ditto.
6851 (rs6000_emit_epilogue): Ditto.
6852 (rs6000_elf_file_end): Ditto.
6853 (rs6000_function_value): Ditto.
6854 (rs6000_libcall_value): Ditto.
6855 * config/rs6000/rs6000.h: Ditto.
6856 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6857 (TARGET_MINMAX): ... this. New.
6858 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6859 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6860 (*movsi_internal1_single): Delete.
6861 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6862 mfpu=, mxilinx-fpu): Delete.
6863 * config/rs6000/singlefp.h: Delete.
6864 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6865 * config/rs6000/t-rs6000: Ditto.
6866 * config/rs6000/t-xilinx: Delete.
6867 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6868 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6869 (VStype_simple): Delete.
6870 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6871 * config/rs6000/xfpu.h: Delete.
6872 * config/rs6000/xfpu.md: Delete.
6873 * config/rs6000/xilinx.h: Delete.
6874 * config/rs6000/xilinx.opt: Delete.
6875 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6876 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6878 2018-05-04 Tom de Vries <tom@codesourcery.com>
6881 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6884 2018-05-04 Richard Biener <rguenther@suse.de>
6887 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6888 (expand_complex_div_wide): Likewise.
6889 (expand_complex_operations_1): Likewise.
6890 (expand_complex_libcall): Preserve EH info of the original stmt.
6891 (tree_lower_complex): Handle removed blocks.
6892 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6893 on complex multiplication and division libcall builtins.
6895 2018-05-04 Richard Biener <rguenther@suse.de>
6898 * fold-const.c (negate_expr_p): Restrict negation of operand
6899 zero of a division to when we know that can happen without
6901 (fold_negate_expr_1): Likewise.
6903 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6906 * real.h (real_nextafter): Declare.
6907 * real.c (real_nextafter): New function.
6908 * fold-const-call.c (fold_const_nextafter): New function.
6909 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6910 CASE_CFN_NEXTTOWARD.
6911 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6912 even when arg1_mode is different from arg0_mode.
6914 2018-05-03 Nathan Sidwell <nathan@acm.org>
6916 * doc/extend.texi (Deprecated Features): Remove
6918 (Backwards Compatibility): Likewise.
6919 * doc/invoke.texi (C++ Language Options): Likewise.
6920 (C++ Dialect Options): Likewise.
6922 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6925 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6926 _mm512_mask_mullox_epi64): New intrinsics.
6928 2018-05-03 Tom de Vries <tom@codesourcery.com>
6931 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6932 dump files): Add offload-tree.
6934 2018-05-03 Richard Biener <rguenther@suse.de>
6936 PR tree-optimization/85615
6937 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6938 to loops not nested in BBs loop father to avoid creating multi-entry
6941 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6943 PR tree-optimization/70291
6944 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6945 arguments. Change return type to tree. Emit libcall as a new
6946 statement rather than replacing existing one when inplace_p is true.
6947 (expand_complex_multiplication_components): New function.
6948 (expand_complex_multiplication): Expand floating-point complex
6949 multiplication using the above.
6950 (expand_complex_division): Rename inner_type parameter to type.
6951 Update expand_complex_libcall call-site.
6952 (expand_complex_operations_1): Update expand_complex_multiplication
6953 and expand_complex_division call-sites.
6955 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6958 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6959 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6960 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6961 the highest significant bit of the shift count mask is clear. In
6962 check whether and[sq]i3 is needed verify that all significant bits
6963 of the shift count other than the highest are set.
6965 2018-05-02 Tom de Vries <tom@codesourcery.com>
6968 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6969 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6970 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6971 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6972 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6973 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6974 __builtin_goacc_parlevel_size.
6976 2018-05-02 Richard Biener <rguenther@suse.de>
6978 PR tree-optimization/85597
6979 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6980 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6983 2018-05-02 Tom de Vries <tom@codesourcery.com>
6986 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6987 dump files): Add ltrans-tree.
6989 2018-05-02 Tom de Vries <tom@codesourcery.com>
6992 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6993 dump files): Add wpa-ipa.
6995 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6997 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6998 powerpc*-*-linux*paired* target.
6999 * config/rs6000/750cl.h: Delete.
7000 * config/rs6000/paired.h: Delete.
7001 * config/rs6000/paired.md: Delete.
7002 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7004 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7005 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7006 comment. Remove paired float support.
7007 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7008 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7010 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7011 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7013 * config/rs6000/rs6000.c: Remove paired float support.
7014 (paired_expand_vector_init, paired_expand_vector_move,
7015 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7016 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7017 paired_expand_builtin, paired_expand_predicate_builtin,
7018 paired_init_builtins): Delete.
7019 * config/rs6000/rs6000.h: Remove paired float support.
7020 * config/rs6000/rs6000.md: Remove paired float support.
7021 (move_from_CR_ov_bit): Delete.
7022 * config/rs6000/rs6000.opt (mpaired): Delete.
7023 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7024 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7026 2018-05-02 Richard Biener <rguenther@suse.de>
7029 * gimplify.c (gimplify_save_expr): When in SSA form allow
7030 SAVE_EXPRs to compute to SSA vars.
7032 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7035 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7036 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7037 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7038 clobber operands[2], instead use a new pseudo. Formatting fixes.
7040 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7042 PR tree-optimization/85586
7043 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7044 exit early for statements in the same group if the accesses are
7047 2018-05-02 Tom de Vries <tom@codesourcery.com>
7050 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7053 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7055 PR tree-optimization/85143
7056 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7058 2018-05-01 Tom de Vries <tom@codesourcery.com>
7061 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7062 not found" error message.
7064 2018-05-01 Tom de Vries <tom@codesourcery.com>
7067 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7068 * vec.c (test_ordered_remove_if): New function.
7069 (vec_c_tests): Call test_ordered_remove_if.
7070 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7071 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7072 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7073 VEC_ORDERED_REMOVE_IF.
7075 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7077 PR tree-optimization/82665
7078 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7079 pointer subtraction where arguments come from a memchr call.
7081 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7083 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7084 --push-state --as-needed and --pop-state instead of --as-needed and
7085 --no-as-needed if ld supports it.
7086 * configure: Regenerated.
7089 * doc/install.texi2html: Replace _002d with - and _002a with * in
7090 generated html files using sed.
7092 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7095 * gcc-rich-location.c (blank_line_before_p): New function.
7096 (use_new_line): New function.
7097 (gcc_rich_location::add_fixit_insert_formatted): New function.
7098 * gcc-rich-location.h
7099 (gcc_rich_location::add_fixit_insert_formatted): New function.
7101 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7103 * selftest.c (assert_streq): Rename "expected" and "actual" to
7104 "val1" and "val2". Extend NULL-handling to cover both inputs
7105 symmetrically, while still requiring both to be non-NULL for a pass.
7106 * selftest.h (assert_streq): Rename "expected" and "actual" to
7108 (ASSERT_EQ): Likewise.
7109 (ASSERT_EQ_AT): Likewise.
7110 (ASSERT_KNOWN_EQ): Likewise.
7111 (ASSERT_KNOWN_EQ_AT): Likewise.
7112 (ASSERT_NE): Likewise.
7113 (ASSERT_MAYBE_NE): Likewise.
7114 (ASSERT_MAYBE_NE_AT): Likewise.
7115 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7116 the assertion to pass.
7117 (ASSERT_STREQ_AT): Likewise.
7119 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7121 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7122 interaction with -pie.
7124 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7126 * selftest.h: Fix alphabetization of per-source-file selftest
7129 2018-04-30 Jason Merrill <jason@redhat.com>
7131 PR c++/61982 - dead stores to destroyed objects.
7132 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7135 2018-04-30 Jason Merrill <jason@redhat.com>
7137 * tree.c (build_clobber): New.
7138 * tree.h: Declare it.
7139 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7141 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7143 * diagnostic-show-locus.c (layout::layout): Update for
7144 location_get_source_line returning a char_span.
7145 (struct char_span): Move to input.h.
7146 (struct correction): Update for fields in char_span becoming
7148 (struct source_line): Update for location_get_source_line
7149 returning a char_span.
7150 (layout::print_line): Likewise.
7151 * edit-context.c (edited_file::print_content): Likewise.
7152 (edited_file::print_diff_hunk): Likewise.
7153 (edited_file::print_run_of_changed_lines): Likewise.
7154 (edited_file::get_num_lines): Likewise.
7155 (edited_line::edited_line): Likewise.
7156 * final.c (asm_show_source): Likewise.
7157 * input.c (location_get_source_line): Convert return type
7158 from const char * to char_span, losing the final "line_len"
7160 (dump_location_info): Update for the above.
7161 (get_substring_ranges_for_loc): Likewise. Use a char_span
7162 when handling the literal within the line.
7163 (test_reading_source_line): Update for location_get_source_line
7164 returning a char_span.
7165 * input.h (class char_span): Move here from
7166 diagnostic-show-locus.c, converting from a struct to a class.
7167 Make data members private.
7168 (char_span::operator bool): New.
7169 (char_span::length): New.
7170 (char_span::get_buffer): New.
7171 (char_span::operator[]): New.
7172 (char_span::subspan): Make const.
7173 (char_span::xstrdup): New.
7174 (location_get_source_line): Convert return type from const char *
7175 to char_span, losing the final "line_size" param.
7177 2018-04-30 Jan Hubicka <jh@suse.cz>
7179 * lto-wrapper.c (ltrans_priorities): New static var.
7180 (cmp_priority): New.
7181 (run_gcc): Read priorities and if doing parallel build order
7182 the Makefile by them.
7184 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7186 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7188 2018-04-30 Richard Biener <rguenther@suse.de>
7190 * tree-cfg.c (verify_address): Remove base argument, add
7191 flag whether to check TREE_ADDRESSABLE and do that.
7192 (verify_expr): Remove.
7193 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7194 (verify_gimple_assign_single): Likewise.
7195 (verify_gimple_switch): Likewise.
7196 (verify_expr_location_1): Dereference tp once. Add (disabled)
7197 piece from verify_expr.
7198 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7200 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7202 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7204 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7206 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7207 (small_data_pattern): Likewise.
7208 (arc_rewrite_small_data): Likewise.
7209 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7210 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7211 (get_symbol_alignment): New function.
7212 (legitimate_small_data_address_p): Likewise.
7213 (legitimate_scaled_address): Update, call
7214 legitimate_small_data_address_p.
7215 (output_sdata): New static variable.
7216 (arc_print_operand): Update how we handle small data operands.
7217 (arc_print_operand_address): Likewise.
7218 (arc_legitimate_address_p): Update, use
7219 legitimate_small_data_address_p.
7220 (arc_rewrite_small_data_p): Remove.
7221 (arc_rewrite_small_data_1): Likewise.
7222 (arc_rewrite_small_data): Likewise.
7223 (small_data_pattern): Likewise.
7224 (compact_sda_memory_operand): Update to use
7225 legitimate_small_data_address_p and get_symbol_alignment.
7226 (prepare_move_operands): Don't rewite sdata pattern.
7227 (prepare_extend_operands): Remove.
7228 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7230 (zero_extendqisi2): Likewise.
7231 (zero_extendhisi2): Likewise.
7232 (extendqihi2): Likewise.
7233 (extendqisi2): Likewise.
7234 (extendhisi2): Likewise.
7238 * config/arc/constraints.md (Usd): Change it to memory constraint.
7240 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7242 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7243 as source of std instructions.
7244 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7245 allow 6-bit constants as source for store instructions.
7246 (movdi_insn): Update instruction pattern to allow 6-bit constants
7247 as source for store instructions.
7249 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7251 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7253 2018-04-30 Nathan Sidwell <nathan@acm.org>
7254 Sandra Loosemore <sandra@codesourcery.com>
7256 * dumpfile.c (dump_open): Allow '-' for stdout.
7257 * doc/invoke.texi (Developer Options): Document dump filename
7258 determination early. Document stdin/stdout selection.
7260 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7262 Microblaze Target: PIC data text relative
7264 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7265 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7267 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7268 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7269 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7270 New addressing mode for data-text relative position indepenedent code.
7271 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7272 'ADDRESS_SYMBOLIC_TXT_REL'.
7273 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7274 (microblaze_legitimate_pic_operand): Exclude function calls from
7275 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7276 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7278 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7279 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7280 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7281 for 'address + offset'.
7282 (microblaze_expand_prologue): Add new function prologue call for
7284 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7285 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7286 table in case of TARGET_PIC_DATA_TEXT_REL.
7287 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7288 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7289 Add new macros 'UNSPEC_TEXT',
7290 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7291 + exclude function calls from 'UNSPEC_PLT' in case of data text
7293 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7294 new target hook for generating address diff vector tables in case of
7296 * doc/tm.texi : Regenerate.
7297 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7298 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7299 of addr diff vector generation.
7300 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7301 target hook definition.
7302 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7303 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7304 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7305 option for data text relative.
7307 2018-04-30 Richard Biener <rguenther@suse.de>
7309 * tree-chrec.h (evolution_function_is_constant_p): Remove
7311 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7313 2018-04-30 Richard Biener <rguenther@suse.de>
7316 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7318 2018-04-30 Richard Biener <rguenther@suse.de>
7320 PR tree-optimization/28364
7321 PR tree-optimization/85275
7322 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7323 copying first exit test.
7325 2018-04-28 Mark Wielaard <mark@klomp.org>
7327 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7329 (dwarf_AT): Handle DW_AT_addr_base.
7330 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7332 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7335 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7336 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7337 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7338 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7340 2018-04-28 Richard Biener <rguenther@suse.de>
7342 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7343 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7344 to reflect use. Only add interesting stmts.
7346 2018-04-27 Martin Jambor <mjambor@suse.cz>
7349 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7350 the jump function allows for passing through aggregate values.
7352 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7354 * input.h (in_system_header_at): Convert from macro to inline
7356 (from_macro_expansion_at): Likewise.
7357 (from_macro_definition_at): Likewise.
7359 2018-04-27 Jeff Law <law@redhat.com>
7361 * config.gcc: Mark tile* targets as deprecated/obsolete.
7363 2018-04-27 Richard Biener <rguenther@suse.de>
7365 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7368 2018-04-27 Richard Biener <rguenther@suse.de>
7370 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7372 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7374 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7375 with Yd constraint. Set "preferred_for_speed" attribute from
7376 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7378 (*movdi_internal): Ditto.
7379 (movti_interunit splitters): Remove
7380 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7381 (movdi_interunit splitters): Ditto.
7382 * config/i386/constraints.md (Ye): Remove.
7383 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7385 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7388 * config/aarch64/constraints.md (Usg): Limit to 31.
7391 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7393 PR tree-optimization/85529
7394 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7395 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7396 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7397 zero extension or masking of the MSB bit.
7398 (optimize_range_tests): Add FIRST_BB argument, pass it through
7399 to optimize_range_tests_var_bound.
7400 (maybe_optimize_range_tests, reassociate_bb): Adjust
7401 optimize_range_tests callers.
7403 2018-04-26 Richard Biener <rguenther@suse.de>
7404 Jakub Jelinek <jakub@redhat.com>
7406 * cgraph.h (symbol_table): Just declare debug method here.
7407 * symtab.c (symbol_table::debug): Define.
7409 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7411 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7413 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7415 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7416 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7417 (*movdi_internal): Substitute Yi and Yj constraint with x
7418 and Ym and Yn constraint with y constraint. Update "isa"
7419 attribute and set "preferred_for_speed" attribute from
7420 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7421 (*movsi_internal): Ditto.
7422 (*movdf_internal): Ditto.
7423 (*movsf_internal): Ditto.
7424 (*zero_extendsidi2): Ditto.
7425 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7426 (sse2_loadld): Ditto.
7427 (*vec_extract<ssevecmodelower>_0): Ditto.
7428 (*vec_extractv4si_0_zext_sse4): Ditto.
7429 (vec_concatv2di): Ditto.
7430 (*vec_dup<mode>): Ditto.
7431 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7432 * config/i386/constraints.md (Yi): Remove.
7437 2018-04-26 Nathan Sidwell <nathan@acm.org>
7439 * dumpfile.c (dump_open): New.
7440 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7441 (dump_finish): Detect stdio/stderr by value not name.
7443 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7445 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7447 2018-04-26 Tom de Vries <tom@codesourcery.com>
7450 * config/nvptx/nvptx.c (verify_neutering_jumps)
7451 (verify_neutering_labels): New function
7452 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7454 2018-04-26 Tom de Vries <tom@codesourcery.com>
7457 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7458 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7461 2018-04-26 Richard Biener <rguenther@suse.de>
7462 Tom de Vries <tom@codesourcery.com>
7465 * lto-streamer-out.c (output_function): Fixup loops if required to match
7466 discovery done in the reader.
7468 2018-04-26 Richard Biener <rguenther@suse.de>
7470 PR tree-optimization/85116
7471 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7472 have a loop exit from the single latch predecessor. Remove
7473 case of header with just condition.
7474 (ch_base::copy_headers): Exclude infinite loops from any
7476 (pass_ch::execute): Record exits.
7478 2018-04-26 Richard Biener <rguenther@suse.de>
7480 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7481 prologue cost vector and pass it to vect_get_load_cost.
7482 (vect_get_peeling_costs_all_drs): Likewise.
7483 (vect_peeling_hash_get_lowest_cost): Likewise.
7484 (vect_enhance_data_refs_alignment): Likewise.
7486 2018-04-26 Richard Biener <rguenther@suse.de>
7489 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7490 checking of integer<->pointer conversions.
7491 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7492 sign-/zero-extending pointer types.
7493 (expand_omp_for_static_chunk): Likewise.
7495 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7496 Jean Lee <xiaoyur347@gmail.com>
7498 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7499 (TARGET_ASAN_SHADOW_OFFSET): Define.
7500 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7501 true for -fsanitize=address.
7503 2018-04-25 Mark Wielaard <mark@klomp.org>
7505 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7508 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7510 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7511 than "alu", remove explicit "memory" and "imm_disp" attributes.
7512 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7515 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7516 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7517 gen_lowpart_no_emit.
7519 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7522 * config/i386/i386.c (ix86_expand_builtin): Change memory
7523 operand to XI, extend p0 to Pmode.
7524 * config/i386/i386.md: Change unspec volatile and operand
7525 1 mode to XI, change operand 0 mode to P.
7527 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7529 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7530 GET_MODE_MASK before any checking.
7531 (nds32_can_use_bset_p): Likewise.
7532 (nds32_can_use_btgl_p): Likewise.
7534 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7536 * config/nds32/nds32-doubleword.md: New define_split pattern for
7537 illegal register number.
7539 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7541 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7543 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7545 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7547 2018-04-25 Richard Biener <rguenther@suse.de>
7549 * lto-streamer.h (LTO_major_version): Bump to 8.
7551 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7553 * BASE-VER: Set to 9.0.0.
7555 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7557 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7558 in __abskf2 and __powikf2.
7560 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7563 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7564 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7565 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7566 Use the above on operand 2. Reindent.
7567 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7569 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7572 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7574 * config/i386/i386.opt (mcet): Removed.
7575 * doc/install.texi: Remove -mcet documentation.
7576 * doc/invoke.texi: Likewise.
7578 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7581 * doc/install.texi: Remove -mcet from bootstrap-cet.
7583 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7586 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7587 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7591 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7592 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7593 containing a CONST_VECTOR.
7595 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7597 * doc/install.texi: Update newlib dependency for nvptx.
7599 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7602 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7603 instead of INTVAL when shifting x left.
7605 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7607 PR tree-optimization/85478
7608 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7609 vect_grouped_store_supported for single element vectors.
7611 2018-04-24 Richard Biener <rguenther@suse.de>
7614 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7615 load cost increase to the case of non-constant step.
7617 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7620 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7621 destination if any_malformed_asm.
7623 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7626 * expr.c (store_field): In the bitfield case, if the value comes from
7627 a function call and is returned in registers by means of a PARALLEL,
7628 do not change the mode of the temporary unless BLKmode and VOIDmode.
7630 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7632 PR rtl-optimization/85423
7633 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7634 dependencies to debug insns when the previous insn is non-debug.
7636 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7638 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7639 enums into a single definition.
7640 (fls): Fix predicates and printing.
7643 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7645 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7646 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7647 and short u6 immediate.
7648 (check_if_valid_sleep_operand): Remove.
7649 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7651 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7653 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7654 flag_always_save_lp condition.
7655 * config/nds32/nds32.opt (malways-save-lp): New option.
7657 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7659 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7660 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7661 * config/nds32/nds32.h
7662 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7663 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7665 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7667 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7668 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7670 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7671 Chung-Ju Wu <jasonwucj@gmail.com>
7673 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7675 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7676 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7678 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7680 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7682 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7684 * config/nds32/nds32-protos.h (nds32_data_alignment,
7685 nds32_local_alignment): Declare.
7686 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7687 nds32_local_alignment): New functions.
7688 (TARGET_CONSTANT_ALIGNMENT): Define.
7689 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7691 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7693 * config/nds32/nds32.c
7694 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7695 (TARGET_MODES_TIEABLE_P): Likewise.
7697 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7699 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7702 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7703 Chung-Ju Wu <jasonwucj@gmail.com>
7705 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7706 for unaligned access.
7707 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7709 * config/nds32/nds32-intrinsic.md: Likewise.
7710 * config/nds32/nds32_intrinsic.h: Likewise.
7711 * config/nds32/nds32.h (nds32_builtins): Likewise.
7712 * config/nds32/nds32.opt (munaligned-access): New option.
7713 * config/nds32/nds32.c (nds32_asm_file_start): Display
7714 flag_unaligned_access status.
7716 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7718 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7719 -mno-relax is present.
7720 * config/riscv/linux.h (LINK_SPEC): Ditto.
7722 2018-04-20 Martin Sebor <msebor@redhat.com>
7725 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7727 (gimple_fold_builtin_stxcpy_chk): Same.
7728 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7730 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7733 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7734 __powikf2 when long double is IEEE 128-bit.
7736 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7738 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7739 step to make sure stack always aligned.
7741 2018-04-20 Carl Love <cel@us.ibm.com>
7744 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7745 size check for arg0.
7747 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7748 Tom de Vries <tom@codesourcery.com>
7751 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7752 Emit insns for calls too.
7753 (nvptx_find_par): Always look for worker-level predecessor insn.
7754 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7756 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7757 (nvptx_process_pars): Propagate frames for calls.
7759 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7762 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7764 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7765 (ix86_handle_option): Don't handle OPT_mibt.
7766 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7768 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7770 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7771 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7772 (ix86_target_macros): Define __CET__ with flag_cf_protection
7773 for -fcf-protection.
7774 * config/i386/i386.c (isa2_opts): Remove -mibt.
7775 * config/i386/i386.h (TARGET_IBT): Removed.
7776 (TARGET_IBT_P): Likewise.
7777 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7778 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7779 * config/i386/i386.opt (mcet): Update help message.
7782 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7783 -mcet as an alias for -mshstk.
7785 2018-04-20 Richard Biener <rguenther@suse.de>
7788 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7789 complexity by forcing a single use of the multiply operand.
7791 2018-04-20 Martin Jambor <mjambor@suse.cz>
7794 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7795 recursion dependency to only apply to non-clones.
7797 2018-04-20 Martin Jambor <mjambor@suse.cz>
7800 * ipa-cp.c (create_specialized_node): Check that clones of
7801 self-recursive edges exist during IPA-CP.
7803 2018-04-19 Toon Moene <toon@moene.org>
7805 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7808 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7810 PR tree-optimization/85467
7811 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7812 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7813 VECTOR_CST element to type.
7815 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7818 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7819 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7820 (builtin_longjmp): Likewise.
7821 (save_stack_nonlocal): New pattern.
7822 (restore_stack_nonlocal): Likewise.
7824 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7827 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7828 Replace ASM_OUTPUT_LABEL with fprintf.
7830 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7833 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7834 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7835 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7836 define __IBT__ and __SHSTK__ for -fcf-protection.
7837 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7839 (ix86_trampoline_init): Likewise.
7840 (x86_output_mi_thunk): Likewise.
7841 (ix86_notrack_prefixed_insn_p): Likewise.
7842 (ix86_option_override_internal): Don't disallow -fcf-protection.
7843 * config/i386/i386.md (rdssp<mode>): Also enable for
7845 (incssp<mode>): Likewise.
7846 (nop_endbr): Likewise.
7847 * config/i386/i386.opt (mcet): Change help message to built-in
7851 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7852 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7853 enable CET built-in functions.
7855 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7857 * common/config/i386/i386-common.c
7858 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7859 OPTION_MASK_ISA_MOVDIRI_UNSET,
7860 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7861 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7862 * config.gcc (movdirintrin.h): New header.
7863 * config/i386/cpuid.h (bit_MOVDIRI,
7864 bit_MOVDIR64B): New bits.
7865 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7867 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7868 (VOID, PVOID, PCVOID)): New function types.
7869 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7870 __builtin_ia32_directstoreu_u64,
7871 __builtin_ia32_movdir64b): New builtins.
7872 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7873 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7875 (ix86_valid_target_attribute_inner_p): Ditto.
7876 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7877 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7878 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7879 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7880 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7881 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7882 (movdiri<mode>, movdir64b_<mode>): New.
7883 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7884 * config/i386/immintrin.h: Include movdirintrin.h.
7885 * config/i386/movdirintrin.h: New file.
7886 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7888 2018-04-19 Richard Biener <rguenther@suse.de>
7891 * cfg.c (clear_bb_flags): When loop state says we have
7892 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7894 2018-04-19 Richard Biener <rguenther@suse.de>
7896 PR tree-optimization/84737
7897 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7898 copying restrict info.
7899 (vect_setup_realignment): Use it.
7900 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7901 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7902 the first DR to all generated stores.
7903 (vectorizable_load): Likewise for loads.
7905 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7907 PR tree-optimization/85446
7908 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7909 the integral and pointer types to have the same precision.
7911 * doc/install.texi: Document --disable-cet being the default and
7914 2018-04-18 Martin Liska <mliska@suse.cz>
7916 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7919 2018-04-18 Martin Liska <mliska@suse.cz>
7922 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7925 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7926 arguments if they are comparable.
7928 2018-04-18 Martin Liska <mliska@suse.cz>
7931 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7934 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7937 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7940 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7941 ENDBR after calling __morestack.
7943 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7946 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7947 by using gcc_base_ver to generate a gcc_driver_version, and use
7948 it when generating GCC_DRIVER_NAME.
7949 * configure: Regenerate.
7951 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7954 * config.gcc: Obsolete powerpc*-*-*spe*.
7956 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7959 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7960 (stabstr_D): Change type of unum from unsigned int to
7961 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7964 2018-04-17 Jim Wilson <jimw@sifive.com>
7967 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7968 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7969 Set arg_pointer_offset after using pretend_args_size.
7971 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7973 PR rtl-optimization/85431
7974 * dse.c (record_store): Ignore zero width stores.
7977 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7978 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7979 __builtin_stack_restore rather than after it.
7980 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7981 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7982 argument instead of virtual_dynamic_stack_rtx.
7984 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7986 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7988 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7989 Add note to error message to explain internal mapping of overloaded
7990 built-in function name to non-overloaded built-in function name.
7991 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7994 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7997 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7998 where the inputs overlap with the output.
8000 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8002 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8003 (=v, v) alternative and explicit "memory" attribute.
8004 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8005 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8007 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8008 "sselog1" type instead of "sselog".
8009 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8010 "sselog". Remove explicit "memory" attribute.
8011 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8012 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8014 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8015 "sselog1" type instead of "sselog". Remove explicit "memory"
8017 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8018 use "sselog1" type instead of "sselog". Remove explicit "memory"
8020 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8021 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8023 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8024 "sselog1" type instead of "sselog". Remove explicit "memory"
8026 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8027 use "sselog1" type instead of "sselog". Remove explicit "memory"
8031 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8034 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8037 2018-04-17 Martin Jambor <mjambor@suse.cz>
8040 * ipa-cp.c (create_specialized_node): Call
8041 expand_all_artificial_thunks if necessary.
8043 2018-04-17 Martin Liska <mliska@suse.cz>
8046 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8047 in message, remote space in between '_G' and '('.
8049 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8052 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8053 avx512f_vmcmp<mode>3<round_saeonly_name>,
8054 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8055 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8056 avx512f_rndscale<mode><round_saeonly_name>,
8057 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8058 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8059 Use %<iptr>2 instead of %2 for -masm=intel.
8060 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8061 avx512f_vcvttss2usi<round_saeonly_name>,
8062 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8064 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8065 avx512f_vcvttsd2usi<round_saeonly_name>,
8066 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8067 Use %q1 instead of %1 for -masm=intel.
8068 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8069 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8070 of %3 for -masm=intel.
8071 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8073 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8075 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8077 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8078 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8080 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8082 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8083 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8084 %g1 and one with %0 and %1.
8085 (avx512er_vmrcp28<mode><round_saeonly_name>,
8086 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8088 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8089 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8090 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8091 of %0 and %{%4%} for -masm=intel.
8092 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8093 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8094 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8095 order of %0 and %{%5%}%{z%} for -masm=intel.
8097 2018-04-17 Jan Hubicka <jh@suse.cz>
8100 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8102 2018-04-17 Martin Liska <mliska@suse.cz>
8105 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8106 for target_clone error message. Make default implementation
8107 clone to be a local declaration.
8108 (separate_attrs): Add new argument and check for an empty
8110 (expand_target_clones): Handle it.
8111 (ipa_target_clone): Make redirection just for target_clones
8114 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8115 Tom de Vries <tom@codesourcery.com>
8118 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8119 tiled basic blocks without omp continue statements.
8121 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8124 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8125 vec_extract expression as having side effects to make sure it gets
8128 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8131 * config/i386/i386.c (get_builtin_code_for_version): Check
8134 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8137 * gcc/config.gcc: Support "skylake".
8138 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8140 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8141 (processor_target_table): Add "skylake".
8142 (ix86_option_override_internal): Add "skylake".
8143 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8144 PROCESSOR_CANNONLAKE.
8145 (get_builtin_code_for_version): Fix priority for
8146 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8147 PROCESSOR_SKYLAKE-AVX512.
8148 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8149 (processor_type): Add PROCESSOR_SKYLAKE.
8151 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8152 Jason Merrill <jason@redhat.com>
8155 * convert.c (convert_to_integer_1): Use direct recursion for
8156 enumeral types and types with a precision less than the number
8157 of bits in their mode.
8159 2018-04-16 Julia Koval <julia.koval@intel.com>
8162 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8163 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8165 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8168 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8169 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8170 and -mno-direct-move.
8172 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8175 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8176 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8177 Ensure negative shifts result in {0}.
8179 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8181 PR rtl-optimization/79916
8182 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8183 regs (if any) to define how to gnerate SD moves when LRA is in
8186 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8188 PR rtl-optimization/85393
8189 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8190 * except.c (expand_dw2_landing_pad_for_region): Make static.
8191 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8192 a label and unconditional jump to old_bb, rather than
8193 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8196 PR rtl-optimization/85376
8197 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8198 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8199 instead of a specific value.
8201 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8202 Bin Cheng <bin.cheng@arm.com>
8204 PR tree-optimization/82965
8205 PR tree-optimization/83991
8206 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8207 by_profile_only parameter.
8208 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8209 information if the loop was predicted to iterate too many times.
8210 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8212 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8215 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8218 2018-04-13 Martin Liska <mliska@suse.cz>
8219 Jakub Jelinek <jakub@redhat.com>
8222 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8223 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8224 * builtins.c (expand_builtin_memory_copy_args): Use
8225 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8226 handle dest_addr == pc_rtx.
8228 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8231 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8232 asked to not generate direct moves.
8233 (fix_trunc<mode>si2_stfiwx): Similar.
8234 (fix_trunc<mode>si2_internal): Similar.
8236 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8239 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8240 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8241 lookup if dest in some wider mode is known to be const0_rtx and
8242 if so, record permanent equivalence for it to be ZERO_EXTEND of
8243 the narrower mode destination.
8245 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8247 * lto-streamer-out.c (output_function): Revert 259346.
8248 * omp-expand.c (expand_oacc_for): Likewise.
8250 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8252 PR rtl-optimization/85354
8253 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8254 * sel-sched.c (sel_global_init): ... here.
8256 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8259 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8260 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8261 mode for PE-COFF targets.
8262 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8263 (i386_pe_asm_lto_end): Likewise.
8264 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8265 (TARGET_ASM_LTO_END): Likewise.
8266 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8267 (i386_pe_asm_lto_start): New function.
8268 (i386_pe_asm_lto_end): Likewise.
8270 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8271 Richard Biener <rguenther@suse.de>
8274 * lto-streamer-out.c (output_function): Fix CFG loop state before
8276 * omp-expand.c (expand_oacc_for): Handle calls to internal
8277 functions like regular functions.
8279 2018-04-12 Richard Biener <rguenther@suse.de>
8282 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8283 for the early LTO debug to properly generate references to it
8284 during DIE emission. Do not re-use that for the skeleton for
8286 (dwarf2out_early_finish): Likewise.
8288 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8291 * config/i386/sse.md
8292 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8293 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8294 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8295 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8296 and output is a reg, avoid creating invalid lowpart subreg, but
8297 instead split into a 512-bit move. Don't split if not AVX512VL,
8298 input is xmm16+ reg and output is a mem.
8299 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8300 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8301 xmm16+ reg and output is a mem.
8303 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8305 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8306 also for flag_dwarf2_cfi_asm.
8308 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8310 PR rtl-optimization/85342
8311 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8312 a bool scalar var inside of the loop instead. Don't try to update
8313 recog_data.operand after failed apply_change_group.
8315 2018-04-12 Tom de Vries <tom@codesourcery.com>
8318 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8319 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8320 array with flexible array member as array without given dimension.
8321 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8322 argument for undefined param to true.
8324 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8327 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8328 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8329 from PowerPC section.
8330 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8331 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8332 help text that is too long.
8333 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8334 help text that is too long.
8335 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8336 help text that is too long.
8338 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8340 * config/alpha/alpha.md (stack_probe_internal): Rename
8341 from "probe_stack". Update all callers.
8343 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8345 PR rtl-optimization/84566
8346 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8347 sched_macro_fuse_insns.
8349 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8352 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8353 (compute_block_dependences): ... from here.
8355 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8357 PR tree-optimization/85331
8358 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8359 from int to HOST_WIDE_INT.
8361 2018-04-11 Martin Jambor <mjambor@suse.cz>
8364 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8365 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8366 not the same as the source val.
8367 (cgraph_edge_brings_value_p): New parameter.
8368 (gather_edges_for_value): Pass destination value to
8369 cgraph_edge_brings_value_p.
8370 (perhaps_add_new_callers): Likewise.
8371 (get_info_about_necessary_edges): Likewise and exclude values brought
8372 only by self-recursive edges.
8373 (create_specialized_node): Redirect only clones of self-calling edges.
8374 (+self_recursive_pass_through_p): New function.
8375 (find_more_scalar_values_for_callers_subset): Use it.
8376 (find_aggregate_values_for_callers_subset): Likewise.
8377 (known_aggs_to_agg_replacement_list): Removed.
8378 (decide_whether_version_node): Re-calculate known constants for all
8379 remaining context clones.
8381 2018-04-11 Richard Biener <rguenther@suse.de>
8384 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8385 from early DWARF output.
8386 (dwarf2out_early_finish): Output line info unconditionally into
8387 early DWARF and add reference to it.
8389 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8392 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8393 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8394 other than V2DFmode using iptr mode attribute.
8395 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8397 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8399 PR rtl-optimization/84659
8400 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8402 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8405 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8407 (output_loc_list): Pass address of a dummy size variable even in the
8408 locview handling loop.
8409 (index_location_lists): Add comment on why skip_loc_list_entry can't
8412 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8415 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8418 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8421 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8422 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8423 and -mstring-compare-inline-limit.
8425 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8428 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8429 for stack clash protection in a register whenever we need it to be in
8432 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8434 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8435 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8437 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8440 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8442 (mlong-double-): Ditto.
8443 * config/rs6000/sysv4.opt (msdata=): Ditto.
8444 (mtls-size=): Ditto.
8446 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8448 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8449 erroneous entries for
8450 "vector int vec_ldl (int, long int *)", and
8451 "vector unsigned int vec_ldl (int, unsigned long int *)".
8452 Add comments and entries for
8453 "vector bool char vec_ldl (int, bool char *)",
8454 "vector bool short vec_ldl (int, bool short *)",
8455 "vector bool int vec_ldl (int, bool int *)",
8456 "vector bool long long vec_ldl (int, bool long long *)",
8457 "vector pixel vec_ldl (int, pixel *)",
8458 "vector long long vec_ldl (int, long long *)",
8459 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8460 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8461 type tree bool_long_long_type_node and correct definition of
8462 bool_V2DI_type_node to make reference to this new type tree.
8463 (rs6000_mangle_type): Replace erroneous reference to
8464 bool_long_type_node with bool_long_long_type_node.
8465 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8466 comments to emphasize sign distinctions for char and int types and
8467 replace RS6000_BTI_bool_long constant with
8468 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8469 use of RS6000_BTI_pixel.
8470 (bool_long_type_node): Remove this macro definition.
8471 (bool_long_long_type_node): New macro definition
8473 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8475 PR rtl-optimization/85300
8476 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8477 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8478 simplify_unary_operation fails.
8480 2018-04-10 Martin Liska <mliska@suse.cz>
8482 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8483 cgraph_edge and ipa_ref.
8485 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8489 * config/i386/sse.md
8490 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8491 computation of the VEC_MERGE selector from mask.
8492 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8493 Fix decoding of the VEC_MERGE selector into mask.
8495 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8497 PR tree-optimization/85286
8498 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8500 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8502 * final.c (final_1): Set insn_last_address as well as
8503 insn_current_address.
8505 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8508 * explow.c (emit_stack_probe): Call validize_mem on memory location
8509 before passing it to gen_probe_stack. Create address operand and
8510 legitimize it for the probe_stack_address case.
8512 2018-04-09 Jan Hubicka <jh@suse.cz>
8515 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8516 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8517 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8519 (free_lang_data): Rebuild type inheritance graph.
8521 2018-04-09 Martin Sebor <msebor@redhat.com>
8523 * invoke.texi (-finline-small-functions): Mention other optimization
8525 (-findirect-inlining, -fpartial-inlining): Same.
8526 (-finline-functions-called-once): Same.
8527 (-freorder-blocks-and-partition): Same.
8529 2018-04-09 Jan Hubicka <jh@suse.cz>
8532 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8533 jumps; choose last target that matches the criteria (i.e.
8534 no partition changes for non-crossing jumps).
8535 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8536 support for redirecting crossing jumps to non-crossing.
8538 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8540 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8541 also for naked functions.
8543 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8545 * config/arc/arc.md (add_shift): New pattern.
8546 (add_shift2): Likewise.
8547 (sub_shift): Likewise.
8548 (sub_shift_cmp0_noout): Likewise.
8549 (compare_si_ashiftsi): Likewise.
8550 (xbfu_cmp0_noout): New combine pattern.
8551 (xbfu_cmp0"): Likewise.
8552 (movsi_set_cc_insn): Place the predicable variant first.
8553 (commutative_binary_cmp0_noout): Remove clobber.
8554 (commutative_binary_cmp0): New pattern.
8555 (noncommutative_binary_cmp0): Likewise.
8556 (noncommutative_binary_cmp0_noout): Likewise.
8557 (noncommutative_binary_comparison_result_used): Removed.
8558 (rsub_cmp0): New pattern.
8559 (rsub_cmp0_noout): Likewise.
8560 (extzvsi): Changed, keep only meaningful variants.
8561 (SQH, SEZ): New iterators.
8562 (SQH_postfix): New mode attribute.
8563 (SEZ_prefix): New code attribute.
8564 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8565 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8566 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8568 (noncommutative_operator): Check the availability of barrel
8571 2018-04-09 Richard Biener <rguenther@suse.de>
8573 PR tree-optimization/85284
8574 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8575 Only use the niter constraining form of simple_iv when the exit
8578 2018-04-09 Tom de Vries <tom@codesourcery.com>
8581 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8582 (define_expand "*memory_barrier"): New define_expand.
8583 (define_insn "memory_barrier"): New insn.
8585 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8587 PR rtl-optimization/80463
8588 PR rtl-optimization/83972
8589 PR rtl-optimization/83480
8591 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8592 correct producer for the insn.
8593 (tidy_control_flow): Fixup seqnos in case of debug insns.
8595 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8597 PR rtl-optimization/83913
8599 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8600 different sched-times when merging exprs.
8602 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8604 PR rtl-optimization/83962
8606 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8607 tidy_fallthru_edge and tidy_control_flow.
8609 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8611 PR rtl-optimization/83530
8613 * sel-sched.c (force_next_insn): New global variable.
8614 (remove_insn_for_debug): When force_next_insn is true, also leave only
8615 next insn in the ready list.
8616 (sel_sched_region): When the region wasn't scheduled, make another pass
8617 over it with force_next_insn set to 1.
8619 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8621 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8623 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8624 for interrupt control.
8625 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8626 functions for interrupt control.
8627 * config/nds32/nds32-intrinsic.md: Likewise.
8628 * config/nds32/nds32_intrinsic.h: Likewise.
8629 * config/nds32/nds32.h (nds32_builtins): Likewise.
8631 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8633 * config/nds32/nds32.c (nds32_init_machine_status,
8634 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8635 strict_aligned_p field.
8636 (nds32_expand_to_rtl_hook): New function.
8637 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8638 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8640 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8641 Chung-Ju Wu <jasonwucj@gmail.com>
8643 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8644 * config/nds32/nds32-n7.md: New file.
8645 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8646 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8648 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8649 * config/nds32/nds32.md (pipeline_model): Add n7.
8650 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8651 * config/nds32/pipelines.md: Include n7 settings.
8653 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8654 Chung-Ju Wu <jasonwucj@gmail.com>
8656 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8657 * config/nds32/nds32-e8.md: New file.
8658 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8659 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8661 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8662 * config/nds32/nds32.md (pipeline_model): Add e8.
8663 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8664 * config/nds32/pipelines.md: Include e8 settings.
8666 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8667 Chung-Ju Wu <jasonwucj@gmail.com>
8669 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8670 * config/nds32/nds32-n8.md: New file.
8671 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8672 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8674 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8675 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8676 * config/nds32/nds32.md (pipeline_model): Add n8.
8677 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8678 * config/nds32/pipelines.md: Include n8 settings.
8680 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8681 Chung-Ju Wu <jasonwucj@gmail.com>
8683 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8684 * config/nds32/nds32-n9-2r1w.md: New file.
8685 * config/nds32/nds32-n9-3r2w.md: New file.
8686 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8687 nds32_register_ports): New or modify for cpu n9.
8688 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8690 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8691 * config/nds32/nds32-utils.c: New file.
8692 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8693 TARGET_MUL_SLOW): Define.
8694 * config/nds32/nds32.md (pipeline_model): New attribute.
8695 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8696 New options that support cpu n9.
8697 * config/nds32/pipelines.md: Include n9 settings.
8698 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8700 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8702 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8703 information if necessary.
8704 (output_cond_branch_compare_zero): Likewise.
8705 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8706 (nds32_target_alignment): Refine for alignment.
8707 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8708 (FUNCTION_BOUNDARY): Modify.
8709 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8711 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8713 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8715 * config/nds32/constants.md (unspec_volatile_element): Add values for
8716 TLB operation and data prefetch.
8717 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8718 functions for TLB operation and data prefetch.
8719 * config/nds32/nds32-intrinsic.md: Likewise.
8720 * config/nds32/nds32_intrinsic.h: Likewise.
8721 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8722 (nds32_print_operand): Likewise.
8723 * config/nds32/nds32.h (nds32_builtins): Likewise.
8725 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8726 Andrew Pinski <pinsika@gcc.gnu.org>
8729 * match.pd: Use constant_boolean_node of correct type instead of
8730 boolean_true_node or boolean_false_node for simplifying
8731 pointer comparisons to zero.
8733 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8735 PR tree-optimization/80021
8736 * tree.c (verify_type_variant): Make error call in verify_variant_match
8737 translatable and remove final full stop.
8739 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8741 * config/nds32/constants.md (unspec_volatile_element): Add
8742 UNSPEC_VOLATILE_EH_RETURN.
8743 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8744 nds32_output_stack_pop): Support dwarf exception handling process.
8745 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8746 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8747 exception handling process.
8748 (nds32_compute_stack_frame): Likewise.
8749 (nds32_return_addr_rtx): Likewise.
8750 (nds32_initial_elimination_offset): Likewise.
8751 (nds32_expand_prologue): Likewise.
8752 (nds32_expand_epilogue): Likewise.
8753 (nds32_dynamic_chain_address): New function.
8754 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8756 (DYNAMIC_CHAIN_ADDRESS): Define.
8757 (EH_RETURN_DATA_REGNO): Define.
8758 (EH_RETURN_STACKADJ_RTX): Define.
8759 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8760 patterns for dwarf exception handling.
8762 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8764 * config/nds32/nds32.h: Clean up obsolete macros.
8766 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8768 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8769 Add enum values for particular instructions.
8770 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8771 particular intrinsic functions.
8772 * config/nds32/nds32-intrinsic.md: Likewise.
8773 * config/nds32/nds32_intrinsic.h: Likewise.
8774 * config/nds32/nds32.h (nds32_builtins): Likewise.
8775 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8776 (btst, ave): New patterns for particular instructions.
8778 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8780 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8781 Add enum values for atomic load/store and memory sync.
8782 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8784 * config/nds32/nds32-intrinsic.md: Likewise.
8785 * config/nds32/nds32_intrinsic.h: Likewise.
8786 * config/nds32/nds32.h (nds32_builtins): Likewise.
8788 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8790 PR tree-optimization/85257
8791 * fold-const.c (native_encode_vector): If not all elts could fit
8792 and off is -1, return 0 rather than offset.
8793 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8794 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8795 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8796 adjust buffer in native_interpret_expr call.
8798 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8800 * config/nds32/constants.md (unspec_volatile_element): Add cache
8801 control enum values.
8802 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8803 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8804 * config/nds32/nds32.c (nds32_cctl_names): New.
8805 (nds32_print_operand): Handle cache control register names.
8806 * config/nds32/nds32.h (nds32_builtins): New enum values.
8807 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8809 * config/nds32/nds32.md (type): Add mmu.
8810 * config/nds32/pipelines.md (simple_insn): Add mmu.
8812 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8814 * config/nds32/nds32.md (type): Remove call.
8815 * config/nds32/pipelines.md (simple_insn): Likewise.
8817 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8819 * config/nds32/constants.md (unspec_volatile_element): Add
8820 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8821 UNSPEC_VOLATILE_FMFCFG.
8822 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8823 description for fmfcfg and fmfcsr.
8824 (bdesc_1arg): Add fmtcsr.
8825 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8826 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8827 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8828 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8829 unspec_fmfcfg): New patterns.
8830 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8831 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8832 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8833 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8834 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8835 __nds32__fmfcfg): Define.
8837 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8839 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8840 intrinsic register names.
8841 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8842 intrinsic register enum values and macros.
8844 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8846 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8847 for load/store addressing form.
8848 (nds32_print_operand_address): Likewise.
8850 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8853 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8854 based on LABEL_REF. Remove useless assertion.
8855 (pic_address_needs_scratch): Fix formatting.
8856 (sparc_legitimize_pic_address): Minor tweaks.
8857 (sparc_delegitimize_address): Adjust assertion accordingly.
8858 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8859 into symbolic_operand.
8860 (movsi_high_pic_label_ref): Likewise.
8861 (movsi_lo_sum_pic_label_ref): Likewise.
8862 (movdi_pic_label_ref): Likewise.
8863 (movdi_high_pic_label_ref): Likewise.
8864 (movdi_lo_sum_pic_label_ref): Likewise.
8866 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8868 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8869 custom LIB_SPEC setup.
8871 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8872 Kito Cheng <kito.cheng@gmail.com>
8874 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8875 * config/riscv/freebsd.h: New.
8877 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8879 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8880 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8883 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8884 Kito Cheng <kito.cheng@gmail.com>
8886 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8887 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8888 * config/nds32/nds32-protos.h (nds32_output_call,
8889 nds32_output_return): Declare.
8890 * config/nds32/nds32.md: Refine all the call and return patterns.
8892 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8895 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8896 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8898 PR rtl-optimization/84872
8899 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8900 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8903 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8905 * expr.c (copy_blkmode_to_reg): Revert 254862.
8906 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8908 2018-04-06 Richard Biener <rguenther@suse.de>
8911 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8912 after seeing a component reference with an adjacent field. Treat
8913 refs to arrays at struct end of external decls similar to
8914 refs to unconstrained commons.
8916 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8919 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8920 look through SAVE_EXPRs with non-side-effects argument. Adjust
8922 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8925 2018-04-06 Richard Biener <rguenther@suse.de>
8928 * alias.c (find_base_term): New wrapper around find_base_term
8929 unwinding CSELIB_VAL_PTR changes.
8930 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8933 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8935 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8937 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8938 constant definitions.
8939 ("nop"): lr 0,0 -> nopr r0
8940 ("nop_lr0", "nop_lr1"): New insn definitions.
8942 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8944 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8945 NDS32_V3PUSH_AVAILABLE_P macro.
8947 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8948 Chung-Ju Wu <jasonwucj@gmail.com>
8950 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8951 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8952 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8953 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8954 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8955 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8956 * config/nds32/constraints.md: New constraints and checking for hard
8957 float configuration.
8958 * config/nds32/iterators.md: New mode iterator and attribute for hard
8959 float configuration.
8960 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8962 * config/nds32/nds32-fpu.md: New file.
8963 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8964 deal with hard float code generation.
8965 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8967 (abi_type, float_reg_number): New enum type.
8968 * config/nds32/nds32-predicates.c: New predicates for hard float.
8969 * config/nds32/nds32-protos.h: Declare functions for hard float.
8970 * config/nds32/nds32.c: Implementation for hard float configuration.
8971 * config/nds32/nds32.h: Definitions for hard float configuration.
8972 * config/nds32/nds32.md: Include hard float machine description and
8973 modify patterns for hard float configuration.
8974 * config/nds32/nds32.opt: New options for hard float configuration.
8975 * config/nds32/predicates.md: New predicates for hard float
8978 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8980 * common/config/nds32/nds32-common.c
8981 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8983 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8986 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8987 CONSTRUCTOR_ELT (ctor, ...)->value.
8989 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8992 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8994 2018-04-05 Tom de Vries <tom@codesourcery.com>
8997 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9000 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9001 Kito Cheng <kito.cheng@gmail.com>
9003 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9004 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9005 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9006 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9008 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9009 Kito Cheng <kito.cheng@gmail.com>
9011 * config/nds32/constraints.md (Ufe): New memory constraint.
9012 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9013 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9014 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9016 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9017 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9019 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9021 * config/nds32/nds32.md: Use optimize_size in the condition for
9022 alu-shift instructions.
9024 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9026 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9028 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9030 * config/nds32/nds32.md (negsi2): Refine pattern.
9032 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9033 Chung-Ju Wu <jasonwucj@gmail.com>
9035 * config/nds32/iterators.md (shift_rotate): New code iterator.
9036 (shift): New code attribute.
9037 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9038 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9039 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9040 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9041 bit-wise operations.
9042 (andsi3, *andsi3): Ditto.
9043 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9044 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9045 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9046 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9047 nds32_ior_operand, nds32_xor_operand): New predicates.
9049 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9051 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9052 (addsi3, subsi3): ... this.
9054 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9056 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9058 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9060 * config/nds32/nds32.md: Adjust indention.
9062 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9064 * config/nds32/nds32.md (feature): New attribute.
9066 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9068 * config/nds32/nds32.md (subtype): New attribute.
9070 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9073 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9074 expansion to perform a bitwise AND of the argument followed by a
9075 boolean negation of the result.
9077 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9079 PR rtl-optimization/84878
9080 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9081 the basic block. Assert the use reference is not artificial and that
9082 it has an associated insn.
9084 2018-04-04 Michael Matz <matz@suse.de>
9086 * builtins.c (compute_objsize): Pass correct operand
9087 to array_at_struct_end_p.
9089 2018-04-04 Richard Biener <rguenther@suse.de>
9092 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9093 from contexts for DINFO_LEVEL_TERSE and below.
9095 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9097 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9098 resiter_operand condition.
9099 * config/nds32/nds32.md (*move<mode>): Ditto.
9101 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9102 Monk Chiang <sh.chiang04@gmail.com>
9104 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9106 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9108 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9110 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9111 Kito Cheng <kito.cheng@gmail.com>
9113 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9114 nds32_cond_code_str, output_cond_branch,
9115 output_cond_branch_compare_zero, nds32_expand_cbranch,
9116 nds32_expand_cstore, nds32_expand_movcc,
9117 nds32_output_cbranchsi4_equality_zero,
9118 nds32_output_cbranchsi4_equality_reg,
9119 nds32_output_cbranchsi4_equality_reg_or_const_int,
9120 nds32_output_cbranchsi4_greater_less_zero: New functions.
9121 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9122 nds32_expand_cstore, nds32_expand_movcc,
9123 nds32_output_cbranchsi4_equality_zero,
9124 nds32_output_cbranchsi4_equality_reg,
9125 nds32_output_cbranchsi4_equality_reg_or_const_int,
9126 nds32_output_cbranchsi4_greater_less_zero): Declare.
9127 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9128 nds32_rimm11s_operand): New predicates.
9129 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9130 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9133 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9135 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9136 * config/nds32/nds32.md: Ditto.
9137 * config/nds32/pipelines.md: Ditto.
9139 2018-04-04 Richard Biener <rguenther@suse.de>
9141 PR tree-optimization/85168
9142 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9143 propagating abnormals.
9145 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9147 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9149 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9150 Kito Cheng <kito.cheng@gmail.com>
9152 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9153 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9154 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9155 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9156 * config/nds32/nds32.md (sibcall_internal): New.
9157 (sibcall_register): Remove.
9158 (sibcall_immediate): Remove.
9159 (sibcall_value_internal): New.
9160 (sibcall_value_register): Remove.
9161 (sibcall_value_immediate): Remove.
9162 * config/nds32/predicates.md (nds32_general_register_operand): New.
9163 (nds32_call_address_operand): New.
9165 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9167 PR rtl-optimization/85167
9168 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9169 bb_defs if *split_p, instead preinitialize it to NULL.
9171 PR tree-optimization/85156
9172 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9173 evaluating the argument multiple times.
9175 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9177 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9179 (_mm_cvtpd_ps): Likewise.
9180 (_mm_cvttpd_epi32): Likewise.
9181 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9182 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9183 vector, pixel, and bool following altivec.h include.
9185 2018-04-03 Martin Sebor <msebor@redhat.com>
9187 * doc/extend.texi (Common Function Attributes): Clarify.
9188 (const attribute): Likewise.
9189 (pure attribute): Likewise.
9191 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9194 * config/i386/i386.c (ix86_expand_vector_set): Use
9195 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9197 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9199 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9200 instructions when changing rounding bits to preserve precision bits
9201 in the x87 control word.
9203 2018-04-03 Martin Liska <mliska@suse.cz>
9205 PR tree-optimization/82491
9206 * rtl.h (strip_offset_and_add): Replace += suboffset with
9207 poly_uint64 () + suboffset.
9209 2018-03-29 Martin Liska <mliska@suse.cz>
9210 Martin Jambor <mjambor@suse.cz>
9213 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9214 param_type is not an integral or pointer type.
9216 2018-04-03 Richard Biener <rguenther@suse.de>
9218 * sese.h (recompute_all_dominators): Remove.
9220 2018-04-02 Martin Sebor <msebor@redhat.com>
9222 * doc/invoke.texi (-Wrestrict): Fix typos.
9224 2018-04-02 Jim Wilson <jimw@sifive.com>
9226 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9227 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9228 (<optab>di3, <optab>si3_extend): Likewise.
9229 (<optab>si3_mask, <optab>si3_mask_1): New.
9230 (<optab>di3_mask, <optab>di3_mask_1): New.
9231 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9232 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9233 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9235 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9237 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9240 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9242 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9243 (nds32_canonicalize_comparison): New function.
9245 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9246 Kito Cheng <kito.cheng@gmail.com>
9247 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9249 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9250 * config/nds32/constants.md (unspec_volatile_element): Add
9251 UNSPEC_VOLATILE_RELAX_GROUP.
9252 * config/nds32/nds32-relax-opt.c: New file.
9253 * config/nds32/nds32-predicates.c
9254 (nds32_symbol_load_store_p): New function.
9255 * config/nds32/nds32-protos.h
9256 (nds32_symbol_load_store_p): Declare function.
9257 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9258 * config/nds32/nds32.c
9259 (nds32_register_pass): New function to register pass.
9260 (nds32_register_passes): New function to register passes.
9261 * config/nds32/nds32.md (relax_group): New pattern.
9262 * config/nds32/nds32.opt (mrelax-hint): New option.
9263 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9265 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9267 * config/nds32/t-nds32: Modify files dependency.
9269 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9271 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9272 (PROFILE_HOOK): Define its implementation.
9274 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9276 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9277 type and 32-bit size.
9279 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9282 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9283 (V_128_256): New mode iterator.
9284 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9285 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9286 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9288 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9289 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9291 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9294 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9295 NaN inputs correctly.
9297 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9300 * config/rs6000/vsx.md (??r): New mode attribute.
9301 (*vsx_mov<mode>_64bit): Use it.
9302 (*vsx_mov<mode>_32bit): Likewise.
9304 2018-03-30 Martin Sebor <msebor@redhat.com>
9306 PR tree-optimization/84818
9307 * builtins.c (check_access): Use warning_n.
9309 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9312 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9314 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9317 2018-03-30 Julia Koval <julia.koval@intel.com>
9320 * x86-tune.def (movx, partial_reg_dependency): Enable for
9323 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9326 * lra-constraints.c (process_alt_operands): Move setting
9327 this_alternative_matches below.
9329 2018-03-29 Martin Liska <mliska@suse.cz>
9332 * doc/invoke.texi: Document how LTO works with debug info.
9333 Describe auto-load support of binutils. Mention 'x86-64'
9334 as valid option value of -march option.
9336 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9338 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9341 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9342 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9343 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9346 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9349 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9350 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9351 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9352 for RS6000_BTM_POWERPC64.
9353 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9354 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9355 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9360 2018-03-28 Carl Love <cel@us.ibm.com>
9363 2017-09-27 Carl Love <cel@us.ibm.com>
9365 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9366 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9367 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9370 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9372 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9373 instead of __vector bool.
9374 (_mm_max_pu8): Likewise.
9375 (_mm_min_pi16): Likewise.
9377 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9380 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9381 (DIVWEUO): Likewise.
9383 (DIVDEUO): Likewise.
9384 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9385 DIVWEUO and DIVDEUO.
9386 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9387 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9388 (div_extend): Likewise.
9389 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9391 (__builtin_divweuo): Likewise.
9392 (__builtin_divdeo): Likewise.
9393 (__builtin_divdeuo): Likewise.
9395 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9398 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9399 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9401 PR tree-optimization/82004
9402 * gimple-match-head.c (optimize_pow_to_exp): New function.
9403 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9404 Don't fold to exp if optimize_pow_to_exp is false.
9406 2018-03-28 Martin Liska <mliska@suse.cz>
9409 * calls.c (initialize_argument_information): Fix trailing space.
9410 * common.opt: Fix typo and provide better explanation for
9411 -fsanitize-coverage option.
9412 * config/i386/i386.opt: Fix typo.
9414 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9415 Martin Liska <mliska@suse.cz>
9418 * gimplify.c (asan_poison_variable): Don't do the check for
9419 gimplify_omp_ctxp here.
9420 (gimplify_decl_expr): Do it here.
9421 (gimplify_target_expr): Likewise.
9423 2018-03-28 Martin Liska <mliska@suse.cz>
9426 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9427 chkp_type_bounds_count if MPX is not enabled.
9429 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9431 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9433 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9436 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9437 function to create the function decl for complex long double
9438 multiply and divide for -mabi=ieeelongdouble.
9439 (init_float128_ieee): Call it.
9441 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9444 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9445 -fcf-protection=branch -mibt.
9446 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9448 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9451 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9452 arm_word_relocations.
9454 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9457 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9458 extern array declarations.
9460 2018-03-27 Richard Biener <rguenther@suse.de>
9463 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9464 explicit single_use checks.
9466 2018-03-27 Richard Biener <rguenther@suse.de>
9468 PR tree-optimization/85082
9469 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9472 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9474 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9475 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9476 Turn on fasynchronous-unwind-tables and funwind-tables.
9478 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9481 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9482 (*bmi_blsr_<mode>_ccz): Ditto.
9484 2018-03-26 Tom de Vries <tom@codesourcery.com>
9486 PR tree-optimization/85063
9487 * omp-general.c (offloading_function_p): New function. Factor out
9489 * omp-offload.c (pass_omp_target_link::gate): ... here.
9490 * omp-general.h (offloading_function_p): Declare.
9491 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9492 with attribute omp declare target for offloading functions.
9494 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9496 PR tree-optimization/84005
9497 * tree-data-ref.h (get_base_for_alignment): Declare.
9498 * tree-data-ref.c (get_base_for_alignment_1): New function.
9499 (get_base_for_alignment): Likewise.
9500 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9501 get_base_for_alignment to find a suitable base object, instead
9502 of always using drb->base_address.
9504 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9507 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9508 known size by default.
9510 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9513 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9514 and non BLKmode operands.
9516 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9519 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9520 Clean up attributes.
9522 2018-03-23 Richard Biener <rguenther@suse.de>
9525 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9526 we are going to emit early debug for LTO.
9528 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9531 * function.c (match_asm_constraints_1): Don't optimize if input
9532 doesn't satisfy general_operand predicate for output's mode.
9535 * alias.c (write_dependence_p): Don't require for x_canonicalized
9536 non-VOIDmode if x has VOIDmode.
9539 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9540 just don't try to optimize it rather than assert it never happens.
9542 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9544 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9545 macro expansions for definition of ST_INTERNAL_<mode> and
9546 LD_INTERNAL_<mode> builtins.
9547 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9549 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9551 (altivec_expand_st_builtin): Likewise.
9552 (altivec_expand_builtin): Remove calls to deleted functions.
9553 (rs6000_address_for_altivec): Delete this function.
9554 * config/rs6000/vector.md: Remove expands for
9555 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9557 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9560 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9561 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9562 re-computing once computed.
9563 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9564 (arm_init_machine_status): Initialize
9565 machine->static_chain_stack_bytes.
9567 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9570 * doc/extend.texi: Add four new prototypes for vec_ld.
9571 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9572 definitions for more logical presentation.
9573 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9574 entries for V1TI variants of __builtin_altivec_ld builtin.
9575 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9576 handling of V1TI variant of LVX icode pattern.
9577 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9578 (rs6000_gimple_fold_builtin): Likewise.
9579 (altivec_init_builtins): Add code to define
9580 __builtin_altivec_lvx_v1ti function.
9582 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9585 * function.c (match_asm_constraints_1): Don't do the optimization
9586 if input isn't a REG, SUBREG, MEM or constant.
9588 2018-03-22 Tom de Vries <tom@codesourcery.com>
9590 PR tree-optimization/84956
9591 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9592 bb_has_abnormal_pred.
9594 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9597 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9598 DECL_INITIAL (decl) to decl at the end.
9599 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9602 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9604 * doc/extend.texi (__builtin_tgmath): Document when complex
9605 integer types are treated as _Complex _Float64.
9607 2018-03-21 Tom de Vries <tom@codesourcery.com>
9609 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9611 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9613 PR tree-optimization/84960
9614 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9615 if it is ENTRY block, move them into single succ of ENTRY in that case.
9617 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9619 PR tree-optimization/84811
9620 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9621 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9622 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9623 (known_subrange_p): Update accordingly. Cast each value involved
9624 in the size comparison, rather than casting the result of the
9627 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9629 PR tree-optimization/84982
9630 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9631 by flipping the least significant bit rather than all bits from
9632 bitpos to bitpos + bitsize - 1.
9634 2018-03-21 Nathan Sidwell <nathan@acm.org>
9636 * doc/extend.texi (Deprecated Features): Remove mention of
9637 long-deleted deprecations.
9639 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9642 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9643 * configure: Regenerate.
9645 2018-03-21 Tom de Vries <tom@codesourcery.com>
9647 PR tree-optimization/83126
9648 * tree-parloops.c (num_phis): New function.
9649 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9651 2018-03-21 Nathan Sidwell <nathan@acm.org>
9653 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9654 mention anon-struct/union members and trailing attributes.
9656 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9658 PR tree-optimization/84969
9659 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9660 builtin memset partitions if they set different rhs values.
9662 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9664 PR rtl-optimization/84989
9665 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9666 VEC_DUPLICATE with scalar result mode.
9668 2018-03-21 Martin Liska <mliska@suse.cz>
9671 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9672 not intended return statement.
9674 2018-03-21 Martin Liska <mliska@suse.cz>
9677 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9678 (chkp_find_bound_slots_1): Limit number of iterations.
9680 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9683 * Minor grammar fixes for x86 options.
9685 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9688 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9689 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9691 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9694 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9695 (altivec_lvx_<mode>_1op): Likewise.
9696 (altivec_stvx_<mode>_2op): Likewise.
9697 (altivec_stvx_<mode>_1op): Likewise.
9698 (altivec_lvx_<VM2:mode>): New define_expand.
9699 (altivec_stvx_<VM2:mode>): Likewise.
9700 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9701 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9702 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9703 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9704 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9705 (rs6000_gen_lvx): Likewise.
9706 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9707 (altivec_expand_stv_builtin): Likewise.
9708 (altivec_expand_builtin): Likewise.
9709 * config/rs6000/vector.md: Likewise.
9711 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9714 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9717 2018-03-20 Richard Biener <rguenther@suse.de>
9720 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9721 sign-conversions as zero, fall back to standard scalar_stmt
9724 2018-03-20 Martin Liska <mliska@suse.cz>
9727 * predict.c (rebuild_frequencies): Handle case when we have
9728 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9730 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9733 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9734 flag_section_anchors.
9735 * varasm.c (use_blocks_for_decl_p): Remove hack for
9736 dw2_force_const_mem.
9739 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9741 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9742 be created, use lowpart_subreg of operands[0] rather than operands[0]
9744 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9745 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9746 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9747 and n constraint instead of aarch64_shift_imm_di and Usd.
9748 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9749 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9751 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9754 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9755 to favor GPR over NEON registers.
9756 (<shift>di3_neon): Likewise.
9758 2018-03-20 Tom de Vries <tom@codesourcery.com>
9761 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9762 (nvptx_process_pars): Emit bar.sync asap and alap.
9764 2018-03-20 Tom de Vries <tom@codesourcery.com>
9767 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9768 seen_label if seen_label is already set.
9770 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9773 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9774 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9775 Use 1U instead of 1. Formatting fixes.
9778 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9779 instead of TREE_TYPE (s1) for the return value.
9781 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9783 PR tree-optimization/84946
9784 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9785 bitsize + bitsize in poly_uint64 rather than poly_int64.
9788 * dwarf2asm.c: Include fold-const.c.
9789 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9790 of decl rather than decl itself.
9792 PR rtl-optimization/84643
9793 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9795 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9798 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9799 calling assemble_variable.
9801 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9804 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9805 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9807 2018-03-19 Jim Wilson <jimw@sifive.com>
9810 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9811 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9812 (riscv_first_stack_step): Likewise.
9813 (riscv_option_override): Use STACK_BOUNDARY instead of
9815 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9817 (BIGGEST_ALIGNMENT): Set to 128.
9818 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9819 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9822 2018-03-19 Richard Biener <rguenther@suse.de>
9824 PR tree-optimization/84933
9825 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9826 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9828 2018-03-19 Richard Biener <rguenther@suse.de>
9830 PR tree-optimization/84859
9831 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9832 (cond_if_else_store_replacement): Perform sinking operation on
9833 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9834 Generalize what a BB with a single eligible store is.
9836 2018-03-19 Richard Biener <rguenther@suse.de>
9838 PR tree-optimization/84929
9839 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9840 chrec_is_positive against non-chrec arg.
9842 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9845 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9847 2018-03-18 Martin Liska <mliska@suse.cz>
9849 PR rtl-optimization/84635
9850 * regrename.c (build_def_use): Use matches_mode only when
9853 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9855 PR tree-optimization/84913
9856 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9857 vectorize chains of COND_EXPRs.
9859 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9861 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9863 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9865 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9867 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9869 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9871 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9872 Kito Cheng <kito.cheng@gmail.com>
9874 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9875 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9876 (nds32_adjust_reg_alloc_order): New function.
9877 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9879 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9881 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9882 nds32_print_operand, nds32_print_operand_address): Use
9883 HOST_WIDE_INT_PRINT_DEC instead.
9885 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9887 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9889 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9892 * config/i386/i386.c (initial_ix86_tune_features,
9893 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9895 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9896 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9897 rather than 1u << ix86_tune. Formatting fix.
9898 (ix86_option_override_internal): Change ix86_arch_mask from
9899 unsigned int to unsigned HOST_WIDE_INT, initialize to
9900 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9901 (ix86_function_specific_restore): Likewise.
9903 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9906 * postreload.c (reload_combine_recognize_pattern): Perform
9907 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9908 truncate_int_for_mode the result for the destination's mode.
9911 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9912 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9914 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9917 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9920 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9921 regno_allocno_class_array and sorted_pseudos.
9922 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9923 insns where regno is used.
9925 2018-03-16 Martin Liska <mliska@suse.cz>
9928 * multiple_target.c (create_dispatcher_calls): Redirect
9929 reference in the symbol table.
9931 2018-03-16 Martin Liska <mliska@suse.cz>
9934 * multiple_target.c (create_dispatcher_calls): Redirect also
9937 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9941 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9942 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9943 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9945 2018-03-16 Julia Koval <julia.koval@intel.com>
9947 * doc/invoke.texi (Skylake Server): Add CLWB.
9948 Cannonlake): Remove CLWB.
9950 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9952 PR tree-optimization/84841
9953 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9955 (FLOAT_ONE_CONST_TYPE): Define.
9956 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9957 (sort_by_operand_rank): Put entries with higher constant_type last
9958 rather than first to match comments.
9960 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9962 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9965 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9968 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9969 check and use address of target always.
9971 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9974 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9975 (indirect_thunk_bnd_needed): Likewise.
9976 (indirect_thunks_used): Likewise.
9977 (indirect_thunks_bnd_used): Likewise.
9978 (indirect_return_needed): New.
9979 (indirect_return_bnd_needed): Likewise.
9980 (output_indirect_thunk_function): Add a bool argument for
9982 (output_indirect_thunk_function): Don't generate alias for
9983 function return thunk.
9984 (ix86_code_end): Call output_indirect_thunk_function to generate
9985 function return thunks.
9986 (ix86_output_function_return): Set indirect_return_bnd_needed
9987 and indirect_return_needed instead of indirect_thunk_bnd_needed
9988 and indirect_thunk_needed.
9990 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9992 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9993 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9994 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9996 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9997 Paul Hua <paul.hua.gm@gmail.com>
10000 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10002 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10004 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10005 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10006 resp. SFmode cases.
10008 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10011 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10012 instead of GET_MODE_SIZE when comparing Units.
10014 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10017 * varasm.c (hash_section): Return an unchangeble hash value
10018 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10019 Return !aarch64_can_use_per_function_literal_pools_p ().
10021 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10024 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10025 rather than address of cmode as last argument to prepare_cmp_insn.
10027 2018-03-15 Julia Koval <julia.koval@intel.com>
10029 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10030 F_AVX512VNNI, F_AVX512BITALG): New.
10032 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10035 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10036 insn for floating-point loads and stores.
10038 2018-03-14 Carl Love <cel@us.ibm.com>
10040 * config/rs6000/rs6000-c.c: Add macro definitions for
10041 ALTIVEC_BUILTIN_VEC_PERMXOR.
10042 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10043 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10044 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10045 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10047 * config/doc/extend.texi: Add prototypes for vec_permxor.
10049 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10052 * diagnostic-show-locus.c (class layout_point): Convert m_line
10053 from int to linenum_type.
10054 (line_span::comparator): Use linenum "compare" function when
10055 comparing line numbers.
10056 (test_line_span): New function.
10057 (layout_range::contains_point): Convert param "row" from int to
10059 (layout_range::intersects_line_p): Likewise.
10060 (layout::will_show_line_p): Likewise.
10061 (layout::print_source_line): Likewise.
10062 (layout::should_print_annotation_line_p): Likewise.
10063 (layout::print_annotation_line): Likewise.
10064 (layout::print_leading_fixits): Likewise.
10065 (layout::annotation_line_showed_range_p): Likewise.
10066 (struct line_corrections): Likewise for field m_row.
10067 (line_corrections::line_corrections): Likewise for param "row".
10068 (layout::print_trailing_fixits): Likewise.
10069 (layout::get_state_at_point): Likewise.
10070 (layout::get_x_bound_for_row): Likewise.
10071 (layout::print_line): Likewise.
10072 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10073 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10074 * input.c (selftest::test_linenum_comparisons): New function.
10075 (selftest::input_c_tests): Call it.
10076 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10077 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10078 * selftest.h (ASSERT_GT): New macro.
10079 (ASSERT_GT_AT): New macro.
10080 (ASSERT_LT): New macro.
10081 (ASSERT_LT_AT): New macro.
10083 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10085 PR rtl-optimization/84780
10086 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10088 2018-03-14 Martin Liska <mliska@suse.cz>
10090 * tree.c (record_node_allocation_statistics): Use
10091 get_stats_node_kind.
10092 (get_stats_node_kind): New function extracted from
10093 record_node_allocation_statistics.
10094 (free_node): Use get_stats_node_kind.
10096 2018-03-14 Richard Biener <rguenther@suse.de>
10098 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10099 that the value-set of ANTIC_IN doesn't grow.
10102 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10104 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10105 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10107 2018-03-14 Julia Koval <julia.koval@intel.com>
10109 * config.gcc (icelake-client, icelake-server): New.
10111 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10112 (initial_ix86_arch_features): Ditto.
10113 (PTA_SKYLAKE): Add SGX.
10114 (PTA_ICELAKE): Remove.
10115 (PTA_ICELAKE_CLIENT): New.
10116 (PTA_ICELAKE_SERVER): New.
10117 (ix86_option_override_internal): Split up icelake on icelake client and
10119 (get_builtin_code_for_version): Ditto.
10120 (fold_builtin_cpu): Ditto.
10121 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10122 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10123 * config/i386/i386.h (processor_type): Ditto.
10124 * doc/invoke.texi: Ditto.
10126 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10129 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10130 INTEGER_CST offset, add it together with bitpos / 8 and
10131 sign extend based on POINTER_SIZE.
10135 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10138 * config/i386/constraints.md (Yc): New register constraint.
10139 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10140 Use Yc constraint for alternative 2 of operand 0. Remove
10141 preferred_for_speed attribute.
10143 2018-03-14 Richard Biener <rguenther@suse.de>
10145 PR tree-optimization/84830
10146 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10147 with the old one to avoid oscillations.
10149 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10152 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10154 (assign_by_spills): Return a flag of reload assignment failure.
10155 Do not process the reload assignment failures. Do not spill other
10156 reload pseudos if they has the same reg class. Update n if
10158 (lra_assign): Add a return arg. Set up from the result of
10159 assign_by_spills call.
10160 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10161 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10162 usage_insns if it is not NULL.
10163 (spill_hard_reg_in_range): New function.
10164 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10165 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10166 function prototypes.
10167 (lra_assign): Change prototype.
10168 * lra.c (lra): Add code to deal with fails by splitting hard reg
10171 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10173 * config/riscv/riscv.opt (mrelax): New option.
10174 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10175 "norelax" when riscv_mrelax is disabled.
10176 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10178 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10181 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10182 reassociation for int modes.
10184 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10186 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10187 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10189 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10190 * config/aarch64/aarch64-sve.md
10191 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10192 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10193 (*extend<mode><Vwide>2): Rename to...
10194 (aarch64_sve_extend<mode><Vwide>2): ...this.
10195 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10196 renaming the old pattern to...
10197 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10199 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10200 define_expand, renaming the old pattern to...
10201 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10202 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10203 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10204 account when deciding which SVE instruction the optab should use.
10205 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10207 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10209 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10210 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10211 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10212 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10213 (tlsdesc_small_<mode>): Turn a define_expand and use
10214 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10215 (tlsdesc_small_advsimd_<mode>): ...this.
10216 (tlsdesc_small_sve_<mode>): New pattern.
10218 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10220 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10221 (UNSPEC_UMUL_HIGHPART): New constants.
10222 (MUL_HIGHPART): New int iteraor.
10223 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10224 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10226 (*<su>mul<mode>3_highpart): New define_insn.
10228 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10231 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10234 2018-03-13 Martin Liska <mliska@suse.cz>
10237 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10239 (sem_item_optimizer::~sem_item_optimizer): Release it.
10240 (sem_item_optimizer::merge_classes): Register variable aliases.
10241 (sem_item_optimizer::fixup_pt_set): New function.
10242 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10243 * ipa-icf.h: Declare new variables and functions.
10245 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10247 PR middle-end/84834
10248 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10249 integer_pow2p@2 and test integer_pow2p in condition.
10250 (A < 0 ? C : 0): Similarly for @1.
10252 PR middle-end/84831
10253 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10254 characters starting at p contain '\0' character, don't look beyond
10258 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10259 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10262 * reg-stack.c (change_stack): Change update_end var from int to
10263 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10264 also call set_block_for_insn on the newly added insns and rescan.
10267 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10268 on the last operand.
10271 * tree.c (stabilize_reference_1): Return save_expr (e) for
10272 STATEMENT_LIST even if it doesn't have side-effects.
10274 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10276 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10278 2018-03-12 Renlin Li <renlin.li@arm.com>
10280 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10281 aarch64_output_scalar_simd_mov_immediate.
10283 2018-03-12 Martin Sebor <msebor@redhat.com>
10285 PR tree-optimization/83456
10286 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10287 for perfectly overlapping calls to memcpy.
10288 (gimple_fold_builtin_memory_chk): Same.
10289 (gimple_fold_builtin_strcpy): Handle no-warning.
10290 (gimple_fold_builtin_stxcpy_chk): Same.
10291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10293 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10295 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10296 parameter. Use it for SFmode.
10297 (rs6000_function_arg_advance_1): Adjust.
10298 (rs6000_function_arg): Adjust.
10299 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10301 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10303 PR rtl-optimization/84169
10304 PR rtl-optimization/84780
10305 * combine.c (can_combine_p): Check for a 2-insn combination whether
10306 the destination register is used between the two insns, too.
10308 2018-03-12 Richard Biener <rguenther@suse.de>
10310 PR tree-optimization/84803
10311 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10312 for refs DR analysis didn't process.
10314 2018-03-12 Richard Biener <rguenther@suse.de>
10316 PR tree-optimization/84777
10317 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10318 force-vectorize loops ignore whether we are optimizing for size.
10320 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10322 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10323 (TARGET_MD_ASM_ADJUST): Define.
10325 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10326 Kito Cheng <kito.cheng@gmail.com>
10327 Chung-Ju Wu <jasonwucj@gmail.com>
10329 * config/nds32/nds32.c (nds32_compute_stack_frame,
10330 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10331 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10332 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10333 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10334 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10335 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10336 * config/nds32/nds32.md (prologue, epilogue): Use macro
10337 NDS32_V3PUSH_AVAILABLE_P to do checking.
10339 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10342 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10343 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10344 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10345 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10346 addition of most attributes on !orig_type_die or the attribute not
10347 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10349 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10350 Chung-Ju Wu <jasonwucj@gmail.com>
10352 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10353 __NDS32_VH__ macro.
10354 * config/nds32/nds32.opt (mvh): New option.
10356 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10357 Chung-Ju Wu <jasonwucj@gmail.com>
10359 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10361 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10362 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10365 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10366 Chung-Ju Wu <jasonwucj@gmail.com>
10368 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10370 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10371 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10373 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10374 Kito Cheng <kito.cheng@gmail.com>
10375 Chung-Ju Wu <jasonwucj@gmail.com>
10377 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10378 UNSPEC_FFMISM and UNSPEC_FLMISM.
10379 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10380 for ffb, ffmism and flmism.
10381 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10382 (unspec_ffmism): Ditto.
10383 (unspec_flmism): Ditto.
10384 (nds32_expand_builtin_impl): Check if string extension is available.
10385 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10386 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10388 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10391 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10394 * lra-assigns.c (assign_by_spills): Return a flag of reload
10395 assignment failure. Do not process the reload assignment
10396 failures. Do not spill other reload pseudos if they has the same
10398 (lra_assign): Add a return arg. Set up from the result of
10399 assign_by_spills call.
10400 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10401 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10402 usage_insns if it is not NULL.
10403 (spill_hard_reg_in_range): New function.
10404 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10405 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10406 function prototypes.
10407 (lra_assign): Change prototype.
10408 * lra.c (lra): Add code to deal with fails by splitting hard reg
10411 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10414 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10416 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10419 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10420 (dw_val_node): Add val_symbolic_view.
10421 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10422 (symview_upper_bound): New.
10423 (new_line_info_table): Initialize symviews_since_reset.
10424 (dwarf2out_source_line): Count symviews_since_reset and set
10425 symview_upper_bound.
10426 (dw_val_equal_p): Handle symview.
10427 (add_AT_symview): New.
10428 (print_dw_val): Handle symview.
10429 (attr_checksum, attr_checksum_ordered): Likewise.
10430 (same_dw_val_p, size_of_die): Likewise.
10431 (value_format, output_die): Likewise.
10432 (add_high_low_attributes): Use add_AT_symview for entry_view.
10433 (dwarf2out_finish): Reset symview_upper_bound, clear
10436 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10439 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10440 Add strict argument and use it.
10441 (rs6000_split_multireg_move): Update for new strict argument.
10442 (mem_operand_gpr): Disallow all non-offsettable addresses.
10443 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10445 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10448 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10449 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10450 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10453 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10454 decl, use remap_type if we want to use the type.
10456 2018-03-09 Martin Sebor <msebor@redhat.com>
10458 PR tree-optimization/84526
10459 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10461 (builtin_access::generic_overlap): Be prepared to handle non-array
10464 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10466 PR rtl-optimization/84682
10467 * lra-constraints.c (process_address_1): Check is_address flag
10468 for address constraints.
10469 (process_alt_operands): Likewise.
10470 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10471 preprocess_constraints.
10472 * recog.h (preprocess_constraints): Add oploc parameter.
10474 * recog.c (preprocess_constraints): Test address_operand for
10475 CT_ADDRESS constraints.
10477 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10480 * lra-assigns.c (assign_by_spills): Return a flag of reload
10481 assignment failure. Do not process the reload assignment
10482 failures. Do not spill other reload pseudos if they has the same
10484 (lra_assign): Add a return arg. Set up from the result of
10485 assign_by_spills call.
10486 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10487 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10488 usage_insns if it is not NULL.
10489 (spill_hard_reg_in_range): New function.
10490 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10491 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10492 function prototypes.
10493 (lra_assign): Change prototype.
10494 * lra.c (lra): Add code to deal with fails by splitting hard reg
10497 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10500 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10501 Accept complain bool parameter. Only emit errors if it is true.
10502 (arm_parse_cpu_option_name): Likewise.
10503 (arm_target_thumb_only): Adjust callers of the above.
10504 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10505 prototype to take a default true bool parameter.
10506 (arm_parse_arch_option_name): Likewise.
10508 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10509 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10513 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10514 * configure: Regenerate.
10515 * configure.ac ("linker --version-script option"): New.
10516 ("linker soname option"): New.
10518 2018-03-09 Richard Biener <rguenther@suse.de>
10520 PR tree-optimization/84775
10521 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10522 immediate uses of predicate stmts and mark them modified.
10525 PR tree-optimization/84178
10526 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10528 (version_loop_for_if_conversion): Delay update_ssa call.
10529 (tree_if_conversion): Delay update_ssa until after predicate
10532 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10535 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10536 when the function accesses prior frames.
10538 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10541 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10542 gen_llsym, otherwise call maybe_gen_llsym.
10544 PR inline-asm/84742
10545 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10546 has ',' character inside of it.
10548 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10551 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10552 as clobbering CC_REGNUM.
10554 2018-03-08 Richard Biener <rguenther@suse.de>
10556 PR middle-end/84552
10557 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10558 (follow_copies_to_constant): Do not follow SSA names registered
10561 2018-03-08 Richard Biener <rguenther@suse.de>
10563 PR tree-optimization/84178
10564 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10566 (version_loop_for_if_conversion): Delay update_ssa call.
10567 (tree_if_conversion): Delay update_ssa until after predicate
10570 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10572 PR tree-optimization/84178
10573 * tree-if-conv.c (release_bb_predicate): Remove the
10574 the assertion that the stmts have NULL use_ops.
10575 Discard the statements, asserting that they haven't
10576 yet been added to a BB.
10578 2018-03-08 Richard Biener <rguenther@suse.de>
10580 PR tree-optimization/84746
10581 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10582 (phi_translate): Pass in destination ANTIC_OUT set.
10583 (phi_translate_1): Likewise. For a simplified result lookup
10584 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10585 (phi_translate_set): Adjust.
10586 (do_pre_regular_insertion): Likewise.
10587 (do_pre_partial_partial_insertion): Likewise.
10589 2018-03-08 Martin Liska <mliska@suse.cz>
10591 PR gcov-profile/84735
10592 * doc/gcov.texi: Document usage of profile files.
10593 * gcov-io.h: Document changes in the format.
10595 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10599 * dwarf2out.c (struct dw_line_info_table): Update comments for
10601 (FORCE_RESET_NEXT_VIEW): New.
10602 (FORCE_RESETTING_VIEW_P): New.
10603 (RESETTING_VIEW_P): Check for -1 too.
10604 (ZERO_VIEW_P): Likewise.
10605 (new_line_info_table): Force-reset next view.
10606 (dwarf2out_begin_function): Likewise.
10607 (dwarf2out_source_line): Simplify zero_view_p initialization.
10608 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10609 view directly. Omit view when omitting .loc at line 0.
10611 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10613 PR tree-optimization/84740
10614 * tree-switch-conversion.c (process_switch): Call build_constructors
10615 only if info.phi_count is non-zero.
10617 PR tree-optimization/84739
10618 * tree-tailcall.c (find_tail_calls): Check call arguments against
10619 DECL_ARGUMENTS (current_function_decl) rather than
10620 DECL_ARGUMENTS (func) when checking for tail recursion.
10622 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10624 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10625 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10626 Volker Reichelt's entry and add entries for people that perform
10627 GCC fuzzy testing and report numerous bugs.
10629 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10632 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10633 readonly data in sdata, if that is disabled.
10634 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10635 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10636 -mreadonly-in-sdata option.
10638 2018-03-07 Martin Sebor <msebor@redhat.com>
10640 PR tree-optimization/84468
10641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10642 basic block when looking for nul assignment.
10644 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10647 * except.h (output_function_exception_table): Adjust prototype.
10648 * except.c (output_function_exception_table): Remove FNNAME parameter
10649 and add SECTION parameter. Ouput one part of the table at a time.
10650 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10651 the first part of the exception table and emit unwind directives.
10652 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10653 (i386_pe_seh_cold_init): Likewise.
10654 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10655 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10656 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10657 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10658 * config/i386/winnt.c: Include except.h.
10659 (struct seh_frame_state): Add reg_offset, after_prologue and
10660 in_cold_section fields.
10661 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10662 (i386_pe_seh_cold_init): New function.
10663 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10664 to seh->in_cold_section.
10665 (seh_emit_push): Record the offset of the push.
10666 (seh_emit_save): Record the offet of the save.
10667 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10668 Test seh->after_prologue to disregard the epilogue.
10669 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10670 (i386_pe_end_cold_function): New function.
10672 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10675 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10676 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10679 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10681 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10682 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10684 PR middle-end/84723
10685 * multiple_target.c: Include tree-inline.h and intl.h.
10686 (expand_target_clones): Diagnose and fail if node->definition and
10687 !tree_versionable_function_p (node->decl).
10689 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10691 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10693 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10694 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10695 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10697 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10700 * combine.c (try_combine): Use reg_or_subregno instead of handling
10701 just paradoxical SUBREGs and REGs.
10703 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10705 * config/arc/arc.c (arc_finalize_pic): Remove function.
10706 (arc_must_save_register): We use single base PIC register, remove
10707 checks to save/restore the PIC register.
10708 (arc_expand_prologue): Likewise.
10709 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10711 (arc_verify_short): Likewise.
10712 (arc_attr_type): Likewise.
10713 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10714 (walk_stores): Likewise.
10715 (arc_address_cost): Make it static.
10716 (arc_verify_short): Likewise.
10717 (branch_dest): Likewise.
10718 (arc_attr_type): Likewise.
10719 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10720 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10721 (arc_final_prescan_insn): Remove inserting the nops due to
10722 hardware hazards. It is done in reorg step.
10723 (insn_length_variant_t): Remove.
10724 (insn_length_parameters_t): Likewise.
10725 (arc_insn_length_parameters): Likewise.
10726 (arc_get_insn_variants): Likewise.
10727 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10729 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10731 PR inline-asm/84683
10732 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10735 PR tree-optimization/84687
10736 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10738 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10740 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10742 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10743 Rename to ppc_speculation_barrier.
10744 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10745 __builtin_ppc_speculation_barrier.
10747 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10750 * combine.c (combine_simplify_rtx): Don't try to simplify if
10751 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10754 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10756 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10757 to 32 bytes when compiling for POWER9.
10759 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10762 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10763 regparm >= 3 with no arg reg available also for calls with
10764 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10767 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10769 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10771 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10774 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10776 2018-03-05 Richard Biener <rguenther@suse.de>
10778 PR tree-optimization/84486
10779 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10780 When inserting a __builtin_assume_aligned call set the LHS
10781 SSA name alignment info accordingly.
10783 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10785 PR tree-optimization/84114
10786 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10787 Avoid reassociation of FLOAT_MODE addition.
10789 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10791 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10792 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10793 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10794 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10795 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10796 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10797 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10799 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10800 __builtin_ia32_wbinvd): New builtins.
10801 (SPECIAL_ARGS2): New.
10802 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10803 (SPECIAL_ARGS2): New.
10804 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10805 (ix86_valid_target_attribute_inner_p): Ditto.
10806 (ix86_init_mmx_sse_builtins): Add special_args2.
10807 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10808 TARGET_WBNOINVD_P): New.
10809 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10810 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10811 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10812 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10813 * config/i386/pconfigintrin.h: New file.
10814 * config/i386/wbnoinvdintrin.h: Ditto.
10815 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10817 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10819 2018-03-05 Richard Biener <rguenther@suse.de>
10821 PR tree-optimization/84670
10822 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10824 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10825 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10826 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10827 doesn't grow if all successors (recursively) were visited at least
10830 2018-03-05 Richard Biener <rguenther@suse.de>
10832 PR tree-optimization/84650
10833 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10834 if executed in the loop pipeline.
10836 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10838 * doc/configfiles.texi (Configuration Files): Move info about
10839 conditionalizing $target-protos.h to...
10840 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10841 differs from $target-protos.h.
10843 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10844 Chung-Ju Wu <jasonwucj@gmail.com>
10846 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10847 * config/nds32/nds32-multiple.md (setmemsi): Define.
10848 * config/nds32/nds32-memory-manipulation.c
10849 (nds32_gen_dup_4_byte_to_word_value): New.
10850 (emit_setmem_word_loop): New.
10851 (emit_setmem_byte_loop): New.
10852 (nds32_expand_setmem_loop): New.
10853 (nds32_expand_setmem_loop_v3m): New.
10854 (nds32_expand_setmem_unroll): New.
10855 (nds32_expand_setmem): New.
10857 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10858 Chung-Ju Wu <jasonwucj@gmail.com>
10860 * config/nds32/nds32-memory-manipulation.c
10861 (nds32_emit_load_store): New.
10862 (nds32_emit_post_inc_load_store): New.
10863 (nds32_emit_mem_move): New.
10864 (nds32_emit_mem_move_block): New.
10865 (nds32_expand_movmemsi_loop_unknown_size): New.
10866 (nds32_expand_movmemsi_loop_known_size): New.
10867 (nds32_expand_movmemsi_loop): New.
10868 (nds32_expand_movmemsi_unroll): New.
10869 (nds32_expand_movmemqi): Rename ...
10870 (nds32_expand_movmemsi): ... to this.
10871 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10872 (movmemsi): ... to this.
10873 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10874 (nds32_expand_movmemsi): ... to this.
10876 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10877 Monk Chiang <sh.chiang04@gmail.com>
10878 Chung-Ju Wu <jasonwucj@gmail.com>
10880 * config/nds32/nds32-protos.h
10881 (nds32_expand_load_multiple): New arguments.
10882 (nds32_expand_store_multiple): Ditto.
10883 (nds32_valid_multiple_load_store): Rename ...
10884 (nds32_valid_multiple_load_store_p): ... to this.
10885 * config/nds32/nds32-memory-manipulation.c
10886 (nds32_expand_load_multiple): Refine implementation.
10887 (nds32_expand_store_multiple): Ditto.
10888 * config/nds32/nds32-multiple.md
10889 (load_multiple): Update nds32_expand_load_multiple interface.
10890 (store_multiple): Update nds32_expand_store_multiple interface.
10891 * config/nds32/nds32-predicates.c
10892 (nds32_valid_multiple_load_store): Rename ...
10893 (nds32_valid_multiple_load_store_p): ... to this and refine
10895 * config/nds32/predicates.md
10896 (nds32_load_multiple_and_update_address_operation): New predicate.
10897 (nds32_store_multiple_and_update_address_operation): New predicate.
10899 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10900 Chung-Ju Wu <jasonwucj@gmail.com>
10902 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10903 (combo): New attribute.
10904 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10906 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10908 * config/nds32/nds32.opt: Change -mcmodel= default value.
10910 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10911 Monk Chiang <sh.chiang04@gmail.com>
10912 Chung-Ju Wu <jasonwucj@gmail.com>
10914 * config/nds32/constants.md (unspec_element): New enum.
10915 * config/nds32/constraints.md (Umw): New constraint.
10916 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10917 * config/nds32/nds32-intrinsic.md: Likewise.
10918 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10919 (nds32_valid_smw_lwm_base_p): New.
10920 (nds32_output_smw_single_word): New.
10921 (nds32_output_lmw_single_word): New.
10922 (nds32_expand_unaligned_load): New.
10923 (nds32_expand_unaligned_store): New.
10924 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10925 (nds32_output_smw_single_word): Declare.
10926 (nds32_output_lmw_single_word): Declare.
10927 (nds32_expand_unaligned_load): Declare.
10928 (nds32_expand_unaligned_store): Declare.
10929 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10930 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10931 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10932 NDS32_BUILTIN_UASTORE_DW.
10933 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10936 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10937 Kito Cheng <kito.cheng@gmail.com>
10938 Chung-Ju Wu <jasonwucj@gmail.com>
10940 * config/nds32/nds32-intrinsic.c
10941 (nds32_expand_builtin_null_ftype_reg): Delete.
10942 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10943 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10944 (nds32_read_argument): New.
10945 (nds32_legitimize_target): Ditto.
10946 (nds32_legitimize_argument): Ditto.
10947 (nds32_check_constant_argument): Ditto.
10948 (nds32_expand_unop_builtin): Ditto.
10949 (nds32_expand_unopimm_builtin): Ditto.
10950 (nds32_expand_binop_builtin): Ditto.
10951 (nds32_builtin_decl_impl): Ditto.
10952 (builtin_description): Ditto.
10953 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10954 (nds32_init_builtins_impl): Ditto.
10955 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10956 (nds32_builtin_decl): New.
10957 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10958 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10960 2018-03-02 Jeff Law <law@redhat.com>
10962 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10963 (redundant_insn, fill_simple_delay_slots): Likewise.
10964 (fill_slots_from_thread): Likewise.
10965 * resource.c (mark_referenced_resources): Likewise.
10966 (mark_set_resources, find_dead_or_set_registers): Likewise.
10968 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10970 * substring-locations.h (format_warning_va): Formatting fix for
10971 ATTRIBUTE_GCC_DIAG.
10972 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10974 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10975 * substring-locations.c: Include intl.h.
10976 (format_warning_va): Turned into small wrapper around
10977 format_warning_n_va, renamed to ...
10978 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10979 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10981 (format_warning_at_substring_n): New function.
10982 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10983 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10984 format_warning_at_substring with just a shorter name instead of
10985 const function pointer.
10986 (fmtwarn_n): New function.
10987 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10988 appropriate, get rid of all the fmtstr temporaries, move conditionals
10989 with G_() wrapped string literals directly into fmtwarn arguments,
10990 cast dir.len to (int), formatting fixes.
10992 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10994 * doc/invoke.texi: Remove "Cilk Plus" references.
10996 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10997 Richard Biener <rguenther@suse.de>
11000 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11001 for error or warning attributes if CALL_FROM_THUNK_P is set.
11004 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11007 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11008 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11011 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11012 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11014 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11015 instead of -1U in last predictors element's probability member.
11017 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11020 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11021 arguments if they are comparable.
11023 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11025 PR tree-optimization/84634
11026 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11027 masks and masked_loop_p with a single loop_masks, making sure it's
11028 null for bb vectorization.
11030 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11032 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11033 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11034 loop->force_vectorize to check whether there is no alias.
11036 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11039 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11041 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11043 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11044 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11045 instead of a loop around prev_real_insn.
11046 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11049 PR inline-asm/84625
11050 * config/i386/i386.c (ix86_print_operand): Use conditional
11051 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11054 2018-03-02 Richard Biener <rguenther@suse.de>
11056 PR tree-optimization/84427
11057 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11058 (bitmap_set_subtract_values): Rewrite to handle multiple
11061 (prune_clobbered_mems): Likewise.
11062 (phi_translate): Take edge instead of pred/phiblock.
11063 (phi_translate_1): Likewise.
11064 (phi_translate_set): Likewise. Insert all translated
11065 exprs for a value into the set, keeping possibly multiple
11066 expressions per value.
11067 (compute_antic_aux): Adjust for phi_translate changes.
11068 When intersecting union the expressions and prune those
11069 not in the final value set, keeping possibly multiple
11070 expressions per value. Do not use value-insertion
11071 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11072 all expressions. Add verification that the value-sets
11073 only shrink during iteration.
11074 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11075 (do_pre_regular_insertion): Likewise.
11076 (do_pre_partial_partial_insertion): Likewise.
11078 2018-03-02 Richard Biener <rguenther@suse.de>
11081 * config/darwin.c (saved_debug_info_level): New static global.
11082 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11083 (darwin_asm_lto_end): Restore debug info generation settings.
11085 2018-03-01 Martin Liska <mliska@suse.cz>
11088 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11089 volatile arguments.
11091 2018-03-01 Richard Biener <rguenther@suse.de>
11094 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11095 type annotation with LTO when debug was disabled at compile-time.
11097 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11099 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11101 (mips_final_postscan_insn): Likewise.
11103 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11105 PR rtl-optimization/84528
11106 * alias.c (init_alias_target): Add commentary.
11107 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11108 a unique base value if the frame pointer is not eliminated
11109 to the stack pointer.
11111 2018-03-01 Tom de Vries <tom@codesourcery.com>
11113 PR rtl-optimization/83327
11114 * lra-int.h (hard_regs_spilled_into): Declare.
11115 * lra.c (hard_regs_spilled_into): Define.
11116 (init_reg_info): Init hard_regs_spilled_into.
11117 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11118 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11119 (process_bb_lives): Handle hard_regs_spilled_into.
11120 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11121 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11123 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11125 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11126 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11127 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11128 * config/rs6000/aix72.h: New file.
11130 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11132 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11133 instead of warning_at with conditional singular and plural messages
11137 * stor-layout.c (update_alignment_for_field): For
11138 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11139 && !DECL_PACKED (field), do the alignment update, just use
11140 only desired_align instead of MAX (type_align, desired_align)
11142 (place_field): Don't do known_align < desired_align handling
11143 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11144 is non-NULL, instead do it after rli->prev_field handling and
11145 only if not within a bitfield word. For DECL_PACKED (field)
11146 use type_align of BITS_PER_UNIT.
11148 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11150 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11151 superfluous parentheses and trailing spaces.
11153 2018-02-28 Richard Biener <rguenther@suse.de>
11155 PR tree-optimization/84584
11156 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11157 SCoPs with fake exit edge.
11159 2018-02-28 Martin Liska <mliska@suse.cz>
11162 * timevar.c (timer::print): Fix format to properly print 100%
11165 2018-02-28 Richard Biener <rguenther@suse.de>
11167 PR middle-end/84607
11168 * genmatch.c (capture_info::walk_match): Do not mark
11169 captured expressions without operands as expr_p given
11170 they act more like predicates and should be subject to
11171 "lost tail" side-effect preserving.
11173 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11175 PR rtl-optimization/81611
11176 * auto-inc-dec.c (attempt_change): Move dead note from
11177 mem_insn if it's the next use of regno
11178 (find_address): Take address use of reg holding
11179 non-incremented value. Add parm to limit search to the named
11181 (merge_in_block): Attempt to use a mem insn that is the next
11182 use of the original regno.
11184 2018-02-27 Martin Sebor <msebor@redhat.com>
11187 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11188 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11190 2018-02-27 Martin Sebor <msebor@redhat.com>
11192 PR translation/84207
11193 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11194 n argument to unsigned HOST_WIDE_INT.
11195 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11196 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11197 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11198 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11200 2018-02-27 Richard Biener <rguenther@suse.de>
11202 PR tree-optimization/84512
11203 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11204 Do not use the estimate returned from record_stmt_cost for
11205 the scalar iteration cost but sum properly using add_stmt_cost.
11207 2018-02-27 Richard Biener <rguenther@suse.de>
11209 PR tree-optimization/84466
11210 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11211 Adjust last change to less strictly validate use operands.
11213 2018-02-27 Martin Liska <mliska@suse.cz>
11215 PR gcov-profile/84548
11216 * gcov.c (process_file): Allow partial overlap and consider it
11217 also as group functions.
11218 (output_lines): Properly calculate range of lines for a group.
11220 2018-02-27 Martin Liska <mliska@suse.cz>
11222 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11223 'ggc' suffixes. Change first column width.
11224 (timer::print): Fix formatting of the column.
11226 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11228 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11229 preserve inline entry blocks for the sake of debug inline
11230 entry point markers alone.
11231 (remove_unused_locals): Suggest in comments a better place to
11232 force the preservation of inline entry blocks that are
11233 otherwise unused, but do not preserve them.
11235 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11237 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11239 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11242 * config/i386/constraints.md (Bs): Replace
11243 ix86_indirect_branch_register with
11244 TARGET_INDIRECT_BRANCH_REGISTER.
11246 * config/i386/i386.md (indirect_jump): Likewise.
11247 (tablejump): Likewise.
11248 (*sibcall_memory): Likewise.
11249 (*sibcall_value_memory): Likewise.
11250 Peepholes of indirect call and jump via memory: Likewise.
11251 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11252 (*sibcall_value_GOT_32): Likewise.
11253 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11254 (GOT_memory_operand): Likewise.
11255 (call_insn_operand): Likewise.
11256 (sibcall_insn_operand): Likewise.
11257 (GOT32_symbol_operand): Likewise.
11258 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11260 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11262 PR rtl-optimization/83496
11263 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11264 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11265 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11266 redundant insn, if any.
11267 (relax_delay_slots): Likewise.
11268 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11270 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11272 PR tree-optimization/83965
11273 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11274 that grouped statements are part of a reduction chain. Return
11275 true if the statement is not marked as a reduction itself but
11276 is part of a group.
11277 (vect_recog_dot_prod_pattern): Don't check whether the statement
11278 is part of a group here.
11279 (vect_recog_sad_pattern): Likewise.
11280 (vect_recog_widen_sum_pattern): Likewise.
11282 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11285 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11287 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11290 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11292 (ix86_output_indirect_function_return): New prototype.
11293 (ix86_split_simple_return_pop_internal): Likewise.
11294 * config/i386/i386.c (indirect_return_via_cx): New.
11295 (indirect_return_via_cx_bnd): Likewise.
11296 (indirect_thunk_name): Handle return va CX_REG.
11297 (output_indirect_thunk_function): Create alias for
11298 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11299 (ix86_output_indirect_jmp): Remove the bool argument.
11300 (ix86_output_indirect_function_return): New function.
11301 (ix86_split_simple_return_pop_internal): Likewise.
11302 * config/i386/i386.md (*indirect_jump): Don't pass false
11303 to ix86_output_indirect_jmp.
11304 (*tablejump_1): Likewise.
11305 (simple_return_pop_internal): Change it to define_insn_and_split.
11306 Call ix86_split_simple_return_pop_internal to split it for
11307 -mfunction-return=.
11308 (simple_return_indirect_internal): Call
11309 ix86_output_indirect_function_return instead of
11310 ix86_output_indirect_jmp.
11312 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11315 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11316 memset and value initialization afterwards.
11318 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11320 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11322 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11325 * common/config/aarch64/aarch64-common.c
11326 (aarch_option_optimization_table[]): Switch
11327 off fomit-frame-pointer
11329 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11330 Chung-Ju Wu <jasonwucj@gmail.com>
11332 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11334 (store_multiple): Ditto.
11336 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11338 * config.gcc: Add --with-cpu support for nds32 target.
11339 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11340 * config/nds32/nds32.opt: Add -mcpu= option.
11342 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11344 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11345 isel=yes): Warn for these deprecated options.
11347 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11349 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11350 ISA_2_5_MASKS_EMBEDDED.
11352 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11354 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11355 p->max as pointers rather than using iterative_hash_expr.
11357 2018-02-23 Carl Love <cel@us.ibm.com>
11359 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11360 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11362 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11363 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11364 P8V_BUILTIN_VEC_VUNSIGNED2.
11366 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11369 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11370 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11372 * lra-constraints.c (curr_insn_transform): Set up
11373 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11375 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11376 * lra-eliminations.c (spill_pseudos): Ditto.
11377 (process_insn_for_elimination): Ditto.
11378 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11379 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11380 LRA_NON_CLOBBERED_ALT.
11382 2018-02-22 Martin Sebor <msebor@redhat.com>
11384 PR tree-optimization/84480
11385 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11386 to maybe_diag_stxncpy_trunc. Call it.
11387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11388 from gimple_fold_builtin_strcpy. Print inlining stack.
11389 (handle_builtin_stxncpy): Print inlining stack.
11390 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11392 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11395 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11396 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11397 and -fcheck-pointer-bounds are used together.
11398 (indirect_thunk_prefix): New enum.
11399 (indirect_thunk_need_prefix): New function.
11400 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11401 "_nt" instead of "_bnd" for NOTRACK prefix.
11402 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11403 (output_indirect_thunk_function): Likewise.
11405 (ix86_code_end): Update output_indirect_thunk_function calls.
11406 (ix86_output_indirect_branch_via_reg): Replace
11407 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11408 (ix86_output_indirect_branch_via_push): Likewise.
11409 (ix86_output_function_return): Likewise.
11410 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11411 incompatible with -fcf-protection=branch and
11412 -fcheck-pointer-bounds.
11414 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11417 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11418 Change gcc_assert call to output_operand_lossage.
11420 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11422 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11424 2018-02-22 DJ Delorie <dj@redhat.com>
11425 Sebastian Perta <sebastian.perta@renesas.com>
11426 Oleg Endo <olegendo@gcc.gnu.org>
11428 * config/rx/rx.c (rx_rtx_costs): New function.
11429 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11431 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11433 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11435 2018-02-22 Martin Liska <mliska@suse.cz>
11438 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11439 Add "native" as a possible value.
11441 2018-02-22 Martin Liska <mliska@suse.cz>
11444 * config/i386/i386.c (ix86_option_override_internal):
11445 Add "native" as a possible value for -march and -mtune.
11447 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11450 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11451 to all type variants.
11453 PR tree-optimization/84503
11454 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11455 width as info->bitpos + info->bitsize - start.
11456 (merged_store_group::merge_overlapping): Simplify width computation.
11457 (check_no_overlap): New function.
11458 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11459 start + width and last_order of the group, fail if check_no_overlap
11461 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11462 to group if check_no_overlap fails.
11464 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11466 * config/rs6000/altivec.md: Delete contraint arguments to
11467 define_expand, define_split, and define_peephole2, and in
11468 define_insn_and_split if always unused.
11469 * config/rs6000/darwin.md: Ditto.
11470 * config/rs6000/dfp.md: Ditto.
11471 * config/rs6000/rs6000.md: Ditto.
11472 * config/rs6000/sync.md: Ditto.
11473 * config/rs6000/vector.md: Ditto.
11474 * config/rs6000/vsx.md: Ditto.
11476 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11478 * config/rs6000/altivec.md: Write output control strings as braced
11479 blocks instead of double-quoted strings.
11480 * config/rs6000/darwin.md: Ditto.
11481 * config/rs6000/rs6000.md: Ditto.
11482 * config/rs6000/vector.md: Ditto.
11483 * config/rs6000/vsx.md: Ditto.
11485 2018-02-21 Jason Merrill <jason@redhat.com>
11487 PR c++/84314 - ICE with templates and fastcall attribute.
11488 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11490 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11492 * ipa-cp.c (determine_versionability): Fix comment typos.
11494 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11497 * ipa-cp.c (determine_versionability): Do not version functions caling
11500 2018-02-21 Martin Liska <mliska@suse.cz>
11503 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11504 Add "native" as a possible value.
11505 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11506 the macro when native cpu detection is available.
11508 2018-02-21 Martin Liska <mliska@suse.cz>
11511 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11512 Add "native" as a possible value.
11513 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11514 when native cpu detection is available.
11516 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11517 Martin Sebor <msebor@redhat.com>
11519 PR tree-optimization/84478
11520 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11522 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11523 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11524 support which is conservatively correct, for 2 only stay conservative
11525 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11526 argument to the 2 argument get_range_strlen, adjust 6 arg
11527 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11529 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11530 (gimple_fold_builtin_strlen): Pass true as last argument to
11533 2018-02-20 Martin Sebor <msebor@redhat.com>
11535 PR middle-end/84095
11536 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11537 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11538 (builtin_memref::builtin_memref): Factor out parts into
11539 set_base_and_offset and call it.
11541 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11543 PR middle-end/84406
11544 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11545 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11546 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11547 search at the associated MODE_INT.
11549 2018-02-20 Jeff Law <law@redhat.com>
11551 PR middle-end/82123
11552 PR tree-optimization/81592
11553 PR middle-end/79257
11554 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11555 for range data rather than using global data.
11556 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11557 range data rather than using global data.
11558 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11559 pass it to children as needed.
11560 (struct directive::fmtresult): Similarly.
11561 (struct directive::set_width): Similarly.
11562 (struct directive::set_precision): Similarly.
11563 (format_integer, format_directive, parse_directive): Similarly.
11564 (format_none): Accept unnamed vr_values parameter.
11565 (format_percent, format_floating, format_character): Similarly.
11566 (format_string, format_plain): Similarly.
11567 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11568 the EVRP range analyzer for range data rather than using global data.
11569 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11570 gimple-ssa-evrp-analyze.h
11571 (class sprintf_dom_walker): Add after_dom_children member function.
11572 Add evrp_range_analyzer member.
11573 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11574 range analyzer as needed.
11575 (sprintf_dom_walker::after_dom_children): New member function.
11576 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11578 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11579 (evrp_range_analyzer::pop_to_marker): Likewise.
11581 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11583 PR tree-optimization/84419
11584 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11585 with the required type if its current type is compatible but
11588 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11590 PR middle-end/82004
11591 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11592 after vectorization.
11594 2018-02-20 Martin Liska <mliska@suse.cz>
11597 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11598 possible values if we don't have a hint.
11600 2018-02-20 Martin Liska <mliska@suse.cz>
11604 * final.c (shorten_branches): Build align_tab array with one
11606 * opts.c (finish_options): Add alignment option limit check.
11607 (MAX_CODE_ALIGN): Likewise.
11608 (MAX_CODE_ALIGN_VALUE): Likewise.
11609 * doc/invoke.texi: Document maximum allowed option value for
11610 all -falign-* options.
11612 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11615 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11616 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11617 * var-tracking.c (emit_note_insn_var_location): Remove all references
11618 to NOTE_INSN_CALL_ARG_LOCATION.
11619 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11620 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11621 Use copy_rtx_if_shared.
11622 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11623 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11624 (dwarf2out_var_location): Remove handling of
11625 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11627 * final.c (final_scan_insn): Remove all references to
11628 NOTE_INSN_CALL_ARG_LOCATION.
11629 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11630 before dumping final insns.
11631 * except.c (emit_note_eh_region_end): Remove all references to
11632 NOTE_INSN_CALL_ARG_LOCATION.
11633 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11634 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11635 * config/arc/arc.c (hwloop_optimize): Likewise.
11636 * config/arm/arm.c (create_fix_barrier): Likewise.
11637 * config/s390/s390.c (s390_chunkify_start): Likewise.
11638 * config/sh/sh.c (find_barrier): Likewise.
11639 * config/i386/i386.c (rest_of_insert_endbranch,
11640 ix86_seh_fixup_eh_fallthru): Likewise.
11641 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11642 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11643 * config/frv/frv.c (frv_function_prologue): Likewise.
11644 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11646 (note_outside_basic_block_p): Remove all references to
11647 NOTE_INSN_CALL_ARG_LOCATION.
11648 * gengtype.c (adjust_field_rtx_def): Likewise.
11649 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11651 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11652 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11655 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11658 PR tree-optimization/84452
11659 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11660 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11663 2018-02-19 Martin Liska <mliska@suse.cz>
11666 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11668 2018-02-19 Martin Liska <mliska@suse.cz>
11669 Richard Sandiford <richard.sandiford@linaro.org>
11671 PR tree-optimization/82491
11672 * gimple-fold.c (get_base_constructor): Make earlier bail out
11675 2018-02-19 Carl Love <cel@us.ibm.com>
11677 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11678 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11680 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11681 P8V_BUILTIN_VEC_NEG.
11683 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11685 * config/rl78/rl78.md (movdf): New define expand.
11687 2018-02-19 Martin Liska <mliska@suse.cz>
11690 * doc/invoke.texi: Fix typo.
11691 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11693 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11695 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11696 handle rs6000_single_float and rs6000_double_float specially for
11699 2018-02-16 Jeff Law <law@redhat.com>
11701 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11702 (pop_regs): Use it.
11704 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11707 * ipa-inline.c (inline_small_functions): Fix a typo.
11709 2018-02-16 Nathan Sidwell <nathan@acm.org>
11711 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11713 2018-02-16 Carl Love <cel@us.ibm.com>
11715 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11716 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11717 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11718 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11719 expansion to P8V_BUILTIN_VEC_FLOAT2.
11721 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11723 PR rtl-optimization/70023
11724 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11725 src_regno into account.
11727 2018-02-16 Carl Love <cel@us.ibm.com>
11729 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11730 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11731 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11732 * config/rs6000/rs6000.c: Remove case statements for
11733 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11734 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11735 and P9V_BUILTIN_VEC_VINSERT4B.
11736 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11737 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11738 * config/rs6000/vsx.md:
11739 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11742 2018-02-16 Carl Love <cel@us.ibm.com>
11744 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11746 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11748 * config/rs6000/rs6000-c.c: Add the definitions for
11749 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11750 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11751 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11752 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11753 definition for insert4b and define insn *insert3b_internal.
11754 * doc/extend.texi: Add documentation for vec_extract4b.
11756 2018-02-16 Nathan Sidwell <nathan@acm.org>
11758 * doc/extend.texi (Backwards Compatibility): Mention friend
11759 injection. Note for-scope is deprecated.
11760 * doc/invoke.texi (-ffriend-injection): Deprecate.
11762 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11764 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11765 that moved to I2, also allow destinations that are a paradoxical
11766 subreg (instead of a normal reg).
11768 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11771 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11774 2018-02-16 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/84037
11777 PR tree-optimization/84016
11779 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11780 Adjust vec_construct for the fact we need additional higher latency
11781 128bit inserts for AVX256 and AVX512 vector builds.
11782 (ix86_add_stmt_cost): Scale vector construction cost for
11785 2018-02-16 Richard Biener <rguenther@suse.de>
11787 PR tree-optimization/84417
11788 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11789 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11790 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11792 2018-02-16 Martin Liska <mliska@suse.cz>
11795 * internal-fn.def (ASAN_CHECK): Set proper flags.
11796 (ASAN_MARK): Likewise.
11798 2018-02-16 Julia Koval <julia.koval@intel.com>
11800 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11801 from PTA_CANNONLAKE.
11803 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11806 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11807 Use ++iter rather than iter++ for std::list iterators.
11808 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11809 defer deleting them until all nodes in the forest are processed. Do
11810 free even leaf nodes. Change to_process into auto_vec.
11813 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11814 * vec.h (vec_default_construct): Use memset instead of placement new
11815 if BROKEN_VALUE_INITIALIZATION is defined.
11816 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11817 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11820 PR rtl-optimization/83723
11821 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11822 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11823 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11825 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11827 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11829 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11831 PR rtl-optimization/81443
11832 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11833 from inner REGs to paradoxical SUBREGs.
11835 2018-02-16 Richard Biener <rguenther@suse.de>
11837 PR tree-optimization/84399
11838 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11839 For operands we can analyze at their definition make sure we can
11840 analyze them at each use as well.
11842 2018-02-16 Richard Biener <rguenther@suse.de>
11844 PR tree-optimization/84190
11845 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11846 volatile accesses if the decl isn't volatile.
11848 2018-02-15 Jason Merrill <jason@redhat.com>
11850 PR c++/84314 - ICE with templates and fastcall attribute.
11851 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11852 TYPE_CANONICAL on an existing type.
11854 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11856 PR tree-optimization/84383
11857 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11858 dstoff nor call operand_equal_p if dstbase is NULL.
11860 PR tree-optimization/84334
11861 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11862 also a CONSTANT_CLASS_P, punt.
11864 2018-02-14 Jim Wilson <jimw@sifive.com>
11866 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11867 first SMALL_OPERAND check. New local min_second_step. Move assert
11868 to where locals are set. Add TARGET_RVC support.
11869 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11871 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11873 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11875 2018-02-14 Martin Sebor <msebor@redhat.com>
11877 PR tree-optimization/83698
11878 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11879 arrays constrain the offset range to their bounds.
11880 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11881 (builtin_access::overlap): Avoid setting the size of overlap if it's
11883 (maybe_diag_overlap): Also consider arrays when deciding what values
11884 of offsets to include in diagnostics.
11886 2018-02-14 Martin Sebor <msebor@redhat.com>
11889 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11890 that correspond to the kind of a declaration.
11892 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11895 * config/pa/pa.md: Load address of PIC label using the linkage table
11896 if the label is nonlocal.
11898 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11900 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11901 warning message if user requests -maltivec=be.
11902 * doc/invoke.texi: Document deprecation of -maltivec=be.
11904 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11907 * config/rs6000/rs6000-c.c: Update definitions for
11908 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11909 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11911 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11914 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11915 add _get_ssp intrinsics. Remove argument from
11916 __builtin_ia32_rdssp[d|q].
11917 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11918 * config/i386/i386-builtin.def: Remove argument from
11919 __builtin_ia32_rdssp[d|q].
11920 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11921 ix86_expand_special_args_builtin for _rdssp[d|q].
11922 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11923 Clear register before usage.
11924 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11925 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11927 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11929 PR tree-optimization/84357
11930 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11931 operand 1 of an ARRAY_REF too.
11933 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11936 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11937 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11939 (set_of_reg): New struct.
11940 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11941 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11942 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11944 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11945 Split into bitclr, bitset, bitinvert patterns if appropriate.
11946 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11947 use rx_fuse_in_memory_bitop.
11948 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11949 to named insn, correct maximum insn length.
11951 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11954 * machmode.def: Define a complex mode for PARTIAL_INT.
11955 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11957 * doc/rtl.texi: Document CSPImode.
11958 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11960 (msp430_hard_regno_nregs_with_padding): Likewise.
11962 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11965 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11967 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11969 PR rtl-optimization/84169
11970 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11971 we generated a parallel as new i3 and we split that to new i2 and i3
11972 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11973 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11974 those to i2, not i1. Partially rewrite this scan code.
11976 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11979 * stor-layout.c (place_field): For variable length fields, adjust
11980 offset_align afterwards not just based on the field's alignment,
11981 but also on the size.
11983 PR middle-end/84309
11984 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11985 of exps and logs in the use_exp2 case.
11987 2018-02-13 Jeff Law <law@redhat.com>
11989 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11990 entry for "vector".
11992 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11995 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11999 * common.opt (gas-loc-support, gas-locview-support): New.
12000 (ginline-points, ginternal-reset-location-views): New.
12001 * doc/invoke.texi: Document them. Use @itemx where intended.
12002 (gvariable-location-views): Adjust.
12003 * target.def (reset_location_view): New.
12004 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12005 (TARGET_RESET_LOCATION_VIEW): New.
12006 * doc/tm.texi: Rebuilt.
12007 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12008 (dwarf2out_default_as_locview_support): New.
12009 (output_asm_line_debug_info): Use option variables.
12010 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12011 (output_loc_list): Likewise.
12012 (add_high_low_attributes): Check option variables.
12013 Don't output entry view attribute in strict mode.
12014 (gen_inlined_subroutine_die): Check option variables.
12015 (dwarf2out_inline_entry): Likewise.
12016 (init_sections_and_labels): Likewise.
12017 (dwarf2out_early_finish): Likewise.
12018 (maybe_reset_location_view): New, from...
12019 (dwarf2out_var_location): ... here. Call it.
12020 * debug.h (dwarf2out_default_as_loc_support): Declare.
12021 (dwarf2out_default_as_locview_support): Declare.
12022 * hooks.c (hook_int_rtx_insn_0): New.
12023 * hooks.h (hook_int_rtx_insn_0): Declare.
12024 * toplev.c (process_options): Take -gas-loc-support and
12025 -gas-locview-support from dwarf2out. Enable
12026 -gvariable-location-views by default only with locview
12027 assembler support. Enable -ginternal-reset-location-views by
12028 default only if the target defines the corresponding hook.
12029 Enable -ginline-points by default if location views are
12030 enabled; force it disabled if statement frontiers are
12032 * tree-inline.c (expand_call_inline): Check option variables.
12033 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12035 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12037 PR tree-optimization/84321
12038 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12039 handling. Also check whether the anti-range contains any values
12040 that satisfy the mask; switch to a VR_RANGE if not.
12042 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12045 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12047 2018-02-13 Martin Jambor <mjambor@suse.cz>
12050 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12051 of call statements, also set location of a load to a temporary.
12053 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12055 * config/rl78/rl78.c (add_vector_labels): New function.
12056 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12057 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12058 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12059 which checks that no arguments are passed.
12060 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12061 * doc/extend.texi: Documentation for the new attribute.
12063 2018-02-13 Andreas Schwab <schwab@suse.de>
12065 * config/riscv/linux.h (CPP_SPEC): Define.
12067 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12070 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12071 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12072 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12073 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12074 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12075 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12076 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12077 temporarily for AES and PCLMUL builtins.
12079 PR tree-optimization/84339
12080 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12081 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12084 PR middle-end/84309
12085 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12086 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12087 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12089 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12091 * omp-simd-clone.h: New file.
12092 * omp-simd-clone.c: Include omp-simd-clone.h.
12093 (expand_simd_clones): No longer static.
12094 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12095 cgraph.h and omp-simd-clone.h.
12096 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12097 (vect_recog_widen_shift_pattern): Formatting fix.
12098 (vect_pattern_recog_1): Don't check optab for calls.
12101 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12102 operands[2] into a REG before using gen_lowpart on it.
12104 2018-02-12 Jeff Law <law@redhat.com>
12107 * config/sh/sh.c (find_barrier): Consider a sibling call
12110 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12111 successfully back substituting a reg.
12113 2018-02-12 Richard Biener <rguenther@suse.de>
12115 PR tree-optimization/84037
12116 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12117 parameter, move visited init to caller.
12118 (vect_slp_analyze_operations): Separate cost from validity
12119 check, initialize visited once for all instances.
12120 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12122 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12124 (vect_model_promotion_demotion_cost): Likewise.
12125 (vectorizable_bswap): Guard cost modeling with !slp_node
12126 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12128 (vectorizable_call): Likewise.
12129 (vectorizable_conversion): Likewise.
12130 (vectorizable_assignment): Likewise.
12131 (vectorizable_shift): Likewise.
12132 (vectorizable_operation): Likewise.
12133 (vectorizable_store): Likewise.
12134 (vectorizable_load): Likewise.
12135 (vectorizable_condition): Likewise.
12136 (vectorizable_comparison): Likewise.
12138 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12141 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12142 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12143 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12145 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12147 PR middle-end/83665
12148 * params.def (inline-min-speedup): Increase from 8 to 15.
12149 (max-inline-insns-auto): Decrease from 40 to 30.
12150 * ipa-split.c (consider_split): Add some buffer for function to
12151 be considered inlining candidate.
12152 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12155 2018-02-12 Richard Biener <rguenther@suse.de>
12157 PR tree-optimization/84037
12158 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12159 matched stmts if we cannot swap the non-matched ones.
12161 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12163 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12164 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12165 _mm_maskz_scalef_round_ss): New intrinsics.
12166 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12167 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12168 __builtin_ia32_scalefss_round): Remove.
12169 (__builtin_ia32_scalefsd_mask_round,
12170 __builtin_ia32_scalefss_mask_round): New intrinsics.
12171 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12172 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12173 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12174 "<round_constraint>")): Changed to ...
12175 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12176 "<round_scalar_constraint>")): ... this.
12177 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12178 %0, %1, %2<round_op3>}"): Changed to ...
12179 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12180 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12181 %2<round_scalar_mask_op3>}"): ... this.
12182 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12184 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12186 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12187 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12188 (_mm_maskz_sqrt_round_ss): New intrinsics.
12189 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12190 (__builtin_ia32_sqrtsd_mask_round)
12191 (__builtin_ia32_sqrtss_mask_round): New builtins.
12192 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12193 (__builtin_ia32_sqrtss_round): Remove.
12194 (__builtin_ia32_sqrtsd_mask_round)
12195 (__builtin_ia32_sqrtss_mask_round): New builtins.
12196 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12197 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12198 ((match_operand:VF_128 1 "vector_operand"
12199 "xBm,<round_constraint>")): Changed to ...
12200 ((match_operand:VF_128 1 "vector_operand"
12201 "xBm,<round_scalar_constraint>")): ... this.
12202 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12203 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12204 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12205 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12206 %<iptr>1<round_scalar_mask_op3>}): ... this.
12207 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12208 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12210 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12213 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12214 Cast vec_cmpeq result to correct type.
12215 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12216 Cast vec_cmpgt result to correct type.
12218 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12220 * final.c (final_scan_insn_1): Renamed from...
12221 (final_scan_insn): ... this. New wrapper, to recover
12222 seen from the outermost call in recursive ones.
12223 * config/sparc/sparc.c (output_return): Drop seen from call.
12224 (output_sibcall): Likewise.
12225 * config/visium/visium.c (output_branch): Likewise.
12227 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12229 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12232 2018-02-10 Alan Modra <amodra@gmail.com>
12235 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12236 Specify LR as an input.
12238 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12241 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12242 remove_member_access_dummy_vars): New functions.
12243 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12244 lower_omp_1, execute_lower_omp): Use them.
12246 PR rtl-optimization/84308
12247 * shrink-wrap.c (spread_components): Release todo vector.
12249 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12251 PR rtl-optimization/57193
12252 * ira-color.c (struct allocno_color_data): Add member
12253 conflict_allocno_hard_prefs.
12254 (update_conflict_allocno_hard_prefs): New.
12255 (bucket_allocno_compare_func): Add a preference based on
12256 conflict_allocno_hard_prefs.
12257 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12258 (color_allocnos): Remove a dead code. Initiate
12259 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12261 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12264 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12265 constraint from =wa to wa. Avoid a subreg on the output operand,
12266 instead use a pseudo and subreg it in a move.
12267 (p9_xxbrd_<mode>): Changed to ...
12268 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12269 (p9_xxbrd_v2df): New expander.
12270 (p9_xxbrw_<mode>): Changed to ...
12271 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12272 (p9_xxbrw_v4sf): New expander.
12274 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12276 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12278 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12281 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12282 multiply in 32-bit mode.
12283 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12284 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12287 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12289 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12290 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12291 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12292 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12294 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12297 * dwarf2out.c (is_trivial_indirect_ref): New function.
12298 (dwarf2out_late_global_decl): Do not generate a location
12299 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12300 and that are not defined in the current unit.
12302 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12304 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12305 instead of a libcall for UNORDERED.
12307 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12310 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12311 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12313 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12316 * config/s390/s390.c (s390_set_current_function): Invoke
12317 s390_indirect_branch_settings also if fndecl didn't change.
12319 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12321 * config/rs6000/rs6000.md (blockage): Set length to zero.
12323 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12325 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12327 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12330 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12331 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12335 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12336 PARALLEL incoming that failed vt_get_decl_and_offset check.
12338 PR middle-end/84237
12339 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12340 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12342 (get_variable_section): For decls in named .bss* sections pass true as
12343 second argument to bss_initializer_p.
12345 2018-02-09 Marek Polacek <polacek@redhat.com>
12346 Jakub Jelinek <jakub@redhat.com>
12349 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12350 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12351 Sync some changes from cxx_fold_indirect_ref.
12353 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12355 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12357 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12358 (BLOCK_INLINE_ENTRY_LABEL): New.
12359 (dwarf2out_var_location): Disregard inline entry markers.
12360 (inline_entry_data): New struct.
12361 (inline_entry_data_hasher): New hashtable type.
12362 (inline_entry_data_hasher::hash): New.
12363 (inline_entry_data_hasher::equal): New.
12364 (inline_entry_data_table): New variable.
12365 (add_high_low_attributes): Add DW_AT_entry_pc and
12366 DW_AT_GNU_entry_view attributes if a pending entry is found
12367 in inline_entry_data_table. Add old entry_pc attribute only
12368 if debug nonbinding markers are disabled.
12369 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12370 markers are enabled.
12371 (block_within_block_p, dwarf2out_inline_entry): New.
12372 (dwarf2out_finish): Check that no entries remained in
12373 inline_entry_data_table.
12374 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12375 (final_scan_insn, notice_source_line): Likewise.
12376 (rest_of_clean_state): Skip inline entry markers.
12377 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12379 * gimple.c (gimple_build_debug_inline_entry): New.
12380 * gimple.h (enum gimple_debug_subcode): Add
12381 GIMPLE_DEBUG_INLINE_ENTRY.
12382 (gimple_build_debug_inline_entry): Declare.
12383 (gimple_debug_inline_entry_p): New.
12384 (gimple_debug_nonbind_marker_p): Adjust.
12385 * insn-notes.def (INLINE_ENTRY): New.
12386 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12387 inline entry marker notes.
12388 (print_insn): Likewise.
12389 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12390 (INSN_DEBUG_MARKER_KIND): Likewise.
12391 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12392 * tree-inline.c (expand_call_inline): Build and insert
12393 debug_inline_entry stmt.
12394 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12395 inline entry blocks early, if nonbind markers are enabled.
12396 (dump_scope_block): Dump fragment info.
12397 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12398 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12399 (gimple_build_debug_inline_entry): New.
12400 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12401 Enable/disable inline entry points too.
12402 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12403 (DEBUG_INSN): Describe inline entry markers.
12405 * common.opt (gvariable-location-views): New.
12406 (gvariable-location-views=incompat5): New.
12407 * config.in: Rebuilt.
12408 * configure: Rebuilt.
12409 * configure.ac: Test assembler for view support.
12410 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12411 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12412 * dwarf2out.c (var_loc_view): New typedef.
12413 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12414 (dwarf2out_locviews_in_attribute): New.
12415 (dwarf2out_locviews_in_loclist): New.
12416 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12417 (enum dw_line_info_opcode): Add LI_adv_address.
12418 (struct dw_line_info_table): Add view.
12419 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12420 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12421 (zero_view_p): New variable.
12422 (ZERO_VIEW_P): New macro.
12423 (output_asm_line_debug_info): New.
12424 (struct var_loc_node): Add view.
12425 (add_AT_view_list, AT_loc_list): New.
12426 (add_var_loc_to_decl): Add view param. Test it against last.
12427 (new_loc_list): Add view params. Record them.
12428 (AT_loc_list_ptr): Handle loc and view lists.
12429 (view_list_to_loc_list_val_node): New.
12430 (print_dw_val): Handle dw_val_class_view_list.
12431 (size_of_die): Likewise.
12432 (value_format): Likewise.
12433 (loc_list_has_views): New.
12434 (gen_llsym): Set vl_symbol too.
12435 (maybe_gen_llsym, skip_loc_list_entry): New.
12436 (dwarf2out_maybe_output_loclist_view_pair): New.
12437 (output_loc_list): Output view list or entries too.
12438 (output_view_list_offset): New.
12439 (output_die): Handle dw_val_class_view_list.
12440 (output_dwarf_version): New.
12441 (output_compilation_unit_header): Use it.
12442 (output_skeleton_debug_sections): Likewise.
12443 (output_rnglists, output_line_info): Likewise.
12444 (output_pubnames, output_aranges): Update version comments.
12445 (output_one_line_info_table): Output view numbers in asm comments.
12446 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12447 Call maybe_gen_llsym.
12448 (loc_list_from_tree_1): Adjust.
12449 (add_AT_location_description): Create view list attribute if
12450 needed, check it's absent otherwise.
12451 (convert_cfa_to_fb_loc_list): Adjust.
12452 (maybe_emit_file): Call output_asm_line_debug_info for test.
12453 (dwarf2out_var_location): Reset views as needed. Precompute
12454 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12455 attribute. Set view.
12456 (new_line_info_table): Reset next view.
12457 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12458 (dwarf2out_source_line): Likewise. Output view resets and labels to
12459 the assembler, or select appropriate line info opcodes.
12460 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12461 (optimize_string_length): Catch it. Adjust.
12462 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12463 dw_val_class_view_list, and remove it if no longer needed.
12464 (hash_loc_list): Hash view numbers.
12465 (loc_list_hasher::equal): Compare them.
12466 (optimize_location_lists): Check whether a view list symbol is
12467 needed, and whether the locview attribute is present, and
12468 whether they match. Remove the locview attribute if no longer
12470 (index_location_lists): Call skip_loc_list_entry for test.
12471 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12472 Use output_dwarf_version.
12473 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12474 (struct dw_val_node): Add val_view_list.
12475 * final.c (SEEN_NEXT_VIEW): New.
12476 (set_next_view_needed): New.
12477 (clear_next_view_needed): New.
12478 (maybe_output_next_view): New.
12479 (final_start_function): Rename to...
12480 (final_start_function_1): ... this. Take pointer to FIRST,
12481 add SEEN parameter. Emit param bindings in the initial view.
12482 (final_start_function): Reintroduce SEEN-less interface.
12483 (final): Rename to...
12484 (final_1): ... this. Take SEEN parameter. Output final pending
12485 next view at the end.
12486 (final): Reintroduce seen-less interface.
12487 (final_scan_insn): Output pending next view before switching
12488 sections or ending a block. Mark the next view as needed when
12489 outputting variable locations. Notify debug backend of section
12490 changes, and of location view changes.
12491 (rest_of_handle_final): Adjust.
12492 * toplev.c (process_options): Autodetect value for debug variable
12493 location views option. Warn on incompat5 without -gdwarf-5.
12494 * doc/invoke.texi (gvariable-location-views): New.
12495 (gvariable-location-views=incompat5): New.
12496 (gno-variable-location-views): New.
12498 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12500 PR tree-optimization/84136
12501 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12502 that the result of find_edge is non-NULL.
12504 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12507 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12508 storing integer register in SImode. Fix cost of 256 and 512
12509 byte aligned SSE register store.
12511 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12513 * config/i386/i386.c (ix86_multiplication_cost): Fix
12514 multiplication cost for TARGET_AVX512DQ.
12516 2018-02-08 Marek Polacek <polacek@redhat.com>
12518 PR tree-optimization/84238
12519 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12522 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12524 PR tree-optimization/84265
12525 * tree-vect-stmts.c (vectorizable_store): Don't treat
12526 VMAT_CONTIGUOUS accesses as grouped.
12527 (vectorizable_load): Likewise.
12529 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12531 PR tree-optimization/81635
12532 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12533 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12534 (test_round_for_mask): New functions.
12535 (wide_int_cc_tests): Call test_round_for_mask.
12536 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12537 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12538 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12539 range returned by get_range_info.
12541 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12544 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12545 * symtab.c: Include builtins.h
12546 (symtab_node::output_to_lto_symbol_table_p): Move here
12547 from lto-streamer-out.c:output_symbol_p.
12548 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12549 (output_symbol_p): Move all logic to symtab.c
12550 (produce_symtab): Update.
12552 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12554 * config/s390/s390-opts.h (enum indirect_branch): Define.
12555 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12556 (s390_indirect_branch_via_thunk)
12557 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12558 (enum s390_indirect_branch_type): Define.
12559 * config/s390/s390.c (struct s390_frame_layout, struct
12560 machine_function): Remove.
12561 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12562 (indirect_branch_table_label_no, indirect_branch_table_name):
12564 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12565 (enum s390_indirect_branch_option): Define.
12566 (s390_return_addr_from_memory): New function.
12567 (s390_handle_string_attribute): New function.
12568 (s390_attribute_table): Add new attribute handler.
12569 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12570 (s390_indirect_branch_via_thunk): New function.
12571 (s390_indirect_branch_via_inline_thunk): New function.
12572 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12573 sibling call optimization for non z10 compiles.
12574 (s390_emit_call): Force indirect branch target to be a single
12575 register. Add r1 clobber for non-z10 compiles.
12576 (s390_emit_epilogue): Emit return jump via return_use expander.
12577 (s390_reorg): Handle JUMP_INSNs as execute targets.
12578 (s390_option_override_internal): Perform validity checks for the
12579 new command line options.
12580 (s390_indirect_branch_attrvalue): New function.
12581 (s390_indirect_branch_settings): New function.
12582 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12583 (s390_output_indirect_thunk_function): New function.
12584 (s390_code_end): Implement target hook.
12585 (s390_case_values_threshold): Implement target hook.
12586 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12588 * config/s390/s390.h (struct s390_frame_layout)
12589 (struct machine_function): Move here from s390.c.
12590 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12591 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12592 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12593 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12594 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12595 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12596 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12597 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12598 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12599 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12600 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12601 (mnemonic attribute): Add values which aren't recognized
12603 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12604 pattern for branch conversion. Fix mnemonic attribute.
12605 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12606 indirect branch via thunk if requested.
12607 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12608 ("*indirect_jump"): Disable for branch conversion using out of
12610 ("indirect_jump_via_thunk<mode>_z10")
12611 ("indirect_jump_via_thunk<mode>")
12612 ("indirect_jump_via_inlinethunk<mode>_z10")
12613 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12614 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12615 ("casesi_jump_via_inlinethunk<mode>_z10")
12616 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12617 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12618 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12619 ("*indirect2_jump"): Disable for branch conversion.
12620 ("casesi_jump"): Turn into expander and expand patterns for branch
12622 ("return_use"): New expander.
12623 ("*return"): Emit return via thunk and rename it to ...
12624 ("*return<mode>"): ... this one.
12625 * config/s390/s390.opt: Add new options and and enum for the
12628 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12630 * lra-constraints.c (match_reload): Unconditionally use
12631 gen_lowpart_SUBREG, rather than selecting between that
12632 and equivalent gen_rtx_SUBREG code.
12634 2018-02-08 Richard Biener <rguenther@suse.de>
12636 PR tree-optimization/84233
12637 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12638 changed flag instead of boguously re-using phi_inserted.
12640 2018-02-08 Martin Jambor <mjambor@suse.cz>
12642 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12643 static local variables.
12645 2018-02-08 Richard Biener <rguenther@suse.de>
12647 PR tree-optimization/84278
12648 * tree-vect-stmts.c (vectorizable_store): When looking for
12649 smaller vector types to perform grouped strided loads/stores
12650 make sure the mode is supported by the target.
12651 (vectorizable_load): Likewise.
12653 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12655 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12656 Increase LDP/STP opportunities by adding adjacent callee-saves.
12658 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12660 PR rtl-optimization/84068
12661 PR rtl-optimization/83459
12662 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12664 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12666 PR tree-optimization/84224
12667 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12668 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12669 non-zero arguments.
12671 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12674 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12675 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12676 count, remove one USE.
12678 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12680 * doc/install.texi (Configuration): Document the
12681 --with-long-double-format={ibm,ieee} PowerPC configuration
12685 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12686 Convert from define_expand to be define_insn_and_split. Rework
12687 float/double/_Float128 conversions to QI/HI/SImode to work with
12688 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12689 conversions to QI/HImode types did a store and then a load to
12690 truncate the value. For conversions to VSX registers, don't split
12691 the insn, instead emit the code directly. Use the code iterator
12692 any_fix to combine signed and unsigned conversions.
12693 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12694 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12695 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12696 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12697 (fix_<mode>di2_hw): Likewise.
12698 (fixuns_<mode>di2_hw): Likewise.
12699 (fix_<mode>si2_hw): Likewise.
12700 (fixuns_<mode>si2_hw): Likewise.
12701 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12702 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12703 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12704 fix<uns>_trunc<SFDF:mode>si2_p8.
12705 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12706 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12707 (fix<uns>_<mode>_mem): Likewise.
12708 (fctiw<u>z_<mode>_mem): Likewise.
12709 (fix<uns>_<mode>_mem): Likewise.
12710 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12711 the register allocator from doing a direct move to the GPRs to do
12712 a store, and instead use the ISA 3.0 store byte/half-word from
12713 vector register instruction. For IEEE 128-bit floating point,
12714 also optimize stores of 32-bit ints.
12715 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12717 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12719 * genextract.c (push_pathstr_operand): New function to support
12721 (walk_rtx): Call push_pathstr_operand.
12722 (print_path): Support [a-zA-Z].
12724 2018-02-07 Richard Biener <rguenther@suse.de>
12726 PR tree-optimization/84037
12727 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12728 (cse_and_gimplify_to_preheader): Declare.
12729 (vect_get_place_in_interleaving_chain): Likewise.
12730 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12732 (_loop_vec_info::~_loop_vec_info): Delete it.
12733 (cse_and_gimplify_to_preheader): New function.
12734 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12735 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12736 (vectorizable_load): Likewise. For grouped stores always base
12737 the IV on the first element.
12738 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12739 condition before gimplifying.
12741 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12743 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12744 *DIV_EXPR and *MOD_EXPR.
12746 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12749 * config/i386/i386.c (ix86_option_override_internal): Mask out
12750 the CF_SET bit when checking -fcf-protection.
12752 2018-02-07 Tom de Vries <tom@codesourcery.com>
12755 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12758 2018-02-07 Richard Biener <rguenther@suse.de>
12760 PR tree-optimization/84204
12761 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12764 PR tree-optimization/84205
12765 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12766 special-case isl_ast_op_zdiv_r.
12768 PR tree-optimization/84223
12769 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12770 Only add conditions from within the region.
12771 (gather_bbs::after_dom_children): Adjust.
12773 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12776 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12777 * config/avr/avr.md: Only post-reload split REG-REG moves if
12778 either register is GENERAL_REG_P.
12780 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12782 PR tree-optimization/84235
12783 * tree-ssa-scopedtables.c
12784 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12785 if the subtraction is performed in floating point type where NaNs are
12786 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12787 build 1. Formatting fix.
12789 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12792 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12793 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12794 and skip it regardless of bb boundaries. Use CALL_P macro,
12795 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12796 unnecessarily, formatting fix.
12798 2018-02-06 Michael Collison <michael.collison@arm.com>
12800 * config/arm/thumb2.md:
12801 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12802 (*thumb_mov_notscc): Ditto.
12804 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12807 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12808 unsigned_fix, not "s".
12810 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12812 * configure.ac (gcc_fn_eh_frame_ro): New function.
12813 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12814 correct .eh_frame permissions.
12815 * configure: Regenerate.
12817 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12819 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12820 irrelevant options.
12822 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12824 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12825 Display warning message for -mno-speculate-indirect-jumps.
12827 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12829 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12831 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12833 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12835 PR tree-optimization/84225
12836 * tree-eh.c (find_trapping_overflow): Only call
12837 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12839 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12842 * config/i386/i386.c: Reimplement the check of possible options
12843 -mibt/-mshstk conbination. Change error messages.
12844 * doc/invoke.texi: Fix a typo: remove extra '='.
12846 2018-02-06 Marek Polacek <polacek@redhat.com>
12848 PR tree-optimization/84228
12849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12851 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12854 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12855 emitted arch directives.
12856 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12857 __ARM_FEATURE_COPROC before changing architectures.
12859 2018-02-06 Richard Biener <rguenther@suse.de>
12861 * config/i386/i386.c (print_reg): Fix typo.
12862 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12864 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12866 * configure: Regenerate.
12868 2018-02-05 Martin Sebor <msebor@redhat.com>
12870 PR tree-optimization/83369
12871 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12874 2018-02-05 Martin Liska <mliska@suse.cz>
12876 * doc/invoke.texi: Cherry-pick upstream r323995.
12878 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12880 * ira.c (ira_init_register_move_cost): Adjust comment.
12882 2018-02-05 Martin Liska <mliska@suse.cz>
12884 PR gcov-profile/84137
12885 * doc/gcov.texi: Fix typo in documentation.
12887 2018-02-05 Martin Liska <mliska@suse.cz>
12889 PR gcov-profile/83879
12890 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12891 using dlopen functionality.
12893 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12895 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12896 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12897 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12898 _mm_maskz_range_round_ss): New intrinsics.
12899 (__builtin_ia32_rangesd128_round)
12900 (__builtin_ia32_rangess128_round): Remove.
12901 (__builtin_ia32_rangesd128_mask_round,
12902 __builtin_ia32_rangess128_mask_round): New builtins.
12903 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12904 __builtin_ia32_rangess128_round): Remove.
12905 (__builtin_ia32_rangesd128_mask_round,
12906 __builtin_ia32_rangess128_mask_round): New builtins.
12907 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12908 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12909 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12910 "<round_saeonly_constraint>")): Changed to ...
12911 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12912 "<round_saeonly_scalar_constraint>")): ... this.
12913 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12914 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12915 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12916 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12917 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12919 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12921 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12923 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12924 Remove all values except native, 8540 and 8548.
12926 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12928 * config/i386/i386.c (ix86_output_function_return): Pass
12929 INVALID_REGNUM, instead of -1, as invalid register number to
12930 indirect_thunk_name and output_indirect_thunk.
12932 2018-02-02 Julia Koval <julia.koval@intel.com>
12934 * config.gcc: Add -march=icelake.
12935 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12936 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12937 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12938 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12939 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12940 (processor_target_table): Add icelake.
12941 (ix86_option_override_internal): Handle new PTAs.
12942 (get_builtin_code_for_version): Handle icelake.
12943 (M_INTEL_COREI7_ICELAKE): New.
12944 (fold_builtin_cpu): Handle icelake.
12945 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12946 * doc/invoke.texi: Add -march=icelake.
12948 2018-02-02 Julia Koval <julia.koval@intel.com>
12950 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12951 to wide_int_bitmask.
12952 * wide-int-bitmask.h: New.
12954 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12957 * config/i386/i386.md: Replace Pmode with word_mode in
12958 builtin_setjmp_setup and builtin_longjmp to support x32.
12960 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12964 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12966 (rs6000_supported_cpu_names): New static variable.
12967 (linux_cpu_translation_table): Likewise.
12968 (elf_platform) <cpu>: Define new static variable and use it.
12969 Translate kernel AT_PLATFORM name to canonical name if needed.
12970 Error if platform name is unknown.
12972 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12975 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12977 2018-02-01 Jeff Law <law@redhat.com>
12980 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12981 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12982 the scratch if RELEASE_VIA_POP is false.
12983 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12984 If we have to save a temporary register, decrement SIZE appropriately.
12985 Pass new arguments to release_scratch_register_on_entry.
12986 (ix86_adjust_stack_and_probe): Likewise.
12987 (ix86_emit_probe_stack_range): Pass new arguments to
12988 release_scratch_register_on_entry.
12990 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12992 PR rtl-optimization/84157
12993 * combine.c (change_zero_ext): Use REG_P predicate in
12994 front of HARD_REGISTER_P predicate.
12996 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12998 * config/avr/avr.c (avr_option_override): Move disabling of
12999 -fdelete-null-pointer-checks to...
13000 * common/config/avr/avr-common.c (avr_option_optimization_table):
13003 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13005 PR tree-optimization/81635
13006 * tree-data-ref.c (split_constant_offset_1): For types that
13007 wrap on overflow, try to use range info to prove that wrapping
13010 2018-02-01 Renlin Li <renlin.li@arm.com>
13013 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13014 TAILCALL_ADDR_REGS.
13015 (aarch64_register_move_cost): Likewise.
13016 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13017 TAILCALL_ADDR_REGS.
13018 (REG_CLASS_NAMES): Likewise.
13019 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13020 TAILCALL_ADDR_REGS. Remove IP registers.
13021 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13023 2018-02-01 Richard Biener <rguenther@suse.de>
13025 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13026 for specifying RPO order and allow NULL for that.
13027 * domwalk.c (dom_walker::dom_walker): Likewise.
13028 (dom_walker::walk): Handle NULL RPO order.
13029 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13031 (rewrite_update_dom_walker): Likewise.
13032 (mark_def_dom_walker): Likewise.
13034 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13036 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13037 (aarch64_maybe_expand_sve_subreg_move): Declare.
13038 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13039 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13041 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13042 that are semantically a reverse operation.
13043 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13044 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13045 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13047 (aarch64_can_change_mode_class): For big-endian, forbid changes
13048 between two SVE modes if they have different element sizes.
13050 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13052 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13053 the TImode handling for big-endian targets.
13055 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13057 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13058 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13060 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13061 Remove BSWAP handing for big-endian targets and use the form of
13062 LD1RQ appropariate for the mode.
13064 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13066 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13067 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13068 duplicated element.
13070 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13073 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13074 check for operands that need to go through aarch64_sve_reload_be.
13076 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13078 PR tree-optimization/81661
13079 PR tree-optimization/84117
13080 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13081 * tree-eh.c: Include gimplify.h.
13082 (find_trapping_overflow, replace_trapping_overflow,
13083 rewrite_to_non_trapping_overflow): New functions.
13084 * tree-vect-loop.c: Include tree-eh.h.
13085 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13086 * tree-data-ref.c: Include tree-eh.h.
13087 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13089 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13091 PR rtl-optimization/84123
13092 * combine.c (change_zero_ext): Check if hard register satisfies
13093 can_change_dest_mode before calling gen_lowpart_SUBREG.
13095 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13098 * ira.c (ira_init_register_move_cost): Remove assert.
13100 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13102 PR rtl-optimization/84071
13103 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13104 * doc/tm.texi: Regenerate.
13106 2018-01-31 Richard Biener <rguenther@suse.de>
13108 PR tree-optimization/84132
13109 * tree-data-ref.c (analyze_miv_subscript): Properly
13110 check whether evolution_function_is_affine_multivariate_p
13111 before calling gcd_of_steps_may_divide_p.
13113 2018-01-31 Julia Koval <julia.koval@intel.com>
13116 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13117 * config/i386/i386.md (rdpid_rex64) New.
13118 (rdpid): Make 32bit only.
13120 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13123 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13124 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13126 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13129 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13131 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13133 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13135 PR rtl-optimization/84071
13136 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13137 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13139 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13141 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13142 (arc_attribute_table): Add 'aux' attribute.
13143 (arc_in_small_data_p): Consider aux like variables.
13144 (arc_is_aux_reg_p): New function.
13145 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13146 (arc_get_aux_arg): New function.
13147 (prepare_move_operands): Handle aux-register access.
13148 (arc_handle_aux_attribute): New function.
13149 * doc/extend.texi (ARC Variable attributes): Add subsection.
13151 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13153 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13154 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13155 (arc_attribute_table): Add 'uncached' attribute.
13156 (arc_print_operand): Print '.di' flag for uncached memory
13158 (arc_in_small_data_p): Do not consider for small data the uncached
13160 (arc_is_uncached_mem_p): New function.
13161 * config/arc/predicates.md (compact_store_memory_operand): Check
13162 for uncached memory accesses.
13163 (nonvol_nonimm_operand): Likewise.
13164 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13166 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13169 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13170 falign-loops=): Add Optimization flag.
13172 2018-01-30 Jeff Law <law@redhat.com>
13175 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13176 INT_REGISTERS_SAVED. Check it prior to calling
13177 get_scratch_register_on_entry.
13178 (ix86_adjust_stack_and_probe): Similarly.
13179 (ix86_emit_probe_stack_range): Similarly.
13180 (ix86_expand_prologue): Corresponding changes.
13182 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13185 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13186 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13188 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13191 * lra-constraints.c (curr_insn_transform): Process AND in the
13194 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13196 PR rtl-optimization/83986
13197 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13198 dependence against last_pending_memory_flush in addition to
13199 pending_jump_insns.
13201 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13203 PR tree-optimization/81611
13204 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13207 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13210 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13213 2018-01-30 Richard Biener <rguenther@suse.de>
13214 Jakub Jelinek <jakub@redhat.com>
13216 PR tree-optimization/84111
13217 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13218 inner loops added during recursion, as they don't have up-to-date
13221 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13224 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13225 (can_inline_edge_by_limits_p): ... here.
13226 (can_early_inline_edge_p, check_callers,
13227 update_caller_keys, update_callee_keys, recursive_inlining,
13228 add_new_edges_to_heap, speculation_useful_p,
13229 inline_small_functions,
13230 inline_small_functions, flatten_function,
13231 inline_to_all_callers_1): Update.
13233 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13235 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13238 2018-01-30 Richard Biener <rguenther@suse.de>
13240 PR tree-optimization/83008
13241 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13242 invariant and constant vector uses in stmts when they need
13243 more than one stmt.
13245 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13248 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13249 * configure: Regenerate.
13251 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13253 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13255 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13256 Use gen_rtx_REG rather than gen_lowpart.
13258 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13260 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13261 rather than 0 when creating partial subregs.
13263 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13265 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13268 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13271 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13272 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13273 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13274 flags. This restores the settings used before the 2017-07-24.
13275 Turning off pre increment/decrement/modify allows IVOPTS to
13276 optimize DF/SF loops where the index is an int.
13278 2018-01-29 Richard Biener <rguenther@suse.de>
13279 Kelvin Nilsen <kelvin@gcc.gnu.org>
13282 * tree-vect-stmts.c (vectorizable_call): Don't call
13283 targetm.vectorize_builtin_md_vectorized_function if callee is
13286 2018-01-22 Carl Love <cel@us.ibm.com>
13288 * doc/extend.tex: Fix typo in second arg in
13289 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13291 2018-01-29 Richard Biener <rguenther@suse.de>
13293 PR tree-optimization/84086
13294 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13295 (flush_ssaname_freelist): When SSA names were released reset
13296 the SCEV hash table.
13298 2018-01-29 Richard Biener <rguenther@suse.de>
13300 PR tree-optimization/84057
13301 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13302 removed paths when removing edges.
13304 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13306 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13307 -mfunction-return=@var{choice}.
13309 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13311 PR diagnostic/84034
13312 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13313 Handle CR like TAB.
13314 (layout::print_source_line): Likewise.
13315 (test_get_line_width_without_trailing_whitespace): Add test cases.
13317 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13319 PR middle-end/84040
13320 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13323 2018-01-26 Jim Wilson <jimw@sifive.com>
13325 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13327 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13330 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13332 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13333 and CMP + SUB-immediate -> SUBS.
13335 2018-01-26 Martin Sebor <msebor@redhat.com>
13337 PR tree-optimization/83896
13338 * tree-ssa-strlen.c (get_string_len): Rename...
13339 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13340 Avoid assuming length is constant.
13341 (handle_char_store): Use HOST_WIDE_INT for string length.
13343 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13346 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13347 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13349 2018-01-26 Richard Biener <rguenther@suse.de>
13351 PR rtl-optimization/84003
13352 * dse.c (record_store): Only record redundant stores when
13353 the earlier store aliases at least all accesses the later one does.
13355 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13357 PR rtl-optimization/83985
13358 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13359 REG_CFA_RESTORE insns.
13360 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13361 REG_CFA_RESTORE insns here.
13364 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13365 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13367 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13369 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13370 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13371 (arc_init): Likewise.
13372 (arc_override_options): Likewise.
13373 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13375 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13377 * config/arc/arc.h (TARGET_DBNZ): Define.
13378 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13379 properly set the tune attribute.
13380 (dbnz): Use TARGET_DBNZ guard.
13381 * config/arc/arc.opt (mtune): Add core3 option.
13383 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13385 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13386 recognize new pic like addresses.
13387 (arc_delegitimize_address): Clean up.
13389 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13391 * config/arc/arc-arches.def: Option mrf16 valid for all
13393 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13394 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13395 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13396 * config/arc/arc-tables.opt: Regenerate.
13397 * config/arc/arc.c (arc_conditional_register_usage): Handle
13398 reduced register file case.
13399 (arc_file_start): Set must have build attributes.
13400 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13401 mrf16 option value.
13402 * config/arc/arc.opt (mrf16): Add new option.
13403 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13404 * config/arc/genmultilib.awk: Handle new mrf16 option.
13405 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13406 * config/arc/t-multilib: Regenerate.
13407 * doc/invoke.texi (ARC Options): Document mrf16 option.
13409 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13411 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13412 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13413 (arc_attribute_table): Add 'secure_call' attribute.
13414 (arc_print_operand): Print secure call operand.
13415 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13417 (arc_is_secure_call_p): New function. * config/arc/arc.md
13418 (call_i): Add support for sjli instruction.
13419 (call_value_i): Likewise.
13420 * config/arc/constraints.md (Csc): New constraint.
13422 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13423 John Eric Martin <John.Martin@emmicro-us.com>
13425 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13426 * config/arc/arc.c (_arc_jli_section): New struct.
13427 (arc_jli_section): New type.
13428 (rc_jli_sections): New static variable.
13429 (arc_handle_jli_attribute): New function.
13430 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13431 (arc_file_end): New function.
13432 (TARGET_ASM_FILE_END): Define.
13433 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13434 (arc_add_jli_section): New function.
13435 (jli_call_scan): Likewise.
13436 (arc_reorg): Call jli_call_scan.
13437 (arc_output_addsi): Remove 'S' from printing asm operand.
13438 (arc_is_jli_call_p): New function.
13439 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13441 (movhi_insn): Likewise.
13442 (movsi_insn): Likewise.
13443 (movsi_set_cc_insn): Likewise.
13444 (loadqi_update): Likewise.
13445 (load_zeroextendqisi_update): Likewise.
13446 (load_signextendqisi_update): Likewise.
13447 (loadhi_update): Likewise.
13448 (load_zeroextendhisi_update): Likewise.
13449 (load_signextendhisi_update): Likewise.
13450 (loadsi_update): Likewise.
13451 (loadsf_update): Likewise.
13452 (movsicc_insn): Likewise.
13453 (bset_insn): Likewise.
13454 (bxor_insn): Likewise.
13455 (bclr_insn): Likewise.
13456 (bmsk_insn): Likewise.
13457 (bicsi3_insn): Likewise.
13458 (cmpsi_cc_c_insn): Likewise.
13459 (movsi_ne): Likewise.
13460 (movsi_cond_exec): Likewise.
13461 (clrsbsi2): Likewise.
13462 (norm_f): Likewise.
13472 (call_i): Remove 'S' asm letter, add jli instruction.
13473 (call_value_i): Likewise.
13474 * config/arc/arc.op (mjli-always): New option.
13475 * config/arc/constraints.md (Cji): New constraint.
13476 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13478 (subsf3_fpx): Likewise.
13479 (mulsf3_fpx): Likewise.
13480 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13482 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13483 function attrbutes.
13484 * doc/invoke.texi (ARC): Document mjli-always option.
13486 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13488 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13489 avoid addition with 0 and use incw and decw where possible.
13491 2018-01-26 Richard Biener <rguenther@suse.de>
13493 PR tree-optimization/81082
13494 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13495 association if it requires casting to unsigned.
13496 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13497 from fold_plusminus_mult_expr to catch important cases late when
13498 range info is available.
13500 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13502 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13503 * configure.ac (hidden_linkonce): New test.
13504 * configure: Regenerate.
13505 * config.in: Regenerate.
13507 2018-01-26 Julia Koval <julia.koval@intel.com>
13509 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13510 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13511 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13512 _mm_mask_bitshuffle_epi64_mask): Fix type.
13513 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13514 USI_FTYPE_V4DI_V4DI_USI): Remove.
13515 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13516 __builtin_ia32_vpshufbitqmb256_mask,
13517 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13518 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13519 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13521 2018-01-26 Alan Modra <amodra@gmail.com>
13524 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13525 UNSPEC_VBPERMQ. Sort other unspecs.
13527 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13529 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13531 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13533 PR middle-end/83055
13534 * predict.c (drop_profile): Do not push/pop cfun; update also
13536 (handle_missing_profiles): Fix logic looking for zero profiles.
13538 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13540 PR middle-end/83977
13541 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13542 on functions with #pragma omp declare simd or functions with simd
13544 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13545 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13546 Remove trailing \n from warning_at calls.
13548 2018-01-25 Tom de Vries <tom@codesourcery.com>
13551 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13552 for neutered workers.
13554 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13557 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13558 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13560 2018-01-24 Jeff Law <law@redhat.com>
13563 * i386.c (get_probe_interval): Move to earlier point.
13564 (ix86_compute_frame_layout): If -fstack-clash-protection and
13565 the frame is larger than the probe interval, then use pushes
13566 to save registers rather than reg->mem moves.
13567 (ix86_expand_prologue): Remove conditional for int_registers_saved
13570 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13573 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13574 min/max for never referenced object.
13576 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13578 PR middle-end/83977
13579 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13581 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13582 attributes from DECL_ATTRIBUTES (decl) without affecting
13583 DECL_ATTRIBUTES (current_function_decl).
13584 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13585 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13587 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13589 PR tree-optimization/83979
13590 * fold-const.c (fold_comparison): Use constant_boolean_node
13591 instead of boolean_{true,false}_node.
13593 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13595 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13598 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13600 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13601 Simplify the clause that sets the length attribute.
13602 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13603 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13604 clause that sets the length attribute.
13605 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13607 2018-01-24 Tom de Vries <tom@codesourcery.com>
13610 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13611 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13612 Add strict parameter.
13613 (prevent_branch_around_nothing): Insert dummy insn between branch to
13614 label and label with no ptx insn inbetween.
13615 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13617 2018-01-24 Tom de Vries <tom@codesourcery.com>
13620 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13621 for neutered threads in warp.
13622 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13624 2018-01-24 Richard Biener <rguenther@suse.de>
13626 PR tree-optimization/83176
13627 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13630 2018-01-24 Richard Biener <rguenther@suse.de>
13632 PR tree-optimization/82819
13633 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13634 code generating pluses that are no-ops in the target precision.
13636 2018-01-24 Richard Biener <rguenther@suse.de>
13638 PR middle-end/84000
13639 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13641 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13643 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13644 to merge probabilities.
13645 * predict.c (probably_never_executed): Also mark as cold functions
13646 with global 0 profile and guessed local profile.
13647 * profile-count.c (profile_probability::combine_with_count): New
13649 * profile-count.h (profile_probability::operator*,
13650 profile_probability::operator*=, profile_probability::operator/,
13651 profile_probability::operator/=): Reduce precision to adjusted
13652 and set value to guessed on contradictory divisions.
13653 (profile_probability::combine_with_freq): Remove.
13654 (profile_probability::combine_wiht_count): Declare.
13655 (profile_count::force_nonzero):: Set to adjusted.
13656 (profile_count::probability_in):: Set quality to adjusted.
13657 * tree-ssa-tail-merge.c (replace_block_by): Use
13658 combine_with_count.
13660 2018-01-23 Andrew Waterman <andrew@sifive.com>
13661 Jim Wilson <jimw@sifive.com>
13663 * config/riscv/riscv.c (riscv_stack_boundary): New.
13664 (riscv_option_override): Set riscv_stack_boundary. Handle
13665 riscv_preferred_stack_boundary_arg.
13666 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13667 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13668 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13669 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13670 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13671 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13673 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13676 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13677 of struct ix86_frame.
13678 (ix86_expand_epilogue): Likewise. Add a local variable for
13679 the reg_save_offset field in struct ix86_frame.
13681 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13683 PR tree-optimization/82604
13684 * tree-loop-distribution.c (enum partition_kind): New enum item
13685 PKIND_PARTIAL_MEMSET.
13686 (partition_builtin_p): Support above new enum item.
13687 (generate_code_for_partition): Ditto.
13688 (compute_access_range): Differentiate cases that equality can be
13689 proven at all loops, the innermost loops or no loops.
13690 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13691 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13692 (finalize_partitions, distribute_loop): Don't fuse partition of
13693 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13694 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13695 parloop is enabled.
13697 2018-01-23 Martin Liska <mliska@suse.cz>
13699 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13700 order to ignore the predictor.
13701 (PRED_POLYMORPHIC_CALL): Likewise.
13702 (PRED_RECURSIVE_CALL): Likewise.
13704 2018-01-23 Martin Liska <mliska@suse.cz>
13706 * tree-profile.c (tree_profiling): Print function header to
13707 aware reader which function we are working on.
13708 * value-prof.c (gimple_find_values_to_profile): Do not print
13709 not interesting value histograms.
13711 2018-01-23 Martin Liska <mliska@suse.cz>
13713 * profile-count.h (enum profile_quality): Add
13714 profile_uninitialized as the first value. Do not number values
13715 as they are zero based.
13716 (profile_count::verify): Update sanity check.
13717 (profile_probability::verify): Likewise.
13719 2018-01-23 Nathan Sidwell <nathan@acm.org>
13721 * doc/invoke.texi (ffor-scope): Deprecate.
13723 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13725 PR tree-optimization/83510
13726 * domwalk.c (set_all_edges_as_executable): New function.
13727 (dom_walker::dom_walker): Convert bool param
13728 "skip_unreachable_blocks" to enum reachability. Move setup of
13729 edge flags to set_all_edges_as_executable and only do it when
13730 reachability is REACHABLE_BLOCKS.
13731 * domwalk.h (enum dom_walker::reachability): New enum.
13732 (dom_walker::dom_walker): Convert bool param
13733 "skip_unreachable_blocks" to enum reachability.
13734 (set_all_edges_as_executable): New decl.
13735 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13736 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13738 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13739 but converting true to REACHABLE_BLOCKS.
13740 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13742 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13743 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13744 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13746 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13747 if check_all_array_refs will be called.
13749 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13751 * tree.c (selftest::test_location_wrappers): Add more test
13754 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13756 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13757 (selftest::test_bit_in_range): Likewise.
13759 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13762 * doc/sourcebuild.texi (vect_float): Say that the selector
13763 only describes the situation when -funsafe-math-optimizations is on.
13764 (vect_float_strict): Document.
13766 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13768 PR tree-optimization/83965
13769 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13770 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13771 instead of checking only for a reduction.
13772 (vect_recog_widen_sum_pattern): Likewise.
13774 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13776 * predict.c (probably_never_executed): Only use precise profile info.
13777 (compute_function_frequency): Skip after inlining hack since we now
13778 have quality checking.
13780 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13782 * profile-count.h (profile_probability::very_unlikely,
13783 profile_probability::unlikely, profile_probability::even): Set
13784 precision to guessed.
13786 2018-01-23 Richard Biener <rguenther@suse.de>
13788 PR tree-optimization/83963
13789 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13790 Properly terminate dominator walk when crossing the exit edge not
13791 when visiting its source block.
13793 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13796 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13797 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13799 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13801 PR tree-optimization/83957
13802 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13803 semicolon after for body surrounded by braces.
13805 PR tree-optimization/83081
13806 * profile-count.h (profile_probability::split): New method.
13807 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13808 Use profile_probability::split.
13809 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13810 when splitting a single conditional jump into 2.
13812 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13814 PR tree-optimization/69452
13815 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13818 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13820 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13821 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13822 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13824 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13826 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13828 * config/rl78/rl78.md (movdi): New define_expand.
13829 * config/rl78/rl78.c (rl78_split_movdi): New function.
13831 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13834 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13836 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13837 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13838 128-bit to produce an UNSPEC move to get the double word with the
13839 signbit and then a shift directly to do signbit.
13840 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13841 implementation with a new version that just does either a direct
13842 move or a regular move. Move memory interface to separate insns.
13843 Move insns so they are next to the expander.
13844 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13845 with signbit move. Split big and little endian case.
13846 (signbit<mode>2_dm_mem_le): Likewise.
13847 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13848 (signbit<mode>2_dm2): Likewise.
13850 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13852 * config/rl78/rl78.md (anddi3): New define_expand.
13854 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13856 * config/rl78/rl78.md (umindi3): New define_expand.
13858 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13860 * config/rl78/rl78.md (smindi3): New define_expand.
13862 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13864 * config/rl78/rl78.md (smaxdi3): New define_expand.
13866 2018-01-22 Carl Love <cel@us.ibm.com>
13868 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13869 LVX_V1TI): Add macro expansion.
13870 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13871 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13872 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13873 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13874 Change check to determine if the instruction is a byte reversing
13875 entry. Fix typo in comment.
13876 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13877 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13878 Add def_builtin calls for new builtins.
13879 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13880 Add define_insn expansion.
13882 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13884 * config/rl78/rl78.md (umaxdi3): New define_expand.
13886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13888 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13889 for non-QImode registers.
13891 2018-01-22 Richard Biener <rguenther@suse.de>
13893 PR tree-optimization/83963
13894 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13895 including the loop exit block.
13896 (scop_detection::merge_sese): Likewise.
13897 (scop_detection::add_scop): Do it here instead.
13899 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13901 * doc/sourcebuild.texi (arm_softfloat): Document.
13903 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13906 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13907 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13908 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13910 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13911 David Edelsohn <dje.gcc@gmail.com>
13914 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13915 Change "crset eq" to "crset 2".
13916 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13917 (*call_indirect_aix<mode>_nospec): Likewise.
13918 (*call_value_indirect_aix<mode>_nospec): Likewise.
13919 (*call_indirect_elfv2<mode>_nospec): Likewise.
13920 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13921 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13922 change assembly output from . to $.
13923 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13924 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13925 (*tablejump<mode>_internal1_nospec): Likewise.
13927 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13930 * config/sh/sh_optimize_sett_clrt.cc:
13931 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13933 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13935 PR tree-optimization/83940
13936 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13937 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13938 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13939 use it to pass back the definition type.
13940 (vect_check_store_rhs): Likewise rhs_dt_out.
13941 (vect_build_gather_load_calls): Add a mask_dt argument and use
13942 it instead of a call to vect_is_simple_use.
13943 (vectorizable_store): Update calls to vect_check_load_store_mask
13944 and vect_check_store_rhs. Use the dt returned by the latter instead
13945 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13946 instead of calls to vect_is_simple_use. Pass the scalar rather
13947 than the vector operand to vect_is_simple_use when handling
13948 second and subsequent copies of an rhs value.
13949 (vectorizable_load): Update calls to vect_check_load_store_mask
13950 and vect_build_gather_load_calls. Use the cached mask_dt and
13951 gs_info.offset_dt instead of calls to vect_is_simple_use.
13953 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13955 PR middle-end/83945
13956 * tree-emutls.c: Include gimplify.h.
13957 (lower_emutls_2): New function.
13958 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13959 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13960 it before further processing.
13963 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13964 UINTVAL (trueop1) instead of INTVAL (op1).
13966 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13970 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13971 INCOMING_FRAME_SP_OFFSET if not defined.
13972 (scan_trace): Add ENTRY argument. If true and
13973 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13974 emit a note to adjust the CFA offset.
13975 (create_cfi_notes): Adjust scan_trace callers.
13976 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13977 INCOMING_FRAME_SP_OFFSET in the CIE.
13978 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13979 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13981 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13982 * doc/tm.texi: Regenerated.
13984 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13986 PR rtl-optimization/83147
13987 * lra-constraints.c (remove_inheritance_pseudos): Use
13988 lra_substitute_pseudo_within_insn.
13990 2018-01-19 Tom de Vries <tom@codesourcery.com>
13991 Cesar Philippidis <cesar@codesourcery.com>
13994 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13996 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13999 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14000 spaces for function labels.
14002 2018-01-19 Martin Liska <mliska@suse.cz>
14004 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14005 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14006 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14007 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14008 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14009 (PRED_CONST_RETURN): Change from 69 to 65.
14010 (PRED_NULL_RETURN): Change from 91 to 71.
14011 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14012 (PRED_LOOP_GUARD): Change from 66 to 73.
14014 2018-01-19 Martin Liska <mliska@suse.cz>
14016 * predict.c (predict_insn_def): Add new assert.
14017 (struct branch_predictor): Change type to signed integer.
14018 (test_prediction_value_range): Amend test to cover
14019 PROB_UNINITIALIZED.
14020 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14021 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14022 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14023 (PRED_LOOP_IV_COMPARE): Likewise.
14024 * predict.h (PROB_UNINITIALIZED): Define new constant.
14026 2018-01-19 Martin Liska <mliska@suse.cz>
14028 * predict.c (dump_prediction): Add new format for
14029 analyze_brprob.py script which is enabled with -details
14031 * profile-count.h (precise_p): New function.
14033 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14035 PR tree-optimization/83922
14036 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14037 there are no statements that need masking.
14038 (vect_active_double_reduction_p): New function.
14039 (vect_analyze_loop_operations): Use it when handling phis that
14040 are not in the loop header.
14042 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14044 PR tree-optimization/83914
14045 * tree-vect-loop.c (vectorizable_induction): Don't convert
14046 init_expr or apply the peeling adjustment for inductions
14047 that are nested within the vectorized loop.
14049 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14051 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14054 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14058 * function.h (gimplify_parameters): Add gimple_seq * argument.
14059 * function.c: Include gimple.h and options.h.
14060 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14061 for the added local temporaries if needed.
14062 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14063 if there are any parameter cleanups, wrap whole body into a
14064 try/finally with the cleanups.
14066 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14069 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14070 Use GET_MODE_CLASS for scalar floating point.
14072 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14076 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14077 Fix call of call_cgraph_insertion_hooks.
14079 2018-01-18 Martin Sebor <msebor@redhat.com>
14081 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14083 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14086 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14089 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14092 * common.opt: (-ffile-prefix-map): New option.
14093 * opts.c (common_handle_option): Defer it.
14094 * opts-global.c (handle_common_deferred_options): Handle it.
14095 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14096 * file-prefix-map.h: New file.
14097 (remap_debug_filename, add_debug_prefix_map): ...here.
14098 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14099 * final.c (debug_prefix_map, add_debug_prefix_map
14100 remap_debug_filename): Move to...
14101 * file-prefix-map.c: New file.
14102 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14103 generalize, get rid of alloca(), use strrchr() instead of strchr().
14104 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14105 Implement in terms of add_prefix_map().
14106 (remap_macro_filename, remap_debug_filename): Implement in term of
14108 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14109 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14110 * dbxout.c: Include file-prefix-map.h.
14111 * varasm.c: Likewise.
14112 * vmsdbgout.c: Likewise.
14113 * xcoffout.c: Likewise.
14114 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14115 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14116 * doc/invoke.texi (-ffile-prefix-map): Document.
14117 (-fdebug-prefix-map): Update description.
14119 2018-01-18 Martin Liska <mliska@suse.cz>
14121 * config/i386/i386.c (indirect_thunk_name): Document that also
14123 (output_indirect_thunk): Document why both instructions
14124 (pause and lfence) are generated.
14126 2018-01-18 Richard Biener <rguenther@suse.de>
14128 PR tree-optimization/83887
14129 * graphite-scop-detection.c
14130 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14131 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14132 (scop_detection::merge_sese): Re-implement with a flood-fill
14133 algorithm that properly finds a SESE region if it exists.
14135 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14138 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14139 pointer_diff optimizations use view_convert instead of convert.
14141 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14144 Generate different code for -mno-speculate-indirect-jumps.
14145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14146 (*call_indirect_aix<mode>): Disable for
14147 -mno-speculate-indirect-jumps.
14148 (*call_indirect_aix<mode>_nospec): New define_insn.
14149 (*call_value_indirect_aix<mode>): Disable for
14150 -mno-speculate-indirect-jumps.
14151 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14152 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14153 -mno-speculate-indirect-jumps.
14154 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14156 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14158 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14159 long double type, set the flags for noting the default long double
14160 type, even if we don't pass or return a long double type.
14162 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14165 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14168 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14170 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14171 support for merge[hl].
14172 (fold_mergehl_helper): New helper function.
14173 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14174 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14175 (altivec_vmrglw_direct): Add xxmrglw insn.
14177 2018-01-17 Andrew Waterman <andrew@sifive.com>
14179 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14180 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14182 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14185 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14186 call the lto_location_cache before reading the
14187 DECL_SOURCE_LOCATION of the types.
14189 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14190 Richard Sandiford <richard.sandiford@linaro.org>
14192 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14193 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14194 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14195 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14196 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14198 * config/aarch64/constraints.md (aarch64_movti_operand):
14200 * config/aarch64/predicates.md (Uti): Add new constraint.
14202 2018-01-17 Carl Love <cel@us.ibm.com>
14204 * config/rs6000/vsx.md (define_expand xl_len_r,
14205 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14206 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14208 (define_expand, define_insn): Move the shift left from the
14209 define_insn to the define_expand for lxvl and stxvl instructions.
14210 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14211 and XL_LEN_R definitions to PURE.
14213 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14215 * config/i386/i386.c (indirect_thunk_name): Declare regno
14216 as unsigned int. Compare regno with INVALID_REGNUM.
14217 (output_indirect_thunk): Ditto.
14218 (output_indirect_thunk_function): Ditto.
14219 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14220 in the call to output_indirect_thunk_function.
14222 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14224 PR middle-end/83884
14225 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14226 rather than the size of inner_type to determine the stack slot size
14227 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14229 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14232 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14235 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14237 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14238 endian Linux systems to optionally enable multilibs for selecting
14239 the long double type if the user configured an explicit type.
14240 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14241 have no long double multilibs if not defined.
14242 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14243 warn if the user used -mabi={ieee,ibm}longdouble and we built
14244 multilibs for long double.
14245 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14246 appropriate multilib option.
14247 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14249 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14250 for building long double multilibs.
14251 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14253 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14255 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14258 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14260 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14263 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14266 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14269 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14271 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14272 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14274 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14276 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14277 different rtl trees depending on TARGET_64BIT.
14278 (rs6000_gen_lvx): Likewise.
14280 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14282 * config/visium/visium.md (nop): Tweak comment.
14283 (hazard_nop): Likewise.
14285 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14287 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14288 -mspeculate-indirect-jumps.
14289 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14290 for -mno-speculate-indirect-jumps.
14291 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14292 (*call_value_indirect_elfv2<mode>): Disable for
14293 -mno-speculate-indirect-jumps.
14294 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14295 (indirect_jump): Emit different RTL for
14296 -mno-speculate-indirect-jumps.
14297 (*indirect_jump<mode>): Disable for
14298 -mno-speculate-indirect-jumps.
14299 (*indirect_jump<mode>_nospec): New define_insn.
14300 (tablejump): Emit different RTL for
14301 -mno-speculate-indirect-jumps.
14302 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14303 (tablejumpsi_nospec): New define_expand.
14304 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14305 (tablejumpdi_nospec): New define_expand.
14306 (*tablejump<mode>_internal1): Disable for
14307 -mno-speculate-indirect-jumps.
14308 (*tablejump<mode>_internal1_nospec): New define_insn.
14309 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14312 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14314 * caller-save.c (insert_save): Drop unnecessary parameter. All
14317 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14318 Richard Biener <rguenth@suse.de>
14321 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14322 return early, inline manually is_gimple_sizepos. Make sure if we
14323 call gimplify_expr we don't end up with a gimple constant.
14324 * tree.c (variably_modified_type_p): Don't return true for
14325 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14326 * gimplify.h (is_gimple_sizepos): Remove.
14328 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14330 PR tree-optimization/83857
14331 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14332 vectorizable_live_operation for pure SLP statements.
14333 (vectorizable_live_operation): Handle PHIs.
14335 2018-01-16 Richard Biener <rguenther@suse.de>
14337 PR tree-optimization/83867
14338 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14339 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14341 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14344 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14345 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14346 If off is not INTEGER_CST, issue a may not be aligned warning
14347 rather than isn't aligned. Use isn%'t rather than isn't.
14348 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14350 <case MULT_EXPR>: Improve the case when bottom and one of the
14351 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14352 operand, in that case check if the other operand is multiple of
14353 bottom divided by the INTEGER_CST operand.
14355 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14358 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14359 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14360 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14361 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14362 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14363 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14364 (pa_function_arg_size): New function.
14366 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14368 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14369 in a separate statement.
14371 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14373 PR tree-optimization/83847
14374 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14375 group gathers and scatters.
14377 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14379 PR rtl-optimization/86620
14380 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14382 PR rtl-optimization/83213
14383 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14384 to last if both are JUMP_INSNs.
14386 PR tree-optimization/83843
14387 * gimple-ssa-store-merging.c
14388 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14389 store_immediate_info for bswap/nop orig_stores.
14391 2018-01-15 Andrew Waterman <andrew@sifive.com>
14393 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14395 <UDIV>: Increase cost if !TARGET_DIV.
14397 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14399 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14400 (define_attr "cr_logical_3op"): New.
14401 (cceq_ior_compare): Adjust.
14402 (cceq_ior_compare_complement): Adjust.
14403 (*cceq_rev_compare): Adjust.
14404 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14405 (is_cracked_insn): Adjust.
14406 (insn_must_be_first_in_group): Adjust.
14407 * config/rs6000/40x.md: Adjust.
14408 * config/rs6000/440.md: Adjust.
14409 * config/rs6000/476.md: Adjust.
14410 * config/rs6000/601.md: Adjust.
14411 * config/rs6000/603.md: Adjust.
14412 * config/rs6000/6xx.md: Adjust.
14413 * config/rs6000/7450.md: Adjust.
14414 * config/rs6000/7xx.md: Adjust.
14415 * config/rs6000/8540.md: Adjust.
14416 * config/rs6000/cell.md: Adjust.
14417 * config/rs6000/e300c2c3.md: Adjust.
14418 * config/rs6000/e500mc.md: Adjust.
14419 * config/rs6000/e500mc64.md: Adjust.
14420 * config/rs6000/e5500.md: Adjust.
14421 * config/rs6000/e6500.md: Adjust.
14422 * config/rs6000/mpc.md: Adjust.
14423 * config/rs6000/power4.md: Adjust.
14424 * config/rs6000/power5.md: Adjust.
14425 * config/rs6000/power6.md: Adjust.
14426 * config/rs6000/power7.md: Adjust.
14427 * config/rs6000/power8.md: Adjust.
14428 * config/rs6000/power9.md: Adjust.
14429 * config/rs6000/rs64.md: Adjust.
14430 * config/rs6000/titan.md: Adjust.
14432 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14434 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14435 ix86_indirect_branch_register logic.
14437 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14439 * config/i386/constraints.md (Bs): Update
14440 ix86_indirect_branch_register check. Don't check
14441 ix86_indirect_branch_register with GOT_memory_operand.
14443 * config/i386/predicates.md (GOT_memory_operand): Don't check
14444 ix86_indirect_branch_register here.
14445 (GOT32_symbol_operand): Likewise.
14447 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14449 * config/i386/predicates.md (constant_call_address_operand):
14450 Rewrite ix86_indirect_branch_register logic.
14451 (sibcall_insn_operand): Likewise.
14453 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14455 * config/i386/constraints.md (Bs): Replace
14456 ix86_indirect_branch_thunk_register with
14457 ix86_indirect_branch_register.
14459 * config/i386/i386.md (indirect_jump): Likewise.
14460 (tablejump): Likewise.
14461 (*sibcall_memory): Likewise.
14462 (*sibcall_value_memory): Likewise.
14463 Peepholes of indirect call and jump via memory: Likewise.
14464 * config/i386/i386.opt: Likewise.
14465 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14466 (GOT_memory_operand): Likewise.
14467 (call_insn_operand): Likewise.
14468 (sibcall_insn_operand): Likewise.
14469 (GOT32_symbol_operand): Likewise.
14471 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14473 PR middle-end/83837
14474 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14475 type rather than type addr's type points to.
14476 (expand_omp_atomic_mutex): Likewise.
14477 (expand_omp_atomic): Likewise.
14479 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14482 * config/i386/i386.c (output_indirect_thunk_function): Use
14483 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14484 for __x86_return_thunk.
14486 2018-01-15 Richard Biener <rguenther@suse.de>
14488 PR middle-end/83850
14489 * expmed.c (extract_bit_field_1): Fix typo.
14491 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14494 * config/arm/iterators.md (VF): New mode iterator.
14495 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14496 Remove integer-related logic from pattern.
14497 (neon_vabd<mode>_3): Likewise.
14499 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14501 PR middle-end/82694
14502 * common.opt (fstrict-overflow): No longer an alias.
14503 (fwrapv-pointer): New option.
14504 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14505 also for pointer types based on flag_wrapv_pointer.
14506 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14507 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14508 opts->x_flag_wrapv got set.
14509 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14510 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14511 POINTER_TYPE_OVERFLOW_UNDEFINED.
14512 * match.pd: Likewise in address comparison pattern.
14513 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14515 2018-01-15 Richard Biener <rguenther@suse.de>
14518 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14519 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14520 Reset type names to their identifier if their TYPE_DECL doesn't
14521 have linkage (and thus is used for ODR and devirt).
14522 (save_debug_info_for_decl): Remove.
14523 (save_debug_info_for_type): Likewise.
14524 (add_tree_to_fld_list): Adjust.
14525 * tree-pretty-print.c (dump_generic_node): Make dumping of
14526 type names more robust.
14528 2018-01-15 Richard Biener <rguenther@suse.de>
14530 * BASE-VER: Bump to 8.0.1.
14532 2018-01-14 Martin Sebor <msebor@redhat.com>
14535 * builtins.c (check_access): Avoid warning when the no-warning bit
14538 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14540 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14541 * ira-color (allocno_hard_regs_compare): Likewise.
14543 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14546 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14547 Use .pushsection/.popsection.
14549 2018-01-14 Martin Sebor <msebor@redhat.com>
14552 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14554 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14556 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14557 entry from extra_headers.
14558 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14559 extra_headers, make the list bitwise identical to the i?86-*-* one.
14561 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14563 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14564 -mcmodel=large with -mindirect-branch=thunk,
14565 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14566 -mfunction-return=thunk-extern.
14567 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14568 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14569 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14571 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14573 * config/i386/i386.c (print_reg): Print the name of the full
14574 integer register without '%'.
14575 (ix86_print_operand): Handle 'V'.
14576 * doc/extend.texi: Document 'V' modifier.
14578 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14580 * config/i386/constraints.md (Bs): Disallow memory operand for
14581 -mindirect-branch-register.
14583 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14584 (GOT_memory_operand): Likewise.
14585 (call_insn_operand): Likewise.
14586 (sibcall_insn_operand): Likewise.
14587 (GOT32_symbol_operand): Likewise.
14588 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14589 for -mindirect-branch-register.
14590 (tablejump): Likewise.
14591 (*sibcall_memory): Likewise.
14592 (*sibcall_value_memory): Likewise.
14593 Disallow peepholes of indirect call and jump via memory for
14594 -mindirect-branch-register.
14595 (*call_pop): Replace m with Bw.
14596 (*call_value_pop): Likewise.
14597 (*sibcall_pop_memory): Replace m with Bs.
14598 * config/i386/i386.opt (mindirect-branch-register): New option.
14599 * doc/invoke.texi: Document -mindirect-branch-register option.
14601 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14603 * config/i386/i386-protos.h (ix86_output_function_return): New.
14604 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14605 set function_return_type.
14606 (indirect_thunk_name): Add ret_p to indicate thunk for function
14608 (output_indirect_thunk_function): Pass false to
14609 indirect_thunk_name.
14610 (ix86_output_indirect_branch_via_reg): Likewise.
14611 (ix86_output_indirect_branch_via_push): Likewise.
14612 (output_indirect_thunk_function): Create alias for function
14613 return thunk if regno < 0.
14614 (ix86_output_function_return): New function.
14615 (ix86_handle_fndecl_attribute): Handle function_return.
14616 (ix86_attribute_table): Add function_return.
14617 * config/i386/i386.h (machine_function): Add
14618 function_return_type.
14619 * config/i386/i386.md (simple_return_internal): Use
14620 ix86_output_function_return.
14621 (simple_return_internal_long): Likewise.
14622 * config/i386/i386.opt (mfunction-return=): New option.
14623 (indirect_branch): Mention -mfunction-return=.
14624 * doc/extend.texi: Document function_return function attribute.
14625 * doc/invoke.texi: Document -mfunction-return= option.
14627 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14629 * config/i386/i386-opts.h (indirect_branch): New.
14630 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14631 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14632 with local indirect jump when converting indirect call and jump.
14633 (ix86_set_indirect_branch_type): New.
14634 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14635 (indirectlabelno): New.
14636 (indirect_thunk_needed): Likewise.
14637 (indirect_thunk_bnd_needed): Likewise.
14638 (indirect_thunks_used): Likewise.
14639 (indirect_thunks_bnd_used): Likewise.
14640 (INDIRECT_LABEL): Likewise.
14641 (indirect_thunk_name): Likewise.
14642 (output_indirect_thunk): Likewise.
14643 (output_indirect_thunk_function): Likewise.
14644 (ix86_output_indirect_branch_via_reg): Likewise.
14645 (ix86_output_indirect_branch_via_push): Likewise.
14646 (ix86_output_indirect_branch): Likewise.
14647 (ix86_output_indirect_jmp): Likewise.
14648 (ix86_code_end): Call output_indirect_thunk_function if needed.
14649 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14651 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14652 (ix86_attribute_table): Add indirect_branch.
14653 * config/i386/i386.h (machine_function): Add indirect_branch_type
14654 and has_local_indirect_jump.
14655 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14657 (tablejump): Likewise.
14658 (*indirect_jump): Use ix86_output_indirect_jmp.
14659 (*tablejump_1): Likewise.
14660 (simple_return_indirect_internal): Likewise.
14661 * config/i386/i386.opt (mindirect-branch=): New option.
14662 (indirect_branch): New.
14665 (thunk-inline): Likewise.
14666 (thunk-extern): Likewise.
14667 * doc/extend.texi: Document indirect_branch function attribute.
14668 * doc/invoke.texi: Document -mindirect-branch= option.
14670 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14673 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14675 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14677 * ipa-inline.c (want_inline_small_function_p): Return false if
14678 inlining has already failed with CIF_FINAL_ERROR.
14679 (update_caller_keys): Call want_inline_small_function_p before
14681 (update_callee_keys): Likewise.
14683 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14685 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14687 (rs6000_quadword_masked_address_p): Likewise.
14688 (quad_aligned_load_p): Likewise.
14689 (quad_aligned_store_p): Likewise.
14690 (const_load_sequence_p): Add comment to describe the outer-most loop.
14691 (mimic_memory_attributes_and_flags): New function.
14692 (rs6000_gen_stvx): Likewise.
14693 (replace_swapped_aligned_store): Likewise.
14694 (rs6000_gen_lvx): Likewise.
14695 (replace_swapped_aligned_load): Likewise.
14696 (replace_swapped_load_constant): Capitalize argument name in
14697 comment describing this function.
14698 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14699 and stores that access quad-word aligned addresses and replace
14700 with stvx or lvx instructions when appropriate.
14701 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14702 New function prototype.
14703 (rs6000_quadword_masked_address_p): Likewise.
14704 (rs6000_gen_lvx): Likewise.
14705 (rs6000_gen_stvx): Likewise.
14706 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14707 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14708 when memory address is aligned.
14709 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14710 this split to select lvx instruction when memory address is aligned.
14711 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14712 instruction when memory address is aligned.
14713 (*vsx_le_perm_load_v16qi): Likewise.
14714 (four unnamed splitters): Modify to select the stvx instruction
14715 when memory is aligned.
14717 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14719 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14720 which appears in the queue multiple times.
14722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14723 Alan Hayward <alan.hayward@arm.com>
14724 David Sherwood <david.sherwood@arm.com>
14726 * tree-vectorizer.h (vec_lower_bound): New structure.
14727 (_loop_vec_info): Add check_nonzero and lower_bounds.
14728 (LOOP_VINFO_CHECK_NONZERO): New macro.
14729 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14730 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14731 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14732 fields. Make seg_len the distance travelled, not including the
14734 (dr_direction_indicator): Declare.
14735 (dr_zero_step_indicator): Likewise.
14736 (dr_known_forward_stride_p): Likewise.
14737 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14739 (runtime_alias_check_p): Allow runtime alias checks with
14741 (operator ==): Compare access_size and align.
14742 (prune_runtime_alias_test_list): Rework for new distinction between
14743 the access_size and seg_len.
14744 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14746 (get_segment_min_max): New function.
14747 (create_intersect_range_checks): Use it.
14748 (dr_step_indicator): New function.
14749 (dr_direction_indicator): Likewise.
14750 (dr_zero_step_indicator): Likewise.
14751 (dr_known_forward_stride_p): Likewise.
14752 * tree-loop-distribution.c (data_ref_segment_size): Return
14753 DR_STEP * (niters - 1).
14754 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14756 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14757 (vect_preserves_scalar_order_p): New function, split out from...
14758 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14759 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14760 (vect_vfa_access_size): New function.
14761 (vect_vfa_align): Likewise.
14762 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14763 (dump_lower_bound): New function.
14764 (vect_check_lower_bound): Likewise.
14765 (vect_small_gap_p): Likewise.
14766 (vectorizable_with_step_bound_p): Likewise.
14767 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14768 depencies if the vectorization factor is 1. Convert the checks
14769 for nonzero steps into checks on the bounds of DR_STEP. Try using
14770 a bunds check for variable steps if the minimum required step is
14771 relatively small. Update calls to the dr_with_seg_len
14772 constructor and to vect_compile_time_alias.
14773 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14775 (vect_loop_versioning): Call it.
14776 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14778 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14781 Alan Hayward <alan.hayward@arm.com>
14782 David Sherwood <david.sherwood@arm.com>
14784 * doc/sourcebuild.texi (vect_scatter_store): Document.
14785 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14787 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14789 * genopinit.c (main): Add supports_vec_scatter_store and
14790 supports_vec_scatter_store_cached to target_optabs.
14791 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14792 IFN_MASK_SCATTER_STORE.
14793 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14795 * internal-fn.h (internal_store_fn_p): Declare.
14796 (internal_fn_stored_value_index): Likewise.
14797 * internal-fn.c (scatter_store_direct): New macro.
14798 (expand_scatter_store_optab_fn): New function.
14799 (direct_scatter_store_optab_supported_p): New macro.
14800 (internal_store_fn_p): New function.
14801 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14802 IFN_MASK_SCATTER_STORE.
14803 (internal_fn_mask_index): Likewise.
14804 (internal_fn_stored_value_index): New function.
14805 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14806 for scatter stores.
14807 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14808 * optabs-query.c (supports_vec_scatter_store_p): New function.
14809 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14810 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14811 true for scatter stores.
14812 (vect_gather_scatter_fn_p): Handle scatter stores too.
14813 (vect_check_gather_scatter): Consider using scatter stores if
14814 supports_vec_scatter_store_p.
14815 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14816 scatter stores too.
14817 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14818 internal_fn_stored_value_index.
14819 (check_load_store_masking): Handle scatter stores too.
14820 (vect_get_store_rhs): Make public.
14821 (vectorizable_call): Use internal_store_fn_p.
14822 (vectorizable_store): Handle scatter store internal functions.
14823 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14824 when deciding whether the end of the group has been reached.
14825 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14826 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14827 (mask_scatter_store<mode>): New insns.
14829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14830 Alan Hayward <alan.hayward@arm.com>
14831 David Sherwood <david.sherwood@arm.com>
14833 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14834 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14835 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14837 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14838 Use vect_truncate_gather_scatter_offset if we can't treat the
14839 operation as a normal gather load or scatter store.
14840 (get_group_load_store_type): Take the gather_scatter_info
14841 as argument. Try using a gather load or scatter store for
14842 single-element groups.
14843 (get_load_store_type): Update calls to get_group_load_store_type
14844 and vect_use_strided_gather_scatters_p.
14846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14847 Alan Hayward <alan.hayward@arm.com>
14848 David Sherwood <david.sherwood@arm.com>
14850 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14851 optional tree argument.
14852 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14854 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14855 but continue to use the current value as a fallback.
14856 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14857 to compare the updates.
14858 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14859 (get_load_store_type): Use it when handling a strided access.
14860 (vect_get_strided_load_store_ops): New function.
14861 (vect_get_data_ptr_increment): Likewise.
14862 (vectorizable_load): Handle strided gather loads. Always pass
14863 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14866 Alan Hayward <alan.hayward@arm.com>
14867 David Sherwood <david.sherwood@arm.com>
14869 * doc/md.texi (gather_load@var{m}): Document.
14870 (mask_gather_load@var{m}): Likewise.
14871 * genopinit.c (main): Add supports_vec_gather_load and
14872 supports_vec_gather_load_cached to target_optabs.
14873 * optabs-tree.c (init_tree_optimization_optabs): Use
14874 ggc_cleared_alloc to allocate target_optabs.
14875 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14876 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14878 * internal-fn.h (internal_load_fn_p): Declare.
14879 (internal_gather_scatter_fn_p): Likewise.
14880 (internal_fn_mask_index): Likewise.
14881 (internal_gather_scatter_fn_supported_p): Likewise.
14882 * internal-fn.c (gather_load_direct): New macro.
14883 (expand_gather_load_optab_fn): New function.
14884 (direct_gather_load_optab_supported_p): New macro.
14885 (direct_internal_fn_optab): New function.
14886 (internal_load_fn_p): Likewise.
14887 (internal_gather_scatter_fn_p): Likewise.
14888 (internal_fn_mask_index): Likewise.
14889 (internal_gather_scatter_fn_supported_p): Likewise.
14890 * optabs-query.c (supports_at_least_one_mode_p): New function.
14891 (supports_vec_gather_load_p): Likewise.
14892 * optabs-query.h (supports_vec_gather_load_p): Declare.
14893 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14894 and memory_type field.
14895 (NUM_PATTERNS): Bump to 15.
14896 * tree-vect-data-refs.c: Include internal-fn.h.
14897 (vect_gather_scatter_fn_p): New function.
14898 (vect_describe_gather_scatter_call): Likewise.
14899 (vect_check_gather_scatter): Try using internal functions for
14900 gather loads. Recognize existing calls to a gather load function.
14901 (vect_analyze_data_refs): Consider using gather loads if
14902 supports_vec_gather_load_p.
14903 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14904 (vect_get_gather_scatter_offset_type): Likewise.
14905 (vect_convert_mask_for_vectype): Likewise.
14906 (vect_add_conversion_to_patterm): Likewise.
14907 (vect_try_gather_scatter_pattern): Likewise.
14908 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14909 (vect_vect_recog_func_ptrs): Add it.
14910 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14911 internal_fn_mask_index and internal_gather_scatter_fn_p.
14912 (check_load_store_masking): Take the gather_scatter_info as an
14913 argument and handle gather loads.
14914 (vect_get_gather_scatter_ops): New function.
14915 (vectorizable_call): Check internal_load_fn_p.
14916 (vectorizable_load): Likewise. Handle gather load internal
14918 (vectorizable_store): Update call to check_load_store_masking.
14919 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14920 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14921 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14922 (aarch64_gather_scale_operand_d): New predicates.
14923 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14924 (mask_gather_load<mode>): New insns.
14926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14927 Alan Hayward <alan.hayward@arm.com>
14928 David Sherwood <david.sherwood@arm.com>
14930 * optabs.def (fold_left_plus_optab): New optab.
14931 * doc/md.texi (fold_left_plus_@var{m}): Document.
14932 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14933 * internal-fn.c (fold_left_direct): Define.
14934 (expand_fold_left_optab_fn): Likewise.
14935 (direct_fold_left_optab_supported_p): Likewise.
14936 * fold-const-call.c (fold_const_fold_left): New function.
14937 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14938 * tree-parloops.c (valid_reduction_p): New function.
14939 (gather_scalar_reductions): Use it.
14940 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14941 (vect_finish_replace_stmt): Declare.
14942 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14943 (needs_fold_left_reduction_p): New function, split out from...
14944 (vect_is_simple_reduction): ...here. Accept reductions that
14945 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14946 (vect_force_simple_reduction): Also store the reduction type in
14947 the assignment's STMT_VINFO_REDUC_TYPE.
14948 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14949 (merge_with_identity): New function.
14950 (vect_expand_fold_left): Likewise.
14951 (vectorize_fold_left_reduction): Likewise.
14952 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14953 scalar phi in place for it. Check for target support and reject
14954 cases that would reassociate the operation. Defer the transform
14955 phase to vectorize_fold_left_reduction.
14956 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14957 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14958 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14962 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14963 call to ifc_temp_var.
14965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14966 Alan Hayward <alan.hayward@arm.com>
14967 David Sherwood <david.sherwood@arm.com>
14969 * target.def (legitimize_address_displacement): Take the original
14970 offset as a poly_int.
14971 * targhooks.h (default_legitimize_address_displacement): Update
14973 * targhooks.c (default_legitimize_address_displacement): Likewise.
14974 * doc/tm.texi: Regenerate.
14975 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14976 as an argument, moving assert of ad->disp == ad->disp_term to...
14977 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14978 Try calling targetm.legitimize_address_displacement before expanding
14979 the address rather than afterwards, and adjust for the new interface.
14980 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14981 Match the new hook interface. Handle SVE addresses.
14982 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14983 new hook interface.
14985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14987 * Makefile.in (OBJS): Add early-remat.o.
14988 * target.def (select_early_remat_modes): New hook.
14989 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14990 * doc/tm.texi: Regenerate.
14991 * targhooks.h (default_select_early_remat_modes): Declare.
14992 * targhooks.c (default_select_early_remat_modes): New function.
14993 * timevar.def (TV_EARLY_REMAT): New timevar.
14994 * passes.def (pass_early_remat): New pass.
14995 * tree-pass.h (make_pass_early_remat): Declare.
14996 * early-remat.c: New file.
14997 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14999 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15002 Alan Hayward <alan.hayward@arm.com>
15003 David Sherwood <david.sherwood@arm.com>
15005 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15006 vfm1 with a bound_epilog parameter.
15007 (vect_do_peeling): Update calls accordingly, and move the prologue
15008 call earlier in the function. Treat the base bound_epilog as 0 for
15009 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15010 this base when peeling for gaps.
15011 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15012 with fully-masked loops.
15013 (vect_estimate_min_profitable_iters): Handle the single peeled
15014 iteration in that case.
15016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15017 Alan Hayward <alan.hayward@arm.com>
15018 David Sherwood <david.sherwood@arm.com>
15020 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15021 single-element interleaving even if the size is not a power of 2.
15022 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15023 accesses for single-element interleaving if the group size is
15026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15027 Alan Hayward <alan.hayward@arm.com>
15028 David Sherwood <david.sherwood@arm.com>
15030 * doc/md.texi (fold_extract_last_@var{m}): Document.
15031 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15032 * optabs.def (fold_extract_last_optab): New optab.
15033 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15034 * internal-fn.c (fold_extract_direct): New macro.
15035 (expand_fold_extract_optab_fn): Likewise.
15036 (direct_fold_extract_optab_supported_p): Likewise.
15037 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15038 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15039 EXTRACT_LAST_REDUCTION.
15040 (get_initial_def_for_reduction): Do not create an initial vector
15041 for EXTRACT_LAST_REDUCTION reductions.
15042 (vectorizable_reduction): Leave the scalar phi in place for
15043 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15044 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15045 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15046 transform phase to vectorizable_condition.
15047 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15049 (vect_finish_stmt_generation): ...here.
15050 (vect_finish_replace_stmt): New function.
15051 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15052 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15054 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15057 Alan Hayward <alan.hayward@arm.com>
15058 David Sherwood <david.sherwood@arm.com>
15060 * doc/md.texi (extract_last_@var{m}): Document.
15061 * optabs.def (extract_last_optab): New optab.
15062 * internal-fn.def (EXTRACT_LAST): New internal function.
15063 * internal-fn.c (cond_unary_direct): New macro.
15064 (expand_cond_unary_optab_fn): Likewise.
15065 (direct_cond_unary_optab_supported_p): Likewise.
15066 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15067 loops using EXTRACT_LAST.
15068 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15069 (extract_last_<mode>): ...this optab.
15070 (vec_extract<mode><Vel>): Update accordingly.
15072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15073 Alan Hayward <alan.hayward@arm.com>
15074 David Sherwood <david.sherwood@arm.com>
15076 * target.def (empty_mask_is_expensive): New hook.
15077 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15078 * doc/tm.texi: Regenerate.
15079 * targhooks.h (default_empty_mask_is_expensive): Declare.
15080 * targhooks.c (default_empty_mask_is_expensive): New function.
15081 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15082 if the target says that empty masks are expensive.
15083 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15085 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15088 Alan Hayward <alan.hayward@arm.com>
15089 David Sherwood <david.sherwood@arm.com>
15091 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15092 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15093 (vect_use_loop_mask_for_alignment_p): New function.
15094 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15095 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15096 niters_skip argument. Make sure that the first niters_skip elements
15097 of the first iteration are inactive.
15098 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15099 Update call to vect_set_loop_masks_directly.
15100 (get_misalign_in_elems): New function, split out from...
15101 (vect_gen_prolog_loop_niters): ...here.
15102 (vect_update_init_of_dr): Take a code argument that specifies whether
15103 the adjustment should be added or subtracted.
15104 (vect_update_init_of_drs): Likewise.
15105 (vect_prepare_for_masked_peels): New function.
15106 (vect_do_peeling): Skip prologue peeling if we're using a mask
15107 instead. Update call to vect_update_inits_of_drs.
15108 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15110 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15111 alignment. Do not include the number of peeled iterations in
15112 the minimum threshold in that case.
15113 (vectorizable_induction): Adjust the start value down by
15114 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15115 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15116 Take the number of skipped iterations into account when calculating
15118 * tree-vect-stmts.c (vect_gen_while_not): New function.
15120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15121 Alan Hayward <alan.hayward@arm.com>
15122 David Sherwood <david.sherwood@arm.com>
15124 * doc/sourcebuild.texi (vect_fully_masked): Document.
15125 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15126 default value to 0.
15127 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15129 (vect_analyze_loop_2): ...here. Don't check the vectorization
15130 factor against the number of loop iterations if the loop is
15133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15134 Alan Hayward <alan.hayward@arm.com>
15135 David Sherwood <david.sherwood@arm.com>
15137 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15138 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15139 (dump_groups): Update accordingly.
15140 (iv_use::mem_type): New member variable.
15141 (address_p): New function.
15142 (record_use): Add a mem_type argument and initialize the new
15144 (record_group_use): Add a mem_type argument. Use address_p.
15145 Remove obsolete null checks of base_object. Update call to record_use.
15146 (find_interesting_uses_op): Update call to record_group_use.
15147 (find_interesting_uses_cond): Likewise.
15148 (find_interesting_uses_address): Likewise.
15149 (get_mem_type_for_internal_fn): New function.
15150 (find_address_like_use): Likewise.
15151 (find_interesting_uses_stmt): Try find_address_like_use before
15152 calling find_interesting_uses_op.
15153 (addr_offset_valid_p): Use the iv mem_type field as the type
15154 of the addressed memory.
15155 (add_autoinc_candidates): Likewise.
15156 (get_address_cost): Likewise.
15157 (split_small_address_groups_p): Use address_p.
15158 (split_address_groups): Likewise.
15159 (add_iv_candidate_for_use): Likewise.
15160 (autoinc_possible_for_pair): Likewise.
15161 (rewrite_groups): Likewise.
15162 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15163 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15164 (get_alias_ptr_type_for_ptr_address): New function.
15165 (rewrite_use_address): Rewrite address uses in calls that were
15166 identified by find_address_like_use.
15168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15169 Alan Hayward <alan.hayward@arm.com>
15170 David Sherwood <david.sherwood@arm.com>
15172 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15174 * gimple-expr.h (is_gimple_addressable: Likewise.
15175 * gimple-expr.c (is_gimple_address): Likewise.
15176 * internal-fn.c (expand_call_mem_ref): New function.
15177 (expand_mask_load_optab_fn): Use it.
15178 (expand_mask_store_optab_fn): Likewise.
15180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15181 Alan Hayward <alan.hayward@arm.com>
15182 David Sherwood <david.sherwood@arm.com>
15184 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15185 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15186 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15187 (cond_umax@var{mode}): Document.
15188 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15189 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15190 (cond_umin_optab, cond_umax_optab): New optabs.
15191 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15192 (COND_IOR, COND_XOR): New internal functions.
15193 * internal-fn.h (get_conditional_internal_fn): Declare.
15194 * internal-fn.c (cond_binary_direct): New macro.
15195 (expand_cond_binary_optab_fn): Likewise.
15196 (direct_cond_binary_optab_supported_p): Likewise.
15197 (get_conditional_internal_fn): New function.
15198 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15199 Cope with reduction statements that are vectorized as calls rather
15201 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15202 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15203 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15204 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15205 (UNSPEC_COND_EOR): New unspecs.
15206 (optab): Add mappings for them.
15207 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15208 (sve_int_op, sve_fp_op): New int attributes.
15210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15211 Alan Hayward <alan.hayward@arm.com>
15212 David Sherwood <david.sherwood@arm.com>
15214 * optabs.def (while_ult_optab): New optab.
15215 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15216 * internal-fn.def (WHILE_ULT): New internal function.
15217 * internal-fn.h (direct_internal_fn_supported_p): New override
15218 that takes two types as argument.
15219 * internal-fn.c (while_direct): New macro.
15220 (expand_while_optab_fn): New function.
15221 (convert_optab_supported_p): Likewise.
15222 (direct_while_optab_supported_p): New macro.
15223 * wide-int.h (wi::udiv_ceil): New function.
15224 * tree-vectorizer.h (rgroup_masks): New structure.
15225 (vec_loop_masks): New typedef.
15226 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15227 and fully_masked_p.
15228 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15229 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15230 (vect_max_vf): New function.
15231 (slpeel_make_loop_iterate_ntimes): Delete.
15232 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15233 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15234 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15235 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15236 internal-fn.h, stor-layout.h and optabs-query.h.
15237 (vect_set_loop_mask): New function.
15238 (add_preheader_seq): Likewise.
15239 (add_header_seq): Likewise.
15240 (interleave_supported_p): Likewise.
15241 (vect_maybe_permute_loop_masks): Likewise.
15242 (vect_set_loop_masks_directly): Likewise.
15243 (vect_set_loop_condition_masked): Likewise.
15244 (vect_set_loop_condition_unmasked): New function, split out from
15245 slpeel_make_loop_iterate_ntimes.
15246 (slpeel_make_loop_iterate_ntimes): Rename to..
15247 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15248 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15249 (vect_do_peeling): Update call accordingly.
15250 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15252 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15253 mask_compare_type, can_fully_mask_p and fully_masked_p.
15254 (release_vec_loop_masks): New function.
15255 (_loop_vec_info): Use it to free the loop masks.
15256 (can_produce_all_loop_masks_p): New function.
15257 (vect_get_max_nscalars_per_iter): Likewise.
15258 (vect_verify_full_masking): Likewise.
15259 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15260 retries, and free the mask rgroups before retrying. Check loop-wide
15261 reasons for disallowing fully-masked loops. Make the final decision
15262 about whether use a fully-masked loop or not.
15263 (vect_estimate_min_profitable_iters): Do not assume that peeling
15264 for the number of iterations will be needed for fully-masked loops.
15265 (vectorizable_reduction): Disable fully-masked loops.
15266 (vectorizable_live_operation): Likewise.
15267 (vect_halve_mask_nunits): New function.
15268 (vect_double_mask_nunits): Likewise.
15269 (vect_record_loop_mask): Likewise.
15270 (vect_get_loop_mask): Likewise.
15271 (vect_transform_loop): Handle the case in which the final loop
15272 iteration might handle a partial vector. Call vect_set_loop_condition
15273 instead of slpeel_make_loop_iterate_ntimes.
15274 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15275 (check_load_store_masking): New function.
15276 (prepare_load_store_mask): Likewise.
15277 (vectorizable_store): Handle fully-masked loops.
15278 (vectorizable_load): Likewise.
15279 (supportable_widening_operation): Use vect_halve_mask_nunits for
15281 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15282 (vect_gen_while): New function.
15283 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15284 (aarch64_uqdec<mode>): New insn.
15286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15287 Alan Hayward <alan.hayward@arm.com>
15288 David Sherwood <david.sherwood@arm.com>
15290 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15291 (reduc_xor_scal_optab): New optabs.
15292 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15293 (reduc_xor_scal_@var{m}): Document.
15294 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15295 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15296 internal functions.
15297 * fold-const-call.c (fold_const_call): Handle them.
15298 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15299 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15300 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15301 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15302 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15303 (UNSPEC_XORV): New unspecs.
15304 (optab): Add entries for them.
15305 (BITWISEV): New int iterator.
15306 (bit_reduc_op): New int attributes.
15308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15309 Alan Hayward <alan.hayward@arm.com>
15310 David Sherwood <david.sherwood@arm.com>
15312 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15313 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15314 * optabs.def (vec_shl_insert_optab): New optab.
15315 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15316 (duplicate_and_interleave): Likewise.
15317 * tree-vect-loop.c: Include internal-fn.h.
15318 (neutral_op_for_slp_reduction): New function, split out from
15319 get_initial_defs_for_reduction.
15320 (get_initial_def_for_reduction): Handle option 2 for variable-length
15321 vectors by loading the neutral value into a vector and then shifting
15322 the initial value into element 0.
15323 (get_initial_defs_for_reduction): Replace the code argument with
15324 the neutral value calculated by neutral_op_for_slp_reduction.
15325 Use gimple_build_vector for constant-length vectors.
15326 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15327 but the first group_size elements have a neutral value.
15328 Use duplicate_and_interleave otherwise.
15329 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15330 Update call to get_initial_defs_for_reduction. Handle SLP
15331 reductions for variable-length vectors by creating one vector
15332 result for each scalar result, with the elements associated
15333 with other scalar results stubbed out with the neutral value.
15334 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15335 Require IFN_VEC_SHL_INSERT for double reductions on
15336 variable-length vectors, or SLP reductions that have
15337 a neutral value. Require can_duplicate_and_interleave_p
15338 support for variable-length unchained SLP reductions if there
15339 is no neutral value, such as for MIN/MAX reductions. Also require
15340 the number of vector elements to be a multiple of the number of
15341 SLP statements when doing variable-length unchained SLP reductions.
15342 Update call to vect_create_epilog_for_reduction.
15343 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15344 and remove initial values.
15345 (duplicate_and_interleave): Make public.
15346 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15347 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15350 Alan Hayward <alan.hayward@arm.com>
15351 David Sherwood <david.sherwood@arm.com>
15353 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15354 (can_duplicate_and_interleave_p): New function.
15355 (vect_get_and_check_slp_defs): Take the vector of statements
15356 rather than just the current one. Remove excess parentheses.
15357 Restriction rejectinon of vect_constant_def and vect_external_def
15358 for variable-length vectors to boolean types, or types for which
15359 can_duplicate_and_interleave_p is false.
15360 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15361 (duplicate_and_interleave): New function.
15362 (vect_get_constant_vectors): Use gimple_build_vector for
15363 constant-length vectors and suitable variable-length constant
15364 vectors. Use duplicate_and_interleave for other variable-length
15365 vectors. Don't defer the update when inserting new statements.
15367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15368 Alan Hayward <alan.hayward@arm.com>
15369 David Sherwood <david.sherwood@arm.com>
15371 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15372 min_profitable_iters doesn't go negative.
15374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15375 Alan Hayward <alan.hayward@arm.com>
15376 David Sherwood <david.sherwood@arm.com>
15378 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15379 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15380 * optabs.def (vec_mask_load_lanes_optab): New optab.
15381 (vec_mask_store_lanes_optab): Likewise.
15382 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15383 (MASK_STORE_LANES): Likewise.
15384 * internal-fn.c (mask_load_lanes_direct): New macro.
15385 (mask_store_lanes_direct): Likewise.
15386 (expand_mask_load_optab_fn): Handle masked operations.
15387 (expand_mask_load_lanes_optab_fn): New macro.
15388 (expand_mask_store_optab_fn): Handle masked operations.
15389 (expand_mask_store_lanes_optab_fn): New macro.
15390 (direct_mask_load_lanes_optab_supported_p): Likewise.
15391 (direct_mask_store_lanes_optab_supported_p): Likewise.
15392 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15394 (vect_load_lanes_supported): Likewise.
15395 * tree-vect-data-refs.c (strip_conversion): New function.
15396 (can_group_stmts_p): Likewise.
15397 (vect_analyze_data_ref_accesses): Use it instead of checking
15398 for a pair of assignments.
15399 (vect_store_lanes_supported): Take a masked_p parameter.
15400 (vect_load_lanes_supported): Likewise.
15401 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15402 vect_store_lanes_supported and vect_load_lanes_supported.
15403 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15404 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15405 parameter. Don't allow gaps for masked accesses.
15406 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15407 and vect_load_lanes_supported.
15408 (get_load_store_type): Take a masked_p parameter and update
15409 call to get_group_load_store_type.
15410 (vectorizable_store): Update call to get_load_store_type.
15411 Handle IFN_MASK_STORE_LANES.
15412 (vectorizable_load): Update call to get_load_store_type.
15413 Handle IFN_MASK_LOAD_LANES.
15415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15416 Alan Hayward <alan.hayward@arm.com>
15417 David Sherwood <david.sherwood@arm.com>
15419 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15421 * config/aarch64/aarch64-protos.h
15422 (aarch64_sve_struct_memory_operand_p): Declare.
15423 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15424 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15425 (VPRED, vpred): Handle SVE structure modes.
15426 * config/aarch64/constraints.md (Utx): New constraint.
15427 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15428 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15429 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15430 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15431 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15432 structure modes. Split into pieces after RA.
15433 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15434 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15436 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15437 SVE structure modes.
15438 (aarch64_classify_address): Likewise.
15439 (sizetochar): Move earlier in file.
15440 (aarch64_print_operand): Handle SVE register lists.
15441 (aarch64_array_mode): New function.
15442 (aarch64_sve_struct_memory_operand_p): Likewise.
15443 (TARGET_ARRAY_MODE): Redefine.
15445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15446 Alan Hayward <alan.hayward@arm.com>
15447 David Sherwood <david.sherwood@arm.com>
15449 * target.def (array_mode): New target hook.
15450 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15451 * doc/tm.texi: Regenerate.
15452 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15453 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15454 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15455 targetm.array_mode.
15456 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15460 Alan Hayward <alan.hayward@arm.com>
15461 David Sherwood <david.sherwood@arm.com>
15463 * fold-const.c (fold_binary_loc): Check the argument types
15464 rather than the result type when testing for a vector operation.
15466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15468 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15469 * doc/tm.texi: Regenerate.
15471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15472 Alan Hayward <alan.hayward@arm.com>
15473 David Sherwood <david.sherwood@arm.com>
15475 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15476 (sve): Document new AArch64 extension.
15477 * doc/md.texi (w): Extend the description of the AArch64
15478 constraint to include SVE vectors.
15479 (Upl, Upa): Document new AArch64 predicate constraints.
15480 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15482 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15483 (msve-vector-bits=): New option.
15484 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15485 SVE when these are disabled.
15486 (sve): New extension.
15487 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15488 modes. Adjust their number of units based on aarch64_sve_vg.
15489 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15490 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15491 aarch64_addr_query_type.
15492 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15493 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15494 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15495 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15496 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15497 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15498 (aarch64_simd_imm_zero_p): Delete.
15499 (aarch64_check_zero_based_sve_index_immediate): Declare.
15500 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15501 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15502 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15503 (aarch64_sve_float_mul_immediate_p): Likewise.
15504 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15505 rather than an rtx.
15506 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15507 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15508 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15509 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15510 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15511 (aarch64_regmode_natural_size): Likewise.
15512 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15513 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15515 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15516 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15517 for VG and the SVE predicate registers.
15518 (V_ALIASES): Add a "z"-prefixed alias.
15519 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15520 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15521 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15522 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15523 (REG_CLASS_NAMES): Add entries for them.
15524 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15525 and the predicate registers.
15526 (aarch64_sve_vg): Declare.
15527 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15528 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15529 (REGMODE_NATURAL_SIZE): Define.
15530 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15532 * config/aarch64/aarch64.c: Include cfgrtl.h.
15533 (simd_immediate_info): Add a constructor for series vectors,
15534 and an associated step field.
15535 (aarch64_sve_vg): New variable.
15536 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15537 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15538 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15539 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15540 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15541 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15542 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15543 (aarch64_get_mask_mode): New functions.
15544 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15545 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15546 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15547 predicate modes and predicate registers. Explicitly restrict
15548 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15549 to store a vector mode if it is recognized by
15550 aarch64_classify_vector_mode.
15551 (aarch64_regmode_natural_size): New function.
15552 (aarch64_hard_regno_caller_save_mode): Return the original mode
15554 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15555 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15556 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15557 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15559 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15560 does not overlap dest if the function is frame-related. Handle
15562 (aarch64_split_add_offset): New function.
15563 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15564 them aarch64_add_offset.
15565 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15566 and update call to aarch64_sub_sp.
15567 (aarch64_add_cfa_expression): New function.
15568 (aarch64_expand_prologue): Pass extra temporary registers to the
15569 functions above. Handle the case in which we need to emit new
15570 DW_CFA_expressions for registers that were originally saved
15571 relative to the stack pointer, but now have to be expressed
15572 relative to the frame pointer.
15573 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15575 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15576 IP0 and IP1 values for SVE frames.
15577 (aarch64_expand_vec_series): New function.
15578 (aarch64_expand_sve_widened_duplicate): Likewise.
15579 (aarch64_expand_sve_const_vector): Likewise.
15580 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15581 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15582 into the register, rather than emitting a SET directly.
15583 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15584 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15585 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15586 (offset_9bit_signed_scaled_p): New functions.
15587 (aarch64_replicate_bitmask_imm): New function.
15588 (aarch64_bitmask_imm): Use it.
15589 (aarch64_cannot_force_const_mem): Reject expressions involving
15590 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15591 (aarch64_classify_index): Handle SVE indices, by requiring
15592 a plain register index with a scale that matches the element size.
15593 (aarch64_classify_address): Handle SVE addresses. Assert that
15594 the mode of the address is VOIDmode or an integer mode.
15595 Update call to aarch64_classify_symbol.
15596 (aarch64_classify_symbolic_expression): Update call to
15597 aarch64_classify_symbol.
15598 (aarch64_const_vec_all_in_range_p): New function.
15599 (aarch64_print_vector_float_operand): Likewise.
15600 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15601 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15602 and the FP immediates 1.0 and 0.5.
15603 (aarch64_print_address_internal): Handle SVE addresses.
15604 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15605 (aarch64_regno_regclass): Handle predicate registers.
15606 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15608 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15609 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15610 (aarch64_convert_sve_vector_bits): New function.
15611 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15612 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15613 rather than an rtx.
15614 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15615 Handle SVE vector and predicate modes. Accept VL-based constants
15616 that need only one temporary register, and VL offsets that require
15617 no temporary registers.
15618 (aarch64_conditional_register_usage): Mark the predicate registers
15619 as fixed if SVE isn't available.
15620 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15621 Return true for SVE vector and predicate modes.
15622 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15623 rather than an unsigned int. Handle SVE modes.
15624 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15626 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15628 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15629 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15630 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15631 (aarch64_sve_float_mul_immediate_p): New functions.
15632 (aarch64_sve_valid_immediate): New function.
15633 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15634 Explicitly reject structure modes. Check for INDEX constants.
15635 Handle PTRUE and PFALSE constants.
15636 (aarch64_check_zero_based_sve_index_immediate): New function.
15637 (aarch64_simd_imm_zero_p): Delete.
15638 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15639 vector modes. Accept constants in the range of CNT[BHWD].
15640 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15641 ask for an Advanced SIMD mode.
15642 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15643 (aarch64_simd_vector_alignment): Handle SVE predicates.
15644 (aarch64_vectorize_preferred_vector_alignment): New function.
15645 (aarch64_simd_vector_alignment_reachable): Use it instead of
15647 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15648 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15650 (MAX_VECT_LEN): Delete.
15651 (expand_vec_perm_d): Add a vec_flags field.
15652 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15653 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15654 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15656 (aarch64_evpc_rev): Rename to...
15657 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15658 (aarch64_evpc_rev_global): New function.
15659 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15660 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15662 (aarch64_evpc_sve_tbl): New function.
15663 (aarch64_expand_vec_perm_const_1): Update after rename of
15664 aarch64_evpc_rev. Handle SVE permutes too, trying
15665 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15666 than aarch64_evpc_tbl.
15667 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15668 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15669 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15670 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15671 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15672 (aarch64_expand_sve_vcond): New functions.
15673 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15674 of aarch64_vector_mode_p.
15675 (aarch64_dwarf_poly_indeterminate_value): New function.
15676 (aarch64_compute_pressure_classes): Likewise.
15677 (aarch64_can_change_mode_class): Likewise.
15678 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15679 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15680 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15681 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15682 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15683 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15684 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15685 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15687 (Dn, Dl, Dr): Accept const as well as const_vector.
15688 (Dz): Likewise. Compare against CONST0_RTX.
15689 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15690 of "vector" where appropriate.
15691 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15692 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15693 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15694 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15695 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15696 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15697 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15698 (v_int_equiv): Extend to SVE modes.
15699 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15701 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15702 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15703 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15704 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15705 (SVE_COND_FP_CMP): New int iterators.
15706 (perm_hilo): Handle the new unpack unspecs.
15707 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15709 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15710 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15711 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15712 (aarch64_equality_operator, aarch64_constant_vector_operand)
15713 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15714 (aarch64_sve_nonimmediate_operand): Likewise.
15715 (aarch64_sve_general_operand): Likewise.
15716 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15717 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15718 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15719 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15720 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15721 (aarch64_sve_float_arith_immediate): Likewise.
15722 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15723 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15724 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15725 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15726 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15727 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15728 (aarch64_sve_float_arith_operand): Likewise.
15729 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15730 (aarch64_sve_float_mul_operand): Likewise.
15731 (aarch64_sve_vec_perm_operand): Likewise.
15732 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15733 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15734 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15735 as well as const_vector.
15736 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15737 in file. Use CONST0_RTX and CONSTM1_RTX.
15738 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15739 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15740 Use aarch64_simd_imm_zero.
15741 * config/aarch64/aarch64-sve.md: New file.
15742 * config/aarch64/aarch64.md: Include it.
15743 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15744 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15745 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15746 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15747 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15748 (sve): New attribute.
15749 (enabled): Disable instructions with the sve attribute unless
15751 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15752 aarch64_expand_mov_immediate.
15753 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15754 CNT[BHSD] immediates.
15755 (movti): Split CONST_POLY_INT moves into two halves.
15756 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15757 Split additions that need a temporary here if the destination
15758 is the stack pointer.
15759 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15760 (*add<mode>3_poly_1): New instruction.
15761 (set_clobber_cc): New expander.
15763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15765 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15766 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15767 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15768 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15769 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15770 Change innermode from fixed_mode_size to machine_mode.
15771 (simplify_subreg): Update call accordingly. Handle a constant-sized
15772 subreg of a variable-length CONST_VECTOR.
15774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15775 Alan Hayward <alan.hayward@arm.com>
15776 David Sherwood <david.sherwood@arm.com>
15778 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15779 (add_offset_to_base): New function, split out from...
15780 (create_mem_ref): ...here. When handling a scale other than 1,
15781 check first whether the address is valid without the offset.
15782 Add it into the base if so, leaving the index and scale as-is.
15784 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15787 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15788 fold_for_warn before checking if arg2 is INTEGER_CST.
15790 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15792 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15793 (store_multiple_operation): Delete.
15794 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15795 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15796 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15797 guarded by TARGET_STRING.
15798 (rs6000_output_load_multiple): Delete.
15799 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15800 OPTION_MASK_STRING / TARGET_STRING handling.
15801 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15802 (const rs6000_opt_masks) <"string">: Change mask to 0.
15803 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15804 (MASK_STRING): Delete.
15805 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15807 (load_multiple): Delete.
15814 (store_multiple): Delete.
15821 (movmemsi_8reg): Delete.
15822 (corresponding unnamed define_insn): Delete.
15823 (movmemsi_6reg): Delete.
15824 (corresponding unnamed define_insn): Delete.
15825 (movmemsi_4reg): Delete.
15826 (corresponding unnamed define_insn): Delete.
15827 (movmemsi_2reg): Delete.
15828 (corresponding unnamed define_insn): Delete.
15829 (movmemsi_1reg): Delete.
15830 (corresponding unnamed define_insn): Delete.
15831 * config/rs6000/rs6000.opt (mno-string): New.
15832 (mstring): Replace by deprecation warning stub.
15833 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15835 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15837 * regrename.c (regrename_do_replace): If replacing the same
15838 reg multiple times, try to reuse last created gen_raw_REG.
15841 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15842 main to workaround a bug in GDB.
15844 2018-01-12 Tom de Vries <tom@codesourcery.com>
15847 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15849 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15851 PR rtl-optimization/80481
15852 * ira-color.c (get_cap_member): New function.
15853 (allocnos_conflict_by_live_ranges_p): Use it.
15854 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15855 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15857 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15860 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15861 (*saddl_se_1): Ditto.
15862 (*ssubsi_1): Ditto.
15863 (*ssubl_se_1): Ditto.
15865 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15867 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15868 rather than wi::to_widest for DR_INITs.
15869 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15870 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15871 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15873 (vect_analyze_group_access_1): Note that here.
15875 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15877 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15878 polynomial type sizes.
15880 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15882 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15883 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15884 (gimple_add_tmp_var): Likewise.
15886 2018-01-12 Martin Liska <mliska@suse.cz>
15888 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15889 (gimple_alloc_sizes): Likewise.
15890 (dump_gimple_statistics): Use PRIu64 in printf format.
15891 * gimple.h: Change uint64_t to int.
15893 2018-01-12 Martin Liska <mliska@suse.cz>
15895 * tree-core.h: Use uint64_t instead of int.
15896 * tree.c (tree_node_counts): Likewise.
15897 (tree_node_sizes): Likewise.
15898 (dump_tree_statistics): Use PRIu64 in printf format.
15900 2018-01-12 Martin Liska <mliska@suse.cz>
15902 * Makefile.in: As qsort_chk is implemented in vec.c, add
15903 vec.o to linkage of gencfn-macros.
15904 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15905 passing the info to record_node_allocation_statistics.
15906 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15908 * ggc-common.c (struct ggc_usage): Add operator== and use
15909 it in operator< and compare function.
15910 * mem-stats.h (struct mem_usage): Likewise.
15911 * vec.c (struct vec_usage): Remove operator< and compare
15912 function. Can be simply inherited.
15914 2018-01-12 Martin Jambor <mjambor@suse.cz>
15917 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15918 * tree-ssa-math-opts.c: Include domwalk.h.
15919 (convert_mult_to_fma_1): New function.
15920 (fma_transformation_info): New type.
15921 (fma_deferring_state): Likewise.
15922 (cancel_fma_deferring): New function.
15923 (result_of_phi): Likewise.
15924 (last_fma_candidate_feeds_initial_phi): Likewise.
15925 (convert_mult_to_fma): Added deferring logic, split actual
15926 transformation to convert_mult_to_fma_1.
15927 (math_opts_dom_walker): New type.
15928 (math_opts_dom_walker::after_dom_children): New method, body moved
15929 here from pass_optimize_widening_mul::execute, added deferring logic
15931 (pass_optimize_widening_mul::execute): Moved most of code to
15932 math_opts_dom_walker::after_dom_children.
15933 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15934 * config/i386/i386.c (ix86_option_override_internal): Added
15935 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15937 2018-01-12 Richard Biener <rguenther@suse.de>
15940 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15941 inline instance vars.
15943 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15946 * config/rx/rx.c (rx_is_restricted_memory_address):
15947 Handle SUBREG case.
15949 2018-01-12 Richard Biener <rguenther@suse.de>
15951 PR tree-optimization/80846
15952 * target.def (split_reduction): New target hook.
15953 * targhooks.c (default_split_reduction): New function.
15954 * targhooks.h (default_split_reduction): Declare.
15955 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15956 target requests first reduce vectors by combining low and high
15958 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15959 (get_vectype_for_scalar_type_and_size): Export.
15960 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15961 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15962 * doc/tm.texi: Regenerate.
15963 * config/i386/i386.c (ix86_split_reduction): Implement
15964 TARGET_VECTORIZE_SPLIT_REDUCTION.
15966 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15969 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15970 in PIC mode except for TARGET_VXWORKS_RTP.
15971 * config/sparc/sparc.c: Include cfgrtl.h.
15972 (TARGET_INIT_PIC_REG): Define.
15973 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15974 (sparc_pic_register_p): New predicate.
15975 (sparc_legitimate_address_p): Use it.
15976 (sparc_legitimize_pic_address): Likewise.
15977 (sparc_delegitimize_address): Likewise.
15978 (sparc_mode_dependent_address_p): Likewise.
15979 (gen_load_pcrel_sym): Remove 4th parameter.
15980 (load_got_register): Adjust call to above. Remove obsolete stuff.
15981 (sparc_expand_prologue): Do not call load_got_register here.
15982 (sparc_flat_expand_prologue): Likewise.
15983 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15984 (sparc_use_pseudo_pic_reg): New function.
15985 (sparc_init_pic_reg): Likewise.
15986 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15987 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15989 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15991 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15992 Add item for branch_cost.
15994 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15996 PR rtl-optimization/83565
15997 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15998 not extend the result to a larger mode for rotate operations.
15999 (num_sign_bit_copies1): Likewise.
16001 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16004 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16006 Use values-Xc.o for -pedantic.
16007 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16009 2018-01-12 Martin Liska <mliska@suse.cz>
16012 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16014 (possible_polymorphic_call_targets): Use it.
16015 (ipa_devirt): Likewise.
16017 2018-01-12 Martin Liska <mliska@suse.cz>
16019 * profile-count.h (enum profile_quality): Use 0 as invalid
16020 enum value of profile_quality.
16022 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16024 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16025 -mext-string options.
16027 2018-01-12 Richard Biener <rguenther@suse.de>
16029 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16030 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16031 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16033 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16035 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16037 * configure.ac (--with-long-double-format): Add support for the
16038 configuration option to change the default long double format on
16040 * config.gcc (powerpc*-linux*-*): Likewise.
16041 * configure: Regenerate.
16042 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16043 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16044 used without modification.
16046 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16048 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16049 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16050 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16051 MISC_BUILTIN_SPEC_BARRIER.
16052 (rs6000_init_builtins): Likewise.
16053 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16055 (speculation_barrier): New define_insn.
16056 * doc/extend.texi: Document __builtin_speculation_barrier.
16058 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16061 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16062 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16063 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16065 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16066 integral modes instead of "ss" and "sd".
16067 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16068 vectors with 32-bit and 64-bit elements.
16069 (vecdupssescalarmodesuffix): New mode attribute.
16070 (vec_dup<mode>): Use it.
16072 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16075 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16076 frame if argument is passed on stack.
16078 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16081 * ree.c (combine_reaching_defs): Optimize also
16082 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16083 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16085 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16087 PR middle-end/83189
16088 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16090 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16092 PR middle-end/83718
16093 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16094 after they are computed.
16096 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16098 PR tree-optimization/83695
16099 * gimple-loop-linterchange.cc
16100 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16101 reset cached scev information after interchange.
16102 (pass_linterchange::execute): Remove call to scev_reset_htab.
16104 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16106 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16107 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16108 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16109 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16110 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16111 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16112 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16113 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16114 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16115 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16116 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16117 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16118 (V_lane_reg): Likewise.
16119 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16121 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16122 (vfmal_lane_low<mode>_intrinsic,
16123 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16124 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16125 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16126 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16127 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16128 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16130 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16132 * config/arm/arm-cpus.in (fp16fml): New feature.
16133 (ALL_SIMD): Add fp16fml.
16134 (armv8.2-a): Add fp16fml as an option.
16135 (armv8.3-a): Likewise.
16136 (armv8.4-a): Add fp16fml as part of fp16.
16137 * config/arm/arm.h (TARGET_FP16FML): Define.
16138 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16140 * config/arm/arm-modes.def (V2HF): Define.
16141 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16142 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16143 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16144 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16145 vfmsl_low, vfmsl_high): New set of builtins.
16146 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16147 (vfml_op): New code attribute.
16148 (VFMLHALVES): New int iterator.
16149 (VFML, VFMLSEL): New mode attributes.
16150 (V_reg): Define mapping for V2HF.
16151 (V_hi, V_lo): New mode attributes.
16152 (VF_constraint): Likewise.
16153 (vfml_half, vfml_half_selector): New int attributes.
16154 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16156 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16157 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16159 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16160 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16161 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16162 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16164 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16165 Document new effective target and option set.
16167 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16169 * config/arm/arm-cpus.in (armv8_4): New feature.
16170 (ARMv8_4a): New fgroup.
16171 (armv8.4-a): New arch.
16172 * config/arm/arm-tables.opt: Regenerate.
16173 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16174 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16175 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16176 Add matching rules for -march=armv8.4-a and extensions.
16177 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16179 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16182 * config/rx/rx.md (BW): New mode attribute.
16183 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16185 2018-01-11 Richard Biener <rguenther@suse.de>
16187 PR tree-optimization/83435
16188 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16189 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16190 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16192 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16193 Alan Hayward <alan.hayward@arm.com>
16194 David Sherwood <david.sherwood@arm.com>
16196 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16198 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16199 (aarch64_print_address_internal): Use it to check for a zero offset.
16201 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16202 Alan Hayward <alan.hayward@arm.com>
16203 David Sherwood <david.sherwood@arm.com>
16205 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16206 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16207 Return a poly_int64 rather than a HOST_WIDE_INT.
16208 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16209 rather than a HOST_WIDE_INT.
16210 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16211 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16212 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16213 final_offset from HOST_WIDE_INT to poly_int64.
16214 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16215 to_constant when getting the number of units in an Advanced SIMD
16217 (aarch64_builtin_vectorized_function): Check for a constant number
16219 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16221 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16222 attribute instead of GET_MODE_NUNITS.
16223 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16224 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16225 GET_MODE_SIZE for fixed-size registers.
16226 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16227 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16228 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16229 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16230 (aarch64_print_operand, aarch64_print_address_internal)
16231 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16232 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16233 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16234 Handle polynomial GET_MODE_SIZE.
16235 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16236 wider than SImode without modification.
16237 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16238 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16239 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16240 passing and returning SVE modes.
16241 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16242 rather than GEN_INT.
16243 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16244 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16245 (aarch64_allocate_and_probe_stack_space): Likewise.
16246 (aarch64_layout_frame): Cope with polynomial offsets.
16247 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16248 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16249 polynomial offsets.
16250 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16251 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16252 poly_int64 rather than a HOST_WIDE_INT.
16253 (aarch64_get_separate_components, aarch64_process_components)
16254 (aarch64_expand_prologue, aarch64_expand_epilogue)
16255 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16256 (aarch64_anchor_offset): New function, split out from...
16257 (aarch64_legitimize_address): ...here.
16258 (aarch64_builtin_vectorization_cost): Handle polynomial
16259 TYPE_VECTOR_SUBPARTS.
16260 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16262 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16263 number of elements from the PARALLEL rather than the mode.
16264 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16265 rather than GET_MODE_BITSIZE.
16266 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16267 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16268 (aarch64_expand_vec_perm_const_1): Handle polynomial
16269 d->perm.length () and d->perm elements.
16270 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16271 Apply to_constant to d->perm elements.
16272 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16273 polynomial CONST_VECTOR_NUNITS.
16274 (aarch64_move_pointer): Take amount as a poly_int64 rather
16276 (aarch64_progress_pointer): Avoid temporary variable.
16277 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16278 the mode attribute instead of GET_MODE.
16280 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16281 Alan Hayward <alan.hayward@arm.com>
16282 David Sherwood <david.sherwood@arm.com>
16284 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16285 x exists before using it.
16286 (aarch64_add_constant_internal): Rename to...
16287 (aarch64_add_offset_1): ...this. Replace regnum with separate
16288 src and dest rtxes. Handle the case in which they're different,
16289 including when the offset is zero. Replace scratchreg with an rtx.
16290 Use 2 additions if there is no spare register into which we can
16291 move a 16-bit constant.
16292 (aarch64_add_constant): Delete.
16293 (aarch64_add_offset): Replace reg with separate src and dest
16294 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16295 Use aarch64_add_offset_1.
16296 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16297 an rtx rather than an int. Take the delta as a poly_int64
16298 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16299 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16300 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16301 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16302 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16303 and aarch64_add_sp.
16304 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16305 aarch64_add_constant.
16307 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16309 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16310 Use scalar_float_mode.
16312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16314 * config/aarch64/aarch64-simd.md
16315 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16316 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16317 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16318 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16319 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16320 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16321 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16322 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16323 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16324 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16326 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16329 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16330 targ_options->x_arm_arch_string is non NULL.
16332 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16334 * config/aarch64/aarch64.h
16335 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16337 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16340 * expmed.c (emit_store_flag_force): Swap if const op0
16341 and change VOIDmode to mode of op0.
16343 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16345 PR rtl-optimization/83761
16346 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16347 than bytes to mode_for_size.
16349 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16351 PR middle-end/83189
16352 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16353 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16356 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16358 PR middle-end/83575
16359 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16360 when in layout mode.
16361 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16362 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16365 2018-01-10 Michael Collison <michael.collison@arm.com>
16367 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16368 * config/aarch64/aarch64-option-extension.def: Add
16369 AARCH64_OPT_EXTENSION of 'fp16fml'.
16370 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16371 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16372 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16373 * config/aarch64/constraints.md (Ui7): New constraint.
16374 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16375 (VFMLA_SEL_W): Ditto.
16378 (VFMLA16_LOW): New int iterator.
16379 (VFMLA16_HIGH): Ditto.
16380 (UNSPEC_FMLAL): New unspec.
16381 (UNSPEC_FMLSL): Ditto.
16382 (UNSPEC_FMLAL2): Ditto.
16383 (UNSPEC_FMLSL2): Ditto.
16384 (f16mac): New code attribute.
16385 * config/aarch64/aarch64-simd-builtins.def
16386 (aarch64_fmlal_lowv2sf): Ditto.
16387 (aarch64_fmlsl_lowv2sf): Ditto.
16388 (aarch64_fmlalq_lowv4sf): Ditto.
16389 (aarch64_fmlslq_lowv4sf): Ditto.
16390 (aarch64_fmlal_highv2sf): Ditto.
16391 (aarch64_fmlsl_highv2sf): Ditto.
16392 (aarch64_fmlalq_highv4sf): Ditto.
16393 (aarch64_fmlslq_highv4sf): Ditto.
16394 (aarch64_fmlal_lane_lowv2sf): Ditto.
16395 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16396 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16397 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16398 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16399 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16400 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16401 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16402 (aarch64_fmlal_lane_highv2sf): Ditto.
16403 (aarch64_fmlsl_lane_highv2sf): Ditto.
16404 (aarch64_fmlal_laneq_highv2sf): Ditto.
16405 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16406 (aarch64_fmlalq_lane_highv4sf): Ditto.
16407 (aarch64_fmlsl_lane_highv4sf): Ditto.
16408 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16409 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16410 * config/aarch64/aarch64-simd.md:
16411 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16412 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16413 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16414 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16415 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16416 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16417 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16418 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16419 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16420 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16421 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16422 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16423 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16424 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16425 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16426 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16427 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16428 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16429 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16430 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16431 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16432 (vfmlsl_low_u32): Ditto.
16433 (vfmlalq_low_u32): Ditto.
16434 (vfmlslq_low_u32): Ditto.
16435 (vfmlal_high_u32): Ditto.
16436 (vfmlsl_high_u32): Ditto.
16437 (vfmlalq_high_u32): Ditto.
16438 (vfmlslq_high_u32): Ditto.
16439 (vfmlal_lane_low_u32): Ditto.
16440 (vfmlsl_lane_low_u32): Ditto.
16441 (vfmlal_laneq_low_u32): Ditto.
16442 (vfmlsl_laneq_low_u32): Ditto.
16443 (vfmlalq_lane_low_u32): Ditto.
16444 (vfmlslq_lane_low_u32): Ditto.
16445 (vfmlalq_laneq_low_u32): Ditto.
16446 (vfmlslq_laneq_low_u32): Ditto.
16447 (vfmlal_lane_high_u32): Ditto.
16448 (vfmlsl_lane_high_u32): Ditto.
16449 (vfmlal_laneq_high_u32): Ditto.
16450 (vfmlsl_laneq_high_u32): Ditto.
16451 (vfmlalq_lane_high_u32): Ditto.
16452 (vfmlslq_lane_high_u32): Ditto.
16453 (vfmlalq_laneq_high_u32): Ditto.
16454 (vfmlslq_laneq_high_u32): Ditto.
16455 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16456 (AARCH64_FL_FOR_ARCH8_4): New.
16457 (AARCH64_ISA_F16FML): New ISA flag.
16458 (TARGET_F16FML): New feature flag for fp16fml.
16459 (doc/invoke.texi): Document new fp16fml option.
16461 2018-01-10 Michael Collison <michael.collison@arm.com>
16463 * config/aarch64/aarch64-builtins.c:
16464 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16465 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16466 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16467 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16468 (AARCH64_ISA_SHA3): New ISA flag.
16469 (TARGET_SHA3): New feature flag for sha3.
16470 * config/aarch64/iterators.md (sha512_op): New int attribute.
16471 (CRYPTO_SHA512): New int iterator.
16472 (UNSPEC_SHA512H): New unspec.
16473 (UNSPEC_SHA512H2): Ditto.
16474 (UNSPEC_SHA512SU0): Ditto.
16475 (UNSPEC_SHA512SU1): Ditto.
16476 * config/aarch64/aarch64-simd-builtins.def
16477 (aarch64_crypto_sha512hqv2di): New builtin.
16478 (aarch64_crypto_sha512h2qv2di): Ditto.
16479 (aarch64_crypto_sha512su0qv2di): Ditto.
16480 (aarch64_crypto_sha512su1qv2di): Ditto.
16481 (aarch64_eor3qv8hi): Ditto.
16482 (aarch64_rax1qv2di): Ditto.
16483 (aarch64_xarqv2di): Ditto.
16484 (aarch64_bcaxqv8hi): Ditto.
16485 * config/aarch64/aarch64-simd.md:
16486 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16487 (aarch64_crypto_sha512su0qv2di): Ditto.
16488 (aarch64_crypto_sha512su1qv2di): Ditto.
16489 (aarch64_eor3qv8hi): Ditto.
16490 (aarch64_rax1qv2di): Ditto.
16491 (aarch64_xarqv2di): Ditto.
16492 (aarch64_bcaxqv8hi): Ditto.
16493 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16494 (vsha512h2q_u64): Ditto.
16495 (vsha512su0q_u64): Ditto.
16496 (vsha512su1q_u64): Ditto.
16497 (veor3q_u16): Ditto.
16498 (vrax1q_u64): Ditto.
16499 (vxarq_u64): Ditto.
16500 (vbcaxq_u16): Ditto.
16501 * config/arm/types.md (crypto_sha512): New type attribute.
16502 (crypto_sha3): Ditto.
16503 (doc/invoke.texi): Document new sha3 option.
16505 2018-01-10 Michael Collison <michael.collison@arm.com>
16507 * config/aarch64/aarch64-builtins.c:
16508 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16509 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16510 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16511 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16512 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16513 (AARCH64_ISA_SM4): New ISA flag.
16514 (TARGET_SM4): New feature flag for sm4.
16515 * config/aarch64/aarch64-simd-builtins.def
16516 (aarch64_sm3ss1qv4si): Ditto.
16517 (aarch64_sm3tt1aq4si): Ditto.
16518 (aarch64_sm3tt1bq4si): Ditto.
16519 (aarch64_sm3tt2aq4si): Ditto.
16520 (aarch64_sm3tt2bq4si): Ditto.
16521 (aarch64_sm3partw1qv4si): Ditto.
16522 (aarch64_sm3partw2qv4si): Ditto.
16523 (aarch64_sm4eqv4si): Ditto.
16524 (aarch64_sm4ekeyqv4si): Ditto.
16525 * config/aarch64/aarch64-simd.md:
16526 (aarch64_sm3ss1qv4si): Ditto.
16527 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16528 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16529 (aarch64_sm4eqv4si): Ditto.
16530 (aarch64_sm4ekeyqv4si): Ditto.
16531 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16532 (sm3part_op): Ditto.
16533 (CRYPTO_SM3TT): Ditto.
16534 (CRYPTO_SM3PART): Ditto.
16535 (UNSPEC_SM3SS1): New unspec.
16536 (UNSPEC_SM3TT1A): Ditto.
16537 (UNSPEC_SM3TT1B): Ditto.
16538 (UNSPEC_SM3TT2A): Ditto.
16539 (UNSPEC_SM3TT2B): Ditto.
16540 (UNSPEC_SM3PARTW1): Ditto.
16541 (UNSPEC_SM3PARTW2): Ditto.
16542 (UNSPEC_SM4E): Ditto.
16543 (UNSPEC_SM4EKEY): Ditto.
16544 * config/aarch64/constraints.md (Ui2): New constraint.
16545 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16546 * config/arm/types.md (crypto_sm3): New type attribute.
16547 (crypto_sm4): Ditto.
16548 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16549 (vsm3tt1aq_u32): Ditto.
16550 (vsm3tt1bq_u32): Ditto.
16551 (vsm3tt2aq_u32): Ditto.
16552 (vsm3tt2bq_u32): Ditto.
16553 (vsm3partw1q_u32): Ditto.
16554 (vsm3partw2q_u32): Ditto.
16555 (vsm4eq_u32): Ditto.
16556 (vsm4ekeyq_u32): Ditto.
16557 (doc/invoke.texi): Document new sm4 option.
16559 2018-01-10 Michael Collison <michael.collison@arm.com>
16561 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16562 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16563 (AARCH64_FL_FOR_ARCH8_4): New.
16564 (AARCH64_FL_V8_4): New flag.
16565 (doc/invoke.texi): Document new armv8.4-a option.
16567 2018-01-10 Michael Collison <michael.collison@arm.com>
16569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16570 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16571 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16572 * config/aarch64/aarch64-option-extension.def: Add
16573 AARCH64_OPT_EXTENSION of 'sha2'.
16574 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16575 (crypto): Disable sha2 and aes if crypto disabled.
16576 (crypto): Enable aes and sha2 if enabled.
16577 (simd): Disable sha2 and aes if simd disabled.
16578 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16580 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16581 (TARGET_SHA2): New feature flag for sha2.
16582 (TARGET_AES): New feature flag for aes.
16583 * config/aarch64/aarch64-simd.md:
16584 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16585 conditional on TARGET_AES.
16586 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16587 (aarch64_crypto_sha1hsi): Make pattern conditional
16589 (aarch64_crypto_sha1hv4si): Ditto.
16590 (aarch64_be_crypto_sha1hv4si): Ditto.
16591 (aarch64_crypto_sha1su1v4si): Ditto.
16592 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16593 (aarch64_crypto_sha1su0v4si): Ditto.
16594 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16595 (aarch64_crypto_sha256su0v4si): Ditto.
16596 (aarch64_crypto_sha256su1v4si): Ditto.
16597 (doc/invoke.texi): Document new aes and sha2 options.
16599 2018-01-10 Martin Sebor <msebor@redhat.com>
16601 PR tree-optimization/83781
16602 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16605 2018-01-11 Martin Sebor <msebor@gmail.com>
16606 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16608 PR tree-optimization/83501
16609 PR tree-optimization/81703
16611 * tree-ssa-strlen.c (get_string_cst): Rename...
16612 (get_string_len): ...to this. Handle global constants.
16613 (handle_char_store): Adjust.
16615 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16616 Jim Wilson <jimw@sifive.com>
16618 * config/riscv/riscv-protos.h (riscv_output_return): New.
16619 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16620 (riscv_attribute_table, riscv_output_return),
16621 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16622 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16623 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16625 (riscv_expand_prologue): Add early return for naked function.
16626 (riscv_expand_epilogue): Likewise.
16627 (riscv_function_ok_for_sibcall): Return false for naked function.
16628 (riscv_set_current_function): New.
16629 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16630 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16631 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16632 * doc/extend.texi (RISC-V Function Attributes): New.
16634 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16636 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16637 check for 128-bit long double before checking TCmode.
16638 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16639 128-bit long doubles before checking TFmode or TCmode.
16640 (FLOAT128_IBM_P): Likewise.
16642 2018-01-10 Martin Sebor <msebor@redhat.com>
16644 PR tree-optimization/83671
16645 * builtins.c (c_strlen): Unconditionally return zero for the empty
16647 Use -Warray-bounds for warnings.
16648 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16649 for non-constant array indices with COMPONENT_REF, arrays of
16650 arrays, and pointers to arrays.
16651 (gimple_fold_builtin_strlen): Determine and set length range for
16652 non-constant character arrays.
16654 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16656 PR middle-end/81897
16657 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16660 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16662 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16664 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16667 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16668 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16669 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16670 indexed_or_indirect_operand predicate.
16671 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16672 (*vsx_le_perm_load_v8hi): Likewise.
16673 (*vsx_le_perm_load_v16qi): Likewise.
16674 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16675 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16676 (*vsx_le_perm_store_v8hi): Likewise.
16677 (*vsx_le_perm_store_v16qi): Likewise.
16678 (eight unnamed splitters): Likewise.
16680 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16682 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16683 * config/rs6000/emmintrin.h: Likewise.
16684 * config/rs6000/mmintrin.h: Likewise.
16685 * config/rs6000/xmmintrin.h: Likewise.
16687 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16690 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16692 * tree.c (tree_nop_conversion): Return true for location wrapper
16694 (maybe_wrap_with_location): New function.
16695 (selftest::check_strip_nops): New function.
16696 (selftest::test_location_wrappers): New function.
16697 (selftest::tree_c_tests): Call it.
16698 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16699 (maybe_wrap_with_location): New decl.
16700 (EXPR_LOCATION_WRAPPER_P): New macro.
16701 (location_wrapper_p): New inline function.
16702 (tree_strip_any_location_wrapper): New inline function.
16704 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16707 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16708 stack_realign_offset for the largest alignment of stack slot
16710 (ix86_find_max_used_stack_alignment): New function.
16711 (ix86_finalize_stack_frame_flags): Use it. Set
16712 max_used_stack_alignment if we don't realign stack.
16713 * config/i386/i386.h (machine_function): Add
16714 max_used_stack_alignment.
16716 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16718 * config/arm/arm.opt (-mbranch-cost): New option.
16719 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16722 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16725 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16726 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16728 2018-01-10 Richard Biener <rguenther@suse.de>
16731 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16732 early out so it also covers the case where we have a non-NULL
16735 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16737 PR tree-optimization/83753
16738 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16739 for non-strided grouped accesses if the number of elements is 1.
16741 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16744 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16745 * i386.h (TARGET_USE_GATHER): Define.
16746 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16748 2018-01-10 Martin Liska <mliska@suse.cz>
16751 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16752 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16754 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16755 CLEANUP_NO_PARTITIONING is not set.
16757 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16759 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16760 for vectors, as a partial revert of r254296.
16761 * rtl.h (const_vec_p): Delete.
16762 (const_vec_duplicate_p): Don't test for vector CONSTs.
16763 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16764 * expmed.c (make_tree): Likewise.
16767 * common.md (E, F): Use CONSTANT_P instead of checking for
16769 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16770 checking for CONST_VECTOR.
16772 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16774 PR middle-end/83575
16775 * predict.c (force_edge_cold): Handle in more sane way edges
16776 with no prediction.
16778 2018-01-09 Carl Love <cel@us.ibm.com>
16780 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16782 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16783 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16784 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16785 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16786 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16787 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16788 * config/rs6000/rs6000-protos.h: Add extern defition for
16789 rs6000_generate_float2_double_code.
16790 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16792 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16793 (float2_v2df): Add define_expand.
16795 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16798 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16799 op_mode in the force_to_mode call.
16801 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16803 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16804 instead of checking each element individually.
16805 (aarch64_evpc_uzp): Likewise.
16806 (aarch64_evpc_zip): Likewise.
16807 (aarch64_evpc_ext): Likewise.
16808 (aarch64_evpc_rev): Likewise.
16809 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16810 instead of checking each element individually. Return true without
16812 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16813 whether all selected elements come from the same input, instead of
16814 checking each element individually. Remove calls to gen_rtx_REG,
16815 start_sequence and end_sequence and instead assert that no rtl is
16818 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16820 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16821 order of HIGH and CONST checks.
16823 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16825 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16826 if the destination isn't an SSA_NAME.
16828 2018-01-09 Richard Biener <rguenther@suse.de>
16830 PR tree-optimization/83668
16831 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16833 (canonicalize_loop_form): ... here, renamed from ...
16834 (canonicalize_loop_closed_ssa_form): ... this and amended to
16835 swap successor edges for loop exit blocks to make us use
16836 the RPO order we need for initial schedule generation.
16838 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16840 PR tree-optimization/64811
16841 * match.pd: When optimizing comparisons with Inf, avoid
16842 introducing or losing exceptions from comparisons with NaN.
16844 2018-01-09 Martin Liska <mliska@suse.cz>
16847 * asan.c (shadow_mem_size): Add gcc_assert.
16849 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16851 Don't save registers in main().
16854 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16855 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16856 * config/avr/avr.c (avr_set_current_function): Don't error if
16857 naked, OS_task or OS_main are specified at the same time.
16858 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16860 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16862 * common/config/avr/avr-common.c (avr_option_optimization_table):
16863 Switch on -mmain-is-OS_task for optimizing compilations.
16865 2018-01-09 Richard Biener <rguenther@suse.de>
16867 PR tree-optimization/83572
16868 * graphite.c: Include cfganal.h.
16869 (graphite_transform_loops): Connect infinite loops to exit
16870 and remove fake edges at the end.
16872 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16874 * ipa-inline.c (edge_badness): Revert accidental checkin.
16876 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16879 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16880 symbols; not inline clones.
16882 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16885 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16886 hard registers. Formatting fixes.
16888 PR preprocessor/83722
16889 * gcc.c (try_generate_repro): Pass
16890 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16891 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16894 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16895 Kito Cheng <kito.cheng@gmail.com>
16897 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16898 (riscv_leaf_function_p): Delete.
16899 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16901 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16903 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16905 (do_ifelse): New function.
16906 (do_isel): New function.
16907 (do_sub3): New function.
16908 (do_add3): New function.
16909 (do_load_mask_compare): New function.
16910 (do_overlap_load_compare): New function.
16911 (expand_compare_loop): New function.
16912 (expand_block_compare): Call expand_compare_loop() when appropriate.
16913 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16914 option description.
16915 (-mblock-compare-inline-loop-limit): New option.
16917 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16920 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16921 Reverse order of second and third operands in first alternative.
16922 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16923 of first and second elements in UNSPEC_VPERMR vector.
16924 (altivec_expand_vec_perm_le): Likewise.
16926 2018-01-08 Jeff Law <law@redhat.com>
16928 PR rtl-optimizatin/81308
16929 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16930 (process_switch): If group_case_labels makes a change, then set
16932 (pass_convert_switch::execute): If a switch is converted, then
16933 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16935 PR rtl-optimization/81308
16936 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16939 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16941 PR target/83663 - Revert r255946
16942 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16943 generation for cases where splatting a value is not useful.
16944 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16945 across a vec_duplicate and a paradoxical subreg forming a vector
16946 mode to a vec_concat.
16948 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16950 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16951 -march=armv8.3-a variants.
16952 * config/arm/t-multilib: Likewise.
16953 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16955 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16957 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16959 (cceq_ior_compare_complement): Give it a name so I can use it, and
16960 change boolean_or_operator predicate to boolean_operator so it can
16961 be used to generate a crand.
16962 (eqne): New code iterator.
16963 (bd/bd_neg): New code_attrs.
16964 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16965 a single define_insn.
16966 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16967 decrement (bdnzt/bdnzf/bdzt/bdzf).
16968 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16969 with the new names of the branch decrement patterns, and added the
16970 names of the branch decrement conditional patterns.
16972 2018-01-08 Richard Biener <rguenther@suse.de>
16974 PR tree-optimization/83563
16975 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16978 2018-01-08 Richard Biener <rguenther@suse.de>
16980 PR middle-end/83713
16981 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16983 2018-01-08 Richard Biener <rguenther@suse.de>
16985 PR tree-optimization/83685
16986 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16987 references to abnormals.
16989 2018-01-08 Richard Biener <rguenther@suse.de>
16992 * dwarf2out.c (output_indirect_strings): Handle empty
16993 skeleton_debug_str_hash.
16994 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16996 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16998 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16999 (emit_store_direct): Likewise.
17000 (arc_trampoline_adjust_address): Likewise.
17001 (arc_asm_trampoline_template): New function.
17002 (arc_initialize_trampoline): Use asm_trampoline_template.
17003 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17004 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17005 * config/arc/arc.md (flush_icache): Delete pattern.
17007 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17009 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17010 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17013 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17016 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17017 by not USED_FOR_TARGET.
17018 (make_pass_resolve_sw_modes): Likewise.
17020 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17022 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17025 2018-01-08 Richard Biener <rguenther@suse.de>
17027 PR middle-end/83580
17028 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17030 2018-01-08 Richard Biener <rguenther@suse.de>
17032 PR middle-end/83517
17033 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17035 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17037 PR middle-end/81897
17038 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17039 basic blocks with a small number of successors.
17040 (convert_control_dep_chain_into_preds): Improve handling of
17042 (dump_predicates): Split apart into...
17043 (dump_pred_chain): ...here...
17044 (dump_pred_info): ...and here.
17045 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17046 (can_chain_union_be_invalidated_p): Improve check for invalidation
17048 (uninit_uses_cannot_happen): Avoid unnecessary if
17049 convert_control_dep_chain_into_preds yielded nothing.
17051 2018-01-06 Martin Sebor <msebor@redhat.com>
17053 PR tree-optimization/83640
17054 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17055 subtracting negative offset from size.
17056 (builtin_access::overlap): Adjust offset bounds of the access to fall
17057 within the size of the object if possible.
17059 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17061 PR rtl-optimization/83699
17062 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17063 extract_bit_field_as_subreg to cases in which the extracted
17064 value is also a vector.
17066 * lra-constraints.c (process_alt_operands): Test for the equivalence
17067 substitutions when detecting a possible reload cycle.
17069 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17072 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17073 by default if flag_selective_schedling{,2}. Formatting fixes.
17075 PR rtl-optimization/83682
17076 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17077 if it has non-VECTOR_MODE element mode.
17078 (vec_duplicate_p): Likewise.
17080 PR middle-end/83694
17081 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17082 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17084 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17087 * config/i386/i386-builtin.def
17088 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17089 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17090 Require also OPTION_MASK_ISA_AVX512F in addition to
17091 OPTION_MASK_ISA_GFNI.
17092 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17093 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17094 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17095 to OPTION_MASK_ISA_GFNI.
17096 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17097 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17098 OPTION_MASK_ISA_AVX512BW.
17099 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17100 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17101 addition to OPTION_MASK_ISA_GFNI.
17102 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17103 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17104 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17105 to OPTION_MASK_ISA_GFNI.
17106 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17107 a requirement for all ISAs rather than any of them with a few
17109 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17111 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17112 bitmasks to be enabled with 3 exceptions, instead of requiring any
17113 enabled ISA with lots of exceptions.
17114 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17115 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17116 Change avx512bw in isa attribute to avx512f.
17117 * config/i386/sgxintrin.h: Add license boilerplate.
17118 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17119 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17120 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17121 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17123 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17124 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17125 temporarily sse2 rather than sse if not enabled already.
17128 * config/i386/sse.md (VI248_VLBW): Rename to ...
17129 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17130 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17131 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17132 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17133 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17134 mode iterator instead of VI248_VLBW.
17136 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17138 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17139 (record_modified): Skip clobbers; add debug output.
17140 (param_change_prob): Use sreal frequencies.
17142 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17144 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17145 punt for user-aligned variables.
17147 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17149 * tree-chrec.c (chrec_contains_symbols): Return true for
17152 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17155 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17156 of (x|y) == x for BICS pattern.
17158 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17160 PR tree-optimization/83605
17161 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17162 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17165 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17167 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17168 * config/epiphany/rtems.h: New file.
17170 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17171 Uros Bizjak <ubizjak@gmail.com>
17174 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17175 QIreg_operand instead of register_operand predicate.
17176 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17177 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17178 comments instead of -fmitigate[-_]rop.
17180 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17183 * cgraphunit.c (symbol_table::compile): Switch to text_section
17184 before calling assembly_start debug hook.
17185 * run-rtl-passes.c (run_rtl_passes): Likewise.
17188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17190 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17191 range_int_cst_p rather than !symbolic_range_p before calling
17192 extract_range_from_multiplicative_op_1.
17194 2018-01-04 Jeff Law <law@redhat.com>
17196 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17197 redundant test in assertion.
17199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17201 * doc/rtl.texi: Document machine_mode wrapper classes.
17203 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17205 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17206 using tree_to_uhwi.
17208 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17210 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17211 the VEC_PERM_EXPR fold to fail.
17213 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17216 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17217 to switched_sections.
17219 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17222 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17223 test for d.testing.
17225 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17228 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17229 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17231 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17234 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17235 is BLKmode and bitpos not zero or mode change is needed.
17237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17240 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17243 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17246 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17247 instead of MULT rtx. Update all corresponding splitters.
17248 (*saddl_se): Ditto.
17249 (*ssub<modesuffix>): Ditto.
17250 (*ssubl_se): Ditto.
17251 (*cmp_sadd_di): Update split patterns.
17252 (*cmp_sadd_si): Ditto.
17253 (*cmp_sadd_sidi): Ditto.
17254 (*cmp_ssub_di): Ditto.
17255 (*cmp_ssub_si): Ditto.
17256 (*cmp_ssub_sidi): Ditto.
17257 * config/alpha/predicates.md (const23_operand): New predicate.
17258 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17259 Look for ASHIFT, not MULT inner operand.
17260 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17262 2018-01-04 Martin Liska <mliska@suse.cz>
17264 PR gcov-profile/83669
17265 * gcov.c (output_intermediate_file): Add version to intermediate
17267 * doc/gcov.texi: Document new field 'version' in intermediate
17268 file format. Fix location of '-k' option of gcov command.
17270 2018-01-04 Martin Liska <mliska@suse.cz>
17273 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17275 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17277 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17279 2018-01-03 Martin Sebor <msebor@redhat.com>
17281 PR tree-optimization/83655
17282 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17283 checking calls with invalid arguments.
17285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17287 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17288 (vectorizable_mask_load_store): Delete.
17289 (vectorizable_call): Return false for masked loads and stores.
17290 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17291 instead of gimple_assign_rhs1.
17292 (vectorizable_load): Handle IFN_MASK_LOAD.
17293 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17297 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17299 (vectorizable_mask_load_store): ...here.
17300 (vectorizable_load): ...and here.
17302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17304 * tree-vect-stmts.c (vect_build_all_ones_mask)
17305 (vect_build_zero_merge_argument): New functions, split out from...
17306 (vectorizable_load): ...here.
17308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17310 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17312 (vectorizable_mask_load_store): ...here.
17313 (vectorizable_store): ...and here.
17315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17317 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17319 (vectorizable_mask_load_store): ...here.
17321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17323 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17324 (vect_model_store_cost): Take a vec_load_store_type instead of a
17326 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17327 (vect_model_store_cost): Take a vec_load_store_type instead of a
17329 (vectorizable_mask_load_store): Update accordingly.
17330 (vectorizable_store): Likewise.
17331 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17335 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17336 IFN_MASK_LOAD calls here rather than...
17337 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17340 Alan Hayward <alan.hayward@arm.com>
17341 David Sherwood <david.sherwood@arm.com>
17343 * expmed.c (extract_bit_field_1): For vector extracts,
17344 fall back to extract_bit_field_as_subreg if vec_extract
17347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17348 Alan Hayward <alan.hayward@arm.com>
17349 David Sherwood <david.sherwood@arm.com>
17351 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17352 they are variable or constant sized.
17353 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17354 slots for constant-sized data.
17356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17357 Alan Hayward <alan.hayward@arm.com>
17358 David Sherwood <david.sherwood@arm.com>
17360 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17361 handling COND_EXPRs with boolean comparisons, try to find a better
17362 basis for the mask type than the boolean itself.
17364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17366 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17367 is calculated and how it can be overridden.
17368 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17369 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17371 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17375 Alan Hayward <alan.hayward@arm.com>
17376 David Sherwood <david.sherwood@arm.com>
17378 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17379 Remove the mode argument.
17380 (aarch64_simd_valid_immediate): Remove the mode and inverse
17382 * config/aarch64/iterators.md (bitsize): New iterator.
17383 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17384 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17385 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17386 aarch64_simd_valid_immediate.
17387 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17388 (aarch64_reg_or_bic_imm): Likewise.
17389 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17390 with an insn_type enum and msl with a modifier_type enum.
17391 Replace element_width with a scalar_mode. Change the shift
17392 to unsigned int. Add constructors for scalar_float_mode and
17393 scalar_int_mode elements.
17394 (aarch64_vect_float_const_representable_p): Delete.
17395 (aarch64_can_const_movi_rtx_p)
17396 (aarch64_simd_scalar_immediate_valid_for_move)
17397 (aarch64_simd_make_constant): Update call to
17398 aarch64_simd_valid_immediate.
17399 (aarch64_advsimd_valid_immediate_hs): New function.
17400 (aarch64_advsimd_valid_immediate): Likewise.
17401 (aarch64_simd_valid_immediate): Remove mode and inverse
17402 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17403 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17404 and aarch64_float_const_representable_p on the result.
17405 (aarch64_output_simd_mov_immediate): Remove mode argument.
17406 Update call to aarch64_simd_valid_immediate and use of
17407 simd_immediate_info.
17408 (aarch64_output_scalar_simd_mov_immediate): Update call
17411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17412 Alan Hayward <alan.hayward@arm.com>
17413 David Sherwood <david.sherwood@arm.com>
17415 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17416 (mode_nunits): Likewise CONST_MODE_NUNITS.
17417 * machmode.def (ADJUST_NUNITS): Document.
17418 * genmodes.c (mode_data::need_nunits_adj): New field.
17419 (blank_mode): Update accordingly.
17420 (adj_nunits): New variable.
17421 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17423 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17424 listed in adj_nunits.
17425 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17426 listed in adj_nunits. Don't emit case statements for such modes.
17427 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17428 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17429 nothing if adj_nunits is nonnull.
17430 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17431 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17432 (emit_mode_fbit): Update use of print_maybe_const_decl.
17433 (emit_move_size): Likewise. Treat the array as non-const
17435 (emit_mode_adjustments): Handle adj_nunits.
17437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17439 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17440 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17441 (VECTOR_MODES): Use it.
17442 (make_vector_modes): Take the prefix as an argument.
17444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17445 Alan Hayward <alan.hayward@arm.com>
17446 David Sherwood <david.sherwood@arm.com>
17448 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17449 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17450 for MODE_VECTOR_BOOL.
17451 * machmode.def (VECTOR_BOOL_MODE): Document.
17452 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17453 (make_vector_bool_mode): New function.
17454 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17456 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17457 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17459 * stor-layout.c (int_mode_for_mode): Likewise.
17460 * tree.c (build_vector_type_for_mode): Likewise.
17461 * varasm.c (output_constant_pool_2): Likewise.
17462 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17463 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17464 for MODE_VECTOR_BOOL.
17465 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17466 of mode class checks.
17467 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17468 instead of a list of mode class checks.
17469 (expand_vector_scalar_condition): Likewise.
17470 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17473 Alan Hayward <alan.hayward@arm.com>
17474 David Sherwood <david.sherwood@arm.com>
17476 * machmode.h (mode_size): Change from unsigned short to
17478 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17479 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17480 or if measurement_type is not polynomial.
17481 (fixed_size_mode::includes_p): Check for constant-sized modes.
17482 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17483 return a poly_uint16 rather than an unsigned short.
17484 (emit_mode_size): Change the type of mode_size from unsigned short
17485 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17486 (emit_mode_adjustments): Cope with polynomial vector sizes.
17487 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17489 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17491 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17492 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17493 * caller-save.c (setup_save_areas): Likewise.
17494 (replace_reg_with_saved_mem): Likewise.
17495 * calls.c (emit_library_call_value_1): Likewise.
17496 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17497 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17498 (gen_lowpart_for_combine): Likewise.
17499 * convert.c (convert_to_integer_1): Likewise.
17500 * cse.c (equiv_constant, cse_insn): Likewise.
17501 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17502 (cselib_subst_to_values): Likewise.
17503 * dce.c (word_dce_process_block): Likewise.
17504 * df-problems.c (df_word_lr_mark_ref): Likewise.
17505 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17506 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17507 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17508 (rtl_for_decl_location): Likewise.
17509 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17510 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17511 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17512 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17513 (expand_expr_real_1): Likewise.
17514 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17515 (pad_below): Likewise.
17516 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17517 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17518 * ira.c (get_subreg_tracking_sizes): Likewise.
17519 * ira-build.c (ira_create_allocno_objects): Likewise.
17520 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17521 (ira_sort_regnos_for_alter_reg): Likewise.
17522 * ira-costs.c (record_operand_costs): Likewise.
17523 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17524 (resolve_simple_move): Likewise.
17525 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17526 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17527 (lra_constraints): Likewise.
17528 (CONST_POOL_OK_P): Reject variable-sized modes.
17529 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17530 (add_pseudo_to_slot, lra_spill): Likewise.
17531 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17532 * optabs-query.c (get_best_extraction_insn): Likewise.
17533 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17534 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17535 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17536 * recog.c (offsettable_address_addr_space_p): Likewise.
17537 * regcprop.c (maybe_mode_change): Likewise.
17538 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17539 * regrename.c (build_def_use): Likewise.
17540 * regstat.c (dump_reg_info): Likewise.
17541 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17542 (find_reloads, find_reloads_subreg_address): Likewise.
17543 * reload1.c (eliminate_regs_1): Likewise.
17544 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17545 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17546 (simplify_binary_operation_1, simplify_subreg): Likewise.
17547 * targhooks.c (default_function_arg_padding): Likewise.
17548 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17549 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17550 (verify_gimple_assign_ternary): Likewise.
17551 * tree-inline.c (estimate_move_cost): Likewise.
17552 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17553 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17554 (get_address_cost_ainc): Likewise.
17555 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17556 (vect_supportable_dr_alignment): Likewise.
17557 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17558 (vectorizable_reduction): Likewise.
17559 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17560 (vectorizable_operation, vectorizable_load): Likewise.
17561 * tree.c (build_same_sized_truth_vector_type): Likewise.
17562 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17563 * var-tracking.c (emit_note_insn_var_location): Likewise.
17564 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17565 (ADDR_VEC_ALIGN): Likewise.
17567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17568 Alan Hayward <alan.hayward@arm.com>
17569 David Sherwood <david.sherwood@arm.com>
17571 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17573 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17574 or if measurement_type is polynomial.
17575 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17576 * combine.c (make_extraction): Likewise.
17577 * dse.c (find_shift_sequence): Likewise.
17578 * dwarf2out.c (mem_loc_descriptor): Likewise.
17579 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17580 (extract_bit_field, extract_low_bits): Likewise.
17581 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17582 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17583 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17584 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17585 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17586 * reload.c (find_reloads): Likewise.
17587 * reload1.c (alter_reg): Likewise.
17588 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17589 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17590 * tree-if-conv.c (predicate_mem_writes): Likewise.
17591 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17592 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17593 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17594 * valtrack.c (dead_debug_insert_temp): Likewise.
17595 * varasm.c (mergeable_constant_section): Likewise.
17596 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17599 Alan Hayward <alan.hayward@arm.com>
17600 David Sherwood <david.sherwood@arm.com>
17602 * expr.c (expand_assignment): Cope with polynomial mode sizes
17603 when assigning to a CONCAT.
17605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17606 Alan Hayward <alan.hayward@arm.com>
17607 David Sherwood <david.sherwood@arm.com>
17609 * machmode.h (mode_precision): Change from unsigned short to
17611 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17613 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17614 or if measurement_type is not polynomial.
17615 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17616 in which the mode is already known to be a scalar_int_mode.
17617 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17618 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17620 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17621 for GET_MODE_PRECISION.
17622 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17623 for GET_MODE_PRECISION.
17624 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17626 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17627 (expand_field_assignment, make_extraction): Likewise.
17628 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17629 (get_last_value): Likewise.
17630 * convert.c (convert_to_integer_1): Likewise.
17631 * cse.c (cse_insn): Likewise.
17632 * expr.c (expand_expr_real_1): Likewise.
17633 * lra-constraints.c (simplify_operand_subreg): Likewise.
17634 * optabs-query.c (can_atomic_load_p): Likewise.
17635 * optabs.c (expand_atomic_load): Likewise.
17636 (expand_atomic_store): Likewise.
17637 * ree.c (combine_reaching_defs): Likewise.
17638 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17639 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17640 * tree.h (type_has_mode_precision_p): Likewise.
17641 * ubsan.c (instrument_si_overflow): Likewise.
17643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17644 Alan Hayward <alan.hayward@arm.com>
17645 David Sherwood <david.sherwood@arm.com>
17647 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17648 polynomial numbers of units.
17649 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17650 (valid_vector_subparts_p): New function.
17651 (build_vector_type): Remove temporary shim and take the number
17652 of units as a poly_uint64 rather than an int.
17653 (build_opaque_vector_type): Take the number of units as a
17654 poly_uint64 rather than an int.
17655 * tree.c (build_vector_from_ctor): Handle polynomial
17656 TYPE_VECTOR_SUBPARTS.
17657 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17658 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17659 (build_vector_from_val): If the number of units is variable,
17660 use build_vec_duplicate_cst for constant operands and
17661 VEC_DUPLICATE_EXPR otherwise.
17662 (make_vector_type): Remove temporary is_constant ().
17663 (build_vector_type, build_opaque_vector_type): Take the number of
17664 units as a poly_uint64 rather than an int.
17665 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17667 * cfgexpand.c (expand_debug_expr): Likewise.
17668 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17669 (store_constructor, expand_expr_real_1): Likewise.
17670 (const_scalar_mask_from_tree): Likewise.
17671 * fold-const-call.c (fold_const_reduction): Likewise.
17672 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17673 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17674 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17675 (fold_relational_const): Likewise.
17676 (native_interpret_vector): Likewise. Change the size from an
17677 int to an unsigned int.
17678 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17679 TYPE_VECTOR_SUBPARTS.
17680 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17681 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17682 duplicating a non-constant operand into a variable-length vector.
17683 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17684 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17685 * ipa-icf.c (sem_variable::equals): Likewise.
17686 * match.pd: Likewise.
17687 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17688 * print-tree.c (print_node): Likewise.
17689 * stor-layout.c (layout_type): Likewise.
17690 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17691 * tree-cfg.c (verify_gimple_comparison): Likewise.
17692 (verify_gimple_assign_binary): Likewise.
17693 (verify_gimple_assign_ternary): Likewise.
17694 (verify_gimple_assign_single): Likewise.
17695 * tree-pretty-print.c (dump_generic_node): Likewise.
17696 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17697 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17698 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17699 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17700 (vect_shift_permute_load_chain): Likewise.
17701 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17702 (expand_vector_condition, optimize_vector_constructor): Likewise.
17703 (lower_vec_perm, get_compute_type): Likewise.
17704 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17705 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17706 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17707 (vect_recog_mask_conversion_pattern): Likewise.
17708 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17709 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17710 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17711 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17712 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17713 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17714 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17715 (supportable_widening_operation): Likewise.
17716 (supportable_narrowing_operation): Likewise.
17717 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17719 * varasm.c (output_constant): Likewise.
17721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17722 Alan Hayward <alan.hayward@arm.com>
17723 David Sherwood <david.sherwood@arm.com>
17725 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17726 so that both the length == 3 and length != 3 cases set up their
17727 own permute vectors. Add comments explaining why we know the
17728 number of elements is constant.
17729 (vect_permute_load_chain): Likewise.
17731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17732 Alan Hayward <alan.hayward@arm.com>
17733 David Sherwood <david.sherwood@arm.com>
17735 * machmode.h (mode_nunits): Change from unsigned char to
17737 (ONLY_FIXED_SIZE_MODES): New macro.
17738 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17739 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17740 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17742 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17743 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17744 or if measurement_type is not polynomial.
17745 * genmodes.c (ZERO_COEFFS): New macro.
17746 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17748 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17749 Use ZERO_COEFFS when emitting initializers.
17750 * data-streamer.h (bp_pack_poly_value): New function.
17751 (bp_unpack_poly_value): Likewise.
17752 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17753 for GET_MODE_NUNITS.
17754 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17755 for GET_MODE_NUNITS.
17756 * tree.c (make_vector_type): Remove temporary shim and make
17757 the real function take the number of units as a poly_uint64
17758 rather than an int.
17759 (build_vector_type_for_mode): Handle polynomial nunits.
17760 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17761 * emit-rtl.c (const_vec_series_p_1): Likewise.
17762 (gen_rtx_CONST_VECTOR): Likewise.
17763 * fold-const.c (test_vec_duplicate_folding): Likewise.
17764 * genrecog.c (validate_pattern): Likewise.
17765 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17766 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17767 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17768 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17769 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17770 * rtlanal.c (subreg_get_info): Likewise.
17771 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17772 (vect_grouped_load_supported): Likewise.
17773 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17774 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17775 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17776 (simplify_const_unary_operation, simplify_binary_operation_1)
17777 (simplify_const_binary_operation, simplify_ternary_operation)
17778 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17779 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17780 instead of CONST_VECTOR_NUNITS.
17781 * varasm.c (output_constant_pool_2): Likewise.
17782 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17783 explicit-encoded elements in the XVEC for variable-length vectors.
17785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17787 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17790 Alan Hayward <alan.hayward@arm.com>
17791 David Sherwood <david.sherwood@arm.com>
17793 * coretypes.h (fixed_size_mode): Declare.
17794 (fixed_size_mode_pod): New typedef.
17795 * builtins.h (target_builtins::x_apply_args_mode)
17796 (target_builtins::x_apply_result_mode): Change type to
17797 fixed_size_mode_pod.
17798 * builtins.c (apply_args_size, apply_result_size, result_vector)
17799 (expand_builtin_apply_args_1, expand_builtin_apply)
17800 (expand_builtin_return): Update accordingly.
17802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17804 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17805 * cselib.c (cselib_hash_rtx): Likewise.
17806 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17807 CONST_VECTOR encoding.
17809 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17810 Jeff Law <law@redhat.com>
17813 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17814 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17815 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17816 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17819 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17820 explicitly probe *sp in a noreturn function if there were any callee
17821 register saves or frame pointer is needed.
17823 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17826 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17827 BLKmode for ternary, binary or unary expressions.
17830 * var-tracking.c (delete_vta_debug_insn): New inline function.
17831 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17832 insns from get_insns () to NULL instead of each bb separately.
17833 Use delete_vta_debug_insn. No longer static.
17834 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17835 delete_vta_debug_insns callers.
17836 * rtl.h (delete_vta_debug_insns): Declare.
17837 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17838 instead of variable_tracking_main.
17840 2018-01-03 Martin Sebor <msebor@redhat.com>
17842 PR tree-optimization/83603
17843 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17844 arguments past the endof the argument list in functions declared
17845 without a prototype.
17846 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17847 Avoid checking when arguments are null.
17849 2018-01-03 Martin Sebor <msebor@redhat.com>
17852 * doc/extend.texi (attribute const): Fix a typo.
17853 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17854 issuing -Wsuggest-attribute for void functions.
17856 2018-01-03 Martin Sebor <msebor@redhat.com>
17858 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17859 offset_int::from instead of wide_int::to_shwi.
17860 (maybe_diag_overlap): Remove assertion.
17861 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17862 * gimple-ssa-sprintf.c (format_directive): Same.
17863 (parse_directive): Same.
17864 (sprintf_dom_walker::compute_format_length): Same.
17865 (try_substitute_return_value): Same.
17867 2018-01-03 Jeff Law <law@redhat.com>
17869 PR middle-end/83654
17870 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17871 non-constant residual for zero at runtime and avoid probing in
17872 that case. Reorganize code for trailing problem to mirror handling
17875 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17877 PR tree-optimization/83501
17878 * tree-ssa-strlen.c (get_string_cst): New.
17879 (handle_char_store): Call get_string_cst.
17881 2018-01-03 Martin Liska <mliska@suse.cz>
17883 PR tree-optimization/83593
17884 * tree-ssa-strlen.c: Include tree-cfg.h.
17885 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17886 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17887 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17889 (strlen_dom_walker::before_dom_children): Call
17890 gimple_purge_dead_eh_edges. Dump tranformation with details
17892 (strlen_dom_walker::before_dom_children): Update call by adding
17893 new argument cleanup_eh.
17894 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17896 2018-01-03 Martin Liska <mliska@suse.cz>
17899 * cif-code.def (VARIADIC_THUNK): New enum value.
17900 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17903 2018-01-03 Jan Beulich <jbeulich@suse.com>
17905 * sse.md (mov<mode>_internal): Tighten condition for when to use
17906 vmovdqu<ssescalarsize> for TI and OI modes.
17908 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17910 Update copyright years.
17912 2018-01-03 Martin Liska <mliska@suse.cz>
17915 * ipa-visibility.c (function_and_variable_visibility): Skip
17916 functions with noipa attribure.
17918 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17920 * gcc.c (process_command): Update copyright notice dates.
17921 * gcov-dump.c (print_version): Ditto.
17922 * gcov.c (print_version): Ditto.
17923 * gcov-tool.c (print_version): Ditto.
17924 * gengtype.c (create_file): Ditto.
17925 * doc/cpp.texi: Bump @copying's copyright year.
17926 * doc/cppinternals.texi: Ditto.
17927 * doc/gcc.texi: Ditto.
17928 * doc/gccint.texi: Ditto.
17929 * doc/gcov.texi: Ditto.
17930 * doc/install.texi: Ditto.
17931 * doc/invoke.texi: Ditto.
17933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17935 * vector-builder.h (vector_builder::m_full_nelts): Change from
17936 unsigned int to poly_uint64.
17937 (vector_builder::full_nelts): Update prototype accordingly.
17938 (vector_builder::new_vector): Likewise.
17939 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17940 (vector_builder::operator ==): Likewise.
17941 (vector_builder::finalize): Likewise.
17942 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17943 Take the number of elements as a poly_uint64 rather than an
17945 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17946 from unsigned int to poly_uint64.
17947 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17948 (vec_perm_indices::new_vector): Likewise.
17949 (vec_perm_indices::length): Likewise.
17950 (vec_perm_indices::nelts_per_input): Likewise.
17951 (vec_perm_indices::input_nelts): Likewise.
17952 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17953 number of elements per input as a poly_uint64 rather than an
17954 unsigned int. Use the original encoding for variable-length
17955 vectors, rather than clamping each individual element.
17956 For the second and subsequent elements in each pattern,
17957 clamp the step and base before clamping their sum.
17958 (vec_perm_indices::series_p): Handle polynomial element counts.
17959 (vec_perm_indices::all_in_range_p): Likewise.
17960 (vec_perm_indices_to_tree): Likewise.
17961 (vec_perm_indices_to_rtx): Likewise.
17962 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17963 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17964 (tree_vector_builder::new_binary_operation): Handle polynomial
17965 element counts. Return false if we need to know the number
17966 of elements at compile time.
17967 * fold-const.c (fold_vec_perm): Punt if the number of elements
17968 isn't known at compile time.
17970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17972 * vec-perm-indices.h (vec_perm_builder): Change element type
17973 from HOST_WIDE_INT to poly_int64.
17974 (vec_perm_indices::element_type): Update accordingly.
17975 (vec_perm_indices::clamp): Handle polynomial element_types.
17976 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17977 (vec_perm_indices::all_in_range_p): Likewise.
17978 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17980 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17981 polynomial vec_perm_indices element types.
17982 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17983 * fold-const.c (fold_vec_perm): Likewise.
17984 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17985 * tree-vect-generic.c (lower_vec_perm): Likewise.
17986 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17987 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17988 element type to HOST_WIDE_INT.
17990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17991 Alan Hayward <alan.hayward@arm.com>
17992 David Sherwood <david.sherwood@arm.com>
17994 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17995 rather than an int. Use plus_constant.
17996 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17997 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18000 Alan Hayward <alan.hayward@arm.com>
18001 David Sherwood <david.sherwood@arm.com>
18003 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18004 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
18011 mode sizes. Require a constant size for BLKmode parameters
18012 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18013 forces a parameter to be padded at the lsb end in order to
18014 fill a complete number of words, require the parameter size
18015 to be ordered wrt UNITS_PER_WORD.
18017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18018 Alan Hayward <alan.hayward@arm.com>
18019 David Sherwood <david.sherwood@arm.com>
18021 * reload1.c (spill_stack_slot_width): Change element type
18022 from unsigned int to poly_uint64_pod.
18023 (alter_reg): Treat mode sizes as polynomial.
18025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18026 Alan Hayward <alan.hayward@arm.com>
18027 David Sherwood <david.sherwood@arm.com>
18029 * reload.c (complex_word_subreg_p): New function.
18030 (reload_inner_reg_of_subreg, push_reload): Use it.
18032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18033 Alan Hayward <alan.hayward@arm.com>
18034 David Sherwood <david.sherwood@arm.com>
18036 * lra-constraints.c (process_alt_operands): Reject matched
18037 operands whose sizes aren't ordered.
18038 (match_reload): Refer to this check here.
18040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18041 Alan Hayward <alan.hayward@arm.com>
18042 David Sherwood <david.sherwood@arm.com>
18044 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18045 that the mode size is in the set {1, 2, 4, 8, 16}.
18047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18048 Alan Hayward <alan.hayward@arm.com>
18049 David Sherwood <david.sherwood@arm.com>
18051 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18052 Use plus_constant instead of gen_rtx_PLUS.
18054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18055 Alan Hayward <alan.hayward@arm.com>
18056 David Sherwood <david.sherwood@arm.com>
18058 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18059 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18060 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18061 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18062 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18063 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18064 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18065 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18066 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18067 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18069 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18070 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18071 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18072 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18073 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18074 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18075 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18076 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18077 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18078 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18080 * expr.c (emit_move_resolve_push): Treat the input and result
18081 of PUSH_ROUNDING as a poly_int64.
18082 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18083 (emit_push_insn): Likewise.
18084 * lra-eliminations.c (mark_not_eliminable): Likewise.
18085 * recog.c (push_operand): Likewise.
18086 * reload1.c (elimination_effects): Likewise.
18087 * rtlanal.c (nonzero_bits1): Likewise.
18088 * calls.c (store_one_arg): Likewise. Require the padding to be
18089 known at compile time.
18091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18092 Alan Hayward <alan.hayward@arm.com>
18093 David Sherwood <david.sherwood@arm.com>
18095 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18096 Use plus_constant instead of gen_rtx_PLUS.
18098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18099 Alan Hayward <alan.hayward@arm.com>
18100 David Sherwood <david.sherwood@arm.com>
18102 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18103 rather than an int.
18105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18106 Alan Hayward <alan.hayward@arm.com>
18107 David Sherwood <david.sherwood@arm.com>
18109 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18110 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18111 via stack temporaries. Treat the mode size as polynomial too.
18113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18114 Alan Hayward <alan.hayward@arm.com>
18115 David Sherwood <david.sherwood@arm.com>
18117 * expr.c (expand_expr_real_2): When handling conversions involving
18118 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18119 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18120 as a poly_uint64 too.
18122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18123 Alan Hayward <alan.hayward@arm.com>
18124 David Sherwood <david.sherwood@arm.com>
18126 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18129 Alan Hayward <alan.hayward@arm.com>
18130 David Sherwood <david.sherwood@arm.com>
18132 * combine.c (can_change_dest_mode): Handle polynomial
18133 REGMODE_NATURAL_SIZE.
18134 * expmed.c (store_bit_field_1): Likewise.
18135 * expr.c (store_constructor): Likewise.
18136 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18137 and polynomial REGMODE_NATURAL_SIZE.
18138 (gen_lowpart_common): Likewise.
18139 * reginfo.c (record_subregs_of_mode): Likewise.
18140 * rtlanal.c (read_modify_subreg_p): Likewise.
18142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18143 Alan Hayward <alan.hayward@arm.com>
18144 David Sherwood <david.sherwood@arm.com>
18146 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18147 numbers of elements.
18149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18150 Alan Hayward <alan.hayward@arm.com>
18151 David Sherwood <david.sherwood@arm.com>
18153 * match.pd: Cope with polynomial numbers of vector elements.
18155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18156 Alan Hayward <alan.hayward@arm.com>
18157 David Sherwood <david.sherwood@arm.com>
18159 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18160 in a POINTER_PLUS_EXPR.
18162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18163 Alan Hayward <alan.hayward@arm.com>
18164 David Sherwood <david.sherwood@arm.com>
18166 * omp-simd-clone.c (simd_clone_subparts): New function.
18167 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18168 (ipa_simd_modify_function_body): Likewise.
18170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18171 Alan Hayward <alan.hayward@arm.com>
18172 David Sherwood <david.sherwood@arm.com>
18174 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18175 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18176 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18177 (expand_vector_condition, vector_element): Likewise.
18178 (subparts_gt): New function.
18179 (get_compute_type): Use subparts_gt.
18180 (count_type_subparts): Delete.
18181 (expand_vector_operations_1): Use subparts_gt instead of
18182 count_type_subparts.
18184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18185 Alan Hayward <alan.hayward@arm.com>
18186 David Sherwood <david.sherwood@arm.com>
18188 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18189 (vect_compile_time_alias): ...this new function. Do the calculation
18190 on poly_ints rather than trees.
18191 (vect_prune_runtime_alias_test_list): Update call accordingly.
18193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18194 Alan Hayward <alan.hayward@arm.com>
18195 David Sherwood <david.sherwood@arm.com>
18197 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18199 (vect_schedule_slp_instance): Likewise.
18201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18202 Alan Hayward <alan.hayward@arm.com>
18203 David Sherwood <david.sherwood@arm.com>
18205 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18206 constant and extern definitions for variable-length vectors.
18207 (vect_get_constant_vectors): Note that the number of units
18208 is known to be constant.
18210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18211 Alan Hayward <alan.hayward@arm.com>
18212 David Sherwood <david.sherwood@arm.com>
18214 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18215 of units as polynomial. Choose between WIDE and NARROW based
18218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18219 Alan Hayward <alan.hayward@arm.com>
18220 David Sherwood <david.sherwood@arm.com>
18222 * tree-vect-stmts.c (simd_clone_subparts): New function.
18223 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18226 Alan Hayward <alan.hayward@arm.com>
18227 David Sherwood <david.sherwood@arm.com>
18229 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18230 vectors as polynomial. Use build_index_vector for
18231 IFN_GOMP_SIMD_LANE.
18233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18234 Alan Hayward <alan.hayward@arm.com>
18235 David Sherwood <david.sherwood@arm.com>
18237 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18238 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18239 for variable-length vectors.
18240 (vectorizable_mask_load_store): Treat the number of units as
18241 polynomial, asserting that it is constant if the condition has
18242 already been enforced.
18243 (vectorizable_store, vectorizable_load): Likewise.
18245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18246 Alan Hayward <alan.hayward@arm.com>
18247 David Sherwood <david.sherwood@arm.com>
18249 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18250 of units as polynomial. Punt if we can't tell at compile time
18251 which vector contains the final result.
18253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18254 Alan Hayward <alan.hayward@arm.com>
18255 David Sherwood <david.sherwood@arm.com>
18257 * tree-vect-loop.c (vectorizable_induction): Treat the number
18258 of units as polynomial. Punt on SLP inductions. Use an integer
18259 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18260 cast of such a series for variable-length floating-point
18263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18264 Alan Hayward <alan.hayward@arm.com>
18265 David Sherwood <david.sherwood@arm.com>
18267 * tree.h (build_index_vector): Declare.
18268 * tree.c (build_index_vector): New function.
18269 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18270 of units as polynomial, forcibly converting it to a constant if
18271 vectorizable_reduction has already enforced the condition.
18272 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18273 to create a {1,2,3,...} vector.
18274 (vectorizable_reduction): Treat the number of units as polynomial.
18275 Choose vectype_in based on the largest scalar element size rather
18276 than the smallest number of units. Enforce the restrictions
18279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18280 Alan Hayward <alan.hayward@arm.com>
18281 David Sherwood <david.sherwood@arm.com>
18283 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18284 number of units as polynomial.
18286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18287 Alan Hayward <alan.hayward@arm.com>
18288 David Sherwood <david.sherwood@arm.com>
18290 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18291 * target.def (autovectorize_vector_sizes): Return the vector sizes
18292 by pointer, using vector_sizes rather than a bitmask.
18293 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18294 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18295 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18297 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18298 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18299 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18300 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18301 * omp-general.c (omp_max_vf): Likewise.
18302 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18303 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18304 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18305 * tree-vect-slp.c (vect_slp_bb): Likewise.
18306 * doc/tm.texi: Regenerate.
18307 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18309 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18310 the vector size as a poly_uint64 rather than an unsigned int.
18311 (current_vector_size): Change from an unsigned int to a poly_uint64.
18312 (get_vectype_for_scalar_type): Update accordingly.
18313 * tree.h (build_truth_vector_type): Take the size and number of
18314 units as a poly_uint64 rather than an unsigned int.
18315 (build_vector_type): Add a temporary overload that takes
18316 the number of units as a poly_uint64 rather than an unsigned int.
18317 * tree.c (make_vector_type): Likewise.
18318 (build_truth_vector_type): Take the number of units as a poly_uint64
18319 rather than an unsigned int.
18321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18322 Alan Hayward <alan.hayward@arm.com>
18323 David Sherwood <david.sherwood@arm.com>
18325 * target.def (get_mask_mode): Take the number of units and length
18326 as poly_uint64s rather than unsigned ints.
18327 * targhooks.h (default_get_mask_mode): Update accordingly.
18328 * targhooks.c (default_get_mask_mode): Likewise.
18329 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18330 * doc/tm.texi: Regenerate.
18332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18333 Alan Hayward <alan.hayward@arm.com>
18334 David Sherwood <david.sherwood@arm.com>
18336 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18337 * omp-general.c (omp_max_vf): Likewise.
18338 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18339 (expand_omp_simd): Handle polynomial safelen.
18340 * omp-low.c (omplow_simd_context): Add a default constructor.
18341 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18342 (lower_rec_simd_input_clauses): Update accordingly.
18343 (lower_rec_input_clauses): Likewise.
18345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18346 Alan Hayward <alan.hayward@arm.com>
18347 David Sherwood <david.sherwood@arm.com>
18349 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18350 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18351 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18352 (vect_analyze_slp_cost): Likewise.
18353 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18354 (vect_model_load_cost): Likewise.
18356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18357 Alan Hayward <alan.hayward@arm.com>
18358 David Sherwood <david.sherwood@arm.com>
18360 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18361 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18362 from an unsigned int * to a poly_uint64_pod *.
18363 (calculate_unrolling_factor): New function.
18364 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18367 Alan Hayward <alan.hayward@arm.com>
18368 David Sherwood <david.sherwood@arm.com>
18370 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18371 from an unsigned int to a poly_uint64.
18372 (_loop_vec_info::slp_unrolling_factor): Likewise.
18373 (_loop_vec_info::vectorization_factor): Change from an int
18375 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18376 (vect_get_num_vectors): New function.
18377 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18378 (vect_get_num_copies): Use vect_get_num_vectors.
18379 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18380 to an unsigned int *.
18381 (vect_analyze_data_refs): Change min_vf from an int * to a
18383 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18384 than an unsigned HOST_WIDE_INT.
18385 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18386 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18387 to an unsigned int *.
18388 (vect_analyze_data_ref_dependences): Likewise.
18389 (vect_compute_data_ref_alignment): Handle polynomial vf.
18390 (vect_enhance_data_refs_alignment): Likewise.
18391 (vect_prune_runtime_alias_test_list): Likewise.
18392 (vect_shift_permute_load_chain): Likewise.
18393 (vect_supportable_dr_alignment): Likewise.
18394 (dependence_distance_ge_vf): Take the vectorization factor as a
18395 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18396 (vect_analyze_data_refs): Change min_vf from an int * to a
18398 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18399 vfm1 as a poly_uint64 rather than an int. Make the same change
18400 for the returned bound_scalar.
18401 (vect_gen_vector_loop_niters): Handle polynomial vf.
18402 (vect_do_peeling): Likewise. Update call to
18403 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18404 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18406 * tree-vect-loop.c (vect_determine_vectorization_factor)
18407 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18408 (vect_get_known_peeling_cost): Likewise.
18409 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18410 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18411 (vect_transform_loop): Likewise. Use the lowest possible VF when
18412 updating the upper bounds of the loop.
18413 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18414 rather than an int.
18415 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18416 polynomial unroll factors.
18417 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18418 (vect_make_slp_decision): Likewise.
18419 (vect_supported_load_permutation_p): Likewise, and polynomial
18421 (vect_analyze_slp_cost): Handle polynomial vf.
18422 (vect_slp_analyze_node_operations): Likewise.
18423 (vect_slp_analyze_bb_1): Likewise.
18424 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18425 than an unsigned HOST_WIDE_INT.
18426 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18427 (vectorizable_load): Handle polynomial vf.
18428 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18430 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18433 Alan Hayward <alan.hayward@arm.com>
18434 David Sherwood <david.sherwood@arm.com>
18436 * match.pd: Handle bit operations involving three constants
18437 and try to fold one pair.
18439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18441 * tree-vect-loop-manip.c: Include gimple-fold.h.
18442 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18443 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18444 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18445 Add a path that uses a step of VF instead of 1, but disable it
18447 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18448 and niters_no_overflow parameters. Update calls to
18449 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18450 Create a new SSA name if the latter choses to use a ste other
18451 than zero, and return it via niters_vector_mult_vf_var.
18452 * tree-vect-loop.c (vect_transform_loop): Update calls to
18453 vect_do_peeling, vect_gen_vector_loop_niters and
18454 slpeel_make_loop_iterate_ntimes.
18455 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18456 (vect_gen_vector_loop_niters): Update declarations after above changes.
18458 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18460 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18461 128-bit round to integer instructions.
18462 (ceil<mode>2): Likewise.
18463 (btrunc<mode>2): Likewise.
18464 (round<mode>2): Likewise.
18466 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18468 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18469 unaligned VSX load/store on P8/P9.
18470 (expand_block_clear): Allow the use of unaligned VSX
18471 load/store on P8/P9.
18473 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18475 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18477 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18478 swap associated with both a load and a store.
18480 2018-01-02 Andrew Waterman <andrew@sifive.com>
18482 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18483 * config/riscv/riscv.md (clear_cache): Use it.
18485 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18487 * web.c: Remove out-of-date comment.
18489 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18491 * expr.c (fixup_args_size_notes): Check that any existing
18492 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18493 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18494 (emit_single_push_insn): ...here.
18496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18498 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18499 (const_vector_encoded_nelts): New function.
18500 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18501 (const_vector_int_elt, const_vector_elt): Declare.
18502 * emit-rtl.c (const_vector_int_elt_1): New function.
18503 (const_vector_elt): Likewise.
18504 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18505 of CONST_VECTOR_ELT.
18507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18509 * expr.c: Include rtx-vector-builder.h.
18510 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18511 directly on the tree encoding.
18512 (const_vector_from_tree): Likewise.
18513 * optabs.c: Include rtx-vector-builder.h.
18514 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18515 sequence of "u" values.
18516 * vec-perm-indices.c: Include rtx-vector-builder.h.
18517 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18518 directly on the vec_perm_indices encoding.
18520 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18522 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18523 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18524 * rtx-vector-builder.h: New file.
18525 * rtx-vector-builder.c: Likewise.
18526 * rtl.h (rtx_def::u2): Add a const_vector field.
18527 (CONST_VECTOR_NPATTERNS): New macro.
18528 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18529 (CONST_VECTOR_DUPLICATE_P): Likewise.
18530 (CONST_VECTOR_STEPPED_P): Likewise.
18531 (CONST_VECTOR_ENCODED_ELT): Likewise.
18532 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18533 (unwrap_const_vec_duplicate): Likewise.
18534 (const_vec_series_p): Check for a non-duplicated vector encoding.
18535 Say that the function only returns true for integer vectors.
18536 * emit-rtl.c: Include rtx-vector-builder.h.
18537 (gen_const_vec_duplicate_1): Delete.
18538 (gen_const_vector): Call gen_const_vec_duplicate instead of
18539 gen_const_vec_duplicate_1.
18540 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18541 (gen_const_vec_duplicate): Use rtx_vector_builder.
18542 (gen_const_vec_series): Likewise.
18543 (gen_rtx_CONST_VECTOR): Likewise.
18544 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18545 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18546 Build a new vector rather than modifying a CONST_VECTOR in-place.
18547 (handle_special_swappables): Update call accordingly.
18548 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18549 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18550 Build a new vector rather than modifying a CONST_VECTOR in-place.
18551 (handle_special_swappables): Update call accordingly.
18553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18555 * simplify-rtx.c (simplify_const_binary_operation): Use
18556 CONST_VECTOR_ELT instead of XVECEXP.
18558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18560 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18561 the selector elements to be different from the data elements
18562 if the selector is a VECTOR_CST.
18563 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18564 ssizetype for the selector.
18566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18568 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18569 before testing each element individually.
18570 * tree-vect-generic.c (lower_vec_perm): Likewise.
18572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18574 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18575 * selftest-run-tests.c (selftest::run_tests): Call it.
18576 * vector-builder.h (vector_builder::operator ==): New function.
18577 (vector_builder::operator !=): Likewise.
18578 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18579 (vec_perm_indices::all_from_input_p): New function.
18580 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18581 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18582 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18583 instead of reading the VECTOR_CST directly. Detect whether both
18584 vector inputs are the same before constructing the vec_perm_indices,
18585 and update the number of inputs argument accordingly. Use the
18586 utility functions added above. Only construct sel2 if we need to.
18588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18590 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18591 the broadcast of the low byte.
18592 (expand_mult_highpart): Use an explicit encoding for the permutes.
18593 * optabs-query.c (can_mult_highpart_p): Likewise.
18594 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18595 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18596 (vectorizable_bswap): Likewise.
18597 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18598 explicit encoding for the power-of-2 permutes.
18599 (vect_permute_store_chain): Likewise.
18600 (vect_grouped_load_supported): Likewise.
18601 (vect_permute_load_chain): Likewise.
18603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18605 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18606 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18607 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18608 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18609 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18610 (vect_gen_perm_mask_any): Likewise.
18612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18614 * int-vector-builder.h: New file.
18615 * vec-perm-indices.h: Include int-vector-builder.h.
18616 (vec_perm_indices): Redefine as an int_vector_builder.
18617 (auto_vec_perm_indices): Delete.
18618 (vec_perm_builder): Redefine as a stand-alone class.
18619 (vec_perm_indices::vec_perm_indices): New function.
18620 (vec_perm_indices::clamp): Likewise.
18621 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18622 (vec_perm_indices::new_vector): New function.
18623 (vec_perm_indices::new_expanded_vector): Update for new
18624 vec_perm_indices class.
18625 (vec_perm_indices::rotate_inputs): New function.
18626 (vec_perm_indices::all_in_range_p): Operate directly on the
18627 encoded form, without computing elided elements.
18628 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18629 encoding. Update for new vec_perm_indices class.
18630 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18631 the given vec_perm_builder.
18632 (expand_vec_perm_var): Update vec_perm_builder constructor.
18633 (expand_mult_highpart): Use vec_perm_builder instead of
18634 auto_vec_perm_indices.
18635 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18636 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18637 or double series encoding as appropriate.
18638 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18639 vec_perm_indices instead of auto_vec_perm_indices.
18640 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18641 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18642 (vect_permute_store_chain): Likewise.
18643 (vect_grouped_load_supported): Likewise.
18644 (vect_permute_load_chain): Likewise.
18645 (vect_shift_permute_load_chain): Likewise.
18646 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18647 (vect_transform_slp_perm_load): Likewise.
18648 (vect_schedule_slp_instance): Likewise.
18649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18650 (vectorizable_mask_load_store): Likewise.
18651 (vectorizable_bswap): Likewise.
18652 (vectorizable_store): Likewise.
18653 (vectorizable_load): Likewise.
18654 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18655 vec_perm_indices instead of auto_vec_perm_indices. Use
18656 tree_to_vec_perm_builder to read the vector from a tree.
18657 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18658 vec_perm_builder instead of a vec_perm_indices.
18659 (have_whole_vector_shift): Use vec_perm_builder and
18660 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18661 truncation to calc_vec_perm_mask_for_shift.
18662 (vect_create_epilog_for_reduction): Likewise.
18663 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18664 from auto_vec_perm_indices to vec_perm_indices.
18665 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18666 instead of changing individual elements.
18667 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18668 the vector in d.perm.
18669 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18670 from auto_vec_perm_indices to vec_perm_indices.
18671 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18672 instead of changing individual elements.
18673 (arm_vectorize_vec_perm_const): Use new_vector to install
18674 the vector in d.perm.
18675 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18676 Update vec_perm_builder constructor.
18677 (rs6000_expand_interleave): Likewise.
18678 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18679 (rs6000_expand_interleave): Likewise.
18681 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18683 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18684 to qimode could truncate the indices.
18685 * optabs.c (expand_vec_perm_var): Likewise.
18687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18689 * Makefile.in (OBJS): Add vec-perm-indices.o.
18690 * vec-perm-indices.h: New file.
18691 * vec-perm-indices.c: Likewise.
18692 * target.h (vec_perm_indices): Replace with a forward class
18694 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18695 * optabs.h: Include vec-perm-indices.h.
18696 (expand_vec_perm): Delete.
18697 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18698 (expand_vec_perm_const): Declare.
18699 * target.def (vec_perm_const_ok): Replace with...
18700 (vec_perm_const): ...this new hook.
18701 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18702 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18703 * doc/tm.texi: Regenerate.
18704 * optabs.def (vec_perm_const): Delete.
18705 * doc/md.texi (vec_perm_const): Likewise.
18706 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18707 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18708 expand_vec_perm for constant permutation vectors. Assert that
18709 the mode of variable permutation vectors is the integer equivalent
18710 of the mode that is being permuted.
18711 * optabs-query.h (selector_fits_mode_p): Declare.
18712 * optabs-query.c: Include vec-perm-indices.h.
18713 (selector_fits_mode_p): New function.
18714 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18715 is defined, instead of checking whether the vec_perm_const_optab
18716 exists. Use targetm.vectorize.vec_perm_const instead of
18717 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18718 fit in the vector mode before using a variable permute.
18719 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18720 vec_perm_indices instead of an rtx.
18721 (expand_vec_perm): Replace with...
18722 (expand_vec_perm_const): ...this new function. Take the selector
18723 as a vec_perm_indices rather than an rtx. Also take the mode of
18724 the selector. Update call to shift_amt_for_vec_perm_mask.
18725 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18726 Use vec_perm_indices::new_expanded_vector to expand the original
18727 selector into bytes. Check whether the indices fit in the vector
18728 mode before using a variable permute.
18729 (expand_vec_perm_var): Make global.
18730 (expand_mult_highpart): Use expand_vec_perm_const.
18731 * fold-const.c: Includes vec-perm-indices.h.
18732 * tree-ssa-forwprop.c: Likewise.
18733 * tree-vect-data-refs.c: Likewise.
18734 * tree-vect-generic.c: Likewise.
18735 * tree-vect-loop.c: Likewise.
18736 * tree-vect-slp.c: Likewise.
18737 * tree-vect-stmts.c: Likewise.
18738 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18740 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18741 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18742 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18743 (aarch64_vectorize_vec_perm_const): ...this new function.
18744 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18745 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18746 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18747 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18748 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18749 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18750 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18752 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18753 check for NEON modes.
18754 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18755 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18756 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18757 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18759 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18760 the old VEC_PERM_CONST conditions.
18761 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18762 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18763 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18764 (ia64_vectorize_vec_perm_const_ok): Merge into...
18765 (ia64_vectorize_vec_perm_const): ...this new function.
18766 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18767 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18768 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18769 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18770 * config/mips/mips.c (mips_expand_vec_perm_const)
18771 (mips_vectorize_vec_perm_const_ok): Merge into...
18772 (mips_vectorize_vec_perm_const): ...this new function.
18773 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18774 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18775 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18776 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18777 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18778 (rs6000_expand_vec_perm_const): Delete.
18779 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18781 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18782 (altivec_expand_vec_perm_const_le): Take each operand individually.
18783 Operate on constant selectors rather than rtxes.
18784 (altivec_expand_vec_perm_const): Likewise. Update call to
18785 altivec_expand_vec_perm_const_le.
18786 (rs6000_expand_vec_perm_const): Delete.
18787 (rs6000_vectorize_vec_perm_const_ok): Delete.
18788 (rs6000_vectorize_vec_perm_const): New function.
18789 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18790 an element count and rtx array.
18791 (rs6000_expand_extract_even): Update call accordingly.
18792 (rs6000_expand_interleave): Likewise.
18793 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18794 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18795 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18796 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18797 (rs6000_expand_vec_perm_const): Delete.
18798 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18799 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18800 (altivec_expand_vec_perm_const_le): Take each operand individually.
18801 Operate on constant selectors rather than rtxes.
18802 (altivec_expand_vec_perm_const): Likewise. Update call to
18803 altivec_expand_vec_perm_const_le.
18804 (rs6000_expand_vec_perm_const): Delete.
18805 (rs6000_vectorize_vec_perm_const_ok): Delete.
18806 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18807 reference to the SPE evmerge intructions.
18808 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18809 an element count and rtx array.
18810 (rs6000_expand_extract_even): Update call accordingly.
18811 (rs6000_expand_interleave): Likewise.
18812 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18813 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18815 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18819 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18820 vector mode and that that mode matches the mode of the data
18822 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18823 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18824 directly using expand_vec_perm_1 when forcing selectors into
18826 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18830 * optabs-query.h (can_vec_perm_p): Delete.
18831 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18832 * optabs-query.c (can_vec_perm_p): Split into...
18833 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18834 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18835 particular selector is valid.
18836 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18837 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18838 (vect_grouped_load_supported): Likewise.
18839 (vect_shift_permute_load_chain): Likewise.
18840 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18841 (vect_transform_slp_perm_load): Likewise.
18842 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18843 (vectorizable_bswap): Likewise.
18844 (vect_gen_perm_mask_checked): Likewise.
18845 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18846 implementations of variable permutation vectors into account
18847 when deciding which selector to use.
18848 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18849 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18850 with a false third argument.
18851 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18852 to test whether the constant selector is valid and can_vec_perm_var_p
18853 to test whether a variable selector is valid.
18855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18857 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18858 * optabs-query.c (can_vec_perm_p): Likewise.
18859 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18860 instead of vec_perm_indices.
18861 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18862 (vect_gen_perm_mask_checked): Likewise,
18863 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18864 (vect_gen_perm_mask_checked): Likewise,
18866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18868 * optabs-query.h (qimode_for_vec_perm): Declare.
18869 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18870 (qimode_for_vec_perm): ...this new function.
18871 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18873 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18875 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18876 does not have a conditional at the top.
18878 2018-01-02 Richard Biener <rguenther@suse.de>
18880 * ipa-inline.c (big_speedup_p): Fix expression.
18882 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18885 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18888 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18892 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18893 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18894 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18895 cond_taken_branch_cost 3->4.
18897 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18899 PR tree-optimization/83581
18900 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18901 TODO_cleanup_cfg if any changes have been made.
18903 PR middle-end/83608
18904 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18905 convert_modes if target mode has the right side, but different mode
18908 PR middle-end/83609
18909 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18910 last argument when extracting from CONCAT. If either from_real or
18911 from_imag is NULL, use expansion through memory. If result is not
18912 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18913 the parts directly to inner mode, if even that fails, use expansion
18916 PR middle-end/83623
18917 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18918 check for bswap in mode rather than HImode and use that in expand_unop
18921 Copyright (C) 2018 Free Software Foundation, Inc.
18923 Copying and distribution of this file, with or without modification,
18924 are permitted in any medium without royalty provided the copyright
18925 notice and this notice are preserved.