1 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
8 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11 2018-07-04 Martin Liska <mliska@suse.cz>
13 * tree-switch-conversion.c: Define
14 max_ratio_for_speed and max_ratio_for_size constants.
16 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
17 Martin Liska <mliska@suse.cz>
22 * common.opt: Rename align options with 'str_' prefix.
23 * common/config/i386/i386-common.c (set_malign_value): New
25 (ix86_handle_option): Use it to set -falign-* options/
26 * config/aarch64/aarch64-protos.h (struct tune_params): Change
27 type from int to string.
28 * config/aarch64/aarch64.c: Update default values from int
30 * config/alpha/alpha.c (alpha_override_options_after_change):
32 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
33 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
34 max skip conditionally.
35 * config/i386/freebsd.h (SUBALIGN_LOG): New.
36 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
37 max skip conditionally.
38 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
39 max skip conditionally.
40 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
41 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
42 max skip conditionally.
43 * config/i386/i386.c (struct ptt): Change type from int to
45 (ix86_default_align): Set default values.
46 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
47 max skip conditionally.
48 * config/i386/iamcu.h (SUBALIGN_LOG): New.
49 (ASM_OUTPUT_MAX_SKIP_ALIGN):
50 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
51 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
52 max skip conditionally.
53 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
54 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
55 * config/i386/x86-64.h (SUBALIGN_LOG): New.
56 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
57 max skip conditionally.
58 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
59 * config/ia64/ia64.c (ia64_option_override): Set default values
60 for alignment options.
61 * config/m68k/m68k.c: Handle new str_align_* options.
62 * config/mips/mips.c (mips_set_compression_mode): Change
64 (mips_option_override): Set default values for options.
65 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
67 * config/rs6000/rs6000.c (rs6000_option_override_internal):
69 * config/rx/rx.c (rx_option_override): Likewise.
70 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
71 (LABEL_ALIGN): Use align_labels_log.
72 (LOOP_ALIGN): Use align_loops_align.
73 * config/s390/s390.c (s390_asm_output_function_label): Use new
75 * config/sh/sh.c (sh_override_options_after_change):
76 Change type of constants.
77 * config/spu/spu.c (spu_sched_init): Likewise.
78 * config/sparc/sparc.c (sparc_option_override): Set default
80 * config/visium/visium.c (visium_option_override): Likewise.
81 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
82 emit p2align format with last argument if it's not needed.
83 * doc/invoke.texi: Document extended format of -falign-*.
84 * final.c: Use align_labels alignment.
85 * flags.h (struct target_flag_state): Change type to use
87 (struct align_flags_tuple): New.
88 (struct align_flags): Likewise.
89 (align_loops_log): Redefine macro to use new types.
90 (align_loops_max_skip): Redefine macro to use new types.
91 (align_jumps_log): Redefine macro to use new types.
92 (align_jumps_max_skip): Redefine macro to use new types.
93 (align_labels_log): Redefine macro to use new types.
94 (align_labels_max_skip): Redefine macro to use new types.
95 (align_functions_log): Redefine macro to use new types.
96 (align_loops): Redefine macro to use new types.
97 (align_jumps): Redefine macro to use new types.
98 (align_labels): Redefine macro to use new types.
99 (align_functions): Redefine macro to use new types.
100 (align_functions_max_skip): Redefine macro to use new types.
101 (align_loops_value): New macro.
102 (align_jumps_value): New macro.
103 (align_labels_value): New macro.
104 (align_functions_value): New macro.
105 * function.c (invoke_set_current_function_hook): Propagate
106 alignment values from flags to global variables default in
108 * ipa-icf.c (sem_function::equals_wpa): Use
109 cl_optimization_option_eq instead of memcmp.
110 * lto-streamer.h (cl_optimization_stream_out): Support streaming
112 (cl_optimization_stream_in): Likewise.
113 * optc-save-gen.awk: Support strings in cl_optimization.
114 * opth-gen.awk: Likewise.
115 * opts.c (finish_options): Remove error checking of invalid
117 (MAX_CODE_ALIGN): Remove.
118 (MAX_CODE_ALIGN_VALUE): Likewise.
119 (parse_and_check_align_values): New function.
120 (check_alignment_argument): Likewise.
121 (common_handle_option): Use check_alignment_argument.
122 * opts.h (parse_and_check_align_values): Declare.
123 * toplev.c (init_alignments): Remove.
124 (read_log_maxskip): New.
125 (parse_N_M): Likewise.
126 (parse_alignment_opts): Likewise.
127 (backend_init_target): Remove usage of init_alignments.
128 * toplev.h (parse_alignment_opts): Declare.
129 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
131 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
132 * tree.c (cl_option_hasher::equal): New.
133 * varasm.c: Use new global macros.
135 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
137 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
138 Use a simpler align directive also if MAXSKIP = ALIGN-1.
139 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
140 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
141 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
142 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
143 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
144 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
145 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
146 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
147 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
148 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
149 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
151 2018-07-04 Martin Liska <mliska@suse.cz>
152 Jonathan Wakely <jwakely@redhat.com>
154 * coverage.c: Use correct type.
155 * doc/invoke.texi: Language correction.
157 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
160 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
161 ENDBRANCH for non-tail call which may return via indirect branch.
162 * doc/extend.texi: Document indirect_return attribute.
164 2018-07-03 Martin Sebor <msebor@redhat.com>
166 PR tree-optimization/86274
167 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
169 (format_floating): Correct handling of infinities and NaNs.
171 2018-07-03 Martin Sebor <msebor@redhat.com>
173 * print-tree.c (print_real_cst): New function.
174 (print_node_brief): Call it.
177 2018-07-03 Jeff Law <law@redhat.com>
179 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
180 into a single pattern.
182 * config/h8300/h8300.md (ors code_iterator): New.
183 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
184 a single pattern and single splitter.
185 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
186 (iorqi3_1, xorqi3_1): Likewise.
187 (iorqi3, xorqi3 expanders): Similarly.
189 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
190 (movmd_internal) into a single pattern using the P mode iterator.
191 (movmd splitters): Similarly.
192 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
193 (movsd splitters): Similarly.
195 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
196 ADDB, ADDW and ADDL into a single ADD attribute which selects the
197 right table based on the size of the operand.
198 * config/h8300/h8300.md (length_table): Corresponding changes. All
199 references to "addb", "addw" and "addl" changed to "add".
200 (btst patterns): Merge two variants into a single pattern.
201 (tstqi, tsthi): Likewise.
202 (addhi3_incdec, addsi3_incdec): Likewise.
203 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
204 (mulhi3, mulsi3): Likewise.
205 (udivhi3, udivsi3): Likewise.
206 (divhi3, divsi3): Likewise.
207 (andorqi3, andorhi3, andorsi3): Likewise.
209 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
212 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
213 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
215 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
217 PR tree-optimization/85694
218 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
219 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
221 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
222 (<u>avg<mode>3_ceil): New patterns.
224 2018-07-03 David Malcolm <dmalcolm@redhat.com>
226 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
227 scan-tree-dump directive.
228 * gcc.dg/vect/slp-perm-2.c: Likewise.
229 * gcc.dg/vect/slp-perm-3.c: Likewise.
230 * gcc.dg/vect/slp-perm-5.c: Likewise.
231 * gcc.dg/vect/slp-perm-6.c: Likewise.
232 * gcc.dg/vect/slp-perm-7.c: Likewise.
233 * gcc.dg/vect/slp-perm-8.c: Likewise.
235 2018-07-03 Marek Polacek <polacek@redhat.com>
238 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
240 2018-07-03 Richard Biener <rguenther@suse.de>
243 * tree-ssa-structalias.c (find_func_clobbers): Properly
244 handle indirect calls.
246 2018-07-03 Jeff Law <law@redhat.com>
248 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
249 (shifts): New code iterator.
250 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
251 expander. Fix HImode handling on H8/SX.
252 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
253 (subqi3, subhi3, subsi3 expanders): Likewise.
254 (andqi3, andhi3, andsi3 expanders): Likewise.
255 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
256 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
257 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
258 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
259 (zero_extendqihi2, zero_extendqisi2): Likewise.
260 (extendqihi2, extendqisi2): Likewise.
261 (rotlqi3, rotlhi3, rotlsi3): Likewise.
262 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
263 (rotlqi3_1, rotlhi3_1): Likewise.
264 (logicalhi3_sn, logicalsi3_sn): Likewise.
265 (logicalhi3, logicalsi3): Likewise.
267 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
269 * tree-vect-patterns.c (vect_recog_rotate_pattern)
270 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
271 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
272 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
273 type to append_pattern_def_seq instead of creating a stmt_vec_info
275 (build_mask_conversion): Likewise. Remove vinfo argument.
276 (vect_add_conversion_to_patterm): Likewise, renaming to...
277 (vect_add_conversion_to_pattern): ...this.
278 (vect_recog_mask_conversion_pattern): Update call to
279 build_mask_conversion. Pass the vector type to
280 append_pattern_def_seq here too.
281 (vect_recog_gather_scatter_pattern): Update call to
282 vect_add_conversion_to_pattern.
284 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
286 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
287 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
288 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
289 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
290 STMT_VINFO_PATTERN_DEF_SEQ to null here.
291 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
292 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
293 append_pattern_def_seq instead of new_pattern_def_seq.
294 (vect_recog_divmod_pattern): Do both of the above.
295 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
298 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
300 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
301 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
302 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
303 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
304 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
305 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
306 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
307 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
308 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
309 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
310 parameter with a single stmt_vec_info.
311 (vect_recog_func_ptr): Likewise.
312 (vect_recog_gather_scatter_pattern): Likewise, folding in...
313 (vect_try_gather_scatter_pattern): ...this.
314 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
315 the stmt_vec_info of the statement to be matched. Don't clear
316 STMT_VINFO_RELATED_STMT.
317 (vect_pattern_recog): Update call accordingly.
319 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
321 PR tree-optimization/85694
322 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
323 (uavgM3_ceil): Document new optabs.
324 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
325 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
327 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
328 (savg_ceil_optab): New optabs.
329 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
330 (vect_vect_recog_func_ptrs): Add it.
331 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
332 constant directly from the associated lhs.
334 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
336 * tree-vect-patterns.c (vect_split_statement): New function.
337 (vect_convert_input): Use it to try to split an existing cast.
339 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
341 * poly-int.h (print_hex): New function.
342 * dumpfile.h (dump_dec, dump_hex): Declare.
343 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
344 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
345 min_input_precision, operation_precision and operation_sign.
346 * tree-vect-patterns.c (vect_get_range_info): New function.
347 (vect_same_loop_or_bb_p, vect_single_imm_use)
348 (vect_operation_fits_smaller_type): Delete.
349 (vect_look_through_possible_promotion): Add an optional
350 single_use_p parameter.
351 (vect_recog_over_widening_pattern): Rewrite to use new
352 stmt_vec_info infomration. Handle one operation at a time.
353 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
354 (vect_truncatable_operation_p, vect_set_operation_type)
355 (vect_set_min_input_precision): New functions.
356 (vect_determine_min_output_precision_1): Likewise.
357 (vect_determine_min_output_precision): Likewise.
358 (vect_determine_precisions_from_range): Likewise.
359 (vect_determine_precisions_from_users): Likewise.
360 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
361 (vect_vect_recog_func_ptrs): Put over_widening first.
363 (vect_pattern_recog): Call vect_determine_precisions.
365 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
367 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
368 statements that have been replaced by further pattern statements.
369 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
371 2018-07-03 Richard Biener <rguenther@suse.de>
373 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
374 always set *dt. Dump vectype in vectype overload.
375 * dumpfile.h (dump_gimple_expr): New function.
376 (dump_gimple_expr_loc): Likewise.
377 * dumpfile.c (dump_gimple_expr): New function.
378 (dump_gimple_expr_loc): Likewise.
380 2018-07-02 Jeff Law <law@redhat.com>
382 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
383 the H8/300, H8/300H and H8/S variants into a single pattern.
384 (movhi_h8300, movqi_h8300hs): Similarly.
385 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
386 (QHI mode iterator): New.
388 * config/h8300/h8300.md: Remove trailing whitespace.
390 2018-07-02 Jim Wilson <jimw@sifive.com>
392 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
393 instead of emit_insn for interrupt returns.
394 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
395 (riscv_sret, riscv_uret): Likewise.
397 2018-07-02 David Malcolm <dmalcolm@redhat.com>
399 * pretty-print.c (selftest::test_pp_format): Move save and restore
400 of quotes to class auto_fix_quotes, and add an instance.
401 * selftest.c: Include "intl.h".
402 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
403 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
404 * selftest.h (selftest::auto_fix_quotes): New class.
406 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
408 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
409 (aarch64_sve_prepare_conditional_op): Remove.
410 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
411 Allow aarch64_simd_reg_or_zero as select operand; remove
412 the aarch64_sve_prepare_conditional_op call.
413 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
414 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
415 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
416 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
417 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
418 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
419 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
420 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
421 and a splitters to match all of the *_any patterns.
422 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
424 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
425 (SVE_COND_FP_BINARY_REV): Remove.
426 (sve_int_op_rev, sve_fp_op_rev): New.
427 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
428 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
429 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
430 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
431 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
432 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
433 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
434 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
435 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
437 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
438 Remove match_dup 1 from the inner unspec.
439 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
441 * config/aarch64/aarch64.md (movprfx): New attr.
442 (length): Default movprfx to 8.
443 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
444 (*madd<SVE_I>, *msub<SVE_I): Likewise.
445 (*<su>mul<SVE_I>3_highpart): Likewise.
446 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
447 (*v<ASHIFT><SVE_I>3): Likewise.
448 (*<su><MAXMIN><SVE_I>3): Likewise.
449 (*<su><MAXMIN><SVE_F>3): Likewise.
450 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
451 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
452 (*div<SVE_F>4): Likewise.
454 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
456 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
459 2018-07-02 Richard Biener <rguenther@suse.de>
461 PR tree-optimization/86363
462 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
463 memset argument refers to a non-variable address.
465 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
467 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
468 {PLUS,MINUS}_EXPR code to...
469 (adjust_symbolic_bound): ...here,
470 (combine_bound): ...here,
471 (set_value_range_with_overflow): ...and here.
473 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
475 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
477 (extract_range_from_abs_expr): ...here.
479 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
481 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
482 -fno-omit-frame-pointer when not optimizing.
484 2018-07-02 Martin Liska <mliska@suse.cz>
487 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
488 (propagate_nothrow): Likewise.
490 2018-07-02 Martin Liska <mliska@suse.cz>
493 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
495 2018-07-02 David Malcolm <dmalcolm@redhat.com>
497 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
498 function in r262149, changing "loc" param from source_location to
499 const dump_location_t &.
500 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
501 declaration, as above.
503 2018-07-01 Paul Koning <ni1d@arrl.net>
505 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
506 -munit-asm, -mgnu-asm, -mdec-asm.
507 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
508 (pdp11_output_labelref): New.
509 (pdp11_output_def): New.
510 (pdp11_output_addr_vec_elt): New.
511 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
512 %# and %@ format codes.
513 (pdp11_option_override): New.
514 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
515 (pdp11_output_ident): New.
516 (pdp11_asm_named_section): New.
517 (pdp11_asm_init_sections): New.
518 (pdp11_file_start): New.
519 (pdp11_file_end): New.
520 (output_ascii): Use .ascii/.asciz for -mdec-asm.
521 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
522 %o, like %c but octal.
523 (pdp11_option_override): New.
524 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
526 (DATA_SECTION_ASM_OP): Ditto.
527 (READONLY_DATA_SECTION_ASM_OP): New.
528 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
529 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
530 (ASM_OUTPUT_LABELREF): Ditto.
531 (ASM_OUTPUT_DEF): Ditto.
532 (ASM_OUTPUT_EXTERNAL): New.
533 (ASM_OUTPUT_SOURCE_FILENAME): New.
534 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
535 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
536 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
537 %# and %@ format codes.
538 * config/pdp11/pdp11.opt (mgnu-asm): New.
539 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
540 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
541 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
543 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
545 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
546 dereferencing path[] beyond its length.
548 (debug_all_paths): New.
549 (rewire_first_differing_edge): New.
550 (adjust_paths_after_duplication): New.
551 (duplicate_thread_path): Call adjust_paths_after_duplication.
553 (thread_through_all_blocks): Add new argument to
554 duplicate_thread_path.
556 2018-06-30 Jim Wilson <jimw@sifive.com>
558 * config/riscv/predicates.md (p2m1_shift_operand): New.
559 (high_mask_shift_operand): New.
560 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
561 pattern using p2m1_shift_operand.
562 (lshsi3_zero_extend_3+2): New combiner pattern using
563 high_mask_shift_operand.
565 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
567 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
569 (vect_recog_rotate_pattern): ...here.
570 (vect_convert_input): Try to insert casts of invariants in the
572 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
573 preheader to be empty.
575 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
577 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
578 vector type. If given, install it in the new statement's
580 (vect_element_precision): New function.
581 (vect_unpromoted_value): New struct.
582 (vect_unpromoted_value::vect_unpromoted_value): New function.
583 (vect_unpromoted_value::set_op): Likewise.
584 (vect_look_through_possible_promotion): Likewise.
585 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
586 (vect_widened_op_tree, vect_convert_input): Likewise.
587 (vect_convert_inputs, vect_convert_output): Likewise.
588 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
589 to handle the optional cast of the multiplication result and
590 vect_widened_op_tree to detect the widened multiplication itself.
591 Do not require the input and output of promotion casts to have
592 the same sign, but base the signedness of the operation on the
593 input rather than the result. If the pattern includes two
594 promotions, check that those promotions have the same sign.
595 Do not restrict the MULT_EXPR handling to a double-width result;
596 handle quadruple-width results and wider. Use vect_convert_inputs
597 to convert the inputs to the common type.
598 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
599 to handle the optional cast of the ABS result. Also allow a sign
600 change or a sign extension between the ABS and MINUS.
601 Use vect_widened_op_tree to detect the widened subtraction and use
602 vect_convert_inputs to convert the inputs to the common type.
603 (vect_handle_widen_op_by_const): Delete.
604 (vect_recog_widen_op_pattern): New function.
605 (vect_recog_widen_mult_pattern): Use it.
606 (vect_recog_widen_shift_pattern): Likewise.
607 (vect_recog_widen_sum_pattern): Use
608 vect_look_through_possible_promotion to handle the promoted
611 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
613 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
614 the containing gimple_seq *.
615 * gimple-iterator.h (gsi_for_stmt): Declare it.
616 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
617 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
618 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
619 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
620 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
622 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
624 (vect_mark_pattern_stmts): ...here. Handle cases in which the
625 statement being replaced is part of an existing pattern
626 definition sequence, inserting the new pattern statements before
628 (vect_pattern_recog_1): Don't return a bool. If the statement
629 is already part of a pattern, instead apply pattern matching
630 to the pattern definition statements. Don't clear the
631 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
632 (vect_pattern_recog): Don't break after the first match;
633 continue processing the pattern definition statements instead.
634 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
636 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
638 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
639 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
640 (vect_recog_widen_sum_pattern): Use it.
642 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
644 * tree-vect-loop.c (vectorizable_reduction): Assert that the
645 phi is not a pattern statement and has not been replaced by
647 * tree-vect-patterns.c (type_conversion_p): Don't check
648 STMT_VINFO_IN_PATTERN_P.
649 (vect_recog_vector_vector_shift_pattern): Likewise.
650 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
651 the pattern statement rather than the original statement; check
652 directly for a WIDEN_MULT_EXPR here.
653 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
654 vect_is_simple_use to return the pattern statement rather
655 than the original statement; use is_pattern_stmt_p to check
656 for such a pattern statement.
657 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
658 to return the pattern statement rather than the original statement;
659 don't do the same transformation here.
660 (vect_is_simple_use): If the defining statement has been replaced
661 by a pattern statement, return the pattern statement instead.
662 Remove the corresponding (local) transformation from the vectype
665 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
667 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
668 end and default to null.
669 * tree-vect-loop.c (vect_create_epilog_for_reduction)
670 (vectorizable_reduction): Update calls accordingly, dropping the
671 gimple ** argument if the passed-back statement isn't needed.
672 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
673 (vect_recog_rotate_pattern): Likewise.
674 (vect_recog_mask_conversion_pattern): Likewise.
675 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
676 (vect_mask_constant_operand_p): Likewise.
677 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
678 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
679 (get_group_load_store_type, get_load_store_type): Likewise.
680 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
681 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
682 (vectorizable_conversion, vectorizable_assignment): Likewise.
683 (vectorizable_shift, vectorizable_operation): Likewise.
684 (vectorizable_store, vect_is_simple_cond): Likewise.
685 (vectorizable_condition, vectorizable_comparison): Likewise.
686 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
687 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
688 and move it to the end. Cope with null def_stmt_outs.
690 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
692 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
694 2018-06-29 Jeff Law <law@redhat.com>
696 * config/v850/v850.c (v850_legitimate_address_p): Handle large
697 displacements for TARGET_V850E2V3 and newer.
698 (TARGET_LRA_P): Remove. Defaults to LRA now.
699 * config/v850/v850.md (sign23byte_load): Remove.
700 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
701 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
703 2018-06-29 Martin Liska <mliska@suse.cz>
706 * coverage.c (coverage_init): Mangle full path name.
707 * doc/invoke.texi: Document the change.
708 * gcov-io.c (mangle_path): New.
709 * gcov-io.h (mangle_path): Likewise.
710 * gcov.c (mangle_name): Use mangle_path for path mangling.
712 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
714 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
715 if starting source register is not even.
717 2018-06-29 Martin Liska <mliska@suse.cz>
719 PR tree-optimization/86263
720 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
721 Make edge redirection.
723 2018-06-29 David Malcolm <dmalcolm@redhat.com>
725 * dumpfile.c (dump_loc): Add indentation based on scope depth.
726 (dump_scope_depth): New variable.
727 (get_dump_scope_depth): New function.
728 (dump_begin_scope): New function.
729 (dump_end_scope): New function.
730 * dumpfile.h (get_dump_scope_depth): New declaration.
731 (dump_begin_scope): New declaration.
732 (dump_end_scope): New declaration.
733 (class auto_dump_scope): New class.
734 (AUTO_DUMP_SCOPE): New macro.
735 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
738 2018-06-29 Richard Biener <rguenther@suse.de>
740 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
741 compute_all_dependences succeeds.
742 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
743 exceed --param loop-max-datarefs-for-datadeps.
745 2018-06-29 Jakub Jelinek <jakub@redhat.com>
747 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
749 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
752 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
753 alternative 0 in preferred_for_speed attribute.
755 2018-06-28 Paul Koning <ni1d@arrl.net>
757 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
758 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
759 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
760 * config/pdp11/pdp11.md: Correct "length" attribute calculation
761 for shift insn patterns.
763 2018-06-28 David Malcolm <dmalcolm@redhat.com>
765 * cgraph.c (cgraph_node::get_body): Replace assignments to
766 "dump_file" with calls to set_dump_file.
767 * dumpfile.c (alt_dump_file): Make static, and group with...
768 (alt_flags): ...this definition.
769 (dumps_are_enabled): New variable.
770 (refresh_dumps_are_enabled): New function.
771 (set_dump_file): New function.
772 (set_alt_dump_file): New function.
773 (gcc::dump_manager::dump_start): Replace assignments to
774 "dump_file" and "alt_dump_file" with calls to set_dump_file and
776 (gcc::dump_manager::dump_finish): Likewise.
777 * dumpfile.h (alt_dump_file): Delete decl.
778 (dumps_are_enabled): New variable decl.
779 (set_dump_file): New function decl.
780 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
782 * tree-nested.c (lower_nested_functions): Replace assignments to
783 "dump_file" with calls to set_dump_file.
785 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
787 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
788 goto_locus of each outgoing edge of each basic block.
790 2018-06-28 Richard Biener <rguenther@suse.de>
792 * dwarf2out.c (decl_scope_table): Remove.
793 (push_decl_scope): Likewise.
794 (pop_decl_scope): Likewise.
795 (gen_type_die_for_member): Do not call push/pop_decl_scope.
796 (gen_struct_or_union_type_die): Likewise.
797 (gen_tagged_type_die): Likewise.
798 (dwarf2out_init): Do not initialize decl_scope_table.
799 (dwarf2out_c_finalize): Do not free it.
801 2018-06-28 Richard Biener <rguenther@suse.de>
803 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
804 deciding whether to not re-use a DIE.
806 2018-06-28 Richard Biener <rguenther@suse.de>
808 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
809 DW_AT_abstract_origin attribute.
811 2018-06-28 Martin Liska <mliska@suse.cz>
813 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
814 Use newly introduced constants.
815 * tree-switch-conversion.h (struct jump_table_cluster):
816 Define max_ratio_for_size and max_ratio_for_speed.
818 2018-06-28 Martin Liska <mliska@suse.cz>
820 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
821 Add new checking assert to catch invalid state.
822 (jump_table_cluster::can_be_handled): Handle single case
824 (jump_table_cluster::is_beneficial): Bail out for such case.
825 (bit_test_cluster::find_bit_tests):
826 Add new checking assert to catch invalid state.
827 (bit_test_cluster::can_be_handled): Handle single case
829 (bit_test_cluster::is_beneficial): Bail out for such case.
830 (switch_decision_tree::analyze_switch_statement):
833 2018-06-28 Martin Liska <mliska@suse.cz>
835 * common.opt: Introduce -completion option.
836 * gcc.c (driver_handle_option): Handle it.
837 (driver::main): Print completions if completion
839 * opt-suggestions.c (option_proposer::get_completions):
841 (option_proposer::suggest_completion): Likewise.
842 (option_proposer::find_param_completions): Likewise.
843 (verify_autocompletions): Likewise.
844 (test_completion_valid_options): Likewise.
845 (test_completion_valid_params): Likewise.
846 (in_completion_p): Likewise.
847 (empty_completion_p): Likewise.
848 (test_completion_partial_match): Likewise.
849 (test_completion_garbage): Likewise.
850 (opt_proposer_c_tests): Likewise.
851 * opt-suggestions.h: Declare new functions.
852 * opts.c (common_handle_option): Handle OPT__completion_.
853 * selftest-run-tests.c (selftest::run_tests): Add
854 opt_proposer_c_tests.
855 * selftest.c (assert_str_startswith): New.
856 * selftest.h (assert_str_startswith): Likewise.
857 (opt_proposer_c_tests): New.
858 (ASSERT_STR_STARTSWITH): Likewise.
860 2018-06-28 Martin Liska <mliska@suse.cz>
862 * Makefile.in: Add opt-suggestions.o.
863 * gcc-main.c: Include opt-suggestions.h.
864 * gcc.c (driver::driver): Likewise.
865 (driver::~driver): Remove m_option_suggestions.
866 (driver::build_option_suggestions): Moved to option_proposer.
867 (driver::suggest_option): Likewise.
868 (driver::handle_unrecognized_options): Use option_proposer.
869 * gcc.h (class driver): Add new memver m_option_proposer.
870 * opt-suggestions.c: New file.
871 * opt-suggestions.h: New file.
873 2018-06-28 Martin Liska <mliska@suse.cz>
875 * vec.h (class auto_string_vec): New (moved from auto_argvec).
876 (auto_string_vec::~auto_string_vec): Likewise.
878 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
880 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
881 prevent_decl_creation_for_types fields up and add reset_location field.
882 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
883 statement if id->reset_location is true.
884 (copy_edges_for_bb): Do not set goto_locus on the new edges if
885 id->reset_location is true.
886 (copy_phis_for_bb): Force input_location on the arguments if
887 id->reset_location is true.
888 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
889 is set on the function to be inlined.
891 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
893 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
895 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
897 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
899 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
900 hard registers for the clobbered pseudo.
902 2018-06-27 Paul Koning <ni1d@arrl.net>
904 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
905 mutually exclusive options.
906 * config/pdp11/constraints.md (h): New constraint.
907 (O): Update definition to match shift code generation.
909 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
911 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
913 (output_jump): Change arguments.
914 (pdp11_fixed_cc_regs): New function.
915 (pdp11_cc_mode): Ditto.
916 (pdp11_expand_shift): Ditto.
917 (pdp11_assemble_shift): Ditto.
918 (pdp11_small_shift): Ditto.
919 (pdp11_branch_cost): Remove.
920 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
922 (pdp11_register_move_cost): Update for CC registers.
923 (pdp11_rtx_costs): Add case for LSHIFTRT.
924 (pdp11_output_jump): Add CCNZ mode conditional branches.
925 (notice_update_cc_on_set): Remove.
926 (pdp11_cc_mode): New function.
927 (simple_memory_operand): Correct pre/post decrement case.
928 (no_side_effect_operand): New function.
929 (pdp11_regno_reg_class): Add CC_REGS class.
930 (pdp11_fixed_cc_regs): New function.
931 (pdp11_small_shift): New function.
932 (pdp11_expand_shift): New function to expand shift insns.
933 (pdp11_assemble_shift): New function to output shifts.
934 (pdp11_branch_cost): Remove.
935 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
936 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
938 (PTRDIFF_TYPE): Ditto.
939 (ADJUST_INSN_LENGTH): New macro.
940 (FIXED_REGISTERS): Add CC registers.
941 (CALL_USED_REGISTERS): Ditto.
943 (REG_CLASS_NAMES): Ditto.
944 (REG_CLASS_CONTENTS): Ditto.
945 (SELECT_CC_MODE): Use new function.
946 (TARGET_FLAGS_REGNUM): New macro.
947 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
948 (cc0_reg_rtx): Remove.
949 (CC_STATUS_MDEP): Remove.
950 (CC_STATUS_MDEFP_INIT): Remove.
952 (NOTICE_UPDATE_CC): Remove.
953 (REGISTER_NAMES): Add CC registers.
954 (BRANCH_COST): Change to constant 1.
955 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
957 * config/pdp11/pdp11.opt (mbcopy): Remove.
958 (mbcopy-builtin): Remove.
959 (mbranch-cheap): Remove.
960 (mbranch-expensive): Remove.
961 * config/pdp11/predicates.md (expand_shift_operand): Update to
962 match shift code generation.
963 (ccnz_operator): New predicate.
964 * doc/invoke.texi (PDP-11 Options): Remove deleted options
965 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
966 Remove non-existent option -mabshi, -mno-abshi. Document mutually
968 * doc/md.texi (PDP-11): Document new D and h constraints. Update
969 description of O constraint.
971 2018-06-27 Jeff Law <law@redhat.com>
972 Austin Law <austinklaw@gmail.com>
974 * config/v850/v850.md (addsi3_set_flags): New pattern.
975 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
976 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
977 (zero_extendhisi2_v850_set_flags): Likewise.
978 (zero_extendqisi2_v850_set_flags): Likewise.
979 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
980 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
981 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
983 * config/v850/v850-protos.h (notice_update_cc): Remove.
984 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
985 (v850_print_operand): Handle 'D' and "d".
986 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
987 Add handling of arithmetic/logical operations compared against zero.
988 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
989 Do not look at v850_compare_op, instead get mode from last argument.
990 (v850_gen_compare): Remove
991 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
992 after reload for prologue insns.
993 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
995 (construct_save_jarl): Likewise.
996 (TARGET_FLAGS_REGNUM): Define.
997 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
998 (NOTICE_UPDATE_CC): Remove.
999 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1000 than cc0. Conditionalize on reload_completed.
1001 (cmpsi_insn, setfcc_insn): Likewise.
1002 (tst1 splitter): Turn into define_and_split which sets the flags
1004 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1005 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1006 (cstoresf4, cstoredf4): Clobber the flags.
1007 (cmpsi, cmpsf, cmpdf): Remove expanders.
1008 (setf_insn): Remove pattern.
1009 (addsi3): Turn into define_and_split which clobbers the flags after
1010 reload and a suitable pattern (addsi3_clobber_flags) for use after
1012 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1013 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1014 (ashrsi3, ashrsi3_v850e2): Likewise.
1015 (bins): Clobber the flags.
1016 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1017 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1018 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1019 (call_value_internal_short, call_value_internal_long): Likewise.
1020 (callt_save_interrupt, callt_return_interrupt): Likewise.
1021 (save_interrupt, return_interrupt): Likewise.
1022 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1023 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1024 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1025 (All FP comparisons): Only allow after reload has completed.
1027 (divh, divhu): Tweak output template.
1028 (branch_z_normal, branch_z_invert): Remove
1029 (branch_nz_normal, branch_nz_invert): Likewise.
1030 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1032 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1033 * config/v850/v850.c (notice_update_cc): Remove.
1034 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1035 (CC_NO_CARRY): Likewise.
1036 (NOTICE_UPDATE_CC): Define to nothing.
1037 * config/v850/v850.md: Remove block comment on cc0 handling
1038 Remove "cc" attribute from all patterns. Remove cc_status handling
1039 from all patterns. Minor formatting fixes.
1041 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1043 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1044 (cortex-a76.cortex-a55): Likewise.
1045 * config/aarch64/aarch64-tune.md: Regenerate.
1046 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1047 cortex-a76.cortex-a55.
1049 2018-06-27 Jeff Law <law@redhat.com>
1051 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1052 (MULTILIB_DIRNAMES): Similarly.
1054 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1056 * gimple.h (gimple_return_retbnd): Delete.
1057 (gimple_return_set_retbnd): Likewise.
1058 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1059 gimple_return_set_retbnd.
1060 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1061 gimple_return_retbnd and adjust.
1062 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1063 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1064 Explicitly return NULL in a couple more cases. Move assertion
1065 on debug statements and remove unreachable code.
1066 (reset_debug_binding): Do not test id->retbnd.
1067 (expand_call_inline): Do not set it.
1069 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1071 * configure.ac: Add --disable-gcov option.
1072 * configure: Regenerate.
1073 * Makefile.in: Honour @enable_gcov@.
1074 * doc/install.texi: Document --disable-gcov.
1076 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1078 * config/arm/arm-cpus.in (cortex-a76): New entry.
1079 (cortex-a76.cortex-a55): Likewise.
1080 * config/arm/arm-tables.opt: Regenerate.
1081 * config/arm/arm-tune.md: Likewise.
1082 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1083 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1084 cortex-a76.cortex-a55.
1086 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1089 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1091 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1093 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1095 (EPILOGUE_USES): Likewise.
1097 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1099 * tree-inline.c (remap_location): New function extracted from...
1100 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1101 (copy_phis_for_bb): ...here. Call remap_location.
1102 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1104 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1106 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1107 unaligned vsx for 16B memset.
1109 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1112 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1113 ieee128_float_type_node to long_double_type_node unless
1114 TARGET_LONG_DOUBLE_128 is set.
1116 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1118 * cfgloop.c (get_loop_location): Convert return type from
1119 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1120 by implicit construction from rtx_insn *, and using
1121 dump_user_location_t::from_function_decl for the fallback case.
1122 * cfgloop.h (get_loop_location): Convert return type from
1123 location_t to dump_user_location_t.
1124 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1125 dump_printf_loc to pass in a dump_location_t rather than a
1126 location_t, via the gimple stmt.
1127 * coverage.c (get_coverage_counts): Update calls to
1128 dump_printf_loc to pass in dump_location_t rather than a
1130 * doc/optinfo.texi (Dump types): Convert example of
1131 dump_printf_loc from taking "locus" to taking "insn". Update
1132 description of the "_loc" calls to cover dump_location_t.
1133 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1135 (dump_user_location_t::dump_user_location_t): New constructors,
1136 from gimple *stmt and rtx_insn *.
1137 (dump_user_location_t::from_function_decl): New function.
1138 (dump_loc): Make static.
1139 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1140 const dump_location_t &.
1141 (dump_generic_expr_loc): Delete.
1142 (dump_printf_loc): Convert param "loc" from location_t to
1143 const dump_location_t &.
1144 (selftest::test_impl_location): New function.
1145 (selftest::dumpfile_c_tests): New function.
1146 * dumpfile.h: Include "profile-count.h".
1147 (class dump_user_location_t): New class.
1148 (struct dump_impl_location_t): New struct.
1149 (class dump_location_t): New class.
1150 (dump_printf_loc): Convert 2nd param from source_location to
1151 const dump_location_t &.
1152 (dump_generic_expr_loc): Delete.
1153 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1154 const dump_location_t &.
1155 * gimple-fold.c (fold_gimple_assign): Update call to
1156 dump_printf_loc to pass in a dump_location_t rather than a
1157 location_t, via the gimple stmt.
1158 (gimple_fold_call): Likewise.
1159 * gimple-loop-interchange.cc
1160 (loop_cand::analyze_iloop_reduction_var): Update for change to
1161 check_reduction_path.
1162 (tree_loop_interchange::interchange): Update for change to
1164 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1165 change in return-type of find_loop_location.
1166 (graphite_regenerate_ast_isl): Likewise.
1167 * graphite-optimize-isl.c (optimize_isl): Likewise.
1168 * graphite.c (graphite_transform_loops): Likewise.
1169 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1170 pass in a dump_location_t rather than a location_t, via the
1172 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1173 * ipa.c (walk_polymorphic_call_targets): Likewise.
1174 * loop-unroll.c (report_unroll): Convert "locus" param from
1175 location_t to dump_location_t.
1176 (decide_unrolling): Update for change to get_loop_location's
1178 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1179 location_t to dump_user_location_t.
1180 (grid_find_single_omp_among_assignments_1): Updates calls to
1181 dump_printf_loc to pass in a dump_location_t rather than a
1182 location_t, via the gimple stmt.
1183 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1184 location_t to dump_location_t. Updates calls to dump_printf_loc
1185 to pass in a dump_location_t rather than a location_t, via the
1187 (grid_inner_loop_gridifiable_p): Likewise.
1188 (grid_dist_follows_simple_pattern): Likewise.
1189 (grid_gfor_follows_tiling_pattern): Likewise.
1190 (grid_target_follows_gridifiable_pattern): Likewise.
1191 (grid_attempt_target_gridification): Convert initialization
1192 of local "grid" from memset to zero-initialization; FIXME: does
1193 this require C++11? Update call to dump_printf_loc to pass in a
1194 optinfo_location rather than a location_t, via the gimple stmt.
1195 * profile.c (read_profile_edge_counts): Updates call to
1196 dump_printf_loc to pass in a dump_location_t rather than a
1198 (compute_branch_probabilities): Likewise.
1199 * selftest-run-tests.c (selftest::run_tests): Call
1201 * selftest.h (dumpfile_c_tests): New decl.
1202 * tree-loop-distribution.c (pass_loop_distribution::execute):
1203 Update for change in return type of find_loop_location.
1204 * tree-parloops.c (parallelize_loops): Likewise.
1205 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1206 "locus" from location_t to dump_user_location_t.
1207 (canonicalize_loop_induction_variables): Likewise.
1208 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1209 for change in return type of find_loop_location.
1210 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1211 to dump_printf_loc to pass in a dump_location_t rather than a
1212 location_t, via the stmt.
1213 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1215 * tree-vect-loop-manip.c (find_loop_location): Convert return
1216 type from source_location to dump_user_location_t.
1217 (vect_do_peeling): Update for above change.
1218 (vect_loop_versioning): Update for change in type of
1220 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1221 from location_t to dump_user_location_t.
1222 (vect_estimate_min_profitable_iters): Update for change in type
1224 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1225 location_t to dump_location_t.
1226 (vect_slp_bb): Update for change in type of vect_location.
1227 * tree-vectorizer.c (vect_location): Convert from source_location
1228 to dump_user_location_t.
1229 (try_vectorize_loop_1): Update for change in vect_location's type.
1230 (vectorize_loops): Likewise.
1231 (increase_alignment): Likewise.
1232 * tree-vectorizer.h (vect_location): Convert from source_location
1233 to dump_user_location_t.
1234 (find_loop_location): Convert return type from source_location to
1235 dump_user_location_t.
1236 (check_reduction_path): Convert 1st param from location_t to
1237 dump_user_location_t.
1238 * value-prof.c (check_counter): Update call to dump_printf_loc to
1239 pass in a dump_user_location_t rather than a location_t; update
1240 call to error_at for change in type of "locus".
1241 (check_ic_target): Update call to dump_printf_loc to
1242 pass in a dump_user_location_t rather than a location_t, via the
1245 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1247 * config/s390/s390.h (enum processor_flags): Do not use
1248 default tune parameter when -march was specified.
1250 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1253 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1254 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1257 2018-06-26 Richard Biener <rguenther@suse.de>
1259 PR tree-optimization/86287
1261 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1262 (vect_analyze_loop): Initialize n_stmts.
1264 2018-06-26 Richard Biener <rguenther@suse.de>
1267 * fold-const.c (fold_convertible_p): Pointer extension
1270 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1273 * dwarf2out.c (loc_list_has_views): Adjust comments.
1274 (dw_loc_list): Split single cross-partition range with
1277 2018-06-25 Jeff Law <law@redhat.com>
1279 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1280 on -mbig-switch by default.
1282 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1284 (const_float_0_operand): Remove unused predicate.
1285 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1286 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1287 (recipsf2): New expander. Original pattern now called
1289 (recipdf2, recipdf2_insn): Similarly.
1290 (rsqrtsf2, rsqrtsf2_insn): Similarly
1291 (rsqrtdf2, rsqrtdf2_insn): Similarly
1293 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1295 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1296 Simplify logic for FreeBSD (twice).
1298 2018-06-25 Martin Sebor <msebor@redhat.com>
1300 PR tree-optimization/86204
1301 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1302 a strnlen result if it's less than the length of the string.
1304 2018-06-25 Martin Sebor <msebor@redhat.com>
1306 PR tree-optimization/85700
1307 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1308 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1309 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1311 2018-06-25 Martin Sebor <msebor@redhat.com>
1313 * doc/extend.texi (Zero-length arrays): Update and clarify.
1315 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1317 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1318 added IEEE/IBM long double multilib support on PowerPC little
1319 endian Linux systems.
1320 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1321 (MULTILIB_DEFAULTS): Likewise.
1322 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1324 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1325 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1326 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1328 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1331 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1332 (REORDER_45): Likewise.
1334 2018-06-25 Jeff Law <law@redhat.com>
1336 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1337 dividend to 32 bits. Adjust length.
1338 (udivmodhi4): Cleanup output template. Fix length.
1340 2018-06-25 Carl Love <cel@us.ibm.com>
1342 * config/rs6000/vsx.md: Change word selector to prefered location.
1344 2018-06-25 Richard Biener <rguenther@suse.de>
1346 PR tree-optimization/86304
1347 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1348 epilogue-if-converted loops as well.
1350 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1352 * lto-section-out.c (lto_begin_section): Do not print section
1353 name for noaddr and unnumbered dumps.
1355 2018-06-25 Richard Biener <rguenther@suse.de>
1357 * tree-vectorizer.h (struct vec_info_shared): New structure
1358 with parts split out from struct vec_info and loop_nest from
1359 struct _loop_vec_info.
1360 (struct vec_info): Adjust accordingly.
1361 (struct _loop_vec_info): Likewise.
1362 (LOOP_VINFO_LOOP_NEST): Adjust.
1363 (LOOP_VINFO_DATAREFS): Likewise.
1364 (LOOP_VINFO_DDRS): Likewise.
1365 (struct _bb_vec_info): Likewise.
1366 (BB_VINFO_DATAREFS): Likewise.
1367 (BB_VINFO_DDRS): Likewise.
1368 (struct _stmt_vec_info): Add dr_aux member.
1369 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1370 (DR_MISALIGNMENT_UNINITIALIZED): New.
1371 (set_dr_misalignment): Adjust.
1372 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1373 (vect_analyze_loop): Adjust prototype.
1374 (vect_analyze_loop_form): Likewise.
1375 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1376 Compute dependences lazily.
1377 (vect_record_base_alignments): Use shared datarefs/ddrs.
1378 (vect_verify_datarefs_alignment): Likewise.
1379 (vect_analyze_data_refs_alignment): Likewise.
1380 (vect_analyze_data_ref_accesses): Likewise.
1381 (vect_analyze_data_refs): Likewise.
1382 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1383 constructor parameter for shared part.
1384 (vect_analyze_loop_form): Pass in shared part and adjust.
1385 (vect_analyze_loop_2): Pass in storage for the number of
1386 stmts. Move loop nest finding to the caller. Compute
1388 (vect_analyze_loop): Pass in shared part.
1389 (vect_transform_loop): Verify shared datarefs are unchanged.
1390 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1391 constructor parameter for shared part.
1392 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1393 (vect_slp_bb): Verify shared datarefs are unchanged before
1395 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1397 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1398 DR_MISALIGNMENT_UNINITIALIZED.
1399 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1400 parameter for shared part.
1401 (vec_info::~vec_info): Adjust.
1402 (vec_info_shared::vec_info_shared): New.
1403 (vec_info_shared::~vec_info_shared): Likewise.
1404 (vec_info_shared::save_datarefs): Likewise.
1405 (vec_info_shared::check_datarefs): Likewise.
1406 (try_vectorize_loop_1): Construct shared part live for analyses
1407 of a single loop for multiple vector sizes.
1408 * tree-parloops.c (gather_scalar_reductions): Adjust.
1410 2018-06-25 Richard Biener <rguenther@suse.de>
1412 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1413 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1414 (vect_analyze_data_refs): Remove similar code from here and
1415 simplify accordingly.
1417 2018-06-25 Richard Biener <rguenther@suse.de>
1419 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1420 for reverse storage order accesses rather than asserting
1421 they cannot happen here.
1423 2018-06-25 Tom de Vries <tdevries@suse.de>
1426 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1427 Use data16 instead of .byte for insn prefix.
1429 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1432 * parser.c (make_char_string_pack): Pass this literal chars
1433 through cpp_interpret_string.
1434 (cp_parser_userdef_numeric_literal): Check the result of
1435 make_char_string_pack.
1437 2018-06-24 Maya Rashish <coypu@sdf.org>
1439 * ginclude/stddef.h: Simplify conditions around avoiding
1440 re-definition of __size_t.
1442 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1444 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1445 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1447 2018-06-22 Maya Rashish <coypu@sdf.org>
1449 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1450 made to OpenBSD and NetBSD through the years.
1452 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1454 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1455 behavior of vec_pack (vector double, vector double) to match
1456 behavior of vec_float2 (vector double, vector double).
1458 2018-06-22 Olivier Hainque <hainque@adacore.com>
1460 * gimplify.c (gimplify_function_tree): Prevent creation
1461 of a trampoline for the address of the current function
1462 passed to entry/exit instrumentation hooks.
1464 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1467 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1470 2018-06-22 Martin Liska <mliska@suse.cz>
1472 PR tree-optimization/86263
1473 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1474 Bail out if is_enabled is false.
1475 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1477 (jump_table_cluster::is_enabled): New function.
1479 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1481 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1482 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1483 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1484 (lto_input_ts_binfo_tree_pointers): Likewise.
1485 * tree-streamer-out.c (streamer_write_tree_bitfields,
1486 write_ts_binfo_tree_pointers): Likewise.
1487 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1489 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1491 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1493 2018-06-22 Martin Liska <mliska@suse.cz>
1495 * symbol-summary.h (get): Make it pure and inline move
1496 functionality from ::get function.
1497 (get): Remove and inline into ::get and ::get_create.
1498 (get_create): Move code from ::get function.
1500 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1503 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1504 -x assembler-with-cpp.
1506 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1508 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1509 _FILE_OFFSET_BITS=64 for C++.
1511 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1513 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1514 conversion insn that shows up when pr85657-3.c is compiled using
1515 IEEE 128-bit long double.
1516 (neg<mode>2_internal): Use the correct mode to check whether the
1517 mode is IBM extended.
1518 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1519 multiply and divide external functions from being created more
1522 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1524 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1526 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1527 the edge can be forwarded.
1528 (cfg_layout_merge_blocks): Likewise.
1530 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1532 * except.c (finish_eh_generation): Commit edge insertions only after
1533 the EH edges have been redirected from post-landing to landing pads.
1535 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1537 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1538 create_tmp_var_for to create the FRAME decl.
1539 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1541 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1543 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1544 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1545 debug statement when resetting its value.
1546 (expand_call_inline): Copy the locus of the call onto the assignment
1547 of the return value, if any. Use local variable in more cases.
1549 2018-06-21 Martin Liska <mliska@suse.cz>
1551 * ipa-pure-const.c (propagate_nothrow): Use
1552 funct_state_summaries->get.
1553 (dump_malloc_lattice): Likewise.
1554 (propagate_malloc): Likewise.
1556 2018-06-21 Richard Biener <rguenther@suse.de>
1558 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1559 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1560 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1562 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1563 BLOCK_ABSTRACT_ORIGIN unconditionally.
1565 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1567 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1569 * ipa-reference.c (ipa_reference_c_finalize): Delete
1570 ipa_ref_opt_sum_summaries and set it to NULL.
1572 2018-06-21 Tom de Vries <tdevries@suse.de>
1574 PR tree-optimization/85859
1575 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1576 test with comment from bb_no_side_effects_p.
1578 2018-06-21 Richard Biener <rguenther@suse.de>
1580 PR tree-optimization/86232
1581 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1582 max for constant niter.
1584 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1586 * config/aarch64/aarch64-simd.md
1587 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1589 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1591 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1592 Make opernads of the unspec commutative.
1594 2018-06-21 Richard Biener <rguenther@suse.de>
1596 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1597 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1598 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1599 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1601 (vect_record_base_alignments): Do not record base alignment
1602 for gathers or scatters.
1603 (vect_compute_data_ref_alignment): Drop return value that is always
1604 true. Bail out early for gathers or scatters.
1605 (vect_enhance_data_refs_alignment): Bail out early for gathers
1607 (vect_find_same_alignment_drs): Likewise.
1608 (vect_analyze_data_refs_alignment): Remove dead code.
1609 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1610 (vect_analyze_data_refs): For possible gathers or scatters do
1611 not create an alternate DR, just check their possible validity
1612 and mark them. Adjust DECL_NONALIASED handling to not rely
1614 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1615 update inits of gathers or scatters.
1616 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1617 Also copy gather/scatter flag to pattern vinfo.
1619 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1621 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1622 behavior of vec_packsu (vector unsigned long long, vector unsigned
1623 long long) to match behavior of vec_packs with same signature.
1625 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1626 Thomas Schwinge <thomas@codesourcery.com>
1627 Cesar Philippidis <cesar@codesourcery.com>
1629 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1630 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1631 (gimplify_adjust_omp_clauses): Likewise.
1632 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1633 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1634 (gimplify_omp_target_update): Update handling of acc update and
1636 * omp-low.c (install_var_field): Remove unused parameter
1637 base_pointers_restrict.
1638 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1639 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1641 (omp_target_base_pointers_restrict_p): Delete.
1642 (scan_omp_target): Update call to scan_sharing_clauses.
1643 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1645 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1646 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1647 (convert_local_omp_clauses): Likewise.
1648 * tree-pretty-print.c (dump_omp_clause): Likewise.
1649 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1651 (omp_clause_code_name): Likewise.
1653 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1656 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1659 PR tree-optimization/86231
1660 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1661 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1663 2018-06-20 Tom de Vries <tdevries@suse.de>
1665 PR tree-optimization/86097
1666 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1667 iv type if signedness of iv type is not the same as that of *nit.
1669 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1671 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1672 EDGE_EH edges, verify they are all EDGE_EH.
1674 2018-06-20 Maya Rashish <coypu@sdf.org>
1676 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1678 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1680 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1681 * config/aarch64/aarch64.c (xgene1_tunings): Add
1682 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1683 (aarch64_mode_valid_for_sched_fusion_p):
1684 Allow 16-byte modes.
1685 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1686 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1688 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1690 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1691 * config/aarch64/iterators.md (VQ2): New mode iterator.
1693 2018-06-20 Martin Liska <mliska@suse.cz>
1695 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1696 Change default ratio from 10 to 8.
1698 2018-06-20 Martin Liska <mliska@suse.cz>
1700 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1702 (bit_test_cluster::find_bit_tests): Likewise.
1703 (switch_decision_tree::analyze_switch_statement): Find clusters.
1704 * tree-switch-conversion.h (struct jump_table_cluster): Document
1707 2018-06-20 Martin Liska <mliska@suse.cz>
1709 * tree-switch-conversion.c (switch_conversion::collect):
1710 Record m_uniq property.
1711 (switch_conversion::expand): Bail out for special conditions.
1712 (group_cluster::~group_cluster): New.
1713 (group_cluster::group_cluster): Likewise.
1714 (group_cluster::dump): Likewise.
1715 (jump_table_cluster::emit): New.
1716 (switch_decision_tree::fix_phi_operands_for_edges): New.
1717 (struct case_node): Remove struct.
1718 (jump_table_cluster::can_be_handled): New.
1719 (case_values_threshold): Moved to header.
1720 (reset_out_edges_aux): Likewise.
1721 (jump_table_cluster::is_beneficial): New.
1722 (bit_test_cluster::can_be_handled): Likewise.
1723 (add_case_node): Remove.
1724 (bit_test_cluster::is_beneficial): New.
1725 (case_bit_test::cmp): New.
1726 (bit_test_cluster::emit): New.
1727 (expand_switch_as_decision_tree_p): Remove.
1728 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1729 (fix_phi_operands_for_edge): Likewise.
1730 (switch_decision_tree::analyze_switch_statement): New.
1731 (compute_cases_per_edge): Move ...
1732 (switch_decision_tree::compute_cases_per_edge): ... here.
1733 (try_switch_expansion): Likewise.
1734 (switch_decision_tree::try_switch_expansion): Likewise.
1735 (record_phi_operand_mapping): Likewise.
1736 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1737 (emit_case_decision_tree): Likewise.
1738 (switch_decision_tree::emit): Likewise.
1739 (balance_case_nodes): Likewise.
1740 (switch_decision_tree::balance_case_nodes): Likewise.
1741 (dump_case_nodes): Likewise.
1742 (switch_decision_tree::dump_case_nodes): Likewise.
1743 (emit_jump): Likewise.
1744 (switch_decision_tree::emit_jump): Likewise.
1745 (emit_cmp_and_jump_insns): Likewise.
1746 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1747 (emit_case_nodes): Likewise.
1748 (switch_decision_tree::emit_case_nodes): Likewise.
1749 (conditional_probability): Remove.
1750 * tree-switch-conversion.h (enum cluster_type): New.
1752 (struct cluster): Likewise.
1753 (cluster::cluster): Likewise.
1754 (struct simple_cluster): Likewise.
1755 (simple_cluster::simple_cluster): Likewise.
1756 (struct group_cluster): Likewise.
1757 (struct jump_table_cluster): Likewise.
1758 (struct bit_test_cluster): Likewise.
1759 (struct min_cluster_item): Likewise.
1760 (struct case_tree_node): Likewise.
1761 (case_tree_node::case_tree_node): Likewise.
1762 (jump_table_cluster::case_values_threshold): Likewise.
1763 (struct case_bit_test): Likewise.
1764 (struct switch_decision_tree): Likewise.
1765 (struct switch_conversion): Likewise.
1766 (switch_decision_tree::reset_out_edges_aux): Likewise.
1768 2018-06-20 Martin Liska <mliska@suse.cz>
1770 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1771 (hoist_edge_and_branch_if_true): Likewise.
1772 (expand_switch_using_bit_tests_p): Likewise.
1773 (struct case_bit_test): Likewise.
1774 (case_bit_test_cmp): Likewise.
1775 (emit_case_bit_tests): Likewise.
1776 (switch_conversion::switch_conversion): New class.
1777 (struct switch_conv_info): Remove old struct.
1778 (collect_switch_conv_info): More to ...
1779 (switch_conversion::collect): ... this.
1780 (check_range): Likewise.
1781 (switch_conversion::check_range): Likewise.
1782 (check_all_empty_except_final): Likewise.
1783 (switch_conversion::check_all_empty_except_final): Likewise.
1784 (check_final_bb): Likewise.
1785 (switch_conversion::check_final_bb): Likewise.
1786 (create_temp_arrays): Likewise.
1787 (switch_conversion::create_temp_arrays): Likewise.
1788 (free_temp_arrays): Likewise.
1789 (gather_default_values): Likewise.
1790 (switch_conversion::gather_default_values): Likewise.
1791 (build_constructors): Likewise.
1792 (switch_conversion::build_constructors): Likewise.
1793 (constructor_contains_same_values_p): Likewise.
1794 (switch_conversion::contains_same_values_p): Likewise.
1795 (array_value_type): Likewise.
1796 (switch_conversion::array_value_type): Likewise.
1797 (build_one_array): Likewise.
1798 (switch_conversion::build_one_array): Likewise.
1799 (build_arrays): Likewise.
1800 (switch_conversion::build_arrays): Likewise.
1801 (gen_def_assigns): Likewise.
1802 (switch_conversion::gen_def_assigns): Likewise.
1803 (prune_bbs): Likewise.
1804 (switch_conversion::prune_bbs): Likewise.
1805 (fix_phi_nodes): Likewise.
1806 (switch_conversion::fix_phi_nodes): Likewise.
1807 (gen_inbound_check): Likewise.
1808 (switch_conversion::gen_inbound_check): Likewise.
1809 (process_switch): Use the newly created class.
1810 (switch_conversion::expand): New.
1811 (switch_conversion::~switch_conversion): New.
1812 * tree-switch-conversion.h: New file.
1814 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1816 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1817 tree-vect-patterns.c.
1818 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1819 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1820 (vect_recog_sad_pattern): Likewise.
1821 (vect_recog_widen_sum_pattern): Likewise.
1822 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1823 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1824 (vect_recog_rotate_pattern): Likewise.
1825 (vect_recog_mult_pattern): Likewise.
1826 (vect_recog_vector_vector_shift_pattern): Likewise.
1827 (vect_recog_divmod_pattern): Likewise.
1828 (vect_recog_mixed_size_cond_pattern): Likewise.
1829 (vect_recog_bool_pattern): Likewise.
1830 (vect_recog_mask_conversion_pattern): Likewise.
1831 (vect_try_gather_scatter_pattern): Likewise.
1832 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1833 (vect_recog_over_widening_pattern): Likewise.
1834 (vect_recog_gather_scatter_pattern): Likewise.
1835 (vect_recog_func_ptr): Move from tree-vectorizer.h
1836 (vect_vect_recog_func_ptrs): Move further down the file.
1837 (vect_recog_func): Likewise. Remove the third argument.
1838 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1839 (vect_pattern_recog_1): Expect the pattern function to do any
1840 necessary target tests. Also expect it to provide a vector type.
1841 Remove the type_in handling.
1843 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1845 * tree-vect-patterns.c (vect_pattern_detected): New function.
1846 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1847 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1848 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1849 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1850 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1851 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1852 (vect_recog_mask_conversion_pattern)
1853 (vect_try_gather_scatter_pattern): Likewise.
1855 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1857 * tree-vect-patterns.c (vect_get_internal_def): New function.
1858 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1859 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1860 (search_type_for_mask_1): Use it.
1862 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1864 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1865 redundant WIDEN_SUM_EXPR handling.
1866 (vect_recog_sad_pattern): Likewise.
1868 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1870 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1871 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1872 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1873 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1874 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1876 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1878 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1879 use the stmt_vec_info of the original bb statement for the
1880 new zero assignment, even if the call is part of a pattern.
1882 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1884 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1885 that the sequence is attached to the original statement rather
1886 than the pattern statement.
1887 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1888 PATTERN_DEF_SEQ from the original statement rather than
1889 the main pattern statement.
1890 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1891 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1892 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1894 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1896 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1897 definition statements before the early exit for statements that aren't
1899 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1901 (vect_transform_loop): ...here. Process pattern definition
1902 statements without first checking whether the main pattern
1903 statement is live or relevant.
1905 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1907 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1908 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1910 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1912 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1913 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1914 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1915 (expand_block_compare): Change select_block_compare_mode call.
1916 (expand_strncmp_align_check): Use new functions, fix comment.
1917 (emit_final_str_compare_gpr): New function.
1918 (expand_strn_compare): Refactor and clean up code.
1919 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1921 2018-06-19 Tony Reix <tony.reix@atos.com>
1922 Damien Bergamini <damien.bergamini@atos.com>
1923 David Edelsohn <dje.gcc@gmail.com>
1925 * collect2.c (static_obj): New variable.
1926 (static_libs): New variable.
1927 (is_in_list): Uncomment declaration.
1928 (main): Track AIX libraries linked statically.
1929 (is_in_list): Uncomment definition.
1930 (scan_prog_file): Don't add AIX shared libraries initializer
1931 to constructor list if linking statically.
1933 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1935 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1937 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1939 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1941 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1944 2018-06-19 Martin Liska <mliska@suse.cz>
1946 * config/i386/i386.c (ix86_can_inline_p): Do not use
1947 ipa_fn_summaries::get_create.
1948 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1950 (devirtualization_time_bonus): Likewise.
1951 (ipcp_propagate_stage): Likewise.
1952 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1953 (edge_set_predicate): Likewise.
1954 (evaluate_conditions_for_known_args): Likewise.
1955 (evaluate_properties_for_edge): Likewise.
1956 (ipa_call_summary::reset): Tranform to ...
1957 (ipa_call_summary::~ipa_call_summary): ... this.
1958 (ipa_fn_summary::reset): Transform to ...
1959 (ipa_fn_summary::~ipa_fn_summary): ... this.
1960 (ipa_fn_summary_t::remove): Rename to ...
1961 (ipa_fn_summary_t::remove_callees): ... this.
1962 (ipa_fn_summary_t::duplicate): Use placement new
1963 instead of memory copy.
1964 (ipa_call_summary_t::duplicate): Likewise.
1965 (ipa_call_summary_t::remove): Remove.
1966 (dump_ipa_call_summary): Change get_create to get.
1967 (ipa_dump_fn_summary): Dump only when summary exists.
1968 (analyze_function_body): Use symbol_summary::get instead
1970 (compute_fn_summary): Likewise.
1971 (estimate_edge_devirt_benefit): Likewise.
1972 (estimate_edge_size_and_time): Likewise.
1973 (inline_update_callee_summaries): Likewise.
1974 (remap_edge_change_prob): Likewise.
1975 (remap_edge_summaries): Likewise.
1976 (ipa_merge_fn_summary_after_inlining): Likewise.
1977 (write_ipa_call_summary): Likewise.
1978 (ipa_fn_summary_write): Likewise.
1979 (ipa_free_fn_summary): Likewise.
1980 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1981 (struct ipa_call_summary): Likewise.
1982 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1984 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1985 (estimate_size_after_inlining): Likewise.
1986 (estimate_growth): Likewise.
1987 (growth_likely_positive): Likewise.
1988 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1989 (inline_call): Likewise.
1990 * ipa-inline.c (caller_growth_limits): Likewise.
1991 (can_inline_edge_p): Likewise.
1992 (can_inline_edge_by_limits_p): Likewise.
1993 (compute_uninlined_call_time): Likewise.
1994 (compute_inlined_call_time): Likewise.
1995 (want_inline_small_function_p): Likewise.
1996 (edge_badness): Likewise.
1997 (update_caller_keys): Likewise.
1998 (update_callee_keys): Likewise.
1999 (inline_small_functions): Likewise.
2000 (inline_to_all_callers_1): Likewise.
2001 (dump_overall_stats): Likewise.
2002 (early_inline_small_functions): Likewise.
2003 (early_inliner): Likewise.
2004 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2005 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2006 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2007 * ipa-split.c (execute_split_functions): Likewise.
2008 * symbol-summary.h: Likewise.
2009 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2011 2018-06-19 Richard Biener <rguenther@suse.de>
2013 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2014 (vectorize_loops): ... here. Fix dbgcnt handling.
2015 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2017 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2020 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2021 ieee128 argument takes up only one (vector) register, not two (floating
2024 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2026 * gimplify.c (gimplify_init_constructor): Really never clear for an
2027 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2029 2018-06-19 Richard Biener <rguenther@suse.de>
2031 PR tree-optimization/86179
2032 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2033 after failed recognition.
2035 2018-06-18 Martin Sebor <msebor@redhat.com>
2038 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2039 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2040 Handle integer subtraction.
2041 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2042 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2044 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2046 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2047 param from rtx to rtx_insn *.
2048 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2050 (frv_ifcvt_modify_insn): Likwise.
2051 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2052 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2053 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2054 as_a <rtx_insn *> cast to local "unprotected_region" once
2055 it's been established that it's not NULL or pc_rtx.
2056 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2057 param "sethi" from rtx to rtx_insn *.
2058 (nds32_group_float_insns): Likewise for param "insn".
2059 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2061 (vax_output_int_subtract): Likewise.
2062 * config/vax/vax.c (vax_output_int_add): Likewise for param
2064 (vax_output_int_subtract): Likewise.
2065 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2066 (emit_pattern_after): Likewise for param "after".
2067 (emit_insn_after): Likewise.
2068 (emit_jump_insn_after): Likewise.
2069 (emit_call_insn_after): Likewise.
2070 (emit_debug_insn_after): Likewise.
2071 (emit_pattern_before): Likewise for param "before".
2072 (emit_insn_before): Likewise.
2073 (emit_jump_insn_before): Likewise.
2074 * final.c (get_insn_template): Likewise for param "insn", removing
2076 * output.h (get_insn_template): Likewise for 2nd param.
2077 * rtl.h (emit_insn_before): Likewise.
2078 (emit_jump_insn_before): Likewise.
2079 (emit_debug_insn_before_noloc): Likewise.
2080 (emit_insn_after): Likewise.
2081 (emit_jump_insn_after): Likewise.
2082 (emit_call_insn_after): Likewise.
2083 (emit_debug_insn_after): Likewise.
2084 (set_insn_deleted): Likewise for param.
2086 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2089 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2090 floating point modes, so that IFmode is numerically greater than
2091 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2092 to declare the ordering. This prevents IFmode from being
2093 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2094 machine. Include rs6000-modes.h to share the fractional values
2095 between genmodes* and the rest of the compiler.
2099 * config/rs6000/rs6000-modes.h: New file.
2100 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2101 meaning of rs6000_long_double_size so that 126..128 selects an
2102 appropriate 128-bit floating point type.
2103 (rs6000_option_override_internal): Likewise.
2104 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2105 (TARGET_LONG_DOUBLE_128): Change the meaning of
2106 rs6000_long_double_size so that 126..128 selects an appropriate
2107 128-bit floating point type.
2108 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2109 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2110 source and destination to match the standard usage.
2111 (truncifkf2): Likewise.
2112 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2113 ISA 2.07 to use an explicit clobber, instead of passing in a
2115 (copysign<mode>3_soft): Likewise.
2117 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2119 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2120 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2121 (vect_slp_analyze_instance_dependence): Likewise.
2122 (vect_enhance_data_refs_alignment): Likewise.
2123 (vect_analyze_data_refs_alignment): Likewise.
2124 (vect_slp_analyze_and_verify_instance_alignment
2125 (vect_analyze_data_ref_accesses): Likewise.
2126 (vect_prune_runtime_alias_test_list): Likewise.
2127 (vect_analyze_data_refs): Likewise.
2128 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2129 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2130 (vect_analyze_scalar_cycles_1): Likewise.
2131 (vect_get_loop_niters): Likewise.
2132 (vect_analyze_loop_form_1): Likewise.
2133 (vect_update_vf_for_slp): Likewise.
2134 (vect_analyze_loop_operations): Likewise.
2135 (vect_analyze_loop): Likewise.
2136 (vectorizable_induction): Likewise.
2137 (vect_transform_loop): Likewise.
2138 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2139 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2140 (vect_make_slp_decision): Likewise.
2141 (vect_detect_hybrid_slp): Likewise.
2142 (vect_slp_analyze_operations): Likewise.
2143 (vect_slp_bb): Likewise.
2144 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2145 (vectorizable_bswap): Likewise.
2146 (vectorizable_call): Likewise.
2147 (vectorizable_simd_clone_call): Likewise.
2148 (vectorizable_conversion): Likewise.
2149 (vectorizable_assignment): Likewise.
2150 (vectorizable_shift): Likewise.
2151 (vectorizable_operation): Likewise.
2152 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2154 2018-06-18 Martin Sebor <msebor@redhat.com>
2156 PR tree-optimization/81384
2157 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2158 * builtins.c (expand_builtin_strnlen): New function.
2159 (expand_builtin): Call it.
2160 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2161 * builtins.def (BUILT_IN_STRNLEN): New.
2162 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2163 Warn for bounds in excess of maximum object size.
2164 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2165 single-value ranges. Handle strnlen.
2166 (handle_builtin_strlen): Handle strnlen.
2167 (strlen_check_and_optimize_stmt): Same.
2168 * doc/extend.texi (Other Builtins): Document strnlen.
2170 2018-06-18 Maya Rashish <coypu@sdf.org>
2172 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2173 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2174 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2176 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2178 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2180 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2182 * tree.c (escaped_string::escape): Replace cast to char * by
2183 const_cast<char *> (unescaped).
2185 2018-06-18 Nick Clifton <nickc@redhat.com>
2188 * tree.c (escaped_string): New class. Converts an unescaped
2189 string into its escaped equivalent.
2190 (warn_deprecated_use): Use the new class to convert the
2191 deprecation message, if present.
2192 (test_escaped_strings): New self test.
2193 (test_c_tests): Add test_escaped_strings.
2194 * doc/extend.texi (deprecated): Add a note that the
2195 deprecation message is affected by the -fmessage-length
2196 option, and that control characters will be escaped.
2197 (#pragma GCC error): Document this pragma.
2198 (#pragma GCC warning): Likewise.
2199 * doc/invoke.texi (-fmessage-length): Document this option's
2200 effect on the #warning and #error preprocessor directives and
2201 the deprecated attribute.
2203 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2205 * tree.c (decl_value_expr_lookup): Revert latest change.
2206 (decl_value_expr_insert): Likewise.
2208 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2210 * gimplify.c (nonlocal_vlas): Delete.
2211 (nonlocal_vla_vars): Likewise.
2212 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2214 (gimplify_body): Do not create and destroy nonlocal_vlas.
2215 * tree-nested.c: Include diagnostic.h.
2216 (use_pointer_in_frame): Tweak.
2217 (lookup_field_for_decl): Add assertion and declare the transformation.
2218 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2219 internal error when the reference is in a wrong context. Do not
2220 create a debug decl by default.
2221 (note_nonlocal_block_vlas): Delete.
2222 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2223 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2224 create a debug decl by default.
2225 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2226 call is in a wrong context.
2227 (fixup_vla_decls): New function.
2228 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2229 debug variables were created.
2230 * tree.c (decl_value_expr_lookup): Add checking assertion.
2231 (decl_value_expr_insert): Likewise.
2233 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2236 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2237 * tree-scalar-evolution.c (interpret_expr): Likewise.
2238 (expression_expensive_p): Likewise.
2239 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2240 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2241 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2242 (ssa_defined_by_minus_one_stmt_p): New.
2244 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2247 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2248 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2249 * dojump.c (do_jump): Likewise.
2250 * expr.c (expand_expr_real_2): Check operand type's sign.
2251 * fold-const.c (const_unop): Handle ABSU_EXPR.
2252 (fold_abs_const): Likewise.
2253 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2254 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2255 (strip_sign_op_1): Likesise.
2256 * match.pd: Add new pattern to generate ABSU_EXPR.
2257 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2258 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2259 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2260 * tree-inline.c (estimate_operator_cost): Likewise.
2261 * tree-pretty-print.c (dump_generic_node): Likewise.
2262 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2263 * tree.def (ABSU_EXPR): New.
2265 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2268 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2269 documented as preserved for backward compatibility only.
2270 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2272 PR rtl-optimization/86108
2273 * bb-reorder.c (create_forwarder_block): Renamed to ...
2274 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2275 jump from new landing pad to the second part.
2276 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2279 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2282 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2283 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2284 Only call store_expr for halves if the mode is the same.
2287 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2288 Fix up comment formatting.
2290 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2292 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2293 (typed_splay_tree::closure,
2294 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2295 (typed_splay_tree::typed_splay_tree,
2296 typed_splay_tree::operator =): Declared private.
2297 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2298 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2299 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2300 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2301 typed_splay_tree::splay_tree_splay,
2302 typed_splay_tree::splay_tree_foreach_helper,
2303 typed_splay_tree::splay_tree_insert,
2304 typed_splay_tree::splay_tree_remove,
2305 typed_splay_tree::splay_tree_lookup,
2306 typed_splay_tree::splay_tree_predecessor,
2307 typed_splay_tree::splay_tree_successor,
2308 typed_splay_tree::splay_tree_min,
2309 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2310 (typed_splay_tree::root, typed_splay_tree::comp,
2311 typed_splay_tree::delete_key,
2312 typed_splay_tree::delete_value): New data members.
2313 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2314 typed_splay_tree::remove.
2316 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2318 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2319 -mginv and -mno-ginv to the assembler.
2320 * config/mips/mips.opt (-mcrc): New option.
2322 * doc/invoke.text (-mcrc): Document.
2325 2018-06-15 Nick Clifton <nickc@redhat.com>
2328 * tree.c (escaped_string): New class. Converts an unescaped
2329 string into its escaped equivalent.
2330 (warn_deprecated_use): Use the new class to convert the
2331 deprecation message, if present.
2332 (test_escaped_strings): New self test.
2333 (test_c_tests): Add test_escaped_strings.
2334 * doc/extend.texi (deprecated): Add a note that the
2335 deprecation message is affected by the -fmessage-length
2336 option, and that control characters will be escaped.
2337 (#pragma GCC error): Document this pragma.
2338 (#pragma GCC warning): Likewise.
2339 * doc/invoke.texi (-fmessage-length): Document this option's
2340 effect on the #warning and #error preprocessor directives and
2341 the deprecated attribute.
2343 2018-06-15 Richard Biener <rguenther@suse.de>
2345 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2346 here, also noting vector size used.
2347 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2348 size used in MSG_OPTIMIZED_LOCATIONS dump.
2349 (pass_slp_vectorize::execute): Adjust.
2351 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2354 * config/arc/arc.c (arc_return_address_register): Fix
2357 2018-06-15 Richard Biener <rguenther@suse.de>
2360 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2361 leave useless conversion stripping to force_gimple_operand_gsi.
2362 (gimplify_build2): Likewise.
2363 (gimplify_build1): Likewise.
2365 2018-06-15 Richard Biener <rguenther@suse.de>
2368 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2369 before adjusting their block.
2371 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2373 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2374 multilibs for *-*-rtems*.
2375 * config/riscv/t-rtems: New file.
2377 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2380 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2381 unsigned_type_for returns NULL.
2384 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2385 subregs of multi-word pseudos unless the float mode has word size.
2387 2018-06-14 Richard Biener <rguenther@suse.de>
2390 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2391 duplicate and harmful type_hash_canon.
2392 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2394 2018-06-14 Richard Biener <rguenther@suse.de>
2397 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2400 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2402 * config/rtems.h (STDINT_LONG32): Define.
2404 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2405 Prachi Godbole <prachi.godbole@imgtec.com>
2407 * config/mips/mips-cpus.def: Define P6600.
2408 * config/mips/mips-tables.opt: Regenerate.
2409 * config/mips/mips.c (mips_ucbranch_type): New enum.
2410 (mips_rtx_cost_data): Add support for P6600.
2411 (mips_issue_rate): Likewise.
2412 (mips_multipass_dfa_lookahead): Likewise.
2413 (mips_avoid_hazard): Likewise.
2414 (mips_reorg_process_insns): Likewise.
2415 (mips_classify_branch_p6600): New function.
2416 * config/mips/mips.h (TUNE_P6600): New define.
2417 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2418 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2419 * config/mips/mips.md: Include p6600.md.
2420 (processor): Add p6600.
2421 * config/mips/p6600.md: New file.
2422 * doc/invoke.texi: Add p6600 to supported architectures.
2424 2018-06-13 Martin Sebor <msebor@redhat.com>
2426 PR tree-optimization/86114
2427 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2429 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2431 2018-06-13 Richard Biener <rguenther@suse.de>
2433 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2434 Properly set vector type of the intermediate stmt.
2435 * tree-vect-stmts.c (vectorizable_operation): The destination
2436 var always has vectype_out type.
2438 2018-06-13 Jeff Law <law@redhat.com>
2440 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2441 integer 0 for argument to print_rtl_with_bb.
2442 (rl78_reorg): Likewise.
2444 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2446 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2447 from rtx to rtx_insn *.
2448 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2450 (add_sched_insns_for_speculation): Likewise for local "target",
2451 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2452 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2453 from rtx_insn ** to rtx_code_label **.
2454 (reorg_emit_nops): Likewise.
2455 (c6x_reorg): Likewise for local "call_labels".
2456 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2458 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2459 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2460 the loops over LABEL_REFS.
2461 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2463 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2464 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2465 (split_branches): Strengthen local "olabel" from rtx to
2466 rtx_insn *, adding a safe_as_a cast.
2467 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2469 (add_insn_after): Likewise for first two params.
2470 (add_insn_before): Likewise.
2471 (remove_insn): Likewise for param.
2472 (emit_pattern_before_noloc): Likewise for second and third params.
2473 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2474 (emit_call_insn_before_noloc): Likewise.
2475 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2477 (emit_barrier_before): Likewise.
2478 (emit_label_before): Strengthen "label" param from "rtx" to
2479 "rtx_code_label *". Strengthen "before" param from "rtx" to
2481 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2483 (emit_pattern_after_noloc): Likewise.
2484 (emit_insn_after_noloc): Likewise.
2485 (emit_jump_insn_after_noloc): Likewise.
2486 (emit_call_insn_after_noloc): Likewise.
2487 (emit_debug_insn_after_noloc): Likewise.
2488 (emit_barrier_after): Likewise.
2489 (emit_label_after): Likewise for both params.
2490 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2491 "loc" param from "int" to "location_t".
2492 (emit_insn_after_setloc): Likewise.
2493 (emit_jump_insn_after_setloc): Likewise.
2494 (emit_call_insn_after_setloc): Likewise.
2495 (emit_debug_insn_after_setloc): Likewise.
2496 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2497 "loc" param from "int" to "location_t".
2498 (emit_pattern_before): Convert NULL_RTX to NULL.
2499 (emit_insn_before_setloc): Convert "loc" param from "int" to
2501 (emit_jump_insn_before_setloc): Likewise.
2502 (emit_call_insn_before_setloc): Likewise.
2503 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2504 rtx_insn *. Convert "loc" param from "int" to "location_t".
2505 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2506 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2507 Convert 3rd param from "int" to "location_t".
2508 (emit_barrier_before, emit_barrier_after, next_real_insn):
2509 Strengthen param from rtx to rtx_insn *.
2510 (emit_label_before): Strengthen 1st param from "rtx" to
2511 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2513 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2514 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2515 Strengthen 2nd param from "rtx" to "rtx_insn *".
2516 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2517 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2518 Likewise. Convert 3rd param from "int" to "location_t".
2519 (emit_label_after): Strengthen 1st param from "rtx" to
2521 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2523 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2524 from "rtx" to "rtx_insn *".
2526 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2528 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2529 bodies streamed in with -Q.
2530 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2531 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2532 * gimple-streamer-out.c: Include gimple-pretty-print.h
2533 (output_bb): Dump stmts streamed.
2534 * lto-section-out.c: Include print-tree.h
2535 (lto_begin_section): Dump sections created.
2536 (lto_output_decl_index): Dump decl encoded.
2537 * lto-streamer-out.c: Include print-tree.h
2538 (create_output_block): Dump output block created.
2539 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2540 (output_function): Dump function output.
2541 (output_constructor): Dump constructor streamed.
2542 (write_global_stream): Output indexes encoded.
2543 (produce_asm_for_decls): Dump streams encoded.
2544 * lto-streamer.c (streamer_dump_file): New global var.
2545 * lto-streamer.h (streamer_dump_file): Declare.
2546 * passes.c (ipa_write_summaries): Initialize streamer dump.
2547 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2550 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2553 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2554 offsets for register save directives. Emit a second batch of save
2555 directives, if need be, when the function accesses prior frames.
2557 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2559 * config/arc/fpu.md (fmasf4): Force operand to register.
2560 (fnmasf4): Likewise.
2562 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2564 * config/arc/arc-protos.h (arc_pad_return): Remove.
2565 * config/arc/arc.c (machine_function): Remove force_short_suffix
2567 (arc_print_operand): Adjust printing of '&'.
2568 (arc_verify_short): Remove conditional printing of short suffix.
2569 (arc_final_prescan_insn): Remove reference to size_reason.
2570 (pad_return): New function.
2571 (arc_reorg): Call pad_return.
2572 (arc_pad_return): Remove.
2573 (arc_init_machine_status): Remove reference to force_short_suffix.
2574 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2575 (attr length): When attribute iscompact is true force to 2
2576 regardless; in the case of maybe check if we want to force the
2577 instruction to have 4 bytes length.
2578 (nopv): Change it to generate 4 byte long nop as well.
2579 (blockage): New pattern.
2580 (simple_return): Remove call to arc_pad_return.
2581 (p_return_i): Likewise.
2583 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2585 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2587 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2589 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2592 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2594 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2595 for ARC700 and ARCv2.
2597 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2600 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2601 operands[2] instead of operands[1].
2604 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2606 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2607 case, check whether the outer register overlaps an unallocatable
2608 register, not just whether it fits the required class.
2610 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2612 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2614 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2615 (memrefs_conflict_p): Likewise.
2616 (init_alias_analysis): Likewise.
2617 * cfgexpand.c (expand_debug_expr): Likewise.
2618 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2619 * cse.c (fold_rtx): Likewise.
2620 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2621 * expr.c (emit_block_move_hints): Likewise.
2622 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2623 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2624 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2625 (emit_group_store): Likewise.
2626 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2627 to read the PRE/POST_MODIFY increment.
2628 * calls.c (store_one_arg): Use strip_offset.
2629 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2631 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2633 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2634 (simplify_binary_operation_1): Extend CONST_INT handling to
2636 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2637 than a HOST_WIDE_INT.
2638 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2640 (adjust_mems, add_stores): Update accodingly.
2641 (vt_canonicalize_addr): Track polynomial offsets.
2642 (emit_note_insn_var_location): Likewise.
2643 (vt_add_function_parameter): Likewise.
2644 (vt_initialize): Likewise.
2646 2018-06-12 Jeff Law <law@redhat.com>
2648 * config.gcc (alpha*-*-freebsd*): Remove.
2649 * config/alpha/freebsd.h: Remove.
2651 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2654 * spellcheck-tree.c (levenshtein_distance): Rename to...
2655 (get_edit_distance): ...this, and update for underlying renaming.
2656 * spellcheck-tree.h (levenshtein_distance): Rename to...
2657 (get_edit_distance): ...this.
2658 * spellcheck.c (levenshtein_distance): Rename to...
2659 (get_edit_distance): ...this. Convert from Levenshtein distance
2660 to Damerau-Levenshtein distance by supporting transpositions of
2661 adjacent characters. Rename "v1" to "v_next" and "v0" to
2663 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2664 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2665 update for underlying renaming.
2666 (selftest::levenshtein_distance_unit_test): Rename to...
2667 (selftest::test_get_edit_distance_unit): ...this, and update for
2668 underlying renaming.
2669 (selftest::test_find_closest_string): Add example from PR 69968
2670 where transposition helps
2671 (selftest::test_metric_conditions): Update for renaming.
2672 (selftest::test_metric_conditions): Likewise.
2673 (selftest::spellcheck_c_tests): Likewise.
2674 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2675 (get_edit_distance): ...this.
2676 (best_match::consider): Update for renaming.
2678 2018-06-12 Martin Sebor <msebor@redhat.com>
2680 PR tree-optimization/85259
2681 * builtins.c (compute_objsize): Handle constant offsets.
2682 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2683 true iff a warning has been issued.
2684 * gimple.h (gimple_nonartificial_location): New function.
2685 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2686 gimple_nonartificial_location and handle -Wno-system-headers.
2687 (handle_builtin_stxncpy): Same.
2689 2018-06-12 Martin Sebor <msebor@redhat.com>
2692 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2694 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2696 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2697 BUILTIN_VEC_XST entries for pointer to double and long long.
2699 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2702 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2704 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2707 2018-06-12 Martin Liska <mliska@suse.cz>
2709 * doc/options.texi: Document IntegerRange.
2711 2018-06-12 Martin Liska <mliska@suse.cz>
2713 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2714 * opt-functions.awk: Handle Deprecated flag.
2715 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2717 (read_cmdline_option): Report warning for a deprecated option.
2718 * opts.h (struct cl_option): Add new field cl_deprecated.
2719 (CL_ERR_DEPRECATED): New.
2721 2018-06-12 Martin Liska <mliska@suse.cz>
2723 * doc/options.texi: Document Deprecated option flag.
2725 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2727 * config/arc/arc-arch.h (arc_extras): New enum.
2728 (arc_cpu_t):Add field extra.
2729 (arc_cpu_types): Consider the extras.
2730 * config/arc/arc-cpus.def: Add extras info.
2731 * config/arc/arc-opts.h (processor_type): Consider extra field.
2732 * config/arc/arc.c (arc_override_options): Handle extra field.
2734 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2736 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2737 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2739 * config/arc/arc.c (arc_init): Update pic warning.
2740 (irq_range): Update irq range parsing warnings.
2741 (arc_override_options): Update various warning messages.
2742 (arc_handle_aux_attribute): Likewise.
2744 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2746 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2748 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2750 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2751 compiler messages directives.
2753 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2755 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2756 * config/mips/mips-tables.opt: Regenerate.
2757 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2759 * doc/invoke.texi: Document -march=i6500.
2761 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2763 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2764 (i6400_gpmul): Add cpu_unit.
2765 (i6400_gpdiv): Likewise.
2766 (i6400_msa_add_d): Update reservations.
2767 (i6400_msa_int_add) Likewise.
2768 (i6400_msa_short_logic3) Likewise.
2769 (i6400_msa_short_logic2) Likewise.
2770 (i6400_msa_short_logic) Likewise.
2771 (i6400_msa_move) Likewise.
2772 (i6400_msa_cmp) Likewise.
2773 (i6400_msa_short_float2) Likewise.
2774 (i6400_msa_div_d) Likewise.
2775 (i6400_msa_long_logic1) Likewise.
2776 (i6400_msa_long_logic2) Likewise.
2777 (i6400_msa_mult) Likewise.
2778 (i6400_msa_long_float2) Likewise.
2779 (i6400_msa_long_float4) Likewise.
2780 (i6400_msa_long_float5) Likewise.
2781 (i6400_msa_long_float8) Likewise.
2782 (i6400_fpu_fadd): Include frint type.
2783 (i6400_fpu_store): New define_insn_reservation.
2784 (i6400_fpu_load): Likewise.
2785 (i6400_fpu_move): Likewise.
2786 (i6400_fpu_fcmp): Likewise.
2787 (i6400_fpu_fmadd): Likewise.
2788 (i6400_int_mult): Include imul3nc type and update reservation.
2789 (i6400_int_div): Include idiv3 type and update reservation.
2790 (i6400_int_load): Update to check type not move_type.
2791 (i6400_int_store): Likewise.
2792 (i6400_int_prefetch): Set zero latency.
2794 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2796 * gcc.c: Document new %@{...} sequence.
2797 (LINK_COMMAND_SPEC): Use it for the -L switches.
2798 (cpp_unique_options): Use it for the -I switches.
2799 (at_file_argbuf): New global variable.
2800 (in_at_file): Likewise.
2801 (alloc_args): Create at_file_argbuf.
2802 (clear_args): Truncate at_file_argbuf.
2803 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2804 (open_at_file): New function.
2805 (close_at_file): Likewise.
2806 (create_at_file): Delete.
2807 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2810 (validate_switches_from_spec): Deal with %@{...} sequence.
2811 (validate_switches): Likewise.
2812 (driver::finalize): Call clear_args.
2814 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2816 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2818 2018-06-11 Martin Sebor <msebor@redhat.com>
2820 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2821 (Wno-class-memaccess): Add @opindex.
2822 (Wno-templates, Wno-multiple-inheritance): Same.
2823 (Wno-virtual-inheritance, Wno-namespaces): Same.
2824 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2825 (Wno-format-overflow, Wno-format-truncation): Same.
2826 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2827 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2828 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2829 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2830 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2831 (Wno-misspelled-isr): Same.
2833 2018-06-11 Martin Sebor <msebor@redhat.com>
2835 * PR tree-optimization/86083
2836 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2838 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2840 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2842 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2845 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2846 on the correct operand.
2847 (*movdi_internal64): Ditto.
2849 2018-06-11 Martin Liska <mliska@suse.cz>
2851 PR tree-optimization/86089
2852 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2854 2018-06-11 Julia Koval <julia.koval@intel.com>
2856 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2857 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2858 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2860 2018-06-11 Olivier Hainque <hainque@adacore.com>
2862 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2863 for Ada with strict dwarf2.
2865 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2868 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2871 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2873 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2875 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2877 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2880 2018-06-08 Martin Liska <mliska@suse.cz>
2882 * tree-cfg.h (debug_function): Fix argument type to match
2885 2018-06-08 Martin Liska <mliska@suse.cz>
2887 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2888 Remove usage of MPX-related (and removed) fields.
2889 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2891 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2893 * cfg.c (debug): Use TDF_NONE rather than 0.
2894 * cfghooks.c (debug): Likewise.
2895 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2896 (struct dump_option_value_info): Convert to...
2897 (struct kv_pair): ...this template type.
2898 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2900 (optinfo_verbosity_options): Likewise.
2901 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2903 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2904 than int for "optgroup_flags" param.
2905 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2907 (dump_dec): Likewise.
2908 (dump_finish): Use TDF_NONE rather than 0.
2909 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2910 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2911 than 0. Update for change to option_ptr.
2912 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2913 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2914 0. Update for changes to optinfo_verbosity_options and
2916 (opt_info_switch_p): Convert optgroup_flags from int to
2918 (dump_basic_block): Use dump_flags_t rather than int
2919 for "dump_kind" param.
2920 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2921 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2922 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2923 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2924 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2925 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2926 TDF_NONE): Convert from macros to...
2927 (enum dump_flag): ...this new enum.
2928 (dump_flags_t): Update to use enum.
2929 (operator|, operator&, operator~, operator|=, operator&=):
2930 Implement for dump_flags_t.
2931 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2932 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2933 Convert from macros to...
2934 (enum optgroup_flag): ...this new enum.
2935 (optgroup_flags_t): New typedef.
2936 (operator|, operator|=): Implement for optgroup_flags_t.
2937 (struct dump_file_info): Convert field "alt_flags" to
2938 dump_flags_t. Convert field "optgroup_flags" to
2940 (dump_basic_block): Use dump_flags_t rather than int for param.
2941 (dump_generic_expr_loc): Likewise.
2942 (dump_dec): Likewise.
2943 (dump_register): Convert param "optgroup_flags" to
2945 (opt_info_enable_passes): Likewise.
2946 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2948 * gimple-pretty-print.c (debug): Likewise.
2949 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2950 (merged_store_group::apply_stores): Likewise.
2951 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2952 * gimple.c (verify_gimple_pp): Likewise.
2953 * graphite-poly.c (print_pbb_body): Likewise.
2954 * passes.c (pass_manager::register_one_dump_file): Convert
2955 local "optgroup_flags" to optgroup_flags_t.
2956 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2958 (debug_body): Likewise.
2959 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2960 to optgroup_flags_t.
2961 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2963 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2964 (convert_mult_to_fma): Likewise.
2965 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2966 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2967 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2968 "dump_kind" to dump_flags_t.
2970 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2972 * config/rs6000/rs6000.c (min, max): Delete.
2974 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2976 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2977 -mabi=spe and -mabi=no-spe.
2979 2018-06-08 Martin Liska <mliska@suse.cz>
2981 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2982 where we expect an existing summary.
2984 2018-06-08 Martin Liska <mliska@suse.cz>
2986 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2987 * ipa-inline.h (estimate_edge_growth): Likewise.
2989 2018-06-08 Martin Liska <mliska@suse.cz>
2991 * cgraph.c (function_version_hasher::hash): Use
2992 cgraph_node::get_uid ().
2993 (function_version_hasher::equal):
2994 * cgraph.h (cgraph_node::get_uid): New method.
2995 * ipa-inline.c (update_caller_keys): Use
2996 cgraph_node::get_uid ().
2997 (update_callee_keys): Likewise.
2998 * ipa-utils.c (searchc): Likewise.
2999 (ipa_reduced_postorder): Likewise.
3000 * lto-cgraph.c (input_node): Likewise.
3001 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3002 * symbol-summary.h (symtab_insertion): Likewise.
3003 (symtab_removal): Likewise.
3004 (symtab_duplication): Likewise.
3005 * tree-pretty-print.c (dump_function_header): Likewise.
3006 * tree-sra.c (convert_callers_for_node): Likewise.
3008 2018-06-08 Martin Liska <mliska@suse.cz>
3010 * cgraph.c (symbol_table::create_edge): Always assign a new
3012 (symbol_table::free_edge): Do not recycle numbers.
3013 * cgraph.h (cgraph_edge::get): New method.
3014 * symbol-summary.h (symtab_removal): Use it.
3015 (symtab_duplication): Likewise.
3016 (call_summary::hashable_uid): Remove.
3018 2018-06-08 Martin Liska <mliska@suse.cz>
3020 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3021 (initialize_growth_caches): Remove.
3022 (free_growth_caches): Likewise.
3023 (do_estimate_edge_time): Use edge_growth_cache.
3024 (do_estimate_edge_size): Likewise.
3025 (do_estimate_edge_hints): Likewise.
3026 * ipa-inline.c (reset_edge_caches): Likewise.
3027 (recursive_inlining): Likewise.
3028 (inline_small_functions): Likewise.
3029 * ipa-inline.h (initialize_growth_caches): Remove.
3030 (estimate_edge_size): Likewise.
3031 (estimate_edge_time): Likewise.
3032 (estimate_edge_hints): Likewise.
3033 (reset_edge_growth_cache): Likewise.
3034 * symbol-summary.h (call_summary::remove): New method.
3036 2018-06-08 Martin Liska <mliska@suse.cz>
3038 * ipa-cp.c (class edge_clone_summary): New summary.
3039 (grow_edge_clone_vectors): Remove.
3040 (ipcp_edge_duplication_hook): Remove.
3041 (class edge_clone_summary_t): New call_summary class.
3042 (ipcp_edge_removal_hook): Remove.
3043 (edge_clone_summary_t::duplicate): New function.
3044 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3045 (create_specialized_node): Likewise.
3046 (ipcp_driver): Initialize edge_clone_summaries and do not
3049 2018-06-08 Martin Liska <mliska@suse.cz>
3051 * symbol-summary.h (get): New function.
3052 (call_summary::m_initialize_when_cloning): New class member.
3054 2018-06-08 Martin Liska <mliska@suse.cz>
3056 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3057 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3058 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3059 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3060 (remove_cgraph_node_from_order): Use the removed_nodes set.
3061 (do_per_function_toporder): Likwise.
3062 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3063 instead of summary_uid.
3064 (symtab_removal): Likewise.
3065 (symtab_duplication): Likewise.
3067 2018-06-08 Martin Liska <mliska@suse.cz>
3069 * ipa-cp.c (ipcp_store_bits_results): Use
3070 ipcp_transformation_sum.
3071 (ipcp_store_vr_results): Likewise.
3072 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3074 (ipcp_transformation_initialize): ... this.
3075 (ipa_set_node_agg_value_chain):
3076 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3077 (write_ipcp_transformation_info): Likewise.
3078 (read_ipcp_transformation_info): Likewise.
3079 (ipcp_update_bits): Likewise.
3080 (ipcp_update_vr): Likewise.
3081 (ipcp_transform_function): Likewise.
3082 * ipa-prop.h: Rename ipcp_transformation_summary to
3083 ipcp_transformation.
3084 (class ipcp_transformation_t): New function summary.
3085 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3086 (ipa_get_agg_replacements_for_node): Likewise.
3088 2018-06-08 Martin Liska <mliska@suse.cz>
3090 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3092 (class funct_state_summary_t): New function_summary class.
3093 (has_function_state): Remove.
3094 (get_function_state): Likewise.
3095 (set_function_state): Likewise.
3096 (add_new_function): Likewise.
3097 (funct_state_summary_t::insert): New function.
3098 (duplicate_node_data): Remove.
3099 (remove_node_data): Remove.
3100 (funct_state_summary_t::duplicate): New function.
3101 (register_hooks): Create new funct_state_summaries.
3102 (pure_const_generate_summary): Use it.
3103 (pure_const_write_summary): Likewise.
3104 (pure_const_read_summary): Likewise.
3105 (propagate_pure_const): Likewise.
3106 (propagate_nothrow): Likewise.
3107 (dump_malloc_lattice): Likewise.
3108 (propagate_malloc): Likewise.
3109 (execute): Do not register hooks, just remove summary
3111 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3114 2018-06-08 Martin Liska <mliska@suse.cz>
3116 * ipa-reference.c (remove_node_data): Remove.
3117 (duplicate_node_data): Likewise.
3118 (class ipa_ref_var_info_summary_t): New class.
3119 (class ipa_ref_opt_summary_t): Likewise.
3120 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3121 (get_reference_optimization_summary): Use
3122 ipa_ref_opt_sum_summaries.
3123 (set_reference_vars_info): Remove.
3124 (set_reference_optimization_summary): Likewise.
3125 (ipa_init): Create summaries.
3126 (init_function_info): Use function summary.
3127 (ipa_ref_opt_summary_t::duplicate): New function.
3128 (ipa_ref_opt_summary_t::remove): New function.
3129 (get_read_write_all_from_node): Fix GNU coding style.
3130 (propagate): Use function summary.
3131 (write_node_summary_p): Fix GNU coding style.
3132 (stream_out_bitmap): Likewise.
3133 (ipa_reference_read_optimization_summary): Use function summary.
3134 (ipa_reference_c_finalize): Do not release hooks.
3136 2018-06-08 Martin Liska <mliska@suse.cz>
3138 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3139 (analyze_function_body): Extract multiple calls of get_create.
3140 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3141 * ipa-inline.c (recursive_inlining): Use ::get method.
3142 * ipa-inline.h (estimate_edge_growth): Likewise.
3144 2018-06-08 Martin Liska <mliska@suse.cz>
3146 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3148 (hsa_function_summary::hsa_function_summary): Use the new enum
3150 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3151 * hsa-gen.c (hsa_get_host_function): Likewise.
3152 (get_brig_function_name): Likewise.
3153 * ipa-hsa.c (process_hsa_functions): Likewise.
3154 (ipa_hsa_write_summary): Likewise.
3155 * symbol-summary.h (symtab_duplication): Use ::get function/
3156 (get): New function.
3158 2018-06-08 Martin Liska <mliska@suse.cz>
3160 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3162 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3163 (hsa_register_kernel): Likewise.
3164 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3165 * hsa-gen.c (hsa_get_host_function): Likewise.
3166 (get_brig_function_name): Likewise.
3167 (generate_hsa): Likewise.
3168 (pass_gen_hsail::execute): Likewise.
3169 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3170 (devirtualization_time_bonus): Likewise.
3171 (ipcp_propagate_stage): Likewise.
3172 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3173 (edge_set_predicate): Likewise.
3174 (evaluate_conditions_for_known_args): Likewise.
3175 (evaluate_properties_for_edge): Likewise.
3176 (ipa_fn_summary::reset): Likewise.
3177 (ipa_fn_summary_t::duplicate): Likewise.
3178 (dump_ipa_call_summary): Likewise.
3179 (ipa_dump_fn_summary): Likewise.
3180 (analyze_function_body): Likewise.
3181 (compute_fn_summary): Likewise.
3182 (estimate_edge_devirt_benefit): Likewise.
3183 (estimate_edge_size_and_time): Likewise.
3184 (estimate_calls_size_and_time): Likewise.
3185 (estimate_node_size_and_time): Likewise.
3186 (inline_update_callee_summaries): Likewise.
3187 (remap_edge_change_prob): Likewise.
3188 (remap_edge_summaries): Likewise.
3189 (ipa_merge_fn_summary_after_inlining): Likewise.
3190 (ipa_update_overall_fn_summary): Likewise.
3191 (read_ipa_call_summary): Likewise.
3192 (inline_read_section): Likewise.
3193 (write_ipa_call_summary): Likewise.
3194 (ipa_fn_summary_write): Likewise.
3195 (ipa_free_fn_summary): Likewise.
3196 * ipa-hsa.c (process_hsa_functions): Likewise.
3197 (ipa_hsa_write_summary): Likewise.
3198 (ipa_hsa_read_section): Likewise.
3199 * ipa-icf.c (sem_function::merge): Likewise.
3200 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3201 (do_estimate_edge_time): Likewise.
3202 (estimate_size_after_inlining): Likewise.
3203 (estimate_growth): Likewise.
3204 (growth_likely_positive): Likewise.
3205 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3206 (inline_call): Likewise.
3207 * ipa-inline.c (caller_growth_limits): Likewise.
3208 (can_inline_edge_p): Likewise.
3209 (can_inline_edge_by_limits_p): Likewise.
3210 (compute_uninlined_call_time): Likewise.
3211 (compute_inlined_call_time): Likewise.
3212 (want_inline_small_function_p): Likewise.
3213 (edge_badness): Likewise.
3214 (update_caller_keys): Likewise.
3215 (update_callee_keys): Likewise.
3216 (recursive_inlining): Likewise.
3217 (inline_small_functions): Likewise.
3218 (inline_to_all_callers_1): Likewise.
3219 (dump_overall_stats): Likewise.
3220 (early_inline_small_functions): Likewise.
3221 (early_inliner): Likewise.
3222 * ipa-inline.h (estimate_edge_growth): Likewise.
3223 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3224 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3225 * ipa-prop.h (IPA_NODE_REF): Likewise.
3226 (IPA_EDGE_REF): Likewise.
3227 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3228 (propagate_malloc): Likewise.
3229 * ipa-split.c (execute_split_functions): Likewise.
3230 * symbol-summary.h: Rename get to get_create.
3232 (get_create): Likewise.
3233 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3235 2018-06-08 Martin Liska <mliska@suse.cz>
3237 * symbol-summary.h (release): Move definition out of class
3239 (symtab_removal): Likewise.
3240 (symtab_duplication): Likewise.
3242 2018-06-08 Martin Liska <mliska@suse.cz>
3244 * symbol-summary.h (function_summary): Move constructor
3245 implementation out of class declaration.
3246 (release): Likewise.
3247 (symtab_insertion): Likewise.
3248 (symtab_removal): Likewise.
3249 (symtab_duplication): Likewise.
3252 2018-06-08 Martin Liska <mliska@suse.cz>
3254 * Makefile.in: Remove support for MPX (macros, related functions,
3255 fields in cgraph_node, ...).
3256 * builtin-types.def (BT_BND): Likewise.
3257 (BT_FN_BND_CONST_PTR): Likewise.
3258 (BT_FN_CONST_PTR_BND): Likewise.
3259 (BT_FN_VOID_PTR_BND): Likewise.
3260 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3261 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3262 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3263 (expand_builtin_mempcpy_with_bounds): Likewise.
3264 (expand_builtin_memset_with_bounds): Likewise.
3265 (expand_builtin_memset_args): Likewise.
3266 (std_expand_builtin_va_start): Likewise.
3267 (expand_builtin): Likewise.
3268 (expand_builtin_with_bounds): Likewise.
3269 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3270 (DEF_LIB_BUILTIN_CHKP): Likewise.
3271 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3272 (DEF_CHKP_BUILTIN): Likewise.
3273 (BUILT_IN_MEMCPY): Likewise.
3274 (BUILT_IN_MEMMOVE): Likewise.
3275 (BUILT_IN_MEMPCPY): Likewise.
3276 (BUILT_IN_MEMSET): Likewise.
3277 (BUILT_IN_STPCPY): Likewise.
3278 (BUILT_IN_STRCAT): Likewise.
3279 (BUILT_IN_STRCHR): Likewise.
3280 (BUILT_IN_STRCPY): Likewise.
3281 (BUILT_IN_STRLEN): Likewise.
3282 (BUILT_IN_MEMCPY_CHK): Likewise.
3283 (BUILT_IN_MEMMOVE_CHK): Likewise.
3284 (BUILT_IN_MEMPCPY_CHK): Likewise.
3285 (BUILT_IN_MEMSET_CHK): Likewise.
3286 (BUILT_IN_STPCPY_CHK): Likewise.
3287 (BUILT_IN_STRCAT_CHK): Likewise.
3288 (BUILT_IN_STRCPY_CHK): Likewise.
3289 * calls.c (store_bounds): Likewise.
3290 (emit_call_1): Likewise.
3291 (special_function_p): Likewise.
3292 (maybe_warn_nonstring_arg): Likewise.
3293 (initialize_argument_information): Likewise.
3294 (finalize_must_preallocate): Likewise.
3295 (compute_argument_addresses): Likewise.
3296 (expand_call): Likewise.
3297 * cfgexpand.c (expand_call_stmt): Likewise.
3298 (expand_return): Likewise.
3299 (expand_gimple_stmt_1): Likewise.
3300 (pass_expand::execute): Likewise.
3301 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3302 (cgraph_node::remove): Likewise.
3303 (cgraph_node::dump): Likewise.
3304 (cgraph_node::verify_node): Likewise.
3305 * cgraph.h (chkp_function_instrumented_p): Likewise.
3306 (symtab_node::get_alias_target): Likewise.
3307 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3308 (cgraph_local_p): Likewise.
3309 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3310 (cgraph_edge::rebuild_references): Likewise.
3311 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3312 (walk_polymorphic_call_targets): Likewise.
3313 (cgraph_node::expand_thunk): Likewise.
3314 (symbol_table::output_weakrefs): Likewise.
3315 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3316 (ix86_handle_option): Likewise.
3317 * config/i386/constraints.md: Likewise.
3318 * config/i386/i386-builtin-types.def (BND): Likewise.
3322 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3323 (BDESC_FIRST): Likewise.
3325 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3326 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3327 * config/i386/i386.c (enum reg_class): Likewise.
3328 (ix86_target_string): Likewise.
3329 (ix86_option_override_internal): Likewise.
3330 (ix86_conditional_register_usage): Likewise.
3331 (ix86_valid_target_attribute_inner_p): Likewise.
3332 (ix86_set_indirect_branch_type): Likewise.
3333 (ix86_set_current_function): Likewise.
3334 (ix86_function_arg_regno_p): Likewise.
3335 (init_cumulative_args): Likewise.
3336 (ix86_function_arg_advance): Likewise.
3337 (ix86_function_arg): Likewise.
3338 (ix86_pass_by_reference): Likewise.
3339 (ix86_function_value_regno_p): Likewise.
3340 (ix86_function_value_1): Likewise.
3341 (ix86_function_value_bounds): Likewise.
3342 (ix86_return_in_memory): Likewise.
3343 (ix86_setup_incoming_vararg_bounds): Likewise.
3344 (ix86_va_start): Likewise.
3345 (indirect_thunk_need_prefix): Likewise.
3346 (print_reg): Likewise.
3347 (ix86_print_operand): Likewise.
3348 (ix86_expand_call): Likewise.
3349 (ix86_output_function_return): Likewise.
3350 (reg_encoded_number): Likewise.
3351 (BDESC_VERIFYS): Likewise.
3352 (ix86_init_mpx_builtins): Likewise.
3353 (ix86_init_builtins): Likewise.
3354 (ix86_emit_cmove): Likewise.
3355 (ix86_emit_move_max): Likewise.
3356 (ix86_expand_builtin): Likewise.
3357 (ix86_builtin_mpx_function): Likewise.
3358 (ix86_get_arg_address_for_bt): Likewise.
3359 (ix86_load_bounds): Likewise.
3360 (ix86_store_bounds): Likewise.
3361 (ix86_load_returned_bounds): Likewise.
3362 (ix86_store_returned_bounds): Likewise.
3363 (ix86_class_likely_spilled_p): Likewise.
3364 (ix86_hard_regno_mode_ok): Likewise.
3365 (x86_order_regs_for_local_alloc): Likewise.
3366 (ix86_mitigate_rop): Likewise.
3367 (ix86_bnd_prefixed_insn_p): Likewise.
3368 (ix86_mpx_bound_mode): Likewise.
3369 (ix86_make_bounds_constant): Likewise.
3370 (ix86_initialize_bounds): Likewise.
3371 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3372 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3373 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3374 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3375 (TARGET_CHKP_BOUND_MODE): Likewise.
3376 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3377 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3378 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3379 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3380 * config/i386/i386.h (TARGET_MPX): Likewise.
3381 (TARGET_MPX_P): Likewise.
3382 (VALID_BND_REG_MODE): Likewise.
3383 (FIRST_BND_REG): Likewise.
3384 (LAST_BND_REG): Likewise.
3385 (enum reg_class): Likewise.
3386 (BND_REG_P): Likewise.
3387 (BND_REGNO_P): Likewise.
3388 (BNDmode): Likewise.
3389 (ADJUST_INSN_LENGTH): Likewise.
3390 * config/i386/i386.md: Likewise.
3391 * config/i386/i386.opt: Likewise.
3392 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3393 (defined): Likewise.
3394 (LINK_MPX): Likewise.
3395 (MPX_SPEC): Likewise.
3396 (LIBMPX_SPEC): Likewise.
3397 (LIBMPXWRAPPERS_SPEC): Likewise.
3398 (CHKP_SPEC): Likewise.
3399 * config/i386/predicates.md: Likewise.
3400 * dbxout.c (dbxout_type): Likewise.
3401 * doc/extend.texi: Likewise.
3402 * doc/invoke.texi: Likewise.
3403 * doc/md.texi: Likewise.
3404 * doc/tm.texi: Likewise.
3405 * doc/tm.texi.in: Likewise.
3406 * dwarf2out.c (is_base_type): Likewise.
3407 (gen_formal_types_die): Likewise.
3408 (gen_subprogram_die): Likewise.
3409 (gen_type_die_with_usage): Likewise.
3410 (gen_decl_die): Likewise.
3411 (dwarf2out_late_global_decl): Likewise.
3412 * expr.c (expand_assignment): Likewise.
3413 (emit_storent_insn): Likewise.
3414 (store_expr_with_bounds): Likewise.
3415 (store_expr): Likewise.
3416 (expand_expr_real_1): Likewise.
3417 * expr.h (store_expr_with_bounds): Likewise.
3418 * function.c (use_register_for_decl): Likewise.
3419 (struct bounds_parm_data): Likewise.
3420 (assign_parms_augmented_arg_list): Likewise.
3421 (assign_parm_find_entry_rtl): Likewise.
3422 (assign_parm_is_stack_parm): Likewise.
3423 (assign_parm_load_bounds): Likewise.
3424 (assign_bounds): Likewise.
3425 (assign_parms): Likewise.
3426 (expand_function_start): Likewise.
3427 * gcc.c (CHKP_SPEC): Likewise.
3428 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3429 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3430 (wrestrict_dom_walker::check_call): Likewise.
3431 * gimple.c (gimple_build_call_from_tree): Likewise.
3432 * gimple.h (enum gf_mask): Likewise.
3433 (gimple_call_with_bounds_p): Likewise.
3434 (gimple_call_set_with_bounds): Likewise.
3435 * gimplify.c (gimplify_init_constructor): Likewise.
3436 * ipa-cp.c (initialize_node_lattices): Likewise.
3437 (propagate_constants_across_call): Likewise.
3438 (find_more_scalar_values_for_callers_subset): Likewise.
3439 * ipa-hsa.c (process_hsa_functions): Likewise.
3440 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3441 * ipa-icf.c (sem_function::merge): Likewise.
3442 * ipa-inline.c (early_inliner): Likewise.
3443 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3444 (warn_function_cold): Likewise.
3445 (propagate_pure_const): Likewise.
3446 * ipa-ref.h (enum GTY): Likewise.
3447 * ipa-split.c (find_retbnd): Likewise.
3448 (consider_split): Likewise.
3449 (split_function): Likewise.
3450 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3451 * ipa.c (walk_polymorphic_call_targets): Likewise.
3452 (symbol_table::remove_unreachable_nodes): Likewise.
3453 (process_references): Likewise.
3454 (cgraph_build_static_cdtor_1): Likewise.
3455 * lto-cgraph.c (lto_output_node): Likewise.
3456 (output_refs): Likewise.
3457 (compute_ltrans_boundary): Likewise.
3458 (input_overwrite_node): Likewise.
3459 (input_node): Likewise.
3460 (input_cgraph_1): Likewise.
3461 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3462 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3463 (class pass_chkp_instrumentation_passes): Likewise.
3464 (make_pass_chkp_instrumentation_passes): Likewise.
3465 * passes.def: Likewise.
3466 * rtl.h (struct GTY): Likewise.
3467 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3468 * stor-layout.c (layout_type): Likewise.
3469 * symtab.c: Likewise.
3470 * target.def: Likewise.
3471 * targhooks.c (default_chkp_bound_type): Likewise.
3472 (default_chkp_bound_mode): Likewise.
3473 (default_builtin_chkp_function): Likewise.
3474 (default_chkp_function_value_bounds): Likewise.
3475 (default_chkp_make_bounds_constant): Likewise.
3476 (default_chkp_initialize_bounds): Likewise.
3477 * targhooks.h (default_chkp_bound_type): Likewise.
3478 (default_chkp_bound_mode): Likewise.
3479 (default_builtin_chkp_function): Likewise.
3480 (default_chkp_function_value_bounds): Likewise.
3481 (default_chkp_make_bounds_constant): Likewise.
3482 (default_chkp_initialize_bounds): Likewise.
3483 * toplev.c (compile_file): Likewise.
3484 (process_options): Likewise.
3485 * tree-core.h (DEF_BUILTIN): Likewise.
3486 (DEF_BUILTIN_CHKP): Likewise.
3487 * tree-inline.c (declare_return_variable): Likewise.
3488 (remap_gimple_stmt): Likewise.
3489 (copy_bb): Likewise.
3490 (initialize_inlined_parameters): Likewise.
3491 (expand_call_inline): Likewise.
3492 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3493 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3494 (make_pass_ipa_chkp_produce_thunks): Likewise.
3495 (make_pass_chkp): Likewise.
3496 (make_pass_chkp_opt): Likewise.
3497 (make_pass_chkp_instrumentation_passes): Likewise.
3498 * tree-pretty-print.c (dump_generic_node): Likewise.
3499 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3500 * tree-ssa-dce.c (propagate_necessity): Likewise.
3501 (eliminate_unnecessary_stmts): Likewise.
3502 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3503 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3504 * tree-ssa-sccvn.h: Likewise.
3505 * tree-ssa-strlen.c (get_string_length): Likewise.
3506 (valid_builtin_call): Likewise.
3507 (adjust_last_stmt): Likewise.
3508 (handle_builtin_strchr): Likewise.
3509 (handle_builtin_strcpy): Likewise.
3510 (handle_builtin_stxncpy): Likewise.
3511 (handle_builtin_memcpy): Likewise.
3512 (handle_builtin_strcat): Likewise.
3513 (strlen_check_and_optimize_stmt): Likewise.
3514 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3515 * tree-streamer-in.c: Likewise.
3516 * tree-streamer.c (record_common_node): Likewise.
3517 * tree.c (tree_code_size): Likewise.
3518 (wide_int_to_tree_1): Likewise.
3519 (type_contains_placeholder_1): Likewise.
3520 (build_common_tree_nodes): Likewise.
3521 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3522 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3523 (POINTER_BOUNDS_P): Likewise.
3524 (BOUNDED_TYPE_P): Likewise.
3525 (BOUNDED_P): Likewise.
3526 (CALL_WITH_BOUNDS_P): Likewise.
3527 (pointer_bounds_type_node): Likewise.
3528 * value-prof.c (gimple_ic): Likewise.
3529 * var-tracking.c (vt_add_function_parameters): Likewise.
3530 * varasm.c (make_decl_rtl): Likewise.
3531 (assemble_start_function): Likewise.
3532 (output_constant): Likewise.
3533 (maybe_assemble_visibility): Likewise.
3534 * varpool.c (ctor_for_folding): Likewise.
3535 * chkp-builtins.def: Remove.
3536 * ipa-chkp.c: Remove.
3537 * ipa-chkp.h: Remove.
3538 * rtl-chkp.c: Remove.
3539 * rtl-chkp.h: Remove.
3540 * tree-chkp-opt.c: Remove.
3541 * tree-chkp.c: Remove.
3542 * tree-chkp.h: Remove.
3544 2018-06-07 Carl Love <cel@us.ibm.com>
3546 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3547 vextract_fp_from_shortl): Add BE support.
3549 2018-06-07 Paul Koning <ni1d@arrl.net>
3551 * compare-elim.c (try_merge_compare): Don't merge compare if
3552 address contains a side effect.
3553 (try_eliminate_compare): Likewise.
3555 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3557 * config.gcc: Support "tremont".
3558 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3559 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3561 * config/i386/i386.c (m_TREMONT): Define.
3562 (processor_target_table): Add "tremont".
3563 (PTA_TREMONT): Define.
3564 (ix86_lea_outperforms): Add TARGET_TREMONT.
3565 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3566 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3567 and M_INTEL_GOLDMONT_PLUS.
3568 (fold_builtin_cpu): Add "tremont".
3569 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3570 (ix86_option_override_internal): Add "tremont".
3571 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3572 (processor_type): Add PROCESSOR_TREMONT.
3573 * config/i386/x86-tune.def: Add m_TREMONT.
3574 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3576 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3578 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3579 symbol defined for msp430i* devices to be lower case.
3581 gcc/testsuite/gcc.target/msp430/
3582 * msp430i-device-symbol.c: New test.
3583 * msp430f-device-symbol.c: New test.
3584 * msp430.h: New test header file.
3586 2018-06-07 Richard Biener <rguenther@suse.de>
3588 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3589 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3590 Properly wrap signed arithmetic if overflow wraps.
3592 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3594 PR tree-optimization/69615
3595 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3596 of a cast from a same precision integral SSA_NAME in a bb dominated
3597 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3598 cast to utype if rhs2 has already a compatible type.
3600 2018-06-07 Richard Biener <rguenther@suse.de>
3602 PR tree-optimization/85935
3603 * graphite-scop-detection.c (find_params_in_bb): Analyze
3604 condition operands with respect to the correct loop. Assert
3605 the analysis doesn't fail.
3607 2018-06-04 Carl Love <cel@us.ibm.com>
3609 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3610 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3611 as it is slightly cheaper.
3612 (first_match_or_eos_index_<mode>):
3613 Calculate index using natural element order.
3614 (first_match_index_<mode>):
3615 Calculate index using natural element order.
3616 (first_match_or_eos_index_<mode>):
3617 Calculate index using natural order.
3618 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3619 for BE and LE modes.
3620 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3621 P9V_BUILTIN_VCLZLSBB_V16QI.
3622 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3625 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3627 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3628 indentation and line wrap for many prototypes. Add missing
3629 @smallexample directives around block of prototypes for vec_xl and
3632 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3634 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3635 track if we pass or return IEEE 128-bit floating point.
3636 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3637 C++ mangling that is compatible with GCC 8.1.
3638 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3639 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3640 floating point types.
3641 (rs6000_function_arg_advance_1): Likewise.
3642 (rs6000_mangle_type): Optionally generate mangled names that match
3643 what GCC 8.1 generated for IEEE 128-bit floating point types.
3644 (rs6000_globalize_decl_name): If we have an external function that
3645 passes or returns IEEE 128-bit types, generate a weak reference
3646 from the mangled name used in GCC 8.1 to the current mangled
3648 (rs6000_init_builtins): Make __ibm128 use the long double type if
3649 long double is IBM extended double. Make __float128 use the long
3650 double type if long double is IEEE 128-bit.
3653 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3654 macro for __ibm128 built-in functions.
3655 (PACK_IF): Add __ibm128 pack/unpack functions.
3656 (UNPACK_IF): Likewise.
3657 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3658 enable long double built-in functions if long double is IEEE
3659 128-bit floating point.
3660 (rs6000_invalid_builtin): Update long double built-in function
3662 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3663 functions, adjust the built-in function to use the long double
3664 built-in function if __ibm128 and long double are the same type.
3665 * doc/extend.texi (PowerPC builtins): Update documention for
3666 __builtin_{,un}pack_longdouble. Add documentation for
3667 __builtin_{,un}pack_ibm128.
3669 2018-06-06 Jim Wilson <jimw@sifive.com>
3671 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3672 (struct machine_function): New field interrupt_mode.
3673 (riscv_handle_type_attribute): New function. Add forward declaration.
3674 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3675 (riscv_expand_epilogue): Check interrupt_mode field.
3676 (riscv_set_current_function): Check interrupt attribute args and
3677 set interrupt_mode field.
3678 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3679 (riscv_sret, riscv_uret): New.
3680 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3681 new arguments to interrupt attribute.
3683 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3686 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3687 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3689 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3691 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3692 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3693 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3694 several redundant entries.
3696 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3698 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3699 type from "rtx" to "rtx_insn *".
3700 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3701 for local "call_insn", removing cast.
3702 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3704 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3706 PR tree-optimization/86066
3707 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3708 for BIT_INSERT_EXPR stores.
3710 2018-06-06 Richard Biener <rguenther@suse.de>
3712 PR tree-optimization/86062
3713 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3714 component refs ontop
3715 of to be offsetted base.
3717 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3719 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3720 to be static and remove check on interrupt attribute name.
3722 gcc/testsuite/gcc.target/msp430/
3723 * function-attributes-4.c: New test.
3724 * static-interrupts.c: New test.
3726 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3728 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3729 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3731 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3734 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3736 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3737 Remove second argument, change how error is called.
3738 (aarch64_layout_arg): Remove second argument from
3739 aarch64_err_no_fpadvsimd call.
3740 (aarch64_init_cumulative_args): Ditto.
3741 (aarch64_gimplify_va_arg_expr): Ditto.
3742 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3744 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3746 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3747 (*simple_return_indirect_internal<mode>): Rename from
3748 simple_return_indirect_internal. Use W mode iterator.
3749 (rstorssp): New expander.
3750 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3751 (clrssbsy): New expander.
3752 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3754 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3756 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3758 (cmse_check_pointed_object): Likewise.
3760 2018-06-05 Martin Liska <mliska@suse.cz>
3762 PR gcov-profile/47618
3763 * doc/invoke.texi: Document how -fprofile-dir format
3766 2018-06-05 Richard Biener <rguenther@suse.de>
3768 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3769 removal pretend DOM info isn't available so we do not update
3770 it and only remove edges, not dominated blocks. Actually free
3771 DOM info in case we removed something. Remove unreachable blocks.
3772 (mfb_keep_latches): Work with either DOM info or marked backedges.
3773 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3774 first. Mark backedges if DOM info isn't available.
3775 (Re-)compute DOM info after cleanup_control_flow_pre.
3777 2018-06-05 Richard Biener <rguenther@suse.de>
3779 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3780 (locus_discrim_hasher::hash): Adjust.
3781 (locus_discrim_hasher::equal): Likewise.
3782 (next_discriminator_for_locus): Work on line directly.
3783 (same_line_p): Pass in expanded locus1 as well.
3784 (assign_discriminators): Avoid redundant location expansions.
3786 2018-06-05 Richard Biener <rguenther@suse.de>
3788 PR tree-optimization/86046
3789 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3790 if required after clearing TREE_ADDRESSABLE.
3792 2018-06-05 Richard Biener <rguenther@suse.de>
3794 PR tree-optimization/86047
3795 * tree-ssa-loop.c (for_each_index): Glob handling of all
3796 decls and constants and really handle all of them.
3798 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3801 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3802 qualifier_void_pointer and qualifier_const_void_pointer.
3803 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3804 (arm_init_builtins): Handle the above.
3805 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3806 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3807 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3810 2018-06-05 Martin Liska <mliska@suse.cz>
3812 * auto-profile.c (read_autofdo_file): Do not use
3813 gcov_ctr_summary struct.
3814 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3815 * coverage.c (struct counts_entry): Likewise.
3816 (read_counts_file): Read just single summary entry.
3817 (get_coverage_counts): Use gcov_summary struct.
3818 * coverage.h (get_coverage_counts): Likewise.
3819 * gcov-dump.c (dump_working_sets): Likewise.
3820 (tag_summary): Dump just single summary.
3821 * gcov-io.c (gcov_write_summary): Write just histogram
3823 (gcov_read_summary): Read just single summary.
3824 (compute_working_sets): Use gcov_summary struct.
3825 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3826 of GCOV_COUNTERS_SUMMABLE.
3827 (GCOV_COUNTERS_SUMMABLE): Remove.
3828 (GCOV_FIRST_VALUE_COUNTER): Replace with
3829 GCOV_COUNTER_V_INTERVAL.
3830 (struct gcov_ctr_summary): Remove.
3831 (struct gcov_summary): Directly use fields of former
3833 (compute_working_sets): Use gcov_summary struct.
3834 * gcov.c (read_count_file): Do not use ctrs fields.
3835 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3837 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3839 * profile.c: Likewise.
3840 * profile.h: Likewise.
3842 2018-06-05 Martin Liska <mliska@suse.cz>
3844 PR gcov-profile/84846
3845 * gcov.c (output_lines): Print working directory only
3846 in intermediate format.
3848 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3850 * config/s390/s390-builtin-types.def: Add void function type.
3851 * config/s390/s390-builtins.def: Use the function type for the
3854 2018-06-04 Jim Wilson <jimw@sifive.com>
3856 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3858 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3859 and maybe_eh_return. Change regno to unsigned int. Use new args to
3860 handle EH_RETURN_DATA_REGNO registers properly.
3861 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3862 (riscv_expand_epilogue): Update comment. Change argument name and
3863 type. Update code to use new name and type. Pass new args to
3864 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3866 * config/riscv/riscv.md (NORMAL_RETURN): New.
3867 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3868 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3869 (eh_return): Call gen_eh_return_internal and emit barrier.
3870 (eh_return_internal): Call riscv_expand_epilogue.
3872 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3874 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3875 bit_insertion field and declare can_be_merged_into method.
3876 (merged_store_group::can_be_merged_into): New method.
3877 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3878 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3879 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3881 2018-06-04 Richard Biener <rguenther@suse.de>
3883 PR tree-optimization/85955
3884 * builtins.c (fold_builtin_sincos): Convert pointers to
3885 destination to appropriate type before dereferencing.
3887 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3889 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3891 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3893 * expr.c (expand_expr_real_1): Force the operand into memory if
3894 its TYPE_MODE is BLKmode and if there is no integer mode for
3895 the number of bits being extracted.
3897 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3901 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3902 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3904 2018-06-04 Richard Biener <rguenther@suse.de>
3906 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3907 (cleanup_tree_cfg_noloop): ... single caller. Do
3908 start_recording_case_labels later.
3910 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3912 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3913 to _IMMINTRIN_H_INCLUDED.
3914 * config/i386/pconfigintrin.h: Ditto.
3915 * config/i386/waitpkgintrin.h: Ditto.
3916 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3917 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3918 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3919 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3920 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3921 waitpkgintrin.h and cldemoteintrin.h.
3923 2018-06-04 Richard Biener <rguenther@suse.de>
3925 PR tree-optimization/86038
3926 * tracer.c (find_best_successor): Check probability for
3927 being initialized, bail out if not.
3929 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3932 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3933 of bits to ignore when comparing architectures.
3935 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3937 PR tree-optimization/69615
3938 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3939 maximum or minimum of the type, try to merge it also as if
3940 range1 is + [-, x - 1] or + [x + 1, -].
3943 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3945 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3947 PR tree-optimization/86034
3948 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3949 the unsigned bitfield type in a bit insertion sequence if it does not
3950 have a larger precision than the bitfield size.
3951 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3953 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3955 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3957 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3959 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3960 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3961 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3962 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3964 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3966 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3967 Disable -fdelete-null-pointer-checks for ELF toolchain.
3969 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3970 Kito Cheng <kito.cheng@gmail.com>
3972 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3973 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3974 (nds32*-*-*): Add glibc and uclibc conditions.
3975 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3976 (TARGET_EXCEPT_UNWIND_INFO): Define.
3977 * config/nds32/elf.h: New file.
3978 * config/nds32/linux.h: New file.
3979 * config/nds32/nds32-elf.opt: New file.
3980 * config/nds32/nds32-linux.opt: New file.
3981 * config/nds32/nds32-fp-as-gp.c
3982 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3983 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3985 (nds32_asm_file_end): Ditto.
3986 (nds32_print_operand): Ditto.
3987 (nds32_insert_attributes): Ditto.
3988 (nds32_init_libfuncs): New function.
3989 (TARGET_HAVE_TLS): Define.
3990 (TARGET_INIT_LIBFUNCS): Define.
3991 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3993 (TARGET_ELF): Apply different mcmodel setting.
3994 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3995 been migrated into elf.h and linux.h files.
3996 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3997 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3998 (mcmodel): The content has been migrated into nds32-elf.opt and
3999 nds32-linux.opt files.
4000 * config/nds32/t-elf: New file.
4001 * config/nds32/t-linux: New file.
4003 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4004 Shiva Chen <shiva0217@gmail.com>
4006 * config/nds32/constants.md (unspec_volatile_element): Add
4007 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4008 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4010 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4011 (make_pass_nds32_fp_as_gp): Declare.
4012 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4014 (nds32_asm_function_end_prologue): Remove unused asm output.
4015 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4016 (nds32_asm_file_start): Output necessary fp_as_gp information.
4017 (nds32_option_override): Adjust register usage.
4018 (nds32_expand_prologue): Consider fp_as_gp situation.
4019 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4020 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4023 (simple_return): Ditto.
4024 (omit_fp_begin): Output special directive for fp_as_gp.
4025 (omit_fp_end): Output special directive for fp_as_gp.
4026 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4027 mforbid-fp-as-gp): New options.
4029 2018-06-01 Mark Wielaard <mark@klomp.org>
4031 * dwarf2out.c (dwarf2out_finish): Remove generation of
4032 DW_AT_loclists_base.
4034 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4036 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4037 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4038 (struct merged_store_group): Add bit_insertion field.
4039 (dump_char_array): Use standard hexadecimal format.
4040 (merged_store_group::merged_store_group): Set bit_insertion to false.
4041 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4042 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4043 also print the mask in the dump file.
4044 (pass_store_merging::gate): Minor tweak.
4045 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4046 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4047 stores with INTEGER_CST stores.
4048 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4049 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4050 and use it throughout. Generate bit insertion sequences if need be.
4051 (pass_store_merging::process_store): Remove redundant condition.
4052 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4054 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4056 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4057 the 128-bit floating point types. Fix function comment.
4059 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4061 * config/aarch64/aarch64-simd.md
4062 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4064 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4067 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4069 PR tree-optimization/85989
4070 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4072 (backprop::intersect_uses): Check it when deciding whether this
4073 is a backedge reference.
4074 (backprop::process_block): Add each phi to m_visited_phis
4075 after visiting it, then clear it at the end.
4077 2018-06-01 Richard Biener <rguenther@suse.de>
4079 * tree-vectorizer.h (vect_dr_stmt): New function.
4080 (vect_get_load_cost): Adjust.
4081 (vect_get_store_cost): Likewise.
4082 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4083 Use vect_dr_stmt instead of DR_SMTT.
4084 (vect_record_base_alignments): Likewise.
4085 (vect_calculate_target_alignment): Likewise.
4086 (vect_compute_data_ref_alignment): Likewise and make static.
4087 (vect_update_misalignment_for_peel): Likewise.
4088 (vect_verify_datarefs_alignment): Likewise.
4089 (vector_alignment_reachable_p): Likewise.
4090 (vect_get_data_access_cost): Likewise. Pass down
4091 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4092 (vect_get_peeling_costs_all_drs): Likewise.
4093 (vect_peeling_hash_get_lowest_cost): Likewise.
4094 (vect_enhance_data_refs_alignment): Likewise.
4095 (vect_find_same_alignment_drs): Likewise.
4096 (vect_analyze_data_refs_alignment): Likewise.
4097 (vect_analyze_group_access_1): Likewise.
4098 (vect_analyze_group_access): Likewise.
4099 (vect_analyze_data_ref_access): Likewise.
4100 (vect_analyze_data_ref_accesses): Likewise.
4101 (vect_vfa_segment_size): Likewise.
4102 (vect_small_gap_p): Likewise.
4103 (vectorizable_with_step_bound_p): Likewise.
4104 (vect_prune_runtime_alias_test_list): Likewise.
4105 (vect_analyze_data_refs): Likewise.
4106 (vect_supportable_dr_alignment): Likewise.
4107 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4108 (vect_gen_prolog_loop_niters): Likewise.
4109 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4110 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4112 (vect_recog_mask_conversion_pattern): Likewise.
4113 (vect_try_gather_scatter_pattern): Likewise.
4114 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4115 to vect_get_store_cost.
4116 (vect_get_store_cost): Get stmt_info instead of DR.
4117 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4118 (vect_get_load_cost): Get stmt_info instead of DR.
4120 2018-06-01 Richard Biener <rguenther@suse.de>
4123 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4124 vars wrapped in MEM_REFs.
4126 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4128 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4129 Fix subreg tests so that we only return a choice between
4130 GENERAL_REGS and FP_REGS if the original classes included both.
4132 2018-06-01 Richard Biener <rguenther@suse.de>
4135 * tree-ssa-structalias.c (get_function_part_constraint):
4136 Handle NULL fi->decl.
4137 (find_func_aliases_for_call): Properly handle indirect
4138 fi from direct call.
4139 (find_func_clobbers): Likewise.
4140 (ipa_pta_execute): Likewise.
4141 (create_variable_info_for): For functions that are ifunc_resolver
4142 resolve to a varinfo that contains the result of the resolver call.
4143 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4146 2018-05-31 Michael Collison <michael.collison@arm.com>
4148 * config/aarch64/aarch64.md:
4149 (*fix_to_zero_extenddfdi2): New pattern.
4150 * gcc.target/aarch64/fix_extend1.c: New testcase.
4152 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4156 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4157 and BUILT_IN_STRNCMP_EQ.
4158 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4159 BUILT_IN_STRNCMP_EQ.
4160 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4161 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4162 (gimple_fold_builtin): Likewise.
4163 * tree-ssa-strlen.c (compute_string_length): New function.
4164 (determine_min_obsize): New function.
4165 (handle_builtin_string_cmp): New function to handle calls to
4166 string compare functions.
4167 (strlen_optimize_stmt): Add handling to builtin string compare
4169 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4170 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4171 * tree.c (build_common_builtin_nodes): Add new defines of
4172 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4174 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4177 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4178 functions with naked attribute.
4180 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4182 * config/i386/sse.md (avx_vec_concat<mode>):
4183 Substitute concat_tg_mode mode attribute with xtg_mode.
4184 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4185 (concat_tg_mode): Remove mode attribute.
4187 2018-05-31 Martin Sebor <msebor@redhat.com>
4190 * calls.c (alloc_max_size): Correct a logic error/typo.
4191 Treat excessive arguments as infinite. Warn for invalid arguments.
4192 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4194 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4197 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4198 and movx for Haswell.
4200 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4201 Cesar Philippidis <cesar@codesourcery.com>
4204 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4205 when emitting error on private/firstprivate reductions.
4206 * omp-low.c (lower_omp_target): Avoid reference-type processing
4207 on pointers for firstprivate clause.
4209 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4211 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4214 * config/aarch64/aarch64-simd.md
4215 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4216 (aarch64_ld1_x3_<mode>): Likewise
4217 (aarch64_st1x2<VALLDIF:mode>): Likewise
4218 (aarch64_st1_x2_<mode>): Likewise
4219 (aarch64_st1x3<VALLDIF:mode>): Likewise
4220 (aarch64_st1_x3_<mode>): Likewise
4221 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4222 (vld1_s8_x3): Likewise.
4223 (vld1_u16_x3): Likewise.
4224 (vld1_s16_x3): Likewise.
4225 (vld1_u32_x3): Likewise.
4226 (vld1_s32_x3): Likewise.
4227 (vld1_u64_x3): Likewise.
4228 (vld1_s64_x3): Likewise.
4229 (vld1_f16_x3): Likewise.
4230 (vld1_f32_x3): Likewise.
4231 (vld1_f64_x3): Likewise.
4232 (vld1_p8_x3): Likewise.
4233 (vld1_p16_x3): Likewise.
4234 (vld1_p64_x3): Likewise.
4235 (vld1q_u8_x3): Likewise.
4236 (vld1q_s8_x3): Likewise.
4237 (vld1q_u16_x3): Likewise.
4238 (vld1q_s16_x3): Likewise.
4239 (vld1q_u32_x3): Likewise.
4240 (vld1q_s32_x3): Likewise.
4241 (vld1q_u64_x3): Likewise.
4242 (vld1q_s64_x3): Likewise.
4243 (vld1q_f16_x3): Likewise.
4244 (vld1q_f32_x3): Likewise.
4245 (vld1q_f64_x3): Likewise.
4246 (vld1q_p8_x3): Likewise.
4247 (vld1q_p16_x3): Likewise.
4248 (vld1q_p64_x3): Likewise.
4249 (vst1_s64_x2): Likewise.
4250 (vst1_u64_x2): Likewise.
4251 (vst1_f64_x2): Likewise.
4252 (vst1_s8_x2): Likewise.
4253 (vst1_p8_x2): Likewise.
4254 (vst1_s16_x2): Likewise.
4255 (vst1_p16_x2): Likewise.
4256 (vst1_s32_x2): Likewise.
4257 (vst1_u8_x2): Likewise.
4258 (vst1_u16_x2): Likewise.
4259 (vst1_u32_x2): Likewise.
4260 (vst1_f16_x2): Likewise.
4261 (vst1_f32_x2): Likewise.
4262 (vst1_p64_x2): Likewise.
4263 (vst1q_s8_x2): Likewise.
4264 (vst1q_p8_x2): Likewise.
4265 (vst1q_s16_x2): Likewise.
4266 (vst1q_p16_x2): Likewise.
4267 (vst1q_s32_x2): Likewise.
4268 (vst1q_s64_x2): Likewise.
4269 (vst1q_u8_x2): Likewise.
4270 (vst1q_u16_x2): Likewise.
4271 (vst1q_u32_x2): Likewise.
4272 (vst1q_u64_x2): Likewise.
4273 (vst1q_f16_x2): Likewise.
4274 (vst1q_f32_x2): Likewise.
4275 (vst1q_f64_x2): Likewise.
4276 (vst1q_p64_x2): Likewise.
4277 (vst1_s64_x3): Likewise.
4278 (vst1_u64_x3): Likewise.
4279 (vst1_f64_x3): Likewise.
4280 (vst1_s8_x3): Likewise.
4281 (vst1_p8_x3): Likewise.
4282 (vst1_s16_x3): Likewise.
4283 (vst1_p16_x3): Likewise.
4284 (vst1_s32_x3): Likewise.
4285 (vst1_u8_x3): Likewise.
4286 (vst1_u16_x3): Likewise.
4287 (vst1_u32_x3): Likewise.
4288 (vst1_f16_x3): Likewise.
4289 (vst1_f32_x3): Likewise.
4290 (vst1_p64_x3): Likewise.
4291 (vst1q_s8_x3): Likewise.
4292 (vst1q_p8_x3): Likewise.
4293 (vst1q_s16_x3): Likewise.
4294 (vst1q_p16_x3): Likewise.
4295 (vst1q_s32_x3): Likewise.
4296 (vst1q_s64_x3): Likewise.
4297 (vst1q_u8_x3): Likewise.
4298 (vst1q_u16_x3): Likewise.
4299 (vst1q_u32_x3): Likewise.
4300 (vst1q_u64_x3): Likewise.
4301 (vst1q_f16_x3): Likewise.
4302 (vst1q_f32_x3): Likewise.
4303 (vst1q_f64_x3): Likewise.
4304 (vst1q_p64_x3): Likewise.
4306 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4308 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4309 user_label_prefix to name.
4311 * tree-core.h: Update comment about the format of NAME string
4312 passed to handler in attribute_spec.
4314 * config/msp430/msp430.md: Remove erroneous subreg expression from
4315 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4316 zero_extend{q,h}isi2.
4318 2018-05-30 Borislav Petkov <bp@suse.de>
4320 * doc/extend.texi: Document some architecture specific
4321 constraints and sort entries.
4323 2018-05-30 Martin Sebor <msebor@redhat.com>
4326 * builtins.c (expand_builtin_stpcpy_1): New function.
4327 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4328 only if the former succeeds.
4330 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4332 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4335 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4337 * doc/invoke.texi (-flinker-output): Document
4339 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4341 * passes.c (ipa_write_summaries): Only modify statements if body
4343 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4344 incrementally linking.
4345 (ipa_passes): Likewise.
4346 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4347 pass down resolution info.
4348 * common.opt (flag_incremental_link): Update info.
4349 * gcc.c (plugin specs): Turn flinker-output=* to
4350 -plugin-opt=-linker-output-known
4351 * toplev.c (compile_file): Also cut compilation when doing incremental
4353 * flag-types. (enum lto_partition_model): Add
4354 LTO_LINKER_OUTPUT_NOLTOREL.
4355 (invoke.texi): Add -flinker-output docs.
4356 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4357 link same way as WPA; do not stream in dead initializers.
4359 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4360 darf2out_register_external_decl): Support incremental link.
4362 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4364 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4366 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4368 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4369 it down to simple_object_copy_lto_debug_sections.
4370 (run_gcc): Determine incremental LTO link time and configure
4371 lto1 into non-wpa mode, disable renaming of debug sections.
4373 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4375 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4376 descriptions of various incorrectly documented functions.
4378 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4381 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4382 address check not strict.
4384 2018-05-30 Richard Biener <rguenther@suse.de>
4386 PR tree-optimization/85964
4387 * tracer.c (better_p): Drop initialized count check, we only
4388 call the function with initialized counts now.
4389 (find_best_successor): Do find a best edge if one
4390 has uninitialized count.
4391 (find_best_predecessor): Likewise. Do BB frequency check only
4392 if count is initialized.
4394 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4396 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4397 (aarch64_ldrstr_offset_compare): New.
4398 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4399 load/store orderings.
4400 (aarch64_gen_adjusted_ldpstp): Likewise.
4402 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4404 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4405 Check for subset of GENERAL_REGS and FP_REGS.
4406 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4409 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4411 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4412 and wi::to_poly_offset. Add the current offset and then check
4413 whether the sum fits, rather than using an unchecked addition of
4414 a checked term. Check for a shwi rather than a uhwi.
4415 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4416 (store_constructor): Use poly_int_tree_p.
4417 (expand_expr_real_1): Likewise.
4418 * function.c (assign_temp): Likewise.
4419 * fold-const.c (const_binop): Use poly_int_tree_p and
4421 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4423 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4424 to_poly_offset for MEM offsets.
4425 * ipa-icf.c (sem_variable::equals): Likewise.
4426 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4427 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4428 wi::to_poly_offset for BIT_FIELD_REF offsets.
4429 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4431 * var-tracking.c (emit_note_insn_var_location): Use
4432 tree_to_poly_uint64.
4434 2018-05-29 Jim Wilson <jimw@sifive.com>
4436 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4438 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4441 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4442 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4444 (sse4_1_round<mode>2): Use nonimmediate_operand
4445 for operand 1 predicate.
4447 2018-05-29 Martin Sebor <msebor@redhat.com>
4448 Richard Biener <rguenther@suse.de>
4451 * calls.c (get_size_range): Call determine_value_range instead
4452 of get_value_range..
4453 * tree-vrp.h (determine_value_range): Declared new function.
4454 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4456 2018-05-29 Richard Biener <rguenther@suse.de>
4458 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4459 sure to use non-pattern stmts for get_earlier_stmt arguments.
4460 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4461 called on pattern stmts.
4462 (get_later_stmt): Likewise.
4464 2018-05-29 Martin Liska <mliska@suse.cz>
4466 PR gcov-profile/85759
4467 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4470 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4472 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4474 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4478 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4479 VEC_PACK_FLOAT_EXPR): New tree codes.
4480 * tree-pretty-print.c (op_code_prio): Handle
4481 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4482 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4483 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4484 * tree-inline.c (estimate_operator_cost): Likewise.
4485 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4486 * fold-const.c (const_binop): Likewise.
4487 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4488 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4489 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4490 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4491 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4492 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4493 * expr.c (expand_expr_real_2): Likewise.
4494 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4495 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4496 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4498 * optabs.c (expand_widen_pattern_expr): For
4499 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4500 sign from result type rather than operand's type.
4501 (expand_binop_directly): For vec_packu_float_optab and
4502 vec_packs_float_optab allow result type to be different from operand's
4504 * optabs-tree.c (optab_for_tree_code): Handle
4505 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4506 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4507 * tree-vect-generic.c (expand_vector_operations_1): Handle
4508 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4509 VEC_PACK_FLOAT_EXPR.
4510 * tree-vect-stmts.c (supportable_widening_operation): Handle
4512 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4513 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4514 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4515 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4516 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4518 (vec_pack<floatprefix>_float_<mode>): New expander.
4519 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4521 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4522 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4523 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4524 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4525 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4527 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4528 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4529 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4530 VEC_PACK_FLOAT_EXPR): Document.
4532 2018-05-29 Richard Biener <rguenther@suse.de>
4534 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4536 (stmt_vec_info_vec): Make pointer.
4537 (init_stmt_vec_info_vec): Remove.
4538 (free_stmt_vec_info_vec): Likewise.
4539 (set_stmt_vec_info_vec): New function.
4540 (free_stmt_vec_infos): Likewise.
4541 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4542 (set_vinfo_for_stmt): Likewise.
4543 (get_earlier_stmt): Likewise.
4544 (get_later_stmt): Likewise.
4545 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4546 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4547 (vec_info::~vec_info): Free stmt_vec_infos.
4548 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4549 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4550 (pass_slp_vectorize::execute): Likewise.
4551 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4552 (free_stmt_vec_info_vec): Likewise.
4553 (set_stmt_vec_info_vec): New function.
4554 (free_stmt_vec_infos): Likewise.
4555 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4556 the global stmt_vec_info_vec.
4557 * tree-parloops.c (gather_scalar_reductions): Use
4558 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4561 2018-05-29 Richard Biener <rguenther@suse.de>
4563 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4565 2018-05-29 Martin Liska <mliska@suse.cz>
4566 David Malcolm <dmalcolm@redhat.com>
4568 * vec.c (test_reverse): New.
4569 (vec_c_tests): Add new test.
4570 * vec.h (vl_ptr>::reverse): New function.
4572 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4574 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4576 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4579 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4581 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4583 2018-05-28 Richard Biener <rguenther@suse.de>
4585 PR tree-optimization/85933
4586 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4587 look at stmts marked as vectorizable.
4589 2018-05-28 Richard Biener <rguenther@suse.de>
4591 PR tree-optimization/85934
4592 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4593 vector boolean check before scalar optimization.
4595 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4597 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4600 2018-05-28 Mark Wielaard <mark@klomp.org>
4602 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4603 if it is an expression containing a minus sign.
4605 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4607 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4609 2018-05-27 Paul Koning <ni1d@arrl.net>
4611 * config/pdp11/pdp11.md (truncsihi2): Remove.
4613 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4614 Chung-Ju Wu <jasonwucj@gmail.com>
4616 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4618 (unaligned_store_dw): Ditto.
4619 * config/nds32/nds32-memory-manipulation.c
4620 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4621 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4622 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4623 (emit_setmem_word_loop): Rename to ...
4624 (emit_setmem_doubleword_loop): ... this.
4625 (nds32_gen_dup_4_byte_to_word_value): New function.
4626 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4627 (nds32_expand_setmem_loop): Refine implementation.
4628 (nds32_expand_setmem_loop_v3m): Ditto.
4629 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4632 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4634 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4636 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4638 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4639 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4640 machine->attr_no_prologue_p.
4641 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4642 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4643 (nds32_expand_epilogue): Consider attr_naked_p.
4644 (nds32_expand_epilogue_v3pop): Likewise.
4645 (nds32_can_use_return_insn): Likewise.
4646 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4647 attr_no_prologue_p fields.
4648 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4650 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4653 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4655 * config/i386/sse.md
4656 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4658 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4660 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4662 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4664 (*<floatsuffix>floatv2div2sf2): Rename to ...
4665 (*float<floatunssuffix>v2div2sf2): ... this.
4666 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4667 (float<floatunssuffix>v2div2sf2_mask): ... this.
4668 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4669 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4670 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4672 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4674 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4676 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4678 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4680 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4682 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4683 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4684 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4685 gen_ufix_truncv8dfv8si2.
4686 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4687 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4688 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4689 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4690 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4691 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4692 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4693 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4695 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4699 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4701 2018-05-25 Jim Wilson <jimw@sifive.com>
4703 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4704 * config/riscv/riscv.c (struct machine_function): Add
4705 interrupt_handler_p and attribute_checked_p fields.
4706 (riscv_attribute_table): Add interrupt.
4707 (riscv_interrupt_type_p): New.
4708 (riscv_save_reg_p): Save extra regs for interrupt handler.
4709 (riscv_use_save_libcall): Return false for interrupt handler.
4710 (riscv_first_stack_step): Add forward declaration.
4711 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4712 for interrupt handler with large frame. Use it for saved reg list.
4713 (riscv_expand_prologue): Move flag_stack_usage_info support to
4714 eliminate duplication.
4715 (riscv_expand_epilogue): Generate mret for interrupt handler.
4716 (riscv_epilogue_uses): New.
4717 (riscv_can_use_return_insn): Return false for interrupt handler.
4718 (riscv_function_ok_for_sibcall): Likewise.
4719 (riscv_set_current_function): Add interrupt handler support.
4720 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4721 * config/riscv/riscv.md (UNSPECV_MRET): New.
4723 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4725 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4727 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4729 PR tree-optimization/85712
4730 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4731 this candidate has already been replaced in-situ by a copy.
4733 2018-05-25 Jason Merrill <jason@redhat.com>
4735 PR c++/80485 - inline function non-zero address.
4736 * symtab.c (nonzero_address): Check DECL_COMDAT.
4738 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4741 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4742 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4743 extension of SImode operation. Use const123_operand predicate.
4744 (*saddsi_1): Remove.
4745 (*saddl_se_1): Ditto.
4747 (*ssubl_se_1): Ditto.
4748 * config/alpha/predicates.md (const123_operand): New predicate.
4749 * config/alpha/constraints.md (P): Use IN_RANGE.
4751 2018-05-25 Richard Biener <rguenther@suse.de>
4753 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4755 (ref_maybe_used_by_stmt_p): Likewise.
4756 (stmt_may_clobber_ref_p): Likewise.
4757 (stmt_may_clobber_ref_p_1): Likewise.
4758 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4760 (ref_maybe_used_by_stmt_p): Likewise.
4761 (stmt_may_clobber_ref_p): Likewise.
4762 (stmt_may_clobber_ref_p_1): Likewise.
4763 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4764 the alias oracle to disambiguate DRs with stmts DR analysis
4766 (vect_analyze_data_refs): Do not give up on not analyzable
4767 DRs for BB vectorization. Remove code truncating the dataref
4770 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4773 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4774 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4775 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4777 2018-05-25 Richard Biener <rguenther@suse.de>
4779 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4780 function, combining stmt data ref gathering and fatal analysis
4782 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4783 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4784 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4785 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4786 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4788 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4790 PR tree-optimization/85720
4791 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4792 SCC if all partitions are builtins.
4793 (version_loop_by_alias_check): New parameter. Generate cancelable
4794 runtime alias check if all partitions are builtins.
4795 (distribute_loop): Update call to above function.
4797 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4799 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4800 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4801 (parm_default_def_partition_arg): Ditto.
4802 (set_parm_default_def_partition): Ditto.
4803 (get_parm_default_def_partitions): Ditto and make it static.
4804 (get_undefined_value_partitions): Ditto and make it static.
4805 (remove_ssa_form): Refactor call to init_var_map here.
4806 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4807 computation for loop region.
4808 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4809 (register_default_def): Delete.
4810 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4811 (parm_default_def_partition_arg): Ditto.
4812 (set_parm_default_def_partition): Ditto.
4813 (get_parm_default_def_partitions): Ditto and make it static.
4814 (get_undefined_value_partitions): Ditto and make it static.
4815 (coalesce_with_default, coalesce_with_default): Update comment.
4816 (create_coalesce_list_for_region): New func factored out from
4817 create_outofssa_var_map.
4818 (populate_coalesce_list_for_outofssa): New func factored out from
4819 create_outofssa_var_map and coalesce_ssa_name.
4820 (create_outofssa_var_map): Delete.
4821 (coalesce_ssa_name): Refactor to support live range computation.
4822 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4823 (get_parm_default_def_partitions): Delete.
4824 (get_undefined_value_partitions): Ditto.
4825 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4826 computation for loop region.
4827 (new_tree_live_info, loe_visit_block): Ditto.
4828 (live_worklist, set_var_live_on_entry): Ditto.
4829 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4830 * tree-ssa-live.h (struct _var_map): New fields.
4831 (init_var_map): Change decl.
4832 (region_contains_p): New.
4834 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4836 * tree-ssa-live.h (live_merge_and_clear): Delete.
4838 2018-05-25 Richard Biener <rguenther@suse.de>
4841 * tree-dump.c (dequeue_and_dump): Remove access to removed
4842 operand 2 of a SWITCH_EXPR.
4844 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4846 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4847 multiplication and division.
4848 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4849 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4850 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4851 (cond_udiv_optab, cond_umod_optab): New optabs.
4852 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4853 (IFN_COND_RDIV): New internal functions.
4854 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4855 TRUNC_MOD_EXPR and RDIV_EXPR.
4856 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4857 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4859 (SVE_INT_BINARY): Include mult.
4860 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4861 (optab, sve_int_op): Handle mult.
4862 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4864 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4865 for SVE_INT_BINARY_SD.
4867 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4869 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4870 (optab, sve_int_op): Handle div and udiv.
4871 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4872 for SVE_INT_BINARY_SD.
4873 (*<optab><mode>3): New insn for the same.
4875 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4877 * tree-vect-patterns.c: Include predict.h.
4878 (vect_recog_divmod_pattern): Restrict check for division support
4879 to when optimizing for size.
4881 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4883 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4884 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4885 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4886 (gimple_match_op::set_op): Likewise.
4887 (gimple_resimplify4): Declare.
4888 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4889 (expr::gen_transform): Likewise.
4890 (decision_tree::gen): Generate a simplification routine for 4 operands.
4891 * gimple-match-head.c (gimple_simplify): Add an overload for
4892 4 operands. In the top-level function, handle up to 4 call
4893 arguments and call gimple_resimplify4.
4894 (gimple_resimplify4): New function.
4895 (build_call_internal): Pass a fourth operand.
4896 (maybe_push_to_seq): Likewise.
4897 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4898 Fold VEC_COND_EXPRs of an operation and a default value into
4899 an IFN_COND_* function if possible.
4900 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4902 (SVE_COND_FP_BINARY): Include them.
4903 (optab, sve_fp_op): Handle them.
4904 (SVE_INT_BINARY_REV): New code iterator.
4905 (SVE_COND_FP_BINARY_REV): New int iterator.
4906 (commutative): New int attribute.
4907 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4909 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4911 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4912 (*cond_<optab><mode>): New patterns for reversed operands.
4914 2018-05-25 Richard Biener <rguenther@suse.de>
4916 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4917 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4918 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4919 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4920 * tree-vect-data-refs.c (everywhere): Adjust users.
4921 * tree-vect-loop.c (everywhere): Likewise.
4922 * tree-vect-slp.c (everywhere): Likewise.
4923 * tree-vect-stmts.c (everywhere): Likewise.
4924 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4926 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4928 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4930 (gcc_cv_as_section_exclude): ... this.
4931 Try Solaris as #exclude syntax.
4932 * configure: Regenerate.
4933 * config.in: Regenerate.
4934 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4936 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4937 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4939 * varasm.c (default_elf_asm_named_section): Don't check if
4940 HAVE_GAS_SECTION_EXCLUDE is defined.
4942 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4944 * doc/md.texi: Update the documentation of the cond_* optabs
4945 to mention the new final operand. Fix GET_MODE_NUNITS call.
4946 Describe the scalar case too.
4947 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4948 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4950 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4951 (get_conditional_internal_fn): Update comment.
4952 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4953 accumulator value as a final argument to conditional functions.
4954 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4955 a define_expand and add an "else" operand. Assert for now that
4956 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4957 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4958 (*cond_<optab><mode>): New patterns.
4959 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4960 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4961 (UNSPEC_COND_EOR): Delete.
4962 (optab): Remove associated mappings.
4963 (SVE_INT_BINARY): New code iterator.
4964 (sve_int_op): Remove int attribute and add "minus" to the code
4966 (SVE_COND_INT_OP): Delete.
4967 (SVE_COND_FP_OP): Rename to...
4968 (SVE_COND_FP_BINARY): ...this.
4970 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4972 * optabs.c (can_reuse_operands_p): New function.
4973 (maybe_legitimize_operands): Try to reuse the results for
4976 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4978 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4979 Add {q} suffix to insn mnemonic.
4981 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4983 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4984 (msp430_warn_func_return): New.
4986 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4988 * fold-const.c (tree_nonzero_bits): New function.
4989 * fold-const.h (tree_nonzero_bits): Likewise.
4990 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4991 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4993 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4997 * varasm.c (assemble_alias): Check ifunc_resolver only on
5000 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5003 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5004 when memory input operand is handled.
5006 2018-05-24 Luis Machado <luis.machado@linaro.org>
5008 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5010 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5012 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5014 * match.pd: Delay FMA folds until after vectorization.
5016 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5019 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5020 address check not strict.
5022 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5024 * gimple-match.h (gimple_match_op): New class.
5025 (mprts_hook): Replace parameters with a gimple_match_op *.
5026 (maybe_build_generic_op): Likewise.
5027 (gimple_simplified_result_is_gimple_val): Replace parameters with
5028 a const gimple_match_op *.
5029 (gimple_simplify): Replace code_helper * and tree * parameters with
5030 a gimple_match_op * parameter.
5031 (gimple_resimplify1): Replace code_helper *, tree and tree *
5032 parameters with a gimple_match_op * parameter.
5033 (gimple_resimplify2): Likewise.
5034 (gimple_resimplify3): Likewise.
5035 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5036 parameters with a gimple_match_op * parameter.
5037 * gimple-match-head.c (gimple_simplify): Change prototypes of
5038 auto-generated functions to take a gimple_match_op * instead of
5039 separate code_helper * and tree * parameters. Make the same
5040 change in the top-level overload and update calls to the
5041 gimple_resimplify routines. Update calls to the auto-generated
5042 functions and to maybe_push_res_to_seq in the publicly-facing
5043 operation-specific gimple_simplify overloads.
5044 (gimple_match_op::MAX_NUM_OPS): Define.
5045 (gimple_resimplify1): Replace rcode and ops with a single res_op
5046 parameter. Update call to gimple_simplify.
5047 (gimple_resimplify2): Likewise.
5048 (gimple_resimplify3): Likewise.
5049 (mprts_hook): Replace parameters with a gimple_match_op *.
5050 (maybe_build_generic_op): Likewise.
5051 (build_call_internal): Replace type, nargs and ops with
5052 a gimple_match_op *.
5053 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5054 with a single gimple_match_op *. Update calls to mprts_hook,
5055 build_call_internal and gimple_simplified_result_is_gimple_val.
5056 Factor out code that is common to the tree_code and combined_fn cases.
5057 * genmatch.c (expr::gen_transform): Replace tem_code and
5058 tem_ops with a gimple_match_op called tem_op. Update calls
5059 to the gimple_resimplify functions and maybe_push_res_to_seq.
5060 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5061 res_ops. Update call to the gimple_resimplify functions.
5062 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5063 (decision_tree::gen): Make the functions take a gimple_match_op *
5064 called res_op instead of separate res_code and res_ops parameters.
5065 Update call accordingly.
5066 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5067 and ops with a single res_op parameter. Update calls to
5068 maybe_build_generic_op and maybe_push_res_to_seq.
5069 (fold_stmt_1): Update calls to gimple_simplify and
5070 replace_stmt_with_simplification.
5071 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5072 and gimple_simplified_result_is_gimple_val.
5073 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5075 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5076 with a gimple_match_op *.
5077 (vn_nary_build_or_lookup): Likewise. Update call to
5078 vn_nary_build_or_lookup_1.
5079 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5080 gimple_match_op *. Update calls to the gimple_resimplify routines
5081 and to gimple_simplified_result_is_gimple_val.
5082 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5083 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5084 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5085 (visit_nary_op): Likewise.
5086 (visit_reference_op_load): Likewise.
5088 2018-05-23 Luis Machado <luis.machado@linaro.org>
5090 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5091 modifier for printing the step amount.
5093 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5096 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5099 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5101 * doc/sourcebuild.texi (Endianness): New subsubsection.
5103 2018-05-23 Luis Machado <luis.machado@linaro.org>
5105 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5106 <prefetch_dynamic_strides>: New const bool field.
5107 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5108 prefetch_dynamic_strides.
5109 (exynosm1_prefetch_tune): Likewise.
5110 (thunderxt88_prefetch_tune): Likewise.
5111 (thunderx_prefetch_tune): Likewise.
5112 (thunderx2t99_prefetch_tune): Likewise.
5113 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5115 (aarch64_override_options_internal): Update to set
5116 PARAM_PREFETCH_DYNAMIC_STRIDES.
5117 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5118 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5119 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5120 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5121 prefetch-dynamic-strides setting.
5123 2018-05-23 Luis Machado <luis.machado@linaro.org>
5125 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5126 <minimum_stride>: New const int field.
5127 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5128 minimum_stride field defaulting to -1.
5129 (exynosm1_prefetch_tune): Likewise.
5130 (thunderxt88_prefetch_tune): Likewise.
5131 (thunderx_prefetch_tune): Likewise.
5132 (thunderx2t99_prefetch_tune): Likewise.
5133 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5134 <default_opt_level>: Set to 3.
5135 (aarch64_override_options_internal): Update to set
5136 PARAM_PREFETCH_MINIMUM_STRIDE.
5137 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5138 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5139 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5141 stride is constant and is below the minimum stride threshold.
5143 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5145 * config/arm/arm-cpus.in (mode26): Delete.
5146 (armv4): Delete mode26 reference.
5147 * config/arm/arm.c (arm_configure_build_target): Delete use of
5150 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5152 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5154 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5155 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5157 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5158 (floatunsdidf2): Ditto.
5160 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5162 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5163 (fixuns_trunc<mode>si2_avx512f): Ditto.
5164 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5165 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5166 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5168 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5170 PR rtl-optimization/79985
5171 * df-scan.c (df_insn_refs_collect): Remove special case for
5172 global registers and asm statements.
5174 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5176 * extend.texi (Global Register Variables): Rewrite the bullet list.
5177 Note that the register is available for allocation. Note that access
5178 via inline asm must use constraints. Add note about async-signal
5179 handlers. Remove paragraph about automagic register selection.
5181 2018-05-23 Richard Biener <rguenther@suse.de>
5183 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5184 of fixed offset from memset VN.
5186 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5188 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5190 (alloc_cand_and_find_basis): Initialize first_interp field.
5191 (slsr_process_mul): Modify first_interp field.
5192 (slsr_process_add): Likewise.
5193 (slsr_process_cast): Modify first_interp field for each new
5195 (slsr_process_copy): Likewise.
5196 (dump_candidate): Dump first_interp field.
5197 (replace_mult_candidate): Process all interpretations, not just
5199 (replace_rhs_if_not_dup): Likewise.
5200 (replace_one_candidate): Likewise.
5202 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5204 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5206 (aarch64_needs_frame_chain): New function.
5207 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5209 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5212 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5213 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5214 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5215 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5216 as true for strict-align.
5217 (aarch64_can_inline_p): Perform checks even when callee has no
5218 attributes to check for strict alignment.
5219 * doc/extend.texi (AArch64 Function Attributes): Document
5221 * doc/invoke.texi: (AArch64 Options): Likewise.
5223 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5225 PR tree-optimization/85853
5226 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5227 the handling of the root of the node to...
5228 (vect_slp_analyze_node_operations_1): ...this new function,
5229 and run the whole thing with the child nodes' def types
5230 set according to their SLP node's def type.
5232 2018-05-23 Richard Biener <rguenther@suse.de>
5235 * tree-data-ref.c (create_runtime_alias_checks): Defer
5236 and ignore overflow warnings.
5238 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5240 PR tree-optimization/85822
5241 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5244 2018-05-23 Richard Biener <rguenther@suse.de>
5246 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5247 memset constants via native_interpret_expr.
5249 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5252 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5254 (cgraph_node::create_alias): Likewise.
5255 (cgraph_node::get_availability): Check ifunc_resolver instead
5256 of looking up ifunc attribute.
5257 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5258 * varasm.c (do_assemble_alias): Likewise.
5259 (assemble_alias): Likewise.
5260 (default_binds_local_p_3): Likewise.
5261 * cgraph.h (cgraph_node): Add ifunc_resolver.
5262 (cgraph_node::only_called_directly_or_aliased_p): Return false
5264 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5266 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5267 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5268 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5269 instead of looking up ifunc attribute.
5271 2018-05-22 Luis Machado <luis.machado@linaro.org>
5273 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5275 2018-05-22 Martin Sebor <msebor@redhat.com>
5278 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5279 only when expasion succeeds.
5280 (expand_builtin_strcmp): Same.
5281 (expand_builtin_strncmp): Same.
5283 2018-05-22 Martin Sebor <msebor@redhat.com>
5285 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5287 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5288 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5290 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5291 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5293 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5294 Add check for legitimate_address.
5295 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5296 (aarch64_swap_ldrstr_operands): New.
5297 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5300 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5301 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5303 * config/aarch64/aarch64.md: New patterns to generate stp
5305 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5306 single words and double words.
5307 (load_pair_sw, load_pair_dw): Likewise.
5308 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5310 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5312 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5313 for different mode ldpstp and add peephole for merged zero stores.
5315 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5317 (aarch64_gen_store_pair): Rename calls to match new patterns.
5318 (aarch64_gen_load_pair): Rename calls to match new patterns.
5319 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5320 (load_pair<DREG:mode><DREG2:mode>): ... This.
5321 (store_pair<mode>): Rename to...
5322 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5323 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5325 (V_INT_EQUIV): Handle SImode.
5326 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5329 2018-05-22 Martin Sebor <msebor@redhat.com>
5332 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5333 or ajust the presumed bound on an operation to avoid unnecessary
5336 2018-05-22 Martin Sebor <msebor@redhat.com>
5338 PR tree-optimization/85826
5339 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5340 assuming that a DECL necesarily has a constant size.
5342 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5345 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5347 2018-05-22 Richard Biener <rguenther@suse.de>
5349 PR tree-optimization/85834
5350 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5351 non-constant and non-zero memset arguments.
5353 2018-05-22 Martin Liska <mliska@suse.cz>
5356 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5358 2018-05-22 Richard Biener <rguenther@suse.de>
5360 PR tree-optimization/85863
5361 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5362 comparisons when vectype is specified.
5363 (vectorizable_condition): Do not specify vectype for
5364 vect_is_simple_cond when SLP vectorizing.
5366 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5369 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5370 define __ibm128 as long double.
5371 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5372 as a distinct type when IEEE 128-bit support is enabled.
5373 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5374 128-bit types to use the correct functions.
5375 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5376 convert between 128-bit floating point types that have different
5377 modes but the same representation, instead of using gen_lowpart to
5379 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5381 (IFKF_reg): New attributes to give the register constraints for
5383 (extend<mode>tf2_internal): New insns to mark an explicit
5384 conversion between 128-bit floating point types that have a
5385 different mode but share the same representation.
5387 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5389 PR tree-optimization/85814
5390 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5391 a null return from get_strinfo when unsharing the next
5392 strinfo in the chain.
5394 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5397 * varasm.c (weak_finish): Clean up weak_decls.
5399 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5401 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5402 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5403 UNSPEC_UADALP values.
5404 * config/aarch64/iterators.md (ABAL): New int iterator.
5407 (sur): Add mappings for the above.
5408 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5410 (aarch64_<sur>abal<mode>_4): Likewise.
5411 (aarch64_<sur>adalp<mode>_3): Likewise.
5412 (<sur>sadv16qi): New define_expand.
5414 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5416 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5417 (*movdf_internal): Ditto.
5418 (*rcpsf2_sse): Ditto.
5419 (*rsqrtsf2_sse): Ditto.
5420 (*sqrt<mode>2_sse): Ditto.
5422 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5424 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5426 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5427 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5428 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5429 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5431 * config/aarch64/arm_neon.h: Likewise.
5432 * config/aarch64/iterators.md (VQ_I): New.
5434 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5436 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5438 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5441 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5443 * config/nds32/constraints.md (S): New constraint.
5444 * config/nds32/nds32.md (call_internal): Use constraint S.
5445 (call_value_internal): Likewise.
5446 (sibcall_internal): Likewise.
5447 (sibcall_value_internal): Likewise.
5449 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5450 Chung-Ju Wu <jasonwucj@gmail.com>
5452 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5455 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5456 Chung-Ju Wu <jasonwucj@gmail.com>
5458 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5459 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5460 (nds32_rtx_costs_impl): Simplify.
5461 (nds32_address_cost_impl): Simplify.
5462 (nds32_init_rtx_costs): New function.
5463 (nds32_rtx_costs_speed_prefer): Likewise.
5464 (nds32_rtx_costs_size_prefer): Likewise.
5465 (nds32_address_cost_speed_prefer): Likewise.
5466 (nds32_address_cost_speed_fwprop): Likewise.
5467 (nds32_address_cost_size_prefer): Likewise.
5468 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5469 * config/nds32/nds32.c (nds32_option_override): Use
5470 nds32_init_rtx_costs function.
5472 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5474 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5475 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5476 (TARGET_PIPELINE_N8): Likewise.
5477 (TARGET_PIPELINE_N10): Likewise.
5478 (TARGET_PIPELINE_N13): Likewise.
5479 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5481 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5483 * config/nds32/nds32-fpu.md: Update copyright year.
5485 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5487 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5489 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5491 * config/nds32/nds32.c
5492 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5493 * config/nds32/nds32.opt (minline-asm-r15): New option.
5495 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5497 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5499 * config/nds32/nds32.md (abssi2): New pattern.
5501 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5503 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5504 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5505 Merge insn pattern from sse_cvtsi2ss<round_name> and
5506 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5507 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5508 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5509 using SWI48 mode iterator.
5510 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5511 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5512 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5513 pattern from sse_cvttss2si<round_saeonly_name>
5514 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5515 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5516 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5517 using SWI48 mode iterator.
5518 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5519 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5520 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5521 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5522 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5523 using SWI48 mode iterator.
5524 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5525 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5526 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5527 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5528 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5529 SWI48 mode iterator.
5530 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5531 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5532 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5533 pattern from sse_cvttsd2si<round_saeonly_name>
5534 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5536 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5538 * config/nds32/nds32-md-auxiliary.c
5539 (nds32_valid_smw_lwm_base_p): Refine.
5540 (nds32_output_smw_single_word): Refine.
5541 (nds32_output_smw_double_word): New.
5542 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5544 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5546 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5547 (nds32_output_stack_pop): Refine.
5548 (nds32_expand_unaligned_load): Refine.
5549 (nds32_expand_unaligned_store): Refine.
5551 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5552 Chung-Ju Wu <jasonwucj@gmail.com>
5554 * config/nds32/constants.md: Add TP_REGNUM constant.
5555 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5556 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5558 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5559 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5560 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5561 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5562 and PIC code generation.
5563 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5565 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5567 * config/nds32/nds32.md: Support TLS and PIC.
5568 * config/nds32/nds32.c: Support TLS and PIC.
5569 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5570 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5573 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5575 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5576 mode with E_ prefix.
5578 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5579 Chung-Ju Wu <jasonwucj@gmail.com>
5581 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5582 * config/nds32/nds32-md-auxiliary.c
5583 (symbolic_reference_mentioned_p): New.
5584 (nds32_legitimize_ict_address): New.
5585 (nds32_expand_ict_move): New.
5586 (nds32_indirect_call_referenced_p): New.
5587 (nds32_symbol_binds_local_p): Delete.
5588 (nds32_long_call_p): Modify.
5589 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5590 * config/nds32/nds32-protos.h
5591 (symbolic_reference_mentioned_p): Declare.
5592 (nds32_legitimize_ict_address): Declare.
5593 (nds32_expand_ict_move): Declare.
5594 (nds32_indirect_call_referenced_p): Declare.
5595 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5596 (nds32_relax_group): Use nds32_ict_const_p as condition.
5597 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5598 (nds32_asm_file_start): Output ict_model directive in asm code.
5599 (nds32_legitimate_address_p): Consider indirect call.
5600 (nds32_print_operand): Consider indirect call.
5601 (nds32_print_operand_address): Consider indirect call.
5602 (nds32_insert_attributes): Handle "indirect_call" attribute.
5603 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5604 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5605 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5606 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5607 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5608 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5609 (TARGET_ICT_MODEL_SMALL): Define.
5610 (TARGET_ICT_MODEL_LARGE): Define.
5611 * config/nds32/nds32.md (movsi): Consider ict model.
5612 (call, call_value): Consider ict model.
5613 (sibcall, sibcall_value): Consider ict model.
5614 * config/nds32/nds32.opt (mict-model): New option.
5615 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5618 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5619 Monk Chiang <sh.chiang04@gmail.com>
5620 Jim Wilson <jimw@sifive.com>
5622 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5623 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5624 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5625 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5626 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5627 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5628 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5629 compute save_libcall_adjustment properly.
5630 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5631 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5632 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5633 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5634 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5635 (ABI_SPEC): Handle mabi=ilp32e.
5636 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5637 (RVE): Add RVE mask.
5638 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5639 <-march>: Add rv32e as an example.
5641 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5644 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5645 (intra_create_variable_infos): Handle C++ constructors.
5647 2018-05-18 Martin Liska <mliska@suse.cz>
5649 * passes.def: Remove a redundant pass.
5651 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5654 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5656 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5658 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5660 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5662 (armv2, armv2a, armv3, armv3m): Delete architectures.
5663 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5664 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5665 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5667 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5668 (*mulsidi3adddi): Likewise.
5669 (mulsidi3): Likewise.
5670 (*mulsidi3_nov6): Likewise.
5671 (umulsidi3): Likewise.
5672 (umulsidi3_nov6): Likewise.
5673 (umaddsidi4): Likewise.
5674 (*umulsidi3adddi): Likewise.
5675 (smulsi3_highpart): Likewise.
5676 (*smulsi3_highpart_nov6): Likewise.
5677 (umulsi3_highpart): Likewise.
5678 (*umulsi3_highpart_nov6): Likewise.
5679 * config/arm/arm.h (arm_arch3m): Delete.
5680 * config/arm/arm.c (arm_arch3m): Delete.
5681 (arm_option_override_internal): Update armv3-related comment.
5682 (arm_configure_build_target): Delete use of isa_bit_mode32.
5683 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5684 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5685 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5687 (mulusa3): Likewise.
5688 * config/arm/arm-protos.h (arm_arch3m): Delete.
5689 * config/arm/arm-tables.opt: Regenerate.
5690 * config/arm/arm-tune.md: Likewise.
5691 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5692 deleted architectures.
5694 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5696 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5697 (armv5t, armv5te): New features.
5698 (ARMv5, ARMv5e): Delete fgroups.
5699 (ARMv5t, ARMv5te): Adjust for above changes.
5701 (armv5, armv5e): Delete arches.
5702 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5704 (*call_reg_arm): Likewise.
5705 (*call_value_reg_armv5): Likewise.
5706 (*call_value_reg_arm): Likewise.
5707 (*call_symbol): Likewise.
5708 (*call_value_symbol): Likewise.
5709 (*sibcall_insn): Likewise.
5710 (*sibcall_value_insn): Likewise.
5712 (prefetch): Likewise.
5713 (define_split and define_peephole2 dependent on arm_arch5):
5715 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5717 (TARGET_ARM_QBIT): Likewise.
5718 (TARGET_DSP_MULTIPLY): Likewise.
5719 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5720 (arm_arch5, arm_arch5e): Delete.
5721 (arm_arch5t, arm_arch5te): Declare.
5722 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5723 (arm_arch5t): Declare.
5724 (arm_option_reconfigure_globals): Update for the above.
5725 (arm_options_perform_arch_sanity_checks): Update comment, replace
5726 use of arm_arch5 with arm_arch5t.
5727 (use_return_insn): Likewise.
5728 (arm_emit_call_insn): Likewise.
5729 (output_return_instruction): Likewise.
5730 (arm_final_prescan_insn): Likewise.
5731 (arm_coproc_builtin_available): Likewise.
5732 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5733 arm_arch5e with arm_arch5t and arm_arch5te.
5734 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5735 (arm_arch5t, arm_arch5te): Declare.
5736 * config/arm/arm-tables.opt: Regenerate.
5737 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5738 * config/arm/t-multilib: Likewise.
5739 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5740 instead of arm_arch5.
5741 (*call_reg_thumb1): Likewise.
5742 (*call_value_reg_thumb1_v5): Likewise.
5743 (*call_value_reg_thumb1): Likewise.
5744 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5746 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5748 2018-05-18 Martin Liska <mliska@suse.cz>
5750 PR gcov-profile/84846
5751 * doc/gcov.texi: Document -t option of gcov tool.
5753 2018-05-18 Martin Liska <mliska@suse.cz>
5755 PR gcov-profile/84846
5756 * gcov.c (print_usage): Add new -t option.
5757 (process_args): Handle the option.
5758 (generate_results): Use stdout as output when requested by
5761 2018-05-18 Martin Liska <mliska@suse.cz>
5763 PR gcov-profile/84846
5764 * coverage.c (coverage_init): Write PWD to .gcno file.
5765 * doc/gcov.texi: Document how working directory is printed.
5766 * gcov-dump.c (dump_gcov_file): Print PWD.
5767 * gcov.c (output_intermediate_file): Likewise.
5768 (read_graph_file): Read PWD string.
5769 (output_lines): Print PWD.
5771 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5774 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5775 for retval and return false if all args to phi are zero.
5777 2018-05-18 Richard Biener <rguenther@suse.de>
5779 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5781 (evrp_dom_walker::before_dom_children): Call it.
5783 2018-05-18 Richard Biener <rguenther@suse.de>
5785 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5786 results when processing array refs with variable index.
5788 2018-05-18 Toon Moene <toon@moene.org>
5790 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5791 directly after that of -floop-interchange. Indicate that both
5792 options are enabled by default when specifying -O3.
5794 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5796 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5797 iterator. Delete separate integer-mode vec_set<mode> expander.
5798 (aarch64_simd_vec_setv2di): Delete.
5799 (vec_setv2di): Delete.
5800 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5801 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5802 the "w, r" alternative.
5804 2018-05-18 Martin Liska <mliska@suse.cz>
5806 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5807 * tree-pass.h (make_pass_lower_switch_O0): New function.
5808 * tree-switch-conversion.c (node_has_low_bound): Remove.
5809 (node_has_high_bound): Likewise.
5810 (node_is_bounded): Likewise.
5811 (class pass_lower_switch): Make it a template type and create
5813 (pass_lower_switch::execute): Add template argument.
5814 (make_pass_lower_switch): New function.
5815 (make_pass_lower_switch_O0): New function.
5816 (do_jump_if_equal): Remove.
5817 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5818 all the hard work to tree optimization passes.
5820 2018-05-18 Martin Liska <mliska@suse.cz>
5822 * dbgcnt.c (limit_low): Renamed from limit.
5823 (limit_high): New variable.
5824 (dbg_cnt_is_enabled): Check for upper limit.
5825 (dbg_cnt): Adjust dumping.
5826 (dbg_cnt_set_limit_by_index): Add new argument for high
5828 (dbg_cnt_set_limit_by_name): Likewise.
5829 (dbg_cnt_process_single_pair): Parse new format.
5830 (dbg_cnt_process_opt): Use strtok.
5831 (dbg_cnt_list_all_counters): Remove 'value' and add
5833 * doc/invoke.texi: Document changes.
5835 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5837 * doc/sourcebuild.texi (scalar_all_fma): Document.
5838 * tree.def (FMA_EXPR): Delete.
5839 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5840 * internal-fn.c (ternary_direct): New macro.
5841 (expand_ternary_optab_fn): Likewise.
5842 (direct_ternary_optab_supported_p): Likewise.
5843 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5844 * builtins.c (fold_builtin_fma): Delete.
5845 (fold_builtin_3): Don't call it.
5846 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5847 * expr.c (expand_expr_real_2): Likewise.
5848 * fold-const.c (operand_equal_p): Likewise.
5849 (fold_ternary_loc): Likewise.
5850 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5851 * gimple.c (DEFTREECODE): Likewise.
5852 * gimplify.c (gimplify_expr): Likewise.
5853 * optabs-tree.c (optab_for_tree_code): Likewise.
5854 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5855 * tree-eh.c (operation_could_trap_p): Likewise.
5856 (stmt_could_throw_1_p): Likewise.
5857 * tree-inline.c (estimate_operator_cost): Likewise.
5858 * tree-pretty-print.c (dump_generic_node): Likewise.
5859 (op_code_prio): Likewise.
5860 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5861 * tree-ssa-operands.c (get_expr_operands): Likewise.
5862 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5863 * fold-const-call.h (fold_fma): Delete.
5864 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5865 CFN_FNMA and CFN_FNMS.
5867 * genmatch.c (combined_fn): New enum.
5868 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5869 (commutative_op): New function.
5870 (commutate): Use it. Handle more than 2 operands.
5871 (dt_operand::gen_gimple_expr): Use commutative_op.
5872 (parser::parse_expr): Allow :c to be used with non-binary
5873 operators if the commutative operand is known.
5874 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5875 CFN_FMS, CFN_FNMA and CFN_FNMS.
5876 (backprop::process_assign_use): Remove FMA_EXPR handling.
5877 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5878 (gen_hsa_fma): New function.
5879 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5880 IFN_FNMA and IFN_FNMS.
5881 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5882 * gimple-fold.h (follow_all_ssa_edges): Declare.
5883 * gimple-fold.c (follow_all_ssa_edges): New function.
5884 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5885 gimple_build interface and use follow_all_ssa_edges to fold the result.
5886 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5887 instead of checking for optabs directly.
5888 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5889 rather than FMA_EXPRs.
5890 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5891 call to IFN_FMA instead of an FMA_EXPR.
5893 2018-05-17 Jim Wilson <jimw@sifive.com>
5895 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5896 sign extended promoted subreg, and the range does not have the sign bit
5897 set, then do a sign extend.
5899 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5900 test, check for sign extended subreg and/or constant operands, and
5901 do a sign extend in that case.
5903 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5905 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5906 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5908 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5909 Change logics_shift_reg to logics_shift_imm.
5910 (thunderx2t99_fp_loadpair_basic): Delete.
5911 (thunderx2t99_fp_storepair_basic): Delete.
5912 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5913 (thunderx2t99_asimd_polynomial): Delete.
5914 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5915 and neon_fp_mul_d_scalar_q.
5916 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5917 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5918 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5919 (thunderx2t99_asimd_lut): Add missing tbl types.
5920 (thunderx2t99_asimd_ext): Delete.
5921 (thunderx2t99_asimd_load1_1_mult): Delete.
5922 (thunderx2t99_asimd_load1_2_mult): Delete.
5923 (thunderx2t99_asimd_load1_ldp): New.
5924 (thunderx2t99_asimd_load1): New.
5925 (thunderx2t99_asimd_load2): Add missing *load2* types.
5926 (thunderx2t99_asimd_load3): New.
5927 (thunderx2t99_asimd_load4): New.
5928 (thunderx2t99_asimd_store1_1_mult): Delete.
5929 (thunderx2t99_asimd_store1_2_mult): Delete.
5930 (thunderx2t99_asimd_store2_mult): Delete.
5931 (thunderx2t99_asimd_store2_onelane): Delete.
5932 (thunderx2t99_asimd_store_stp): New.
5933 (thunderx2t99_asimd_store1): New.
5934 (thunderx2t99_asimd_store2): New.
5935 (thunderx2t99_asimd_store3): New.
5936 (thunderx2t99_asimd_store4): New.
5938 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5940 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5941 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5943 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5944 Segher Boessenkool <segher@kernel.crashing.org>
5947 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5950 2018-05-17 Richard Biener <rguenther@suse.de>
5952 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5953 for pruning loop and prune defs feeding only already visited PHIs.
5955 2018-05-17 Richard Biener <rguenther@suse.de>
5957 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5959 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5960 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/85793
5963 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5964 for VMAT_ELEMENTWISE.
5966 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5968 * internal-fn.h (lookup_internal_fn): Declare
5969 * internal-fn.c (lookup_internal_fn): New function.
5970 * gimple.c (gimple_build_call_from_tree): Handle calls to
5972 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5973 internal function names.
5974 * tree-pretty-print.c (dump_generic_node): Likewise.
5975 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5977 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5979 * gimple-fold.h (gimple_build): Make the function forms take
5980 combined_fn rather than built_in_function.
5981 (gimple_simplify): Likewise.
5982 * gimple-match-head.c (gimple_simplify): Likewise.
5983 * gimple-fold.c (gimple_build): Likewise.
5984 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5985 rather than gimple_build_call_internal.
5986 (get_initial_defs_for_reduction): Likewise.
5987 (vect_create_epilog_for_reduction): Likewise.
5988 (vectorizable_live_operation): Likewise.
5990 2018-05-17 Martin Liska <mliska@suse.cz>
5992 * gimple-ssa-sprintf.c (format_directive): Do not use
5993 space in between 'G_' and '('.
5995 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5998 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5999 even if the mask is not all ones.
6002 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6004 (ix86_gimple_fold_builtin): Likewise.
6007 * config/i386/i386.c: Include tree-vector-builder.h.
6008 (ix86_vector_shift_count): New function.
6009 (ix86_fold_builtin): Fold shift builtins by scalar count.
6010 (ix86_gimple_fold_builtin): Likewise.
6012 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6013 _mm512_setzero): New intrinsics.
6015 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6016 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6018 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6019 code generation for cases where splatting a value is not useful.
6020 * simplify-rtx.c (simplify_ternary_operation): Simplify
6021 vec_merge across a vec_duplicate and a paradoxical subreg forming
6022 a vector mode to a vec_concat.
6024 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6026 * config.gcc: Support "goldmont-plus".
6027 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6029 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6030 PROCESSOR_GOLDMONT_PLUS.
6031 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6032 (processor_target_table): Add "goldmont-plus".
6033 (PTA_GOLDMONT_PLUS): Define.
6034 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6035 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6036 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6037 (fold_builtin_cpu): Add "goldmont-plus".
6038 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6039 (ix86_option_override_internal): Add "goldmont-plus".
6040 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6041 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6042 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6043 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6045 2018-05-17 Richard Biener <rguenther@suse.de>
6047 PR tree-optimization/85757
6048 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6049 remove defs that only feed that PHI from further processing.
6051 2018-05-16 Jim Wilson <jimw@sifive.com>
6053 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6055 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6057 2018-05-16 Mark Wielaard <mark@klomp.org>
6059 * dwarf2out.c (count_index_strings): New function.
6060 (output_indirect_strings): Call count_index_strings and generate
6061 header for dwarf_version >= 5.
6063 2018-05-16 Mark Wielaard <mark@klomp.org>
6065 * dwarf2out.c (dwarf_FORM): New function.
6066 (set_indirect_string): Use dwarf_FORM.
6067 (reset_indirect_string): Likewise.
6068 (size_of_die): Likewise.
6069 (value_format): Likewise.
6070 (output_die): Likewise.
6071 (add_skeleton_AT_string): Likewise.
6072 (output_macinfo_op): Likewise.
6073 (index_string): Likewise.
6074 (output_index_string_offset): Likewise.
6075 (output_index_string): Likewise.
6076 (count_index_strings): Likewise.
6078 2018-05-16 Carl Love <cel@us.ibm.com>
6080 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6081 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6083 2018-05-16 Martin Jambor <mjambor@suse.cz>
6085 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6086 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6088 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6090 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6091 (fnma<mode>4): Likewise.
6092 (fms<mode>4): Likewise.
6093 (fnms<mode>4): Likewise.
6094 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6095 (aarch64_fnma<mode>4): Likewise.
6096 (aarch64_fms<mode>4): Likewise.
6097 (aarch64_fnms<mode>4): Likewise.
6098 (aarch64_fnmadd<mode>4): Likewise.
6100 2018-05-16 Jason Merrill <jason@redhat.com>
6102 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6104 2018-05-16 Richard Biener <rguenther@suse.de>
6106 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6107 (dump_stmt_cost): Declare.
6108 (add_stmt_cost): Dump cost we add.
6109 (add_stmt_costs): New function.
6110 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6112 (vect_analyze_stmt): Adjust prototype.
6113 (vectorizable_condition): Likewise.
6114 (vectorizable_live_operation): Likewise.
6115 (vectorizable_reduction): Likewise.
6116 (vectorizable_induction): Likewise.
6117 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6118 cost vector to pass to vectorizable_ and record afterwards.
6119 (vect_model_reduction_cost): Take cost vector argument and adjust.
6120 (vect_model_induction_cost): Likewise.
6121 (vectorizable_reduction): Likewise.
6122 (vectorizable_induction): Likewise.
6123 (vectorizable_live_operation): Likewise.
6124 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6125 SLP_TREE_NUMBER_OF_VEC_STMTS.
6126 (vect_analyze_slp_cost_1): Remove.
6127 (vect_analyze_slp_cost): Likewise.
6128 (vect_slp_analyze_node_operations): Take visited args and
6129 a target cost vector. Avoid processing already visited stmt sets.
6130 (vect_slp_analyze_operations): Use a local cost vector to gather
6131 costs and register those of non-discarded instances.
6132 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6133 (vect_schedule_slp_instance): Remove copying of
6134 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6136 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6137 adding cost. Record cost entry location.
6138 (vect_prologue_cost_for_slp_op): Function to compute cost of
6139 a constant or invariant generated for SLP vect in the prologue,
6140 split out from vect_analyze_slp_cost_1.
6141 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6142 (vect_model_promotion_demotion_cost): Likewise.
6143 (vect_model_store_cost): Likewise, make static.
6144 (vect_model_load_cost): Likewise.
6145 (vectorizable_bswap): Add cost vector arg and adjust.
6146 (vectorizable_call): Likewise.
6147 (vectorizable_simd_clone_call): Likewise.
6148 (vectorizable_conversion): Likewise.
6149 (vectorizable_assignment): Likewise.
6150 (vectorizable_shift): Likewise.
6151 (vectorizable_operation): Likewise.
6152 (vectorizable_store): Likewise.
6153 (vectorizable_load): Likewise.
6154 (vectorizable_condition): Likewise.
6155 (vectorizable_comparison): Likewise.
6156 (can_vectorize_live_stmts): Likewise.
6157 (vect_analyze_stmt): Likewise.
6158 (vect_transform_stmt): Adjust calls to vectorizable_*.
6159 * tree-vectorizer.c: Include gimple-pretty-print.h.
6160 (dump_stmt_cost): New function.
6162 2018-05-16 Richard Biener <rguenther@suse.de>
6164 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6165 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6166 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6167 (check_name): New callback.
6168 (dse_classify_store): Track cycles via a visited bitmap of PHI
6169 defs and simplify handling of in-loop and across loop dead stores
6170 and properly fail for loop-variant refs. Handle byte-tracking with
6171 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6174 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6176 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6177 (vect_get_mask_type_for_stmt): Likewise.
6178 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6180 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6181 to determine the statement's vector type and the vector type that
6182 should be used for calculating nunits. Deal with cases in which
6183 the type has to be deferred.
6184 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6185 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6186 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6187 (vect_determine_vf_for_stmt): New functions, split out from...
6188 (vect_determine_vectorization_factor): ...here.
6189 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6190 (vect_get_mask_type_for_stmt): New functions, split out from
6191 vect_determine_vectorization_factor.
6193 2018-05-16 Richard Biener <rguenther@suse.de>
6195 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6196 verify the [VEC_]COND_EXPR embedded comparison.
6198 2018-05-15 Martin Sebor <msebor@redhat.com>
6200 PR tree-optimization/85753
6201 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6202 RECORD_TYPE in addition to ARRAY_TYPE.
6204 2018-05-15 Martin Sebor <msebor@redhat.com>
6207 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6209 2018-05-15 Richard Biener <rguenther@suse.de>
6211 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6212 add by_clobber_p one. Change algorithm to collect all defs
6213 representing uses we need to walk and try reducing them to
6214 a single one before failing.
6215 (dse_dom_walker::dse_optimize_stmt): Adjust.
6217 2018-05-13 Mark Wielaard <mark@klomp.org>
6219 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6220 (size_of_loc_descr): Likewise.
6221 (output_loc_operands): Likewise.
6222 (output_loc_operands_raw): Likewise.
6223 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6224 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6225 (hash_loc_operands): Likewise.
6226 (compare_loc_operands): Likewise.
6228 2018-05-14 Mark Wielaard <mark@klomp.org>
6230 * dwarf2out.c (count_index_addrs): New function.
6231 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6233 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6235 PR tree-optimization/83648
6236 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6237 return value as malloc candidate.
6239 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6242 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6243 param as true in call to suggest_attribute.
6245 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6247 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6248 -mreadonly-in-sdata.
6250 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6252 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6254 (aarch64_crypto_aesd_fused): Likewise.
6256 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6258 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6259 (movsi_aarch64): Likewise.
6260 (load_pairsi): Likewise.
6261 (load_pairdi): Likewise.
6262 (store_pairsi): Likewise.
6263 (store_pairdi): Likewise.
6264 (load_pairsf): Likewise.
6265 (load_pairdf): Likewise.
6266 (store_pairsf): Likewise.
6267 (store_pairdf): Likewise.
6268 (zero_extend): Likewise.
6269 (trunc): Swap alternatives.
6270 (fcvt_target): Add '?' to prefer w over r.
6272 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6275 * config/i386/i386.md: Disallow non-commutative arithmetics in
6276 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6277 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6278 in the peephole2 before it.
6280 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6282 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6283 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6284 (ix86_handle_option): Handle -mcldemote.
6285 * config.gcc: New header.
6286 * config/i386/cldemoteintrin.h: New file.
6287 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6288 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6290 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6291 OPTION_MASK_ISA_CLDEMOTE.
6292 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6293 (ix86_valid_target_attribute_inner_p): Ditto.
6294 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6295 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6296 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6297 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6298 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6300 * config/i386/i386.opt: Add -mcldemote.
6301 * config/i386/x86intrin.h: New header.
6302 * doc/invoke.texi: Add -mcldemote.
6304 2018-05-14 Richard Biener <rguenther@suse.de>
6306 * doc/match-and-simplify.texi: Adjust :s documentation.
6308 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6310 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6311 intended memcpy size.
6312 (REORDER_45): Likewise.
6314 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6316 * sort.cc: New file.
6317 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6318 * vec.c (qsort_chk): Use gcc_qsort.
6319 * Makefile.in (OBJS-libcommon): Add sort.o.
6320 (build/sort.o): New target. Use it...
6321 (BUILD_RTL): ... here, and...
6322 (build/gencfn-macros): ... here, and...
6323 (build/genmatch): ... here.
6325 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6326 Chung-Ju Wu <jasonwucj@gmail.com>
6328 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6329 * config/nds32/nds32-graywolf.md: New file.
6330 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6331 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6333 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6334 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6335 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6336 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6337 * config/nds32/pipelines.md: Include n15 settings.
6339 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6340 Chung-Ju Wu <jasonwucj@gmail.com>
6342 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6343 * config/nds32/nds32-n13.md: New file.
6344 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6345 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6347 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6348 * config/nds32/nds32.md (pipeline_model): Add n13.
6349 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6350 * config/nds32/pipelines.md: Include n13 settings.
6352 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6353 Chung-Ju Wu <jasonwucj@gmail.com>
6355 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6356 * config/nds32/nds32-n10.md: New file.
6357 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6358 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6360 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6361 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6362 * config/nds32/nds32.md (pipeline_model): Add n10.
6363 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6364 * config/nds32/pipelines.md: Include n10 settings.
6366 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6367 Kito Cheng <kito.cheng@gmail.com>
6368 Chung-Ju Wu <jasonwucj@gmail.com>
6370 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6371 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6372 Add enum values for DSP extension instructions.
6373 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6375 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6376 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6378 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6379 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6380 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6381 * config/nds32/nds32-intrinsic.md: Likewise.
6382 * config/nds32/nds32_intrinsic.h: Likewise.
6383 * config/nds32/nds32-md-auxiliary.c: Likewise.
6384 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6385 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6386 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6387 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6388 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6389 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6390 TYPE_DMAC in switch statement.
6391 * config/nds32/nds32.c: New checking and implementation for DSP
6392 extension instructions.
6393 * config/nds32/nds32.h: Likewise.
6394 * config/nds32/nds32.md: Likewise.
6395 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6396 * config/nds32/predicates.md: Implement new predicates for DSP
6399 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6401 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6402 Reformat alternatives and attributes so it is easier to identify
6403 which constraints/attributes go with which instruction.
6404 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6405 (mov<mode>_softfloat32, FMOVE64): Likewise.
6406 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6407 (mov<mode>_softfloat64, FMOVE64): Likewise.
6409 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6411 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6413 (Basic PowerPC Built-in Functions): The new name of the
6414 subsection previously known as "PowerPC Built-in Functions".
6415 (Basic PowerPC Built-in Functions Available on all Configurations):
6417 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6418 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6419 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6420 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6422 2018-05-11 Martin Jambor <mjambor@suse.cz>
6425 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6428 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6431 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6433 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6435 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6436 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6437 (ix86_handle_option): Handle -mwaitpkg.
6438 * config.gcc: New header.
6439 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6440 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6441 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6443 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6444 OPTION_MASK_ISA_WAITPKG.
6445 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6446 (ix86_option_override_internal): Add PTA_WAITPKG.
6447 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6448 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6449 IX86_BUILTIN_TPAUSE.
6450 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6451 __builtin_ia32_umwait and __builtin_ia32_tpause.
6452 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6453 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6454 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6455 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6456 UNSPECV_TPAUSE): New.
6457 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6458 * config/i386/i386.opt: Add -mwaitpkg.
6459 * config/i386/waitpkgintrin.h: New file.
6460 * config/i386/x86intrin.h: New header.
6461 * doc/invoke.texi: Add -mwaitpkg.
6463 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6466 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6468 (cortex-m0): Use armv6s-m isa.
6469 (cortex-m0plus): Likewise.
6470 (cortex-m1): Likewise.
6471 (cortex-m0.small-multiply): Likewise.
6472 (cortex-m0plus.small-multiply): Likewise.
6473 (cortex-m1.small-multiply): Likewise.
6475 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6476 Jakub Jelinek <jakub@redhat.com>
6478 PR tree-optimization/85692
6479 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6480 source permute as well.
6482 2018-05-11 Martin Liska <mliska@suse.cz>
6485 * doc/extend.texi: Document LLVM style format for no_sanitize
6488 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6490 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6491 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6492 (mode_supports_vsx_dform_quad): Likewise.
6493 (mode_supports_vmx_dform): Move these functions to be next to the
6494 other mode_supports functions.
6495 (mode_supports_dq_form): Likewise.
6496 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6497 mode_supports_dq_form.
6498 (reg_offset_addressing_ok_p): Likewise.
6499 (offsettable_ok_by_alignment): Likewise.
6500 (rs6000_legitimate_offset_address_p): Likewise.
6501 (legitimate_lo_sum_address_p): Likewise.
6502 (rs6000_legitimize_address): Likewise.
6503 (rs6000_legitimize_reload_address): Likewise.
6504 (rs6000_secondary_reload_inner): Likewise.
6505 (rs6000_preferred_reload_class): Likewise.
6506 (rs6000_output_move_128bit): Likewise.
6508 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6510 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6511 Generate SImode target register for null target.
6512 <case IX86_BUILTIN_XGETBV>: Ditto.
6513 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6514 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6516 2018-05-10 Carl Love <cel@us.ibm.com>
6518 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6519 dcbtt and dcbtstt if operands[2] is 0.
6521 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6524 * config/i386/sse.md (usadv64qi): New expander.
6526 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6528 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6529 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6530 -maltivec=be support.
6531 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6532 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6533 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6534 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6535 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6536 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6537 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6538 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6539 altivec_vsumsws): Adjust.
6540 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6541 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6542 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6544 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6545 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6546 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6547 (altivec_lve<VI_char>x): Delete expand.
6548 (*altivec_lve<VI_char>x_internal): Rename to...
6549 (altivec_lve<VI_char>x): ... this.
6550 (altivec_lvxl_<mode>): Delete expand.
6551 (*altivec_lvxl_<mode>_internal): Rename to ...
6552 (altivec_lvxl_<mode>): ... this.
6553 (altivec_stvxl_<mode>): Delete expand.
6554 (*altivec_stvxl_<mode>_internal): Rename to ...
6555 (altivec_stvxl_<mode>): ... this.
6556 (altivec_stve<VI_char>x): Delete expand.
6557 (*altivec_stve<VI_char>x_internal): Rename to ...
6558 (altivec_stve<VI_char>x): ... this.
6559 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6560 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6561 reduc_plus_scal_<mode>): Adjust.
6562 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6564 (rs6000_cpu_cpp_builtins): Adjust.
6565 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6566 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6567 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6568 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6569 -maltivec=be support.
6570 (rs6000_split_vec_extract_var): Adjust.
6571 (rs6000_split_v4si_init): Adjust.
6572 (swap_selector_for_mode): Delete.
6573 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6574 altivec_expand_stvex_be): Delete.
6575 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6576 -maltivec=be support.
6577 (rs6000_gimple_fold_builtin): Ditto.
6578 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6580 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6581 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6582 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6583 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6584 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6585 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6586 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6587 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6588 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6589 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6590 anonymous split): Adjust.
6591 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6592 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6594 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6596 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6597 when --with-gxx-include-dir is also specified.
6598 * configure: Regenerate.
6600 2018-05-09 Jim Wilson <jimw@sifive.com>
6603 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6604 * config/riscv/t-withmultilib: New.
6605 * config/riscv/withmultilib.h: New.
6606 * doc/install.texi: Document RISC-V --with-multilib-list support.
6608 2018-05-09 Richard Biener <rguenther@suse.de>
6610 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6612 (vect_bb_vectorization_profitable_p): Adjust. Compute
6613 actual scalar cost using the cost vector and the add_stmt_cost
6616 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6618 PR rtl-optimization/85645
6619 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6620 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6622 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6624 PR rtl-optimization/85645
6625 * shrink-wrap.c (spread_components): Return a boolean saying if
6626 anything was changed.
6627 (try_shrink_wrapping_separate): Iterate spread_components until
6628 nothing changes anymore.
6630 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6632 PR rtl-optimization/85645
6633 * regrename.c (build_def_use): Also kill the chains that include the
6634 destination of a REG_CFA_REGISTER note.
6636 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6638 PR rtl-optimization/85645
6639 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6640 insn that has a REG_CFA_REGISTER note.
6642 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6644 * cfgexpand.c (expand_clobber): New function.
6645 (expand_gimple_stmt_1): Use it.
6646 * tree-vect-stmts.c (vect_clobber_variable): New function,
6648 (vectorizable_simd_clone_call): ...here.
6649 (vectorizable_store): Emit a clobber either side of an
6650 IFN_STORE_LANES sequence.
6651 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6653 2018-05-09 Tom de Vries <tom@codesourcery.com>
6656 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6657 (define_insn "trap_if_false"): Add exit after trap.
6659 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6661 PR rtl-optimization/85638
6662 * bb-reorder.c: Include common/common-target.h.
6663 (create_forwarder_block): New function extracted from...
6664 (fix_up_crossing_landing_pad): ...here. Rename into...
6665 (dw2_fix_up_crossing_landing_pad): ...this.
6666 (sjlj_fix_up_crossing_landing_pad): New function.
6667 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6668 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6669 from both partitions and exit the loop after one iteration.
6671 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6674 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6676 (Basic PowerPC Built-in Functions): The new name of the
6677 subsection previously known as "PowerPC Built-in Functions".
6678 (Basic PowerPC Built-in Functions Available on all Configurations):
6680 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6682 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6684 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6686 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6689 2018-05-08 Jim Wilson <jimw@sifive.com>
6691 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6692 (LD_EMUL_SUFFIX): New.
6693 (LINK_SPEC): Use it.
6695 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6697 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6699 (Basic PowerPC Built-in Functions): The new name of the
6700 subsection previously known as "PowerPC Built-in Functions".
6701 (Basic PowerPC Built-in Functions Available on all Configurations):
6703 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6705 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6707 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6709 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6712 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6715 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6716 after cmpelim optimization.
6718 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6720 * config.gcc: Support "goldmont".
6721 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6722 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6724 * config/i386/i386.c (m_GOLDMONT): Define.
6725 (processor_target_table): Add "goldmont".
6726 (PTA_GOLDMONT): Define.
6727 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6728 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6729 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6730 (fold_builtin_cpu): Add "goldmont".
6731 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6732 (ix86_option_override_internal): Add "goldmont".
6733 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6734 (processor_type): Add PROCESSOR_GOLDMONT.
6735 * config/i386/i386.md: Add CPU "glm".
6736 * config/i386/glm.md: New file.
6737 * config/i386/x86-tune.def: Add m_GOLDMONT.
6738 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6740 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6743 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6745 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6746 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6747 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6750 * config/i386/i386.c (ix86_fold_builtin): Handle
6751 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6754 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6755 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6757 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6760 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6761 (check_arch): Likewise.
6762 (check_fpu): Return the result rather than printing it.
6763 (end arch): Fix operator precedence.
6764 (end cpu): Likewise.
6765 (END): Print the result from check_fpu.
6767 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6768 Alan Hayward <alan.hayward@arm.com>
6769 David Sherwood <david.sherwood@arm.com>
6771 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6772 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6773 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6774 (*fcmuo<mode>_and): New patterns.
6776 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6778 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6779 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6780 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6781 (cmp_op, sve_imm_con): New code attributes.
6782 (SVE_COND_INT_CMP, imm_con): Delete.
6783 (cmp_op): Remove above unspecs from int attribute.
6784 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6786 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6787 comparison-specific unspecs.
6788 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6789 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6790 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6791 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6792 (*vec_fcm<cmp_op><mode>): Rename to...
6793 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6794 (*vec_fcmuo<mode>): Rename to...
6795 (*fcmuo<mode>): ...this and adjust likewise.
6796 (*pred_fcm<cmp_op><mode>): New pattern.
6797 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6798 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6800 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6802 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6803 (aarch64_emit_sve_predicated_cond): New function.
6804 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6805 (aarch64_emit_unspec_cond_or): Replace with...
6806 (aarch64_emit_sve_or_conds): ...this new function. Use
6807 aarch64_emit_sve_ptrue_op for the individual comparisons and
6808 aarch64_emit_binop to OR them together.
6809 (aarch64_emit_inverted_unspec_cond): Replace with...
6810 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6811 aarch64_emit_sve_ptrue_op for the comparison and
6812 aarch64_emit_unop to invert the result.
6813 (aarch64_expand_sve_vec_cmp_float): Update after the above
6814 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6816 2018-05-07 Nathan Sidwell <nathan@acm.org>
6818 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6819 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6820 (Backwards Compatibility): Likewise.
6822 2018-05-07 Luis Machado <luis.machado@linaro.org>
6826 2018-05-07 Luis Machado <luis.machado@linaro.org>
6828 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6829 <prefetch_dynamic_strides>: New const bool field.
6830 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6831 prefetch_dynamic_strides.
6832 (exynosm1_prefetch_tune): Likewise.
6833 (thunderxt88_prefetch_tune): Likewise.
6834 (thunderx_prefetch_tune): Likewise.
6835 (thunderx2t99_prefetch_tune): Likewise.
6836 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6838 (aarch64_override_options_internal): Update to set
6839 PARAM_PREFETCH_DYNAMIC_STRIDES.
6840 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6841 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6842 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6843 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6844 prefetch-dynamic-strides setting.
6846 2018-05-07 Luis Machado <luis.machado@linaro.org>
6848 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6849 <minimum_stride>: New const int field.
6850 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6851 minimum_stride field.
6852 (exynosm1_prefetch_tune): Likewise.
6853 (thunderxt88_prefetch_tune): Likewise.
6854 (thunderx_prefetch_tune): Likewise.
6855 (thunderx2t99_prefetch_tune): Likewise.
6856 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6857 (aarch64_override_options_internal): Update to set
6858 PARAM_PREFETCH_MINIMUM_STRIDE.
6859 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6860 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6861 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6862 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6863 stride is constant and is below the minimum stride threshold.
6865 2018-05-07 Luis Machado <luis.machado@linaro.org>
6867 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6870 2018-05-07 Luis Machado <luis.machado@linaro.org>
6872 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6873 <prefetch_dynamic_strides>: New const bool field.
6874 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6875 prefetch_dynamic_strides.
6876 (exynosm1_prefetch_tune): Likewise.
6877 (thunderxt88_prefetch_tune): Likewise.
6878 (thunderx_prefetch_tune): Likewise.
6879 (thunderx2t99_prefetch_tune): Likewise.
6880 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6882 (aarch64_override_options_internal): Update to set
6883 PARAM_PREFETCH_DYNAMIC_STRIDES.
6884 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6885 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6886 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6887 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6888 prefetch-dynamic-strides setting.
6890 2018-05-07 Luis Machado <luis.machado@linaro.org>
6892 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6893 <minimum_stride>: New const int field.
6894 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6895 minimum_stride field.
6896 (exynosm1_prefetch_tune): Likewise.
6897 (thunderxt88_prefetch_tune): Likewise.
6898 (thunderx_prefetch_tune): Likewise.
6899 (thunderx2t99_prefetch_tune): Likewise.
6900 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6901 (aarch64_override_options_internal): Update to set
6902 PARAM_PREFETCH_MINIMUM_STRIDE.
6903 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6904 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6905 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6906 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6907 stride is constant and is below the minimum stride threshold.
6909 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6912 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6913 the type is addressable. Don't force op into register if it has
6916 2018-05-05 Roland McGrath <mcgrathr@google.com>
6919 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6920 any section for which we don't know a specific type it should have,
6921 regardless of name. Previously this was done only for the exact
6922 names ".init_array", ".fini_array", and ".preinit_array".
6923 (default_elf_asm_named_section): Add comment about
6924 relationship with default_section_type_flags and SECTION_NOTYPE.
6925 (get_section): Don't consider it a type conflict if one side has
6926 SECTION_NOTYPE and the other doesn't, as long as neither has the
6927 SECTION_BSS et al used in the default_section_type_flags logic.
6929 2018-05-05 Tom de Vries <tom@codesourcery.com>
6932 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6933 (workaround_barsyncs): New function.
6934 (nvptx_reorg): Use workaround_barsyncs.
6935 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6936 (define_expand "nvptx_membar_cta"): New define_expand.
6937 (define_insn "*nvptx_membar_cta"): New insn.
6939 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6941 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6942 To improve optimization opportunities.
6943 * builtin-types.def: The new needed builtin types for the above.
6945 2018-05-04 Richard Biener <rguenther@suse.de>
6947 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6948 * gimple-ssa-store-merging.c
6949 (imm_store_chain_info::output_merged_store): Remove redundant create,
6950 release split_store vector contents on failure.
6951 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6952 scalar stmt vector on cache hit.
6954 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6956 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6958 * config.gcc (powerpc-xilinx-eabi*): Remove.
6959 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6961 (fusion_addis_mem_combo_load): Ditto.
6962 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6964 (rs6000_cpu_cpp_builtins): Ditto.
6965 * config/rs6000/rs6000-linux.c
6966 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6967 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6968 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6970 (rs6000_setup_reg_addr_masks): Ditto.
6971 (rs6000_init_hard_regno_mode_ok): Ditto.
6972 (rs6000_option_override_internal): Ditto.
6973 (legitimate_lo_sum_address_p): Ditto.
6974 (rs6000_legitimize_address): Ditto.
6975 (rs6000_legitimize_reload_address): Ditto.
6976 (rs6000_legitimate_address_p): Ditto.
6977 (abi_v4_pass_in_fpr): Ditto.
6978 (setup_incoming_varargs): Ditto.
6979 (rs6000_gimplify_va_arg): Ditto.
6980 (rs6000_split_multireg_move): Ditto.
6981 (rs6000_savres_strategy): Ditto.
6982 (rs6000_emit_prologue_components): Ditto.
6983 (rs6000_emit_epilogue_components): Ditto.
6984 (rs6000_emit_prologue): Ditto.
6985 (rs6000_emit_epilogue): Ditto.
6986 (rs6000_elf_file_end): Ditto.
6987 (rs6000_function_value): Ditto.
6988 (rs6000_libcall_value): Ditto.
6989 * config/rs6000/rs6000.h: Ditto.
6990 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6991 (TARGET_MINMAX): ... this. New.
6992 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6993 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6994 (*movsi_internal1_single): Delete.
6995 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6996 mfpu=, mxilinx-fpu): Delete.
6997 * config/rs6000/singlefp.h: Delete.
6998 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6999 * config/rs6000/t-rs6000: Ditto.
7000 * config/rs6000/t-xilinx: Delete.
7001 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7002 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7003 (VStype_simple): Delete.
7004 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7005 * config/rs6000/xfpu.h: Delete.
7006 * config/rs6000/xfpu.md: Delete.
7007 * config/rs6000/xilinx.h: Delete.
7008 * config/rs6000/xilinx.opt: Delete.
7009 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7010 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7012 2018-05-04 Tom de Vries <tom@codesourcery.com>
7015 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7018 2018-05-04 Richard Biener <rguenther@suse.de>
7021 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7022 (expand_complex_div_wide): Likewise.
7023 (expand_complex_operations_1): Likewise.
7024 (expand_complex_libcall): Preserve EH info of the original stmt.
7025 (tree_lower_complex): Handle removed blocks.
7026 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7027 on complex multiplication and division libcall builtins.
7029 2018-05-04 Richard Biener <rguenther@suse.de>
7032 * fold-const.c (negate_expr_p): Restrict negation of operand
7033 zero of a division to when we know that can happen without
7035 (fold_negate_expr_1): Likewise.
7037 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7040 * real.h (real_nextafter): Declare.
7041 * real.c (real_nextafter): New function.
7042 * fold-const-call.c (fold_const_nextafter): New function.
7043 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7044 CASE_CFN_NEXTTOWARD.
7045 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7046 even when arg1_mode is different from arg0_mode.
7048 2018-05-03 Nathan Sidwell <nathan@acm.org>
7050 * doc/extend.texi (Deprecated Features): Remove
7052 (Backwards Compatibility): Likewise.
7053 * doc/invoke.texi (C++ Language Options): Likewise.
7054 (C++ Dialect Options): Likewise.
7056 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7059 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7060 _mm512_mask_mullox_epi64): New intrinsics.
7062 2018-05-03 Tom de Vries <tom@codesourcery.com>
7065 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7066 dump files): Add offload-tree.
7068 2018-05-03 Richard Biener <rguenther@suse.de>
7070 PR tree-optimization/85615
7071 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7072 to loops not nested in BBs loop father to avoid creating multi-entry
7075 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7077 PR tree-optimization/70291
7078 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7079 arguments. Change return type to tree. Emit libcall as a new
7080 statement rather than replacing existing one when inplace_p is true.
7081 (expand_complex_multiplication_components): New function.
7082 (expand_complex_multiplication): Expand floating-point complex
7083 multiplication using the above.
7084 (expand_complex_division): Rename inner_type parameter to type.
7085 Update expand_complex_libcall call-site.
7086 (expand_complex_operations_1): Update expand_complex_multiplication
7087 and expand_complex_division call-sites.
7089 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7092 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7093 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7094 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7095 the highest significant bit of the shift count mask is clear. In
7096 check whether and[sq]i3 is needed verify that all significant bits
7097 of the shift count other than the highest are set.
7099 2018-05-02 Tom de Vries <tom@codesourcery.com>
7102 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7103 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7104 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7105 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7106 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7107 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7108 __builtin_goacc_parlevel_size.
7110 2018-05-02 Richard Biener <rguenther@suse.de>
7112 PR tree-optimization/85597
7113 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7114 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7117 2018-05-02 Tom de Vries <tom@codesourcery.com>
7120 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7121 dump files): Add ltrans-tree.
7123 2018-05-02 Tom de Vries <tom@codesourcery.com>
7126 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7127 dump files): Add wpa-ipa.
7129 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7131 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7132 powerpc*-*-linux*paired* target.
7133 * config/rs6000/750cl.h: Delete.
7134 * config/rs6000/paired.h: Delete.
7135 * config/rs6000/paired.md: Delete.
7136 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7138 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7139 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7140 comment. Remove paired float support.
7141 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7142 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7144 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7145 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7147 * config/rs6000/rs6000.c: Remove paired float support.
7148 (paired_expand_vector_init, paired_expand_vector_move,
7149 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7150 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7151 paired_expand_builtin, paired_expand_predicate_builtin,
7152 paired_init_builtins): Delete.
7153 * config/rs6000/rs6000.h: Remove paired float support.
7154 * config/rs6000/rs6000.md: Remove paired float support.
7155 (move_from_CR_ov_bit): Delete.
7156 * config/rs6000/rs6000.opt (mpaired): Delete.
7157 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7158 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7160 2018-05-02 Richard Biener <rguenther@suse.de>
7163 * gimplify.c (gimplify_save_expr): When in SSA form allow
7164 SAVE_EXPRs to compute to SSA vars.
7166 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7169 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7170 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7171 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7172 clobber operands[2], instead use a new pseudo. Formatting fixes.
7174 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7176 PR tree-optimization/85586
7177 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7178 exit early for statements in the same group if the accesses are
7181 2018-05-02 Tom de Vries <tom@codesourcery.com>
7184 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7187 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7189 PR tree-optimization/85143
7190 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7192 2018-05-01 Tom de Vries <tom@codesourcery.com>
7195 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7196 not found" error message.
7198 2018-05-01 Tom de Vries <tom@codesourcery.com>
7201 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7202 * vec.c (test_ordered_remove_if): New function.
7203 (vec_c_tests): Call test_ordered_remove_if.
7204 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7205 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7206 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7207 VEC_ORDERED_REMOVE_IF.
7209 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7211 PR tree-optimization/82665
7212 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7213 pointer subtraction where arguments come from a memchr call.
7215 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7217 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7218 --push-state --as-needed and --pop-state instead of --as-needed and
7219 --no-as-needed if ld supports it.
7220 * configure: Regenerated.
7223 * doc/install.texi2html: Replace _002d with - and _002a with * in
7224 generated html files using sed.
7226 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7229 * gcc-rich-location.c (blank_line_before_p): New function.
7230 (use_new_line): New function.
7231 (gcc_rich_location::add_fixit_insert_formatted): New function.
7232 * gcc-rich-location.h
7233 (gcc_rich_location::add_fixit_insert_formatted): New function.
7235 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7237 * selftest.c (assert_streq): Rename "expected" and "actual" to
7238 "val1" and "val2". Extend NULL-handling to cover both inputs
7239 symmetrically, while still requiring both to be non-NULL for a pass.
7240 * selftest.h (assert_streq): Rename "expected" and "actual" to
7242 (ASSERT_EQ): Likewise.
7243 (ASSERT_EQ_AT): Likewise.
7244 (ASSERT_KNOWN_EQ): Likewise.
7245 (ASSERT_KNOWN_EQ_AT): Likewise.
7246 (ASSERT_NE): Likewise.
7247 (ASSERT_MAYBE_NE): Likewise.
7248 (ASSERT_MAYBE_NE_AT): Likewise.
7249 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7250 the assertion to pass.
7251 (ASSERT_STREQ_AT): Likewise.
7253 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7255 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7256 interaction with -pie.
7258 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7260 * selftest.h: Fix alphabetization of per-source-file selftest
7263 2018-04-30 Jason Merrill <jason@redhat.com>
7265 PR c++/61982 - dead stores to destroyed objects.
7266 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7269 2018-04-30 Jason Merrill <jason@redhat.com>
7271 * tree.c (build_clobber): New.
7272 * tree.h: Declare it.
7273 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7275 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7277 * diagnostic-show-locus.c (layout::layout): Update for
7278 location_get_source_line returning a char_span.
7279 (struct char_span): Move to input.h.
7280 (struct correction): Update for fields in char_span becoming
7282 (struct source_line): Update for location_get_source_line
7283 returning a char_span.
7284 (layout::print_line): Likewise.
7285 * edit-context.c (edited_file::print_content): Likewise.
7286 (edited_file::print_diff_hunk): Likewise.
7287 (edited_file::print_run_of_changed_lines): Likewise.
7288 (edited_file::get_num_lines): Likewise.
7289 (edited_line::edited_line): Likewise.
7290 * final.c (asm_show_source): Likewise.
7291 * input.c (location_get_source_line): Convert return type
7292 from const char * to char_span, losing the final "line_len"
7294 (dump_location_info): Update for the above.
7295 (get_substring_ranges_for_loc): Likewise. Use a char_span
7296 when handling the literal within the line.
7297 (test_reading_source_line): Update for location_get_source_line
7298 returning a char_span.
7299 * input.h (class char_span): Move here from
7300 diagnostic-show-locus.c, converting from a struct to a class.
7301 Make data members private.
7302 (char_span::operator bool): New.
7303 (char_span::length): New.
7304 (char_span::get_buffer): New.
7305 (char_span::operator[]): New.
7306 (char_span::subspan): Make const.
7307 (char_span::xstrdup): New.
7308 (location_get_source_line): Convert return type from const char *
7309 to char_span, losing the final "line_size" param.
7311 2018-04-30 Jan Hubicka <jh@suse.cz>
7313 * lto-wrapper.c (ltrans_priorities): New static var.
7314 (cmp_priority): New.
7315 (run_gcc): Read priorities and if doing parallel build order
7316 the Makefile by them.
7318 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7320 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7322 2018-04-30 Richard Biener <rguenther@suse.de>
7324 * tree-cfg.c (verify_address): Remove base argument, add
7325 flag whether to check TREE_ADDRESSABLE and do that.
7326 (verify_expr): Remove.
7327 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7328 (verify_gimple_assign_single): Likewise.
7329 (verify_gimple_switch): Likewise.
7330 (verify_expr_location_1): Dereference tp once. Add (disabled)
7331 piece from verify_expr.
7332 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7334 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7336 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7338 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7340 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7341 (small_data_pattern): Likewise.
7342 (arc_rewrite_small_data): Likewise.
7343 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7344 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7345 (get_symbol_alignment): New function.
7346 (legitimate_small_data_address_p): Likewise.
7347 (legitimate_scaled_address): Update, call
7348 legitimate_small_data_address_p.
7349 (output_sdata): New static variable.
7350 (arc_print_operand): Update how we handle small data operands.
7351 (arc_print_operand_address): Likewise.
7352 (arc_legitimate_address_p): Update, use
7353 legitimate_small_data_address_p.
7354 (arc_rewrite_small_data_p): Remove.
7355 (arc_rewrite_small_data_1): Likewise.
7356 (arc_rewrite_small_data): Likewise.
7357 (small_data_pattern): Likewise.
7358 (compact_sda_memory_operand): Update to use
7359 legitimate_small_data_address_p and get_symbol_alignment.
7360 (prepare_move_operands): Don't rewite sdata pattern.
7361 (prepare_extend_operands): Remove.
7362 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7364 (zero_extendqisi2): Likewise.
7365 (zero_extendhisi2): Likewise.
7366 (extendqihi2): Likewise.
7367 (extendqisi2): Likewise.
7368 (extendhisi2): Likewise.
7372 * config/arc/constraints.md (Usd): Change it to memory constraint.
7374 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7376 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7377 as source of std instructions.
7378 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7379 allow 6-bit constants as source for store instructions.
7380 (movdi_insn): Update instruction pattern to allow 6-bit constants
7381 as source for store instructions.
7383 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7385 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7387 2018-04-30 Nathan Sidwell <nathan@acm.org>
7388 Sandra Loosemore <sandra@codesourcery.com>
7390 * dumpfile.c (dump_open): Allow '-' for stdout.
7391 * doc/invoke.texi (Developer Options): Document dump filename
7392 determination early. Document stdin/stdout selection.
7394 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7396 Microblaze Target: PIC data text relative
7398 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7399 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7401 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7402 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7403 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7404 New addressing mode for data-text relative position indepenedent code.
7405 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7406 'ADDRESS_SYMBOLIC_TXT_REL'.
7407 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7408 (microblaze_legitimate_pic_operand): Exclude function calls from
7409 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7410 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7412 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7413 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7414 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7415 for 'address + offset'.
7416 (microblaze_expand_prologue): Add new function prologue call for
7418 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7419 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7420 table in case of TARGET_PIC_DATA_TEXT_REL.
7421 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7422 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7423 Add new macros 'UNSPEC_TEXT',
7424 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7425 + exclude function calls from 'UNSPEC_PLT' in case of data text
7427 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7428 new target hook for generating address diff vector tables in case of
7430 * doc/tm.texi : Regenerate.
7431 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7432 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7433 of addr diff vector generation.
7434 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7435 target hook definition.
7436 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7437 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7438 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7439 option for data text relative.
7441 2018-04-30 Richard Biener <rguenther@suse.de>
7443 * tree-chrec.h (evolution_function_is_constant_p): Remove
7445 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7447 2018-04-30 Richard Biener <rguenther@suse.de>
7450 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7452 2018-04-30 Richard Biener <rguenther@suse.de>
7454 PR tree-optimization/28364
7455 PR tree-optimization/85275
7456 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7457 copying first exit test.
7459 2018-04-28 Mark Wielaard <mark@klomp.org>
7461 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7463 (dwarf_AT): Handle DW_AT_addr_base.
7464 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7466 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7469 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7470 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7471 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7472 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7474 2018-04-28 Richard Biener <rguenther@suse.de>
7476 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7477 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7478 to reflect use. Only add interesting stmts.
7480 2018-04-27 Martin Jambor <mjambor@suse.cz>
7483 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7484 the jump function allows for passing through aggregate values.
7486 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7488 * input.h (in_system_header_at): Convert from macro to inline
7490 (from_macro_expansion_at): Likewise.
7491 (from_macro_definition_at): Likewise.
7493 2018-04-27 Jeff Law <law@redhat.com>
7495 * config.gcc: Mark tile* targets as deprecated/obsolete.
7497 2018-04-27 Richard Biener <rguenther@suse.de>
7499 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7502 2018-04-27 Richard Biener <rguenther@suse.de>
7504 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7506 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7508 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7509 with Yd constraint. Set "preferred_for_speed" attribute from
7510 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7512 (*movdi_internal): Ditto.
7513 (movti_interunit splitters): Remove
7514 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7515 (movdi_interunit splitters): Ditto.
7516 * config/i386/constraints.md (Ye): Remove.
7517 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7519 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7522 * config/aarch64/constraints.md (Usg): Limit to 31.
7525 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7527 PR tree-optimization/85529
7528 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7529 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7530 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7531 zero extension or masking of the MSB bit.
7532 (optimize_range_tests): Add FIRST_BB argument, pass it through
7533 to optimize_range_tests_var_bound.
7534 (maybe_optimize_range_tests, reassociate_bb): Adjust
7535 optimize_range_tests callers.
7537 2018-04-26 Richard Biener <rguenther@suse.de>
7538 Jakub Jelinek <jakub@redhat.com>
7540 * cgraph.h (symbol_table): Just declare debug method here.
7541 * symtab.c (symbol_table::debug): Define.
7543 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7545 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7547 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7549 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7550 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7551 (*movdi_internal): Substitute Yi and Yj constraint with x
7552 and Ym and Yn constraint with y constraint. Update "isa"
7553 attribute and set "preferred_for_speed" attribute from
7554 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7555 (*movsi_internal): Ditto.
7556 (*movdf_internal): Ditto.
7557 (*movsf_internal): Ditto.
7558 (*zero_extendsidi2): Ditto.
7559 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7560 (sse2_loadld): Ditto.
7561 (*vec_extract<ssevecmodelower>_0): Ditto.
7562 (*vec_extractv4si_0_zext_sse4): Ditto.
7563 (vec_concatv2di): Ditto.
7564 (*vec_dup<mode>): Ditto.
7565 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7566 * config/i386/constraints.md (Yi): Remove.
7571 2018-04-26 Nathan Sidwell <nathan@acm.org>
7573 * dumpfile.c (dump_open): New.
7574 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7575 (dump_finish): Detect stdio/stderr by value not name.
7577 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7579 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7581 2018-04-26 Tom de Vries <tom@codesourcery.com>
7584 * config/nvptx/nvptx.c (verify_neutering_jumps)
7585 (verify_neutering_labels): New function
7586 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7588 2018-04-26 Tom de Vries <tom@codesourcery.com>
7591 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7592 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7595 2018-04-26 Richard Biener <rguenther@suse.de>
7596 Tom de Vries <tom@codesourcery.com>
7599 * lto-streamer-out.c (output_function): Fixup loops if required to match
7600 discovery done in the reader.
7602 2018-04-26 Richard Biener <rguenther@suse.de>
7604 PR tree-optimization/85116
7605 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7606 have a loop exit from the single latch predecessor. Remove
7607 case of header with just condition.
7608 (ch_base::copy_headers): Exclude infinite loops from any
7610 (pass_ch::execute): Record exits.
7612 2018-04-26 Richard Biener <rguenther@suse.de>
7614 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7615 prologue cost vector and pass it to vect_get_load_cost.
7616 (vect_get_peeling_costs_all_drs): Likewise.
7617 (vect_peeling_hash_get_lowest_cost): Likewise.
7618 (vect_enhance_data_refs_alignment): Likewise.
7620 2018-04-26 Richard Biener <rguenther@suse.de>
7623 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7624 checking of integer<->pointer conversions.
7625 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7626 sign-/zero-extending pointer types.
7627 (expand_omp_for_static_chunk): Likewise.
7629 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7630 Jean Lee <xiaoyur347@gmail.com>
7632 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7633 (TARGET_ASAN_SHADOW_OFFSET): Define.
7634 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7635 true for -fsanitize=address.
7637 2018-04-25 Mark Wielaard <mark@klomp.org>
7639 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7642 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7644 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7645 than "alu", remove explicit "memory" and "imm_disp" attributes.
7646 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7649 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7650 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7651 gen_lowpart_no_emit.
7653 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7656 * config/i386/i386.c (ix86_expand_builtin): Change memory
7657 operand to XI, extend p0 to Pmode.
7658 * config/i386/i386.md: Change unspec volatile and operand
7659 1 mode to XI, change operand 0 mode to P.
7661 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7663 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7664 GET_MODE_MASK before any checking.
7665 (nds32_can_use_bset_p): Likewise.
7666 (nds32_can_use_btgl_p): Likewise.
7668 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7670 * config/nds32/nds32-doubleword.md: New define_split pattern for
7671 illegal register number.
7673 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7675 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7677 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7679 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7681 2018-04-25 Richard Biener <rguenther@suse.de>
7683 * lto-streamer.h (LTO_major_version): Bump to 8.
7685 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7687 * BASE-VER: Set to 9.0.0.
7689 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7691 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7692 in __abskf2 and __powikf2.
7694 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7697 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7698 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7699 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7700 Use the above on operand 2. Reindent.
7701 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7703 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7706 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7708 * config/i386/i386.opt (mcet): Removed.
7709 * doc/install.texi: Remove -mcet documentation.
7710 * doc/invoke.texi: Likewise.
7712 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7715 * doc/install.texi: Remove -mcet from bootstrap-cet.
7717 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7720 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7721 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7725 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7726 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7727 containing a CONST_VECTOR.
7729 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7731 * doc/install.texi: Update newlib dependency for nvptx.
7733 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7736 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7737 instead of INTVAL when shifting x left.
7739 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7741 PR tree-optimization/85478
7742 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7743 vect_grouped_store_supported for single element vectors.
7745 2018-04-24 Richard Biener <rguenther@suse.de>
7748 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7749 load cost increase to the case of non-constant step.
7751 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7754 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7755 destination if any_malformed_asm.
7757 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7760 * expr.c (store_field): In the bitfield case, if the value comes from
7761 a function call and is returned in registers by means of a PARALLEL,
7762 do not change the mode of the temporary unless BLKmode and VOIDmode.
7764 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7766 PR rtl-optimization/85423
7767 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7768 dependencies to debug insns when the previous insn is non-debug.
7770 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7772 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7773 enums into a single definition.
7774 (fls): Fix predicates and printing.
7777 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7779 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7780 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7781 and short u6 immediate.
7782 (check_if_valid_sleep_operand): Remove.
7783 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7785 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7787 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7788 flag_always_save_lp condition.
7789 * config/nds32/nds32.opt (malways-save-lp): New option.
7791 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7793 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7794 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7795 * config/nds32/nds32.h
7796 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7797 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7799 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7801 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7802 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7804 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7805 Chung-Ju Wu <jasonwucj@gmail.com>
7807 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7809 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7810 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7812 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7814 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7816 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7818 * config/nds32/nds32-protos.h (nds32_data_alignment,
7819 nds32_local_alignment): Declare.
7820 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7821 nds32_local_alignment): New functions.
7822 (TARGET_CONSTANT_ALIGNMENT): Define.
7823 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7825 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7827 * config/nds32/nds32.c
7828 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7829 (TARGET_MODES_TIEABLE_P): Likewise.
7831 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7833 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7836 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7837 Chung-Ju Wu <jasonwucj@gmail.com>
7839 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7840 for unaligned access.
7841 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7843 * config/nds32/nds32-intrinsic.md: Likewise.
7844 * config/nds32/nds32_intrinsic.h: Likewise.
7845 * config/nds32/nds32.h (nds32_builtins): Likewise.
7846 * config/nds32/nds32.opt (munaligned-access): New option.
7847 * config/nds32/nds32.c (nds32_asm_file_start): Display
7848 flag_unaligned_access status.
7850 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7852 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7853 -mno-relax is present.
7854 * config/riscv/linux.h (LINK_SPEC): Ditto.
7856 2018-04-20 Martin Sebor <msebor@redhat.com>
7859 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7861 (gimple_fold_builtin_stxcpy_chk): Same.
7862 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7864 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7867 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7868 __powikf2 when long double is IEEE 128-bit.
7870 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7872 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7873 step to make sure stack always aligned.
7875 2018-04-20 Carl Love <cel@us.ibm.com>
7878 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7879 size check for arg0.
7881 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7882 Tom de Vries <tom@codesourcery.com>
7885 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7886 Emit insns for calls too.
7887 (nvptx_find_par): Always look for worker-level predecessor insn.
7888 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7890 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7891 (nvptx_process_pars): Propagate frames for calls.
7893 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7896 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7898 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7899 (ix86_handle_option): Don't handle OPT_mibt.
7900 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7902 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7904 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7905 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7906 (ix86_target_macros): Define __CET__ with flag_cf_protection
7907 for -fcf-protection.
7908 * config/i386/i386.c (isa2_opts): Remove -mibt.
7909 * config/i386/i386.h (TARGET_IBT): Removed.
7910 (TARGET_IBT_P): Likewise.
7911 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7912 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7913 * config/i386/i386.opt (mcet): Update help message.
7916 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7917 -mcet as an alias for -mshstk.
7919 2018-04-20 Richard Biener <rguenther@suse.de>
7922 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7923 complexity by forcing a single use of the multiply operand.
7925 2018-04-20 Martin Jambor <mjambor@suse.cz>
7928 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7929 recursion dependency to only apply to non-clones.
7931 2018-04-20 Martin Jambor <mjambor@suse.cz>
7934 * ipa-cp.c (create_specialized_node): Check that clones of
7935 self-recursive edges exist during IPA-CP.
7937 2018-04-19 Toon Moene <toon@moene.org>
7939 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7942 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7944 PR tree-optimization/85467
7945 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7946 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7947 VECTOR_CST element to type.
7949 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7952 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7953 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7954 (builtin_longjmp): Likewise.
7955 (save_stack_nonlocal): New pattern.
7956 (restore_stack_nonlocal): Likewise.
7958 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7961 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7962 Replace ASM_OUTPUT_LABEL with fprintf.
7964 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7967 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7968 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7969 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7970 define __IBT__ and __SHSTK__ for -fcf-protection.
7971 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7973 (ix86_trampoline_init): Likewise.
7974 (x86_output_mi_thunk): Likewise.
7975 (ix86_notrack_prefixed_insn_p): Likewise.
7976 (ix86_option_override_internal): Don't disallow -fcf-protection.
7977 * config/i386/i386.md (rdssp<mode>): Also enable for
7979 (incssp<mode>): Likewise.
7980 (nop_endbr): Likewise.
7981 * config/i386/i386.opt (mcet): Change help message to built-in
7985 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7986 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7987 enable CET built-in functions.
7989 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7991 * common/config/i386/i386-common.c
7992 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7993 OPTION_MASK_ISA_MOVDIRI_UNSET,
7994 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7995 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7996 * config.gcc (movdirintrin.h): New header.
7997 * config/i386/cpuid.h (bit_MOVDIRI,
7998 bit_MOVDIR64B): New bits.
7999 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8001 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8002 (VOID, PVOID, PCVOID)): New function types.
8003 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8004 __builtin_ia32_directstoreu_u64,
8005 __builtin_ia32_movdir64b): New builtins.
8006 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8007 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8009 (ix86_valid_target_attribute_inner_p): Ditto.
8010 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8011 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8012 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8013 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8014 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8015 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8016 (movdiri<mode>, movdir64b_<mode>): New.
8017 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8018 * config/i386/immintrin.h: Include movdirintrin.h.
8019 * config/i386/movdirintrin.h: New file.
8020 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8022 2018-04-19 Richard Biener <rguenther@suse.de>
8025 * cfg.c (clear_bb_flags): When loop state says we have
8026 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8028 2018-04-19 Richard Biener <rguenther@suse.de>
8030 PR tree-optimization/84737
8031 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8032 copying restrict info.
8033 (vect_setup_realignment): Use it.
8034 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8035 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8036 the first DR to all generated stores.
8037 (vectorizable_load): Likewise for loads.
8039 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8041 PR tree-optimization/85446
8042 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8043 the integral and pointer types to have the same precision.
8045 * doc/install.texi: Document --disable-cet being the default and
8048 2018-04-18 Martin Liska <mliska@suse.cz>
8050 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8053 2018-04-18 Martin Liska <mliska@suse.cz>
8056 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8059 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8060 arguments if they are comparable.
8062 2018-04-18 Martin Liska <mliska@suse.cz>
8065 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8068 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8071 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8074 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8075 ENDBR after calling __morestack.
8077 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8080 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8081 by using gcc_base_ver to generate a gcc_driver_version, and use
8082 it when generating GCC_DRIVER_NAME.
8083 * configure: Regenerate.
8085 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8088 * config.gcc: Obsolete powerpc*-*-*spe*.
8090 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8093 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8094 (stabstr_D): Change type of unum from unsigned int to
8095 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8098 2018-04-17 Jim Wilson <jimw@sifive.com>
8101 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8102 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8103 Set arg_pointer_offset after using pretend_args_size.
8105 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8107 PR rtl-optimization/85431
8108 * dse.c (record_store): Ignore zero width stores.
8111 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8112 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8113 __builtin_stack_restore rather than after it.
8114 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8115 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8116 argument instead of virtual_dynamic_stack_rtx.
8118 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8120 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8122 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8123 Add note to error message to explain internal mapping of overloaded
8124 built-in function name to non-overloaded built-in function name.
8125 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8128 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8131 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8132 where the inputs overlap with the output.
8134 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8136 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8137 (=v, v) alternative and explicit "memory" attribute.
8138 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8139 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8141 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8142 "sselog1" type instead of "sselog".
8143 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8144 "sselog". Remove explicit "memory" attribute.
8145 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8146 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8148 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8149 "sselog1" type instead of "sselog". Remove explicit "memory"
8151 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8152 use "sselog1" type instead of "sselog". Remove explicit "memory"
8154 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8155 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8157 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8158 "sselog1" type instead of "sselog". Remove explicit "memory"
8160 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8161 use "sselog1" type instead of "sselog". Remove explicit "memory"
8165 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8168 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8171 2018-04-17 Martin Jambor <mjambor@suse.cz>
8174 * ipa-cp.c (create_specialized_node): Call
8175 expand_all_artificial_thunks if necessary.
8177 2018-04-17 Martin Liska <mliska@suse.cz>
8180 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8181 in message, remote space in between '_G' and '('.
8183 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8186 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8187 avx512f_vmcmp<mode>3<round_saeonly_name>,
8188 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8189 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8190 avx512f_rndscale<mode><round_saeonly_name>,
8191 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8192 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8193 Use %<iptr>2 instead of %2 for -masm=intel.
8194 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8195 avx512f_vcvttss2usi<round_saeonly_name>,
8196 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8198 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8199 avx512f_vcvttsd2usi<round_saeonly_name>,
8200 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8201 Use %q1 instead of %1 for -masm=intel.
8202 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8203 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8204 of %3 for -masm=intel.
8205 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8207 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8209 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8211 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8212 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8214 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8216 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8217 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8218 %g1 and one with %0 and %1.
8219 (avx512er_vmrcp28<mode><round_saeonly_name>,
8220 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8222 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8223 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8224 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8225 of %0 and %{%4%} for -masm=intel.
8226 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8227 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8228 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8229 order of %0 and %{%5%}%{z%} for -masm=intel.
8231 2018-04-17 Jan Hubicka <jh@suse.cz>
8234 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8236 2018-04-17 Martin Liska <mliska@suse.cz>
8239 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8240 for target_clone error message. Make default implementation
8241 clone to be a local declaration.
8242 (separate_attrs): Add new argument and check for an empty
8244 (expand_target_clones): Handle it.
8245 (ipa_target_clone): Make redirection just for target_clones
8248 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8249 Tom de Vries <tom@codesourcery.com>
8252 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8253 tiled basic blocks without omp continue statements.
8255 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8258 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8259 vec_extract expression as having side effects to make sure it gets
8262 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8265 * config/i386/i386.c (get_builtin_code_for_version): Check
8268 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8271 * gcc/config.gcc: Support "skylake".
8272 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8274 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8275 (processor_target_table): Add "skylake".
8276 (ix86_option_override_internal): Add "skylake".
8277 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8278 PROCESSOR_CANNONLAKE.
8279 (get_builtin_code_for_version): Fix priority for
8280 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8281 PROCESSOR_SKYLAKE-AVX512.
8282 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8283 (processor_type): Add PROCESSOR_SKYLAKE.
8285 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8286 Jason Merrill <jason@redhat.com>
8289 * convert.c (convert_to_integer_1): Use direct recursion for
8290 enumeral types and types with a precision less than the number
8291 of bits in their mode.
8293 2018-04-16 Julia Koval <julia.koval@intel.com>
8296 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8297 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8299 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8302 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8303 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8304 and -mno-direct-move.
8306 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8309 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8310 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8311 Ensure negative shifts result in {0}.
8313 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8315 PR rtl-optimization/79916
8316 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8317 regs (if any) to define how to gnerate SD moves when LRA is in
8320 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8322 PR rtl-optimization/85393
8323 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8324 * except.c (expand_dw2_landing_pad_for_region): Make static.
8325 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8326 a label and unconditional jump to old_bb, rather than
8327 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8330 PR rtl-optimization/85376
8331 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8332 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8333 instead of a specific value.
8335 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8336 Bin Cheng <bin.cheng@arm.com>
8338 PR tree-optimization/82965
8339 PR tree-optimization/83991
8340 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8341 by_profile_only parameter.
8342 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8343 information if the loop was predicted to iterate too many times.
8344 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8346 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8349 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8352 2018-04-13 Martin Liska <mliska@suse.cz>
8353 Jakub Jelinek <jakub@redhat.com>
8356 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8357 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8358 * builtins.c (expand_builtin_memory_copy_args): Use
8359 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8360 handle dest_addr == pc_rtx.
8362 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8365 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8366 asked to not generate direct moves.
8367 (fix_trunc<mode>si2_stfiwx): Similar.
8368 (fix_trunc<mode>si2_internal): Similar.
8370 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8373 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8374 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8375 lookup if dest in some wider mode is known to be const0_rtx and
8376 if so, record permanent equivalence for it to be ZERO_EXTEND of
8377 the narrower mode destination.
8379 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8381 * lto-streamer-out.c (output_function): Revert 259346.
8382 * omp-expand.c (expand_oacc_for): Likewise.
8384 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8386 PR rtl-optimization/85354
8387 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8388 * sel-sched.c (sel_global_init): ... here.
8390 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8393 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8394 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8395 mode for PE-COFF targets.
8396 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8397 (i386_pe_asm_lto_end): Likewise.
8398 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8399 (TARGET_ASM_LTO_END): Likewise.
8400 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8401 (i386_pe_asm_lto_start): New function.
8402 (i386_pe_asm_lto_end): Likewise.
8404 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8405 Richard Biener <rguenther@suse.de>
8408 * lto-streamer-out.c (output_function): Fix CFG loop state before
8410 * omp-expand.c (expand_oacc_for): Handle calls to internal
8411 functions like regular functions.
8413 2018-04-12 Richard Biener <rguenther@suse.de>
8416 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8417 for the early LTO debug to properly generate references to it
8418 during DIE emission. Do not re-use that for the skeleton for
8420 (dwarf2out_early_finish): Likewise.
8422 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8425 * config/i386/sse.md
8426 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8427 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8428 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8429 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8430 and output is a reg, avoid creating invalid lowpart subreg, but
8431 instead split into a 512-bit move. Don't split if not AVX512VL,
8432 input is xmm16+ reg and output is a mem.
8433 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8434 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8435 xmm16+ reg and output is a mem.
8437 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8439 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8440 also for flag_dwarf2_cfi_asm.
8442 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8444 PR rtl-optimization/85342
8445 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8446 a bool scalar var inside of the loop instead. Don't try to update
8447 recog_data.operand after failed apply_change_group.
8449 2018-04-12 Tom de Vries <tom@codesourcery.com>
8452 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8453 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8454 array with flexible array member as array without given dimension.
8455 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8456 argument for undefined param to true.
8458 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8461 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8462 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8463 from PowerPC section.
8464 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8465 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8466 help text that is too long.
8467 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8468 help text that is too long.
8469 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8470 help text that is too long.
8472 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8474 * config/alpha/alpha.md (stack_probe_internal): Rename
8475 from "probe_stack". Update all callers.
8477 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8479 PR rtl-optimization/84566
8480 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8481 sched_macro_fuse_insns.
8483 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8486 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8487 (compute_block_dependences): ... from here.
8489 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8491 PR tree-optimization/85331
8492 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8493 from int to HOST_WIDE_INT.
8495 2018-04-11 Martin Jambor <mjambor@suse.cz>
8498 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8499 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8500 not the same as the source val.
8501 (cgraph_edge_brings_value_p): New parameter.
8502 (gather_edges_for_value): Pass destination value to
8503 cgraph_edge_brings_value_p.
8504 (perhaps_add_new_callers): Likewise.
8505 (get_info_about_necessary_edges): Likewise and exclude values brought
8506 only by self-recursive edges.
8507 (create_specialized_node): Redirect only clones of self-calling edges.
8508 (+self_recursive_pass_through_p): New function.
8509 (find_more_scalar_values_for_callers_subset): Use it.
8510 (find_aggregate_values_for_callers_subset): Likewise.
8511 (known_aggs_to_agg_replacement_list): Removed.
8512 (decide_whether_version_node): Re-calculate known constants for all
8513 remaining context clones.
8515 2018-04-11 Richard Biener <rguenther@suse.de>
8518 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8519 from early DWARF output.
8520 (dwarf2out_early_finish): Output line info unconditionally into
8521 early DWARF and add reference to it.
8523 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8526 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8527 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8528 other than V2DFmode using iptr mode attribute.
8529 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8531 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8533 PR rtl-optimization/84659
8534 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8536 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8539 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8541 (output_loc_list): Pass address of a dummy size variable even in the
8542 locview handling loop.
8543 (index_location_lists): Add comment on why skip_loc_list_entry can't
8546 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8549 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8552 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8555 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8556 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8557 and -mstring-compare-inline-limit.
8559 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8562 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8563 for stack clash protection in a register whenever we need it to be in
8566 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8568 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8569 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8571 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8574 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8576 (mlong-double-): Ditto.
8577 * config/rs6000/sysv4.opt (msdata=): Ditto.
8578 (mtls-size=): Ditto.
8580 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8582 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8583 erroneous entries for
8584 "vector int vec_ldl (int, long int *)", and
8585 "vector unsigned int vec_ldl (int, unsigned long int *)".
8586 Add comments and entries for
8587 "vector bool char vec_ldl (int, bool char *)",
8588 "vector bool short vec_ldl (int, bool short *)",
8589 "vector bool int vec_ldl (int, bool int *)",
8590 "vector bool long long vec_ldl (int, bool long long *)",
8591 "vector pixel vec_ldl (int, pixel *)",
8592 "vector long long vec_ldl (int, long long *)",
8593 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8594 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8595 type tree bool_long_long_type_node and correct definition of
8596 bool_V2DI_type_node to make reference to this new type tree.
8597 (rs6000_mangle_type): Replace erroneous reference to
8598 bool_long_type_node with bool_long_long_type_node.
8599 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8600 comments to emphasize sign distinctions for char and int types and
8601 replace RS6000_BTI_bool_long constant with
8602 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8603 use of RS6000_BTI_pixel.
8604 (bool_long_type_node): Remove this macro definition.
8605 (bool_long_long_type_node): New macro definition
8607 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8609 PR rtl-optimization/85300
8610 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8611 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8612 simplify_unary_operation fails.
8614 2018-04-10 Martin Liska <mliska@suse.cz>
8616 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8617 cgraph_edge and ipa_ref.
8619 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8623 * config/i386/sse.md
8624 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8625 computation of the VEC_MERGE selector from mask.
8626 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8627 Fix decoding of the VEC_MERGE selector into mask.
8629 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8631 PR tree-optimization/85286
8632 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8634 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8636 * final.c (final_1): Set insn_last_address as well as
8637 insn_current_address.
8639 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8642 * explow.c (emit_stack_probe): Call validize_mem on memory location
8643 before passing it to gen_probe_stack. Create address operand and
8644 legitimize it for the probe_stack_address case.
8646 2018-04-09 Jan Hubicka <jh@suse.cz>
8649 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8650 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8651 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8653 (free_lang_data): Rebuild type inheritance graph.
8655 2018-04-09 Martin Sebor <msebor@redhat.com>
8657 * invoke.texi (-finline-small-functions): Mention other optimization
8659 (-findirect-inlining, -fpartial-inlining): Same.
8660 (-finline-functions-called-once): Same.
8661 (-freorder-blocks-and-partition): Same.
8663 2018-04-09 Jan Hubicka <jh@suse.cz>
8666 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8667 jumps; choose last target that matches the criteria (i.e.
8668 no partition changes for non-crossing jumps).
8669 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8670 support for redirecting crossing jumps to non-crossing.
8672 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8674 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8675 also for naked functions.
8677 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8679 * config/arc/arc.md (add_shift): New pattern.
8680 (add_shift2): Likewise.
8681 (sub_shift): Likewise.
8682 (sub_shift_cmp0_noout): Likewise.
8683 (compare_si_ashiftsi): Likewise.
8684 (xbfu_cmp0_noout): New combine pattern.
8685 (xbfu_cmp0"): Likewise.
8686 (movsi_set_cc_insn): Place the predicable variant first.
8687 (commutative_binary_cmp0_noout): Remove clobber.
8688 (commutative_binary_cmp0): New pattern.
8689 (noncommutative_binary_cmp0): Likewise.
8690 (noncommutative_binary_cmp0_noout): Likewise.
8691 (noncommutative_binary_comparison_result_used): Removed.
8692 (rsub_cmp0): New pattern.
8693 (rsub_cmp0_noout): Likewise.
8694 (extzvsi): Changed, keep only meaningful variants.
8695 (SQH, SEZ): New iterators.
8696 (SQH_postfix): New mode attribute.
8697 (SEZ_prefix): New code attribute.
8698 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8699 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8700 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8702 (noncommutative_operator): Check the availability of barrel
8705 2018-04-09 Richard Biener <rguenther@suse.de>
8707 PR tree-optimization/85284
8708 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8709 Only use the niter constraining form of simple_iv when the exit
8712 2018-04-09 Tom de Vries <tom@codesourcery.com>
8715 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8716 (define_expand "*memory_barrier"): New define_expand.
8717 (define_insn "memory_barrier"): New insn.
8719 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8721 PR rtl-optimization/80463
8722 PR rtl-optimization/83972
8723 PR rtl-optimization/83480
8725 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8726 correct producer for the insn.
8727 (tidy_control_flow): Fixup seqnos in case of debug insns.
8729 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8731 PR rtl-optimization/83913
8733 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8734 different sched-times when merging exprs.
8736 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8738 PR rtl-optimization/83962
8740 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8741 tidy_fallthru_edge and tidy_control_flow.
8743 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8745 PR rtl-optimization/83530
8747 * sel-sched.c (force_next_insn): New global variable.
8748 (remove_insn_for_debug): When force_next_insn is true, also leave only
8749 next insn in the ready list.
8750 (sel_sched_region): When the region wasn't scheduled, make another pass
8751 over it with force_next_insn set to 1.
8753 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8755 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8757 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8758 for interrupt control.
8759 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8760 functions for interrupt control.
8761 * config/nds32/nds32-intrinsic.md: Likewise.
8762 * config/nds32/nds32_intrinsic.h: Likewise.
8763 * config/nds32/nds32.h (nds32_builtins): Likewise.
8765 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8767 * config/nds32/nds32.c (nds32_init_machine_status,
8768 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8769 strict_aligned_p field.
8770 (nds32_expand_to_rtl_hook): New function.
8771 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8772 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8774 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8775 Chung-Ju Wu <jasonwucj@gmail.com>
8777 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8778 * config/nds32/nds32-n7.md: New file.
8779 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8780 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8782 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8783 * config/nds32/nds32.md (pipeline_model): Add n7.
8784 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8785 * config/nds32/pipelines.md: Include n7 settings.
8787 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8788 Chung-Ju Wu <jasonwucj@gmail.com>
8790 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8791 * config/nds32/nds32-e8.md: New file.
8792 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8793 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8795 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8796 * config/nds32/nds32.md (pipeline_model): Add e8.
8797 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8798 * config/nds32/pipelines.md: Include e8 settings.
8800 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8801 Chung-Ju Wu <jasonwucj@gmail.com>
8803 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8804 * config/nds32/nds32-n8.md: New file.
8805 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8806 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8808 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8809 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8810 * config/nds32/nds32.md (pipeline_model): Add n8.
8811 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8812 * config/nds32/pipelines.md: Include n8 settings.
8814 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8815 Chung-Ju Wu <jasonwucj@gmail.com>
8817 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8818 * config/nds32/nds32-n9-2r1w.md: New file.
8819 * config/nds32/nds32-n9-3r2w.md: New file.
8820 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8821 nds32_register_ports): New or modify for cpu n9.
8822 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8824 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8825 * config/nds32/nds32-utils.c: New file.
8826 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8827 TARGET_MUL_SLOW): Define.
8828 * config/nds32/nds32.md (pipeline_model): New attribute.
8829 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8830 New options that support cpu n9.
8831 * config/nds32/pipelines.md: Include n9 settings.
8832 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8834 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8836 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8837 information if necessary.
8838 (output_cond_branch_compare_zero): Likewise.
8839 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8840 (nds32_target_alignment): Refine for alignment.
8841 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8842 (FUNCTION_BOUNDARY): Modify.
8843 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8845 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8847 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8849 * config/nds32/constants.md (unspec_volatile_element): Add values for
8850 TLB operation and data prefetch.
8851 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8852 functions for TLB operation and data prefetch.
8853 * config/nds32/nds32-intrinsic.md: Likewise.
8854 * config/nds32/nds32_intrinsic.h: Likewise.
8855 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8856 (nds32_print_operand): Likewise.
8857 * config/nds32/nds32.h (nds32_builtins): Likewise.
8859 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8860 Andrew Pinski <pinsika@gcc.gnu.org>
8863 * match.pd: Use constant_boolean_node of correct type instead of
8864 boolean_true_node or boolean_false_node for simplifying
8865 pointer comparisons to zero.
8867 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8869 PR tree-optimization/80021
8870 * tree.c (verify_type_variant): Make error call in verify_variant_match
8871 translatable and remove final full stop.
8873 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8875 * config/nds32/constants.md (unspec_volatile_element): Add
8876 UNSPEC_VOLATILE_EH_RETURN.
8877 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8878 nds32_output_stack_pop): Support dwarf exception handling process.
8879 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8880 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8881 exception handling process.
8882 (nds32_compute_stack_frame): Likewise.
8883 (nds32_return_addr_rtx): Likewise.
8884 (nds32_initial_elimination_offset): Likewise.
8885 (nds32_expand_prologue): Likewise.
8886 (nds32_expand_epilogue): Likewise.
8887 (nds32_dynamic_chain_address): New function.
8888 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8890 (DYNAMIC_CHAIN_ADDRESS): Define.
8891 (EH_RETURN_DATA_REGNO): Define.
8892 (EH_RETURN_STACKADJ_RTX): Define.
8893 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8894 patterns for dwarf exception handling.
8896 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8898 * config/nds32/nds32.h: Clean up obsolete macros.
8900 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8902 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8903 Add enum values for particular instructions.
8904 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8905 particular intrinsic functions.
8906 * config/nds32/nds32-intrinsic.md: Likewise.
8907 * config/nds32/nds32_intrinsic.h: Likewise.
8908 * config/nds32/nds32.h (nds32_builtins): Likewise.
8909 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8910 (btst, ave): New patterns for particular instructions.
8912 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8914 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8915 Add enum values for atomic load/store and memory sync.
8916 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8918 * config/nds32/nds32-intrinsic.md: Likewise.
8919 * config/nds32/nds32_intrinsic.h: Likewise.
8920 * config/nds32/nds32.h (nds32_builtins): Likewise.
8922 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8924 PR tree-optimization/85257
8925 * fold-const.c (native_encode_vector): If not all elts could fit
8926 and off is -1, return 0 rather than offset.
8927 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8928 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8929 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8930 adjust buffer in native_interpret_expr call.
8932 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8934 * config/nds32/constants.md (unspec_volatile_element): Add cache
8935 control enum values.
8936 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8937 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8938 * config/nds32/nds32.c (nds32_cctl_names): New.
8939 (nds32_print_operand): Handle cache control register names.
8940 * config/nds32/nds32.h (nds32_builtins): New enum values.
8941 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8943 * config/nds32/nds32.md (type): Add mmu.
8944 * config/nds32/pipelines.md (simple_insn): Add mmu.
8946 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8948 * config/nds32/nds32.md (type): Remove call.
8949 * config/nds32/pipelines.md (simple_insn): Likewise.
8951 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8953 * config/nds32/constants.md (unspec_volatile_element): Add
8954 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8955 UNSPEC_VOLATILE_FMFCFG.
8956 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8957 description for fmfcfg and fmfcsr.
8958 (bdesc_1arg): Add fmtcsr.
8959 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8960 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8961 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8962 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8963 unspec_fmfcfg): New patterns.
8964 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8965 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8966 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8967 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8968 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8969 __nds32__fmfcfg): Define.
8971 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8973 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8974 intrinsic register names.
8975 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8976 intrinsic register enum values and macros.
8978 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8980 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8981 for load/store addressing form.
8982 (nds32_print_operand_address): Likewise.
8984 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8987 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8988 based on LABEL_REF. Remove useless assertion.
8989 (pic_address_needs_scratch): Fix formatting.
8990 (sparc_legitimize_pic_address): Minor tweaks.
8991 (sparc_delegitimize_address): Adjust assertion accordingly.
8992 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8993 into symbolic_operand.
8994 (movsi_high_pic_label_ref): Likewise.
8995 (movsi_lo_sum_pic_label_ref): Likewise.
8996 (movdi_pic_label_ref): Likewise.
8997 (movdi_high_pic_label_ref): Likewise.
8998 (movdi_lo_sum_pic_label_ref): Likewise.
9000 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9002 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9003 custom LIB_SPEC setup.
9005 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9006 Kito Cheng <kito.cheng@gmail.com>
9008 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9009 * config/riscv/freebsd.h: New.
9011 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9013 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9014 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9017 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9018 Kito Cheng <kito.cheng@gmail.com>
9020 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9021 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9022 * config/nds32/nds32-protos.h (nds32_output_call,
9023 nds32_output_return): Declare.
9024 * config/nds32/nds32.md: Refine all the call and return patterns.
9026 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9029 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9030 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9032 PR rtl-optimization/84872
9033 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9034 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9037 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9039 * expr.c (copy_blkmode_to_reg): Revert 254862.
9040 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9042 2018-04-06 Richard Biener <rguenther@suse.de>
9045 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9046 after seeing a component reference with an adjacent field. Treat
9047 refs to arrays at struct end of external decls similar to
9048 refs to unconstrained commons.
9050 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9053 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9054 look through SAVE_EXPRs with non-side-effects argument. Adjust
9056 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9059 2018-04-06 Richard Biener <rguenther@suse.de>
9062 * alias.c (find_base_term): New wrapper around find_base_term
9063 unwinding CSELIB_VAL_PTR changes.
9064 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9067 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9069 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9071 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9072 constant definitions.
9073 ("nop"): lr 0,0 -> nopr r0
9074 ("nop_lr0", "nop_lr1"): New insn definitions.
9076 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9078 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9079 NDS32_V3PUSH_AVAILABLE_P macro.
9081 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9082 Chung-Ju Wu <jasonwucj@gmail.com>
9084 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9085 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9086 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9087 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9088 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9089 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9090 * config/nds32/constraints.md: New constraints and checking for hard
9091 float configuration.
9092 * config/nds32/iterators.md: New mode iterator and attribute for hard
9093 float configuration.
9094 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9096 * config/nds32/nds32-fpu.md: New file.
9097 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9098 deal with hard float code generation.
9099 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9101 (abi_type, float_reg_number): New enum type.
9102 * config/nds32/nds32-predicates.c: New predicates for hard float.
9103 * config/nds32/nds32-protos.h: Declare functions for hard float.
9104 * config/nds32/nds32.c: Implementation for hard float configuration.
9105 * config/nds32/nds32.h: Definitions for hard float configuration.
9106 * config/nds32/nds32.md: Include hard float machine description and
9107 modify patterns for hard float configuration.
9108 * config/nds32/nds32.opt: New options for hard float configuration.
9109 * config/nds32/predicates.md: New predicates for hard float
9112 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9114 * common/config/nds32/nds32-common.c
9115 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9117 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9120 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9121 CONSTRUCTOR_ELT (ctor, ...)->value.
9123 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9126 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9128 2018-04-05 Tom de Vries <tom@codesourcery.com>
9131 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9134 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9135 Kito Cheng <kito.cheng@gmail.com>
9137 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9138 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9139 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9140 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9142 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9143 Kito Cheng <kito.cheng@gmail.com>
9145 * config/nds32/constraints.md (Ufe): New memory constraint.
9146 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9147 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9148 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9150 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9151 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9153 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9155 * config/nds32/nds32.md: Use optimize_size in the condition for
9156 alu-shift instructions.
9158 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9160 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9162 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9164 * config/nds32/nds32.md (negsi2): Refine pattern.
9166 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9167 Chung-Ju Wu <jasonwucj@gmail.com>
9169 * config/nds32/iterators.md (shift_rotate): New code iterator.
9170 (shift): New code attribute.
9171 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9172 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9173 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9174 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9175 bit-wise operations.
9176 (andsi3, *andsi3): Ditto.
9177 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9178 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9179 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9180 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9181 nds32_ior_operand, nds32_xor_operand): New predicates.
9183 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9185 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9186 (addsi3, subsi3): ... this.
9188 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9190 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9194 * config/nds32/nds32.md: Adjust indention.
9196 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9198 * config/nds32/nds32.md (feature): New attribute.
9200 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9202 * config/nds32/nds32.md (subtype): New attribute.
9204 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9207 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9208 expansion to perform a bitwise AND of the argument followed by a
9209 boolean negation of the result.
9211 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9213 PR rtl-optimization/84878
9214 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9215 the basic block. Assert the use reference is not artificial and that
9216 it has an associated insn.
9218 2018-04-04 Michael Matz <matz@suse.de>
9220 * builtins.c (compute_objsize): Pass correct operand
9221 to array_at_struct_end_p.
9223 2018-04-04 Richard Biener <rguenther@suse.de>
9226 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9227 from contexts for DINFO_LEVEL_TERSE and below.
9229 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9231 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9232 resiter_operand condition.
9233 * config/nds32/nds32.md (*move<mode>): Ditto.
9235 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9236 Monk Chiang <sh.chiang04@gmail.com>
9238 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9240 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9242 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9244 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9245 Kito Cheng <kito.cheng@gmail.com>
9247 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9248 nds32_cond_code_str, output_cond_branch,
9249 output_cond_branch_compare_zero, nds32_expand_cbranch,
9250 nds32_expand_cstore, nds32_expand_movcc,
9251 nds32_output_cbranchsi4_equality_zero,
9252 nds32_output_cbranchsi4_equality_reg,
9253 nds32_output_cbranchsi4_equality_reg_or_const_int,
9254 nds32_output_cbranchsi4_greater_less_zero: New functions.
9255 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9256 nds32_expand_cstore, nds32_expand_movcc,
9257 nds32_output_cbranchsi4_equality_zero,
9258 nds32_output_cbranchsi4_equality_reg,
9259 nds32_output_cbranchsi4_equality_reg_or_const_int,
9260 nds32_output_cbranchsi4_greater_less_zero): Declare.
9261 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9262 nds32_rimm11s_operand): New predicates.
9263 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9264 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9267 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9269 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9270 * config/nds32/nds32.md: Ditto.
9271 * config/nds32/pipelines.md: Ditto.
9273 2018-04-04 Richard Biener <rguenther@suse.de>
9275 PR tree-optimization/85168
9276 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9277 propagating abnormals.
9279 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9281 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9283 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9284 Kito Cheng <kito.cheng@gmail.com>
9286 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9287 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9288 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9289 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9290 * config/nds32/nds32.md (sibcall_internal): New.
9291 (sibcall_register): Remove.
9292 (sibcall_immediate): Remove.
9293 (sibcall_value_internal): New.
9294 (sibcall_value_register): Remove.
9295 (sibcall_value_immediate): Remove.
9296 * config/nds32/predicates.md (nds32_general_register_operand): New.
9297 (nds32_call_address_operand): New.
9299 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9301 PR rtl-optimization/85167
9302 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9303 bb_defs if *split_p, instead preinitialize it to NULL.
9305 PR tree-optimization/85156
9306 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9307 evaluating the argument multiple times.
9309 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9311 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9313 (_mm_cvtpd_ps): Likewise.
9314 (_mm_cvttpd_epi32): Likewise.
9315 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9316 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9317 vector, pixel, and bool following altivec.h include.
9319 2018-04-03 Martin Sebor <msebor@redhat.com>
9321 * doc/extend.texi (Common Function Attributes): Clarify.
9322 (const attribute): Likewise.
9323 (pure attribute): Likewise.
9325 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9328 * config/i386/i386.c (ix86_expand_vector_set): Use
9329 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9331 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9333 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9334 instructions when changing rounding bits to preserve precision bits
9335 in the x87 control word.
9337 2018-04-03 Martin Liska <mliska@suse.cz>
9339 PR tree-optimization/82491
9340 * rtl.h (strip_offset_and_add): Replace += suboffset with
9341 poly_uint64 () + suboffset.
9343 2018-03-29 Martin Liska <mliska@suse.cz>
9344 Martin Jambor <mjambor@suse.cz>
9347 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9348 param_type is not an integral or pointer type.
9350 2018-04-03 Richard Biener <rguenther@suse.de>
9352 * sese.h (recompute_all_dominators): Remove.
9354 2018-04-02 Martin Sebor <msebor@redhat.com>
9356 * doc/invoke.texi (-Wrestrict): Fix typos.
9358 2018-04-02 Jim Wilson <jimw@sifive.com>
9360 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9361 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9362 (<optab>di3, <optab>si3_extend): Likewise.
9363 (<optab>si3_mask, <optab>si3_mask_1): New.
9364 (<optab>di3_mask, <optab>di3_mask_1): New.
9365 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9366 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9367 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9369 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9371 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9374 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9376 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9377 (nds32_canonicalize_comparison): New function.
9379 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9380 Kito Cheng <kito.cheng@gmail.com>
9381 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9383 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9384 * config/nds32/constants.md (unspec_volatile_element): Add
9385 UNSPEC_VOLATILE_RELAX_GROUP.
9386 * config/nds32/nds32-relax-opt.c: New file.
9387 * config/nds32/nds32-predicates.c
9388 (nds32_symbol_load_store_p): New function.
9389 * config/nds32/nds32-protos.h
9390 (nds32_symbol_load_store_p): Declare function.
9391 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9392 * config/nds32/nds32.c
9393 (nds32_register_pass): New function to register pass.
9394 (nds32_register_passes): New function to register passes.
9395 * config/nds32/nds32.md (relax_group): New pattern.
9396 * config/nds32/nds32.opt (mrelax-hint): New option.
9397 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9399 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9401 * config/nds32/t-nds32: Modify files dependency.
9403 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9405 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9406 (PROFILE_HOOK): Define its implementation.
9408 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9410 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9411 type and 32-bit size.
9413 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9416 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9417 (V_128_256): New mode iterator.
9418 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9419 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9420 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9422 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9423 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9425 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9428 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9429 NaN inputs correctly.
9431 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9434 * config/rs6000/vsx.md (??r): New mode attribute.
9435 (*vsx_mov<mode>_64bit): Use it.
9436 (*vsx_mov<mode>_32bit): Likewise.
9438 2018-03-30 Martin Sebor <msebor@redhat.com>
9440 PR tree-optimization/84818
9441 * builtins.c (check_access): Use warning_n.
9443 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9446 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9448 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9451 2018-03-30 Julia Koval <julia.koval@intel.com>
9454 * x86-tune.def (movx, partial_reg_dependency): Enable for
9457 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9460 * lra-constraints.c (process_alt_operands): Move setting
9461 this_alternative_matches below.
9463 2018-03-29 Martin Liska <mliska@suse.cz>
9466 * doc/invoke.texi: Document how LTO works with debug info.
9467 Describe auto-load support of binutils. Mention 'x86-64'
9468 as valid option value of -march option.
9470 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9472 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9475 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9476 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9477 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9480 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9483 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9484 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9485 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9486 for RS6000_BTM_POWERPC64.
9487 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9488 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9489 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9494 2018-03-28 Carl Love <cel@us.ibm.com>
9497 2017-09-27 Carl Love <cel@us.ibm.com>
9499 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9500 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9501 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9504 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9506 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9507 instead of __vector bool.
9508 (_mm_max_pu8): Likewise.
9509 (_mm_min_pi16): Likewise.
9511 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9514 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9515 (DIVWEUO): Likewise.
9517 (DIVDEUO): Likewise.
9518 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9519 DIVWEUO and DIVDEUO.
9520 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9521 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9522 (div_extend): Likewise.
9523 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9525 (__builtin_divweuo): Likewise.
9526 (__builtin_divdeo): Likewise.
9527 (__builtin_divdeuo): Likewise.
9529 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9532 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9533 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9535 PR tree-optimization/82004
9536 * gimple-match-head.c (optimize_pow_to_exp): New function.
9537 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9538 Don't fold to exp if optimize_pow_to_exp is false.
9540 2018-03-28 Martin Liska <mliska@suse.cz>
9543 * calls.c (initialize_argument_information): Fix trailing space.
9544 * common.opt: Fix typo and provide better explanation for
9545 -fsanitize-coverage option.
9546 * config/i386/i386.opt: Fix typo.
9548 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9549 Martin Liska <mliska@suse.cz>
9552 * gimplify.c (asan_poison_variable): Don't do the check for
9553 gimplify_omp_ctxp here.
9554 (gimplify_decl_expr): Do it here.
9555 (gimplify_target_expr): Likewise.
9557 2018-03-28 Martin Liska <mliska@suse.cz>
9560 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9561 chkp_type_bounds_count if MPX is not enabled.
9563 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9565 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9567 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9570 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9571 function to create the function decl for complex long double
9572 multiply and divide for -mabi=ieeelongdouble.
9573 (init_float128_ieee): Call it.
9575 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9578 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9579 -fcf-protection=branch -mibt.
9580 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9582 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9585 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9586 arm_word_relocations.
9588 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9591 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9592 extern array declarations.
9594 2018-03-27 Richard Biener <rguenther@suse.de>
9597 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9598 explicit single_use checks.
9600 2018-03-27 Richard Biener <rguenther@suse.de>
9602 PR tree-optimization/85082
9603 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9606 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9608 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9609 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9610 Turn on fasynchronous-unwind-tables and funwind-tables.
9612 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9615 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9616 (*bmi_blsr_<mode>_ccz): Ditto.
9618 2018-03-26 Tom de Vries <tom@codesourcery.com>
9620 PR tree-optimization/85063
9621 * omp-general.c (offloading_function_p): New function. Factor out
9623 * omp-offload.c (pass_omp_target_link::gate): ... here.
9624 * omp-general.h (offloading_function_p): Declare.
9625 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9626 with attribute omp declare target for offloading functions.
9628 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9630 PR tree-optimization/84005
9631 * tree-data-ref.h (get_base_for_alignment): Declare.
9632 * tree-data-ref.c (get_base_for_alignment_1): New function.
9633 (get_base_for_alignment): Likewise.
9634 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9635 get_base_for_alignment to find a suitable base object, instead
9636 of always using drb->base_address.
9638 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9641 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9642 known size by default.
9644 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9647 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9648 and non BLKmode operands.
9650 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9653 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9654 Clean up attributes.
9656 2018-03-23 Richard Biener <rguenther@suse.de>
9659 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9660 we are going to emit early debug for LTO.
9662 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9665 * function.c (match_asm_constraints_1): Don't optimize if input
9666 doesn't satisfy general_operand predicate for output's mode.
9669 * alias.c (write_dependence_p): Don't require for x_canonicalized
9670 non-VOIDmode if x has VOIDmode.
9673 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9674 just don't try to optimize it rather than assert it never happens.
9676 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9678 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9679 macro expansions for definition of ST_INTERNAL_<mode> and
9680 LD_INTERNAL_<mode> builtins.
9681 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9683 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9685 (altivec_expand_st_builtin): Likewise.
9686 (altivec_expand_builtin): Remove calls to deleted functions.
9687 (rs6000_address_for_altivec): Delete this function.
9688 * config/rs6000/vector.md: Remove expands for
9689 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9691 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9694 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9695 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9696 re-computing once computed.
9697 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9698 (arm_init_machine_status): Initialize
9699 machine->static_chain_stack_bytes.
9701 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9704 * doc/extend.texi: Add four new prototypes for vec_ld.
9705 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9706 definitions for more logical presentation.
9707 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9708 entries for V1TI variants of __builtin_altivec_ld builtin.
9709 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9710 handling of V1TI variant of LVX icode pattern.
9711 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9712 (rs6000_gimple_fold_builtin): Likewise.
9713 (altivec_init_builtins): Add code to define
9714 __builtin_altivec_lvx_v1ti function.
9716 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9719 * function.c (match_asm_constraints_1): Don't do the optimization
9720 if input isn't a REG, SUBREG, MEM or constant.
9722 2018-03-22 Tom de Vries <tom@codesourcery.com>
9724 PR tree-optimization/84956
9725 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9726 bb_has_abnormal_pred.
9728 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9731 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9732 DECL_INITIAL (decl) to decl at the end.
9733 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9736 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9738 * doc/extend.texi (__builtin_tgmath): Document when complex
9739 integer types are treated as _Complex _Float64.
9741 2018-03-21 Tom de Vries <tom@codesourcery.com>
9743 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9745 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9747 PR tree-optimization/84960
9748 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9749 if it is ENTRY block, move them into single succ of ENTRY in that case.
9751 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9753 PR tree-optimization/84811
9754 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9755 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9756 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9757 (known_subrange_p): Update accordingly. Cast each value involved
9758 in the size comparison, rather than casting the result of the
9761 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9763 PR tree-optimization/84982
9764 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9765 by flipping the least significant bit rather than all bits from
9766 bitpos to bitpos + bitsize - 1.
9768 2018-03-21 Nathan Sidwell <nathan@acm.org>
9770 * doc/extend.texi (Deprecated Features): Remove mention of
9771 long-deleted deprecations.
9773 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9776 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9777 * configure: Regenerate.
9779 2018-03-21 Tom de Vries <tom@codesourcery.com>
9781 PR tree-optimization/83126
9782 * tree-parloops.c (num_phis): New function.
9783 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9785 2018-03-21 Nathan Sidwell <nathan@acm.org>
9787 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9788 mention anon-struct/union members and trailing attributes.
9790 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9792 PR tree-optimization/84969
9793 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9794 builtin memset partitions if they set different rhs values.
9796 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9798 PR rtl-optimization/84989
9799 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9800 VEC_DUPLICATE with scalar result mode.
9802 2018-03-21 Martin Liska <mliska@suse.cz>
9805 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9806 not intended return statement.
9808 2018-03-21 Martin Liska <mliska@suse.cz>
9811 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9812 (chkp_find_bound_slots_1): Limit number of iterations.
9814 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9817 * Minor grammar fixes for x86 options.
9819 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9822 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9823 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9825 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9828 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9829 (altivec_lvx_<mode>_1op): Likewise.
9830 (altivec_stvx_<mode>_2op): Likewise.
9831 (altivec_stvx_<mode>_1op): Likewise.
9832 (altivec_lvx_<VM2:mode>): New define_expand.
9833 (altivec_stvx_<VM2:mode>): Likewise.
9834 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9835 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9836 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9837 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9838 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9839 (rs6000_gen_lvx): Likewise.
9840 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9841 (altivec_expand_stv_builtin): Likewise.
9842 (altivec_expand_builtin): Likewise.
9843 * config/rs6000/vector.md: Likewise.
9845 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9848 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9851 2018-03-20 Richard Biener <rguenther@suse.de>
9854 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9855 sign-conversions as zero, fall back to standard scalar_stmt
9858 2018-03-20 Martin Liska <mliska@suse.cz>
9861 * predict.c (rebuild_frequencies): Handle case when we have
9862 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9864 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9867 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9868 flag_section_anchors.
9869 * varasm.c (use_blocks_for_decl_p): Remove hack for
9870 dw2_force_const_mem.
9873 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9875 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9876 be created, use lowpart_subreg of operands[0] rather than operands[0]
9878 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9879 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9880 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9881 and n constraint instead of aarch64_shift_imm_di and Usd.
9882 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9883 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9885 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9888 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9889 to favor GPR over NEON registers.
9890 (<shift>di3_neon): Likewise.
9892 2018-03-20 Tom de Vries <tom@codesourcery.com>
9895 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9896 (nvptx_process_pars): Emit bar.sync asap and alap.
9898 2018-03-20 Tom de Vries <tom@codesourcery.com>
9901 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9902 seen_label if seen_label is already set.
9904 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9907 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9908 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9909 Use 1U instead of 1. Formatting fixes.
9912 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9913 instead of TREE_TYPE (s1) for the return value.
9915 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9917 PR tree-optimization/84946
9918 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9919 bitsize + bitsize in poly_uint64 rather than poly_int64.
9922 * dwarf2asm.c: Include fold-const.c.
9923 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9924 of decl rather than decl itself.
9926 PR rtl-optimization/84643
9927 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9929 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9932 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9933 calling assemble_variable.
9935 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9938 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9939 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9941 2018-03-19 Jim Wilson <jimw@sifive.com>
9944 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9945 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9946 (riscv_first_stack_step): Likewise.
9947 (riscv_option_override): Use STACK_BOUNDARY instead of
9949 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9951 (BIGGEST_ALIGNMENT): Set to 128.
9952 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9953 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9956 2018-03-19 Richard Biener <rguenther@suse.de>
9958 PR tree-optimization/84933
9959 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9960 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9962 2018-03-19 Richard Biener <rguenther@suse.de>
9964 PR tree-optimization/84859
9965 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9966 (cond_if_else_store_replacement): Perform sinking operation on
9967 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9968 Generalize what a BB with a single eligible store is.
9970 2018-03-19 Richard Biener <rguenther@suse.de>
9972 PR tree-optimization/84929
9973 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9974 chrec_is_positive against non-chrec arg.
9976 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9979 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9981 2018-03-18 Martin Liska <mliska@suse.cz>
9983 PR rtl-optimization/84635
9984 * regrename.c (build_def_use): Use matches_mode only when
9987 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9989 PR tree-optimization/84913
9990 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9991 vectorize chains of COND_EXPRs.
9993 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9995 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9997 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9999 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10001 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10003 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10005 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10006 Kito Cheng <kito.cheng@gmail.com>
10008 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10009 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10010 (nds32_adjust_reg_alloc_order): New function.
10011 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10013 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10015 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10016 nds32_print_operand, nds32_print_operand_address): Use
10017 HOST_WIDE_INT_PRINT_DEC instead.
10019 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10021 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10023 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10026 * config/i386/i386.c (initial_ix86_tune_features,
10027 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10028 unsigned long long.
10029 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10030 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10031 rather than 1u << ix86_tune. Formatting fix.
10032 (ix86_option_override_internal): Change ix86_arch_mask from
10033 unsigned int to unsigned HOST_WIDE_INT, initialize to
10034 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10035 (ix86_function_specific_restore): Likewise.
10037 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10040 * postreload.c (reload_combine_recognize_pattern): Perform
10041 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10042 truncate_int_for_mode the result for the destination's mode.
10045 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10046 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10048 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10051 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10054 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10055 regno_allocno_class_array and sorted_pseudos.
10056 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10057 insns where regno is used.
10059 2018-03-16 Martin Liska <mliska@suse.cz>
10062 * multiple_target.c (create_dispatcher_calls): Redirect
10063 reference in the symbol table.
10065 2018-03-16 Martin Liska <mliska@suse.cz>
10068 * multiple_target.c (create_dispatcher_calls): Redirect also
10071 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10075 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10076 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10077 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10079 2018-03-16 Julia Koval <julia.koval@intel.com>
10081 * doc/invoke.texi (Skylake Server): Add CLWB.
10082 Cannonlake): Remove CLWB.
10084 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10086 PR tree-optimization/84841
10087 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10089 (FLOAT_ONE_CONST_TYPE): Define.
10090 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10091 (sort_by_operand_rank): Put entries with higher constant_type last
10092 rather than first to match comments.
10094 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10096 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10099 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10102 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10103 check and use address of target always.
10105 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10108 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10109 (indirect_thunk_bnd_needed): Likewise.
10110 (indirect_thunks_used): Likewise.
10111 (indirect_thunks_bnd_used): Likewise.
10112 (indirect_return_needed): New.
10113 (indirect_return_bnd_needed): Likewise.
10114 (output_indirect_thunk_function): Add a bool argument for
10116 (output_indirect_thunk_function): Don't generate alias for
10117 function return thunk.
10118 (ix86_code_end): Call output_indirect_thunk_function to generate
10119 function return thunks.
10120 (ix86_output_function_return): Set indirect_return_bnd_needed
10121 and indirect_return_needed instead of indirect_thunk_bnd_needed
10122 and indirect_thunk_needed.
10124 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10126 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10127 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10128 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10130 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10131 Paul Hua <paul.hua.gm@gmail.com>
10134 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10136 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10138 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10139 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10140 resp. SFmode cases.
10142 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10145 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10146 instead of GET_MODE_SIZE when comparing Units.
10148 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10151 * varasm.c (hash_section): Return an unchangeble hash value
10152 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10153 Return !aarch64_can_use_per_function_literal_pools_p ().
10155 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10158 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10159 rather than address of cmode as last argument to prepare_cmp_insn.
10161 2018-03-15 Julia Koval <julia.koval@intel.com>
10163 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10164 F_AVX512VNNI, F_AVX512BITALG): New.
10166 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10169 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10170 insn for floating-point loads and stores.
10172 2018-03-14 Carl Love <cel@us.ibm.com>
10174 * config/rs6000/rs6000-c.c: Add macro definitions for
10175 ALTIVEC_BUILTIN_VEC_PERMXOR.
10176 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10177 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10178 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10179 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10181 * config/doc/extend.texi: Add prototypes for vec_permxor.
10183 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10186 * diagnostic-show-locus.c (class layout_point): Convert m_line
10187 from int to linenum_type.
10188 (line_span::comparator): Use linenum "compare" function when
10189 comparing line numbers.
10190 (test_line_span): New function.
10191 (layout_range::contains_point): Convert param "row" from int to
10193 (layout_range::intersects_line_p): Likewise.
10194 (layout::will_show_line_p): Likewise.
10195 (layout::print_source_line): Likewise.
10196 (layout::should_print_annotation_line_p): Likewise.
10197 (layout::print_annotation_line): Likewise.
10198 (layout::print_leading_fixits): Likewise.
10199 (layout::annotation_line_showed_range_p): Likewise.
10200 (struct line_corrections): Likewise for field m_row.
10201 (line_corrections::line_corrections): Likewise for param "row".
10202 (layout::print_trailing_fixits): Likewise.
10203 (layout::get_state_at_point): Likewise.
10204 (layout::get_x_bound_for_row): Likewise.
10205 (layout::print_line): Likewise.
10206 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10207 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10208 * input.c (selftest::test_linenum_comparisons): New function.
10209 (selftest::input_c_tests): Call it.
10210 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10211 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10212 * selftest.h (ASSERT_GT): New macro.
10213 (ASSERT_GT_AT): New macro.
10214 (ASSERT_LT): New macro.
10215 (ASSERT_LT_AT): New macro.
10217 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10219 PR rtl-optimization/84780
10220 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10222 2018-03-14 Martin Liska <mliska@suse.cz>
10224 * tree.c (record_node_allocation_statistics): Use
10225 get_stats_node_kind.
10226 (get_stats_node_kind): New function extracted from
10227 record_node_allocation_statistics.
10228 (free_node): Use get_stats_node_kind.
10230 2018-03-14 Richard Biener <rguenther@suse.de>
10232 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10233 that the value-set of ANTIC_IN doesn't grow.
10236 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10238 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10239 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10241 2018-03-14 Julia Koval <julia.koval@intel.com>
10243 * config.gcc (icelake-client, icelake-server): New.
10245 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10246 (initial_ix86_arch_features): Ditto.
10247 (PTA_SKYLAKE): Add SGX.
10248 (PTA_ICELAKE): Remove.
10249 (PTA_ICELAKE_CLIENT): New.
10250 (PTA_ICELAKE_SERVER): New.
10251 (ix86_option_override_internal): Split up icelake on icelake client and
10253 (get_builtin_code_for_version): Ditto.
10254 (fold_builtin_cpu): Ditto.
10255 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10256 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10257 * config/i386/i386.h (processor_type): Ditto.
10258 * doc/invoke.texi: Ditto.
10260 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10263 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10264 INTEGER_CST offset, add it together with bitpos / 8 and
10265 sign extend based on POINTER_SIZE.
10269 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10272 * config/i386/constraints.md (Yc): New register constraint.
10273 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10274 Use Yc constraint for alternative 2 of operand 0. Remove
10275 preferred_for_speed attribute.
10277 2018-03-14 Richard Biener <rguenther@suse.de>
10279 PR tree-optimization/84830
10280 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10281 with the old one to avoid oscillations.
10283 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10286 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10288 (assign_by_spills): Return a flag of reload assignment failure.
10289 Do not process the reload assignment failures. Do not spill other
10290 reload pseudos if they has the same reg class. Update n if
10292 (lra_assign): Add a return arg. Set up from the result of
10293 assign_by_spills call.
10294 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10295 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10296 usage_insns if it is not NULL.
10297 (spill_hard_reg_in_range): New function.
10298 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10299 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10300 function prototypes.
10301 (lra_assign): Change prototype.
10302 * lra.c (lra): Add code to deal with fails by splitting hard reg
10305 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10307 * config/riscv/riscv.opt (mrelax): New option.
10308 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10309 "norelax" when riscv_mrelax is disabled.
10310 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10312 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10315 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10316 reassociation for int modes.
10318 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10320 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10321 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10323 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10324 * config/aarch64/aarch64-sve.md
10325 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10326 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10327 (*extend<mode><Vwide>2): Rename to...
10328 (aarch64_sve_extend<mode><Vwide>2): ...this.
10329 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10330 renaming the old pattern to...
10331 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10333 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10334 define_expand, renaming the old pattern to...
10335 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10336 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10337 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10338 account when deciding which SVE instruction the optab should use.
10339 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10341 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10343 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10344 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10345 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10346 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10347 (tlsdesc_small_<mode>): Turn a define_expand and use
10348 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10349 (tlsdesc_small_advsimd_<mode>): ...this.
10350 (tlsdesc_small_sve_<mode>): New pattern.
10352 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10354 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10355 (UNSPEC_UMUL_HIGHPART): New constants.
10356 (MUL_HIGHPART): New int iteraor.
10357 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10358 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10360 (*<su>mul<mode>3_highpart): New define_insn.
10362 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10365 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10368 2018-03-13 Martin Liska <mliska@suse.cz>
10371 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10373 (sem_item_optimizer::~sem_item_optimizer): Release it.
10374 (sem_item_optimizer::merge_classes): Register variable aliases.
10375 (sem_item_optimizer::fixup_pt_set): New function.
10376 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10377 * ipa-icf.h: Declare new variables and functions.
10379 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10381 PR middle-end/84834
10382 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10383 integer_pow2p@2 and test integer_pow2p in condition.
10384 (A < 0 ? C : 0): Similarly for @1.
10386 PR middle-end/84831
10387 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10388 characters starting at p contain '\0' character, don't look beyond
10392 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10393 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10396 * reg-stack.c (change_stack): Change update_end var from int to
10397 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10398 also call set_block_for_insn on the newly added insns and rescan.
10401 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10402 on the last operand.
10405 * tree.c (stabilize_reference_1): Return save_expr (e) for
10406 STATEMENT_LIST even if it doesn't have side-effects.
10408 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10410 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10412 2018-03-12 Renlin Li <renlin.li@arm.com>
10414 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10415 aarch64_output_scalar_simd_mov_immediate.
10417 2018-03-12 Martin Sebor <msebor@redhat.com>
10419 PR tree-optimization/83456
10420 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10421 for perfectly overlapping calls to memcpy.
10422 (gimple_fold_builtin_memory_chk): Same.
10423 (gimple_fold_builtin_strcpy): Handle no-warning.
10424 (gimple_fold_builtin_stxcpy_chk): Same.
10425 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10427 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10429 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10430 parameter. Use it for SFmode.
10431 (rs6000_function_arg_advance_1): Adjust.
10432 (rs6000_function_arg): Adjust.
10433 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10435 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10437 PR rtl-optimization/84169
10438 PR rtl-optimization/84780
10439 * combine.c (can_combine_p): Check for a 2-insn combination whether
10440 the destination register is used between the two insns, too.
10442 2018-03-12 Richard Biener <rguenther@suse.de>
10444 PR tree-optimization/84803
10445 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10446 for refs DR analysis didn't process.
10448 2018-03-12 Richard Biener <rguenther@suse.de>
10450 PR tree-optimization/84777
10451 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10452 force-vectorize loops ignore whether we are optimizing for size.
10454 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10456 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10457 (TARGET_MD_ASM_ADJUST): Define.
10459 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10460 Kito Cheng <kito.cheng@gmail.com>
10461 Chung-Ju Wu <jasonwucj@gmail.com>
10463 * config/nds32/nds32.c (nds32_compute_stack_frame,
10464 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10465 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10466 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10467 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10468 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10469 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10470 * config/nds32/nds32.md (prologue, epilogue): Use macro
10471 NDS32_V3PUSH_AVAILABLE_P to do checking.
10473 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10476 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10477 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10478 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10479 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10480 addition of most attributes on !orig_type_die or the attribute not
10481 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10483 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10484 Chung-Ju Wu <jasonwucj@gmail.com>
10486 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10487 __NDS32_VH__ macro.
10488 * config/nds32/nds32.opt (mvh): New option.
10490 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10491 Chung-Ju Wu <jasonwucj@gmail.com>
10493 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10495 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10496 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10499 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10500 Chung-Ju Wu <jasonwucj@gmail.com>
10502 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10504 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10505 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10507 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10508 Kito Cheng <kito.cheng@gmail.com>
10509 Chung-Ju Wu <jasonwucj@gmail.com>
10511 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10512 UNSPEC_FFMISM and UNSPEC_FLMISM.
10513 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10514 for ffb, ffmism and flmism.
10515 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10516 (unspec_ffmism): Ditto.
10517 (unspec_flmism): Ditto.
10518 (nds32_expand_builtin_impl): Check if string extension is available.
10519 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10520 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10522 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10525 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10528 * lra-assigns.c (assign_by_spills): Return a flag of reload
10529 assignment failure. Do not process the reload assignment
10530 failures. Do not spill other reload pseudos if they has the same
10532 (lra_assign): Add a return arg. Set up from the result of
10533 assign_by_spills call.
10534 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10535 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10536 usage_insns if it is not NULL.
10537 (spill_hard_reg_in_range): New function.
10538 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10539 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10540 function prototypes.
10541 (lra_assign): Change prototype.
10542 * lra.c (lra): Add code to deal with fails by splitting hard reg
10545 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10548 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10550 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10553 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10554 (dw_val_node): Add val_symbolic_view.
10555 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10556 (symview_upper_bound): New.
10557 (new_line_info_table): Initialize symviews_since_reset.
10558 (dwarf2out_source_line): Count symviews_since_reset and set
10559 symview_upper_bound.
10560 (dw_val_equal_p): Handle symview.
10561 (add_AT_symview): New.
10562 (print_dw_val): Handle symview.
10563 (attr_checksum, attr_checksum_ordered): Likewise.
10564 (same_dw_val_p, size_of_die): Likewise.
10565 (value_format, output_die): Likewise.
10566 (add_high_low_attributes): Use add_AT_symview for entry_view.
10567 (dwarf2out_finish): Reset symview_upper_bound, clear
10570 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10573 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10574 Add strict argument and use it.
10575 (rs6000_split_multireg_move): Update for new strict argument.
10576 (mem_operand_gpr): Disallow all non-offsettable addresses.
10577 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10579 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10582 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10583 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10584 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10587 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10588 decl, use remap_type if we want to use the type.
10590 2018-03-09 Martin Sebor <msebor@redhat.com>
10592 PR tree-optimization/84526
10593 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10595 (builtin_access::generic_overlap): Be prepared to handle non-array
10598 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10600 PR rtl-optimization/84682
10601 * lra-constraints.c (process_address_1): Check is_address flag
10602 for address constraints.
10603 (process_alt_operands): Likewise.
10604 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10605 preprocess_constraints.
10606 * recog.h (preprocess_constraints): Add oploc parameter.
10608 * recog.c (preprocess_constraints): Test address_operand for
10609 CT_ADDRESS constraints.
10611 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10614 * lra-assigns.c (assign_by_spills): Return a flag of reload
10615 assignment failure. Do not process the reload assignment
10616 failures. Do not spill other reload pseudos if they has the same
10618 (lra_assign): Add a return arg. Set up from the result of
10619 assign_by_spills call.
10620 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10621 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10622 usage_insns if it is not NULL.
10623 (spill_hard_reg_in_range): New function.
10624 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10625 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10626 function prototypes.
10627 (lra_assign): Change prototype.
10628 * lra.c (lra): Add code to deal with fails by splitting hard reg
10631 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10634 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10635 Accept complain bool parameter. Only emit errors if it is true.
10636 (arm_parse_cpu_option_name): Likewise.
10637 (arm_target_thumb_only): Adjust callers of the above.
10638 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10639 prototype to take a default true bool parameter.
10640 (arm_parse_arch_option_name): Likewise.
10642 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10643 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10647 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10648 * configure: Regenerate.
10649 * configure.ac ("linker --version-script option"): New.
10650 ("linker soname option"): New.
10652 2018-03-09 Richard Biener <rguenther@suse.de>
10654 PR tree-optimization/84775
10655 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10656 immediate uses of predicate stmts and mark them modified.
10659 PR tree-optimization/84178
10660 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10662 (version_loop_for_if_conversion): Delay update_ssa call.
10663 (tree_if_conversion): Delay update_ssa until after predicate
10666 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10669 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10670 when the function accesses prior frames.
10672 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10675 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10676 gen_llsym, otherwise call maybe_gen_llsym.
10678 PR inline-asm/84742
10679 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10680 has ',' character inside of it.
10682 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10685 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10686 as clobbering CC_REGNUM.
10688 2018-03-08 Richard Biener <rguenther@suse.de>
10690 PR middle-end/84552
10691 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10692 (follow_copies_to_constant): Do not follow SSA names registered
10695 2018-03-08 Richard Biener <rguenther@suse.de>
10697 PR tree-optimization/84178
10698 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10700 (version_loop_for_if_conversion): Delay update_ssa call.
10701 (tree_if_conversion): Delay update_ssa until after predicate
10704 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10706 PR tree-optimization/84178
10707 * tree-if-conv.c (release_bb_predicate): Remove the
10708 the assertion that the stmts have NULL use_ops.
10709 Discard the statements, asserting that they haven't
10710 yet been added to a BB.
10712 2018-03-08 Richard Biener <rguenther@suse.de>
10714 PR tree-optimization/84746
10715 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10716 (phi_translate): Pass in destination ANTIC_OUT set.
10717 (phi_translate_1): Likewise. For a simplified result lookup
10718 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10719 (phi_translate_set): Adjust.
10720 (do_pre_regular_insertion): Likewise.
10721 (do_pre_partial_partial_insertion): Likewise.
10723 2018-03-08 Martin Liska <mliska@suse.cz>
10725 PR gcov-profile/84735
10726 * doc/gcov.texi: Document usage of profile files.
10727 * gcov-io.h: Document changes in the format.
10729 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10733 * dwarf2out.c (struct dw_line_info_table): Update comments for
10735 (FORCE_RESET_NEXT_VIEW): New.
10736 (FORCE_RESETTING_VIEW_P): New.
10737 (RESETTING_VIEW_P): Check for -1 too.
10738 (ZERO_VIEW_P): Likewise.
10739 (new_line_info_table): Force-reset next view.
10740 (dwarf2out_begin_function): Likewise.
10741 (dwarf2out_source_line): Simplify zero_view_p initialization.
10742 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10743 view directly. Omit view when omitting .loc at line 0.
10745 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10747 PR tree-optimization/84740
10748 * tree-switch-conversion.c (process_switch): Call build_constructors
10749 only if info.phi_count is non-zero.
10751 PR tree-optimization/84739
10752 * tree-tailcall.c (find_tail_calls): Check call arguments against
10753 DECL_ARGUMENTS (current_function_decl) rather than
10754 DECL_ARGUMENTS (func) when checking for tail recursion.
10756 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10758 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10759 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10760 Volker Reichelt's entry and add entries for people that perform
10761 GCC fuzzy testing and report numerous bugs.
10763 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10766 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10767 readonly data in sdata, if that is disabled.
10768 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10769 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10770 -mreadonly-in-sdata option.
10772 2018-03-07 Martin Sebor <msebor@redhat.com>
10774 PR tree-optimization/84468
10775 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10776 basic block when looking for nul assignment.
10778 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10781 * except.h (output_function_exception_table): Adjust prototype.
10782 * except.c (output_function_exception_table): Remove FNNAME parameter
10783 and add SECTION parameter. Ouput one part of the table at a time.
10784 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10785 the first part of the exception table and emit unwind directives.
10786 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10787 (i386_pe_seh_cold_init): Likewise.
10788 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10789 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10790 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10791 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10792 * config/i386/winnt.c: Include except.h.
10793 (struct seh_frame_state): Add reg_offset, after_prologue and
10794 in_cold_section fields.
10795 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10796 (i386_pe_seh_cold_init): New function.
10797 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10798 to seh->in_cold_section.
10799 (seh_emit_push): Record the offset of the push.
10800 (seh_emit_save): Record the offet of the save.
10801 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10802 Test seh->after_prologue to disregard the epilogue.
10803 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10804 (i386_pe_end_cold_function): New function.
10806 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10809 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10810 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10813 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10815 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10816 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10818 PR middle-end/84723
10819 * multiple_target.c: Include tree-inline.h and intl.h.
10820 (expand_target_clones): Diagnose and fail if node->definition and
10821 !tree_versionable_function_p (node->decl).
10823 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10825 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10827 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10828 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10829 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10831 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10834 * combine.c (try_combine): Use reg_or_subregno instead of handling
10835 just paradoxical SUBREGs and REGs.
10837 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10839 * config/arc/arc.c (arc_finalize_pic): Remove function.
10840 (arc_must_save_register): We use single base PIC register, remove
10841 checks to save/restore the PIC register.
10842 (arc_expand_prologue): Likewise.
10843 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10845 (arc_verify_short): Likewise.
10846 (arc_attr_type): Likewise.
10847 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10848 (walk_stores): Likewise.
10849 (arc_address_cost): Make it static.
10850 (arc_verify_short): Likewise.
10851 (branch_dest): Likewise.
10852 (arc_attr_type): Likewise.
10853 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10854 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10855 (arc_final_prescan_insn): Remove inserting the nops due to
10856 hardware hazards. It is done in reorg step.
10857 (insn_length_variant_t): Remove.
10858 (insn_length_parameters_t): Likewise.
10859 (arc_insn_length_parameters): Likewise.
10860 (arc_get_insn_variants): Likewise.
10861 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10863 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10865 PR inline-asm/84683
10866 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10869 PR tree-optimization/84687
10870 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10872 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10874 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10876 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10877 Rename to ppc_speculation_barrier.
10878 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10879 __builtin_ppc_speculation_barrier.
10881 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10884 * combine.c (combine_simplify_rtx): Don't try to simplify if
10885 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10888 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10890 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10891 to 32 bytes when compiling for POWER9.
10893 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10896 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10897 regparm >= 3 with no arg reg available also for calls with
10898 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10901 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10903 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10905 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10908 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10910 2018-03-05 Richard Biener <rguenther@suse.de>
10912 PR tree-optimization/84486
10913 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10914 When inserting a __builtin_assume_aligned call set the LHS
10915 SSA name alignment info accordingly.
10917 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10919 PR tree-optimization/84114
10920 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10921 Avoid reassociation of FLOAT_MODE addition.
10923 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10925 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10926 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10927 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10928 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10929 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10930 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10931 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10933 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10934 __builtin_ia32_wbinvd): New builtins.
10935 (SPECIAL_ARGS2): New.
10936 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10937 (SPECIAL_ARGS2): New.
10938 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10939 (ix86_valid_target_attribute_inner_p): Ditto.
10940 (ix86_init_mmx_sse_builtins): Add special_args2.
10941 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10942 TARGET_WBNOINVD_P): New.
10943 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10944 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10945 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10946 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10947 * config/i386/pconfigintrin.h: New file.
10948 * config/i386/wbnoinvdintrin.h: Ditto.
10949 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10951 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10953 2018-03-05 Richard Biener <rguenther@suse.de>
10955 PR tree-optimization/84670
10956 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10958 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10959 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10960 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10961 doesn't grow if all successors (recursively) were visited at least
10964 2018-03-05 Richard Biener <rguenther@suse.de>
10966 PR tree-optimization/84650
10967 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10968 if executed in the loop pipeline.
10970 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10972 * doc/configfiles.texi (Configuration Files): Move info about
10973 conditionalizing $target-protos.h to...
10974 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10975 differs from $target-protos.h.
10977 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10978 Chung-Ju Wu <jasonwucj@gmail.com>
10980 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10981 * config/nds32/nds32-multiple.md (setmemsi): Define.
10982 * config/nds32/nds32-memory-manipulation.c
10983 (nds32_gen_dup_4_byte_to_word_value): New.
10984 (emit_setmem_word_loop): New.
10985 (emit_setmem_byte_loop): New.
10986 (nds32_expand_setmem_loop): New.
10987 (nds32_expand_setmem_loop_v3m): New.
10988 (nds32_expand_setmem_unroll): New.
10989 (nds32_expand_setmem): New.
10991 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10992 Chung-Ju Wu <jasonwucj@gmail.com>
10994 * config/nds32/nds32-memory-manipulation.c
10995 (nds32_emit_load_store): New.
10996 (nds32_emit_post_inc_load_store): New.
10997 (nds32_emit_mem_move): New.
10998 (nds32_emit_mem_move_block): New.
10999 (nds32_expand_movmemsi_loop_unknown_size): New.
11000 (nds32_expand_movmemsi_loop_known_size): New.
11001 (nds32_expand_movmemsi_loop): New.
11002 (nds32_expand_movmemsi_unroll): New.
11003 (nds32_expand_movmemqi): Rename ...
11004 (nds32_expand_movmemsi): ... to this.
11005 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11006 (movmemsi): ... to this.
11007 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11008 (nds32_expand_movmemsi): ... to this.
11010 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11011 Monk Chiang <sh.chiang04@gmail.com>
11012 Chung-Ju Wu <jasonwucj@gmail.com>
11014 * config/nds32/nds32-protos.h
11015 (nds32_expand_load_multiple): New arguments.
11016 (nds32_expand_store_multiple): Ditto.
11017 (nds32_valid_multiple_load_store): Rename ...
11018 (nds32_valid_multiple_load_store_p): ... to this.
11019 * config/nds32/nds32-memory-manipulation.c
11020 (nds32_expand_load_multiple): Refine implementation.
11021 (nds32_expand_store_multiple): Ditto.
11022 * config/nds32/nds32-multiple.md
11023 (load_multiple): Update nds32_expand_load_multiple interface.
11024 (store_multiple): Update nds32_expand_store_multiple interface.
11025 * config/nds32/nds32-predicates.c
11026 (nds32_valid_multiple_load_store): Rename ...
11027 (nds32_valid_multiple_load_store_p): ... to this and refine
11029 * config/nds32/predicates.md
11030 (nds32_load_multiple_and_update_address_operation): New predicate.
11031 (nds32_store_multiple_and_update_address_operation): New predicate.
11033 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11034 Chung-Ju Wu <jasonwucj@gmail.com>
11036 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11037 (combo): New attribute.
11038 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11040 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11042 * config/nds32/nds32.opt: Change -mcmodel= default value.
11044 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11045 Monk Chiang <sh.chiang04@gmail.com>
11046 Chung-Ju Wu <jasonwucj@gmail.com>
11048 * config/nds32/constants.md (unspec_element): New enum.
11049 * config/nds32/constraints.md (Umw): New constraint.
11050 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11051 * config/nds32/nds32-intrinsic.md: Likewise.
11052 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11053 (nds32_valid_smw_lwm_base_p): New.
11054 (nds32_output_smw_single_word): New.
11055 (nds32_output_lmw_single_word): New.
11056 (nds32_expand_unaligned_load): New.
11057 (nds32_expand_unaligned_store): New.
11058 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11059 (nds32_output_smw_single_word): Declare.
11060 (nds32_output_lmw_single_word): Declare.
11061 (nds32_expand_unaligned_load): Declare.
11062 (nds32_expand_unaligned_store): Declare.
11063 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11064 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11065 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11066 NDS32_BUILTIN_UASTORE_DW.
11067 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11070 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11071 Kito Cheng <kito.cheng@gmail.com>
11072 Chung-Ju Wu <jasonwucj@gmail.com>
11074 * config/nds32/nds32-intrinsic.c
11075 (nds32_expand_builtin_null_ftype_reg): Delete.
11076 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11077 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11078 (nds32_read_argument): New.
11079 (nds32_legitimize_target): Ditto.
11080 (nds32_legitimize_argument): Ditto.
11081 (nds32_check_constant_argument): Ditto.
11082 (nds32_expand_unop_builtin): Ditto.
11083 (nds32_expand_unopimm_builtin): Ditto.
11084 (nds32_expand_binop_builtin): Ditto.
11085 (nds32_builtin_decl_impl): Ditto.
11086 (builtin_description): Ditto.
11087 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11088 (nds32_init_builtins_impl): Ditto.
11089 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11090 (nds32_builtin_decl): New.
11091 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11092 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11094 2018-03-02 Jeff Law <law@redhat.com>
11096 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11097 (redundant_insn, fill_simple_delay_slots): Likewise.
11098 (fill_slots_from_thread): Likewise.
11099 * resource.c (mark_referenced_resources): Likewise.
11100 (mark_set_resources, find_dead_or_set_registers): Likewise.
11102 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11104 * substring-locations.h (format_warning_va): Formatting fix for
11105 ATTRIBUTE_GCC_DIAG.
11106 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11108 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11109 * substring-locations.c: Include intl.h.
11110 (format_warning_va): Turned into small wrapper around
11111 format_warning_n_va, renamed to ...
11112 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11113 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11115 (format_warning_at_substring_n): New function.
11116 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11117 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11118 format_warning_at_substring with just a shorter name instead of
11119 const function pointer.
11120 (fmtwarn_n): New function.
11121 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11122 appropriate, get rid of all the fmtstr temporaries, move conditionals
11123 with G_() wrapped string literals directly into fmtwarn arguments,
11124 cast dir.len to (int), formatting fixes.
11126 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11128 * doc/invoke.texi: Remove "Cilk Plus" references.
11130 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11131 Richard Biener <rguenther@suse.de>
11134 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11135 for error or warning attributes if CALL_FROM_THUNK_P is set.
11138 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11141 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11142 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11145 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11146 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11148 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11149 instead of -1U in last predictors element's probability member.
11151 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11154 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11155 arguments if they are comparable.
11157 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11159 PR tree-optimization/84634
11160 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11161 masks and masked_loop_p with a single loop_masks, making sure it's
11162 null for bb vectorization.
11164 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11166 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11167 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11168 loop->force_vectorize to check whether there is no alias.
11170 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11173 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11175 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11177 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11178 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11179 instead of a loop around prev_real_insn.
11180 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11183 PR inline-asm/84625
11184 * config/i386/i386.c (ix86_print_operand): Use conditional
11185 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11188 2018-03-02 Richard Biener <rguenther@suse.de>
11190 PR tree-optimization/84427
11191 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11192 (bitmap_set_subtract_values): Rewrite to handle multiple
11195 (prune_clobbered_mems): Likewise.
11196 (phi_translate): Take edge instead of pred/phiblock.
11197 (phi_translate_1): Likewise.
11198 (phi_translate_set): Likewise. Insert all translated
11199 exprs for a value into the set, keeping possibly multiple
11200 expressions per value.
11201 (compute_antic_aux): Adjust for phi_translate changes.
11202 When intersecting union the expressions and prune those
11203 not in the final value set, keeping possibly multiple
11204 expressions per value. Do not use value-insertion
11205 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11206 all expressions. Add verification that the value-sets
11207 only shrink during iteration.
11208 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11209 (do_pre_regular_insertion): Likewise.
11210 (do_pre_partial_partial_insertion): Likewise.
11212 2018-03-02 Richard Biener <rguenther@suse.de>
11215 * config/darwin.c (saved_debug_info_level): New static global.
11216 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11217 (darwin_asm_lto_end): Restore debug info generation settings.
11219 2018-03-01 Martin Liska <mliska@suse.cz>
11222 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11223 volatile arguments.
11225 2018-03-01 Richard Biener <rguenther@suse.de>
11228 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11229 type annotation with LTO when debug was disabled at compile-time.
11231 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11233 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11235 (mips_final_postscan_insn): Likewise.
11237 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11239 PR rtl-optimization/84528
11240 * alias.c (init_alias_target): Add commentary.
11241 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11242 a unique base value if the frame pointer is not eliminated
11243 to the stack pointer.
11245 2018-03-01 Tom de Vries <tom@codesourcery.com>
11247 PR rtl-optimization/83327
11248 * lra-int.h (hard_regs_spilled_into): Declare.
11249 * lra.c (hard_regs_spilled_into): Define.
11250 (init_reg_info): Init hard_regs_spilled_into.
11251 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11252 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11253 (process_bb_lives): Handle hard_regs_spilled_into.
11254 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11255 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11257 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11259 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11260 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11261 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11262 * config/rs6000/aix72.h: New file.
11264 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11266 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11267 instead of warning_at with conditional singular and plural messages
11271 * stor-layout.c (update_alignment_for_field): For
11272 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11273 && !DECL_PACKED (field), do the alignment update, just use
11274 only desired_align instead of MAX (type_align, desired_align)
11276 (place_field): Don't do known_align < desired_align handling
11277 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11278 is non-NULL, instead do it after rli->prev_field handling and
11279 only if not within a bitfield word. For DECL_PACKED (field)
11280 use type_align of BITS_PER_UNIT.
11282 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11284 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11285 superfluous parentheses and trailing spaces.
11287 2018-02-28 Richard Biener <rguenther@suse.de>
11289 PR tree-optimization/84584
11290 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11291 SCoPs with fake exit edge.
11293 2018-02-28 Martin Liska <mliska@suse.cz>
11296 * timevar.c (timer::print): Fix format to properly print 100%
11299 2018-02-28 Richard Biener <rguenther@suse.de>
11301 PR middle-end/84607
11302 * genmatch.c (capture_info::walk_match): Do not mark
11303 captured expressions without operands as expr_p given
11304 they act more like predicates and should be subject to
11305 "lost tail" side-effect preserving.
11307 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11309 PR rtl-optimization/81611
11310 * auto-inc-dec.c (attempt_change): Move dead note from
11311 mem_insn if it's the next use of regno
11312 (find_address): Take address use of reg holding
11313 non-incremented value. Add parm to limit search to the named
11315 (merge_in_block): Attempt to use a mem insn that is the next
11316 use of the original regno.
11318 2018-02-27 Martin Sebor <msebor@redhat.com>
11321 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11322 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11324 2018-02-27 Martin Sebor <msebor@redhat.com>
11326 PR translation/84207
11327 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11328 n argument to unsigned HOST_WIDE_INT.
11329 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11330 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11331 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11332 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11334 2018-02-27 Richard Biener <rguenther@suse.de>
11336 PR tree-optimization/84512
11337 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11338 Do not use the estimate returned from record_stmt_cost for
11339 the scalar iteration cost but sum properly using add_stmt_cost.
11341 2018-02-27 Richard Biener <rguenther@suse.de>
11343 PR tree-optimization/84466
11344 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11345 Adjust last change to less strictly validate use operands.
11347 2018-02-27 Martin Liska <mliska@suse.cz>
11349 PR gcov-profile/84548
11350 * gcov.c (process_file): Allow partial overlap and consider it
11351 also as group functions.
11352 (output_lines): Properly calculate range of lines for a group.
11354 2018-02-27 Martin Liska <mliska@suse.cz>
11356 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11357 'ggc' suffixes. Change first column width.
11358 (timer::print): Fix formatting of the column.
11360 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11362 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11363 preserve inline entry blocks for the sake of debug inline
11364 entry point markers alone.
11365 (remove_unused_locals): Suggest in comments a better place to
11366 force the preservation of inline entry blocks that are
11367 otherwise unused, but do not preserve them.
11369 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11371 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11373 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11376 * config/i386/constraints.md (Bs): Replace
11377 ix86_indirect_branch_register with
11378 TARGET_INDIRECT_BRANCH_REGISTER.
11380 * config/i386/i386.md (indirect_jump): Likewise.
11381 (tablejump): Likewise.
11382 (*sibcall_memory): Likewise.
11383 (*sibcall_value_memory): Likewise.
11384 Peepholes of indirect call and jump via memory: Likewise.
11385 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11386 (*sibcall_value_GOT_32): Likewise.
11387 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11388 (GOT_memory_operand): Likewise.
11389 (call_insn_operand): Likewise.
11390 (sibcall_insn_operand): Likewise.
11391 (GOT32_symbol_operand): Likewise.
11392 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11394 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11396 PR rtl-optimization/83496
11397 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11398 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11399 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11400 redundant insn, if any.
11401 (relax_delay_slots): Likewise.
11402 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11404 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11406 PR tree-optimization/83965
11407 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11408 that grouped statements are part of a reduction chain. Return
11409 true if the statement is not marked as a reduction itself but
11410 is part of a group.
11411 (vect_recog_dot_prod_pattern): Don't check whether the statement
11412 is part of a group here.
11413 (vect_recog_sad_pattern): Likewise.
11414 (vect_recog_widen_sum_pattern): Likewise.
11416 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11419 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11421 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11424 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11426 (ix86_output_indirect_function_return): New prototype.
11427 (ix86_split_simple_return_pop_internal): Likewise.
11428 * config/i386/i386.c (indirect_return_via_cx): New.
11429 (indirect_return_via_cx_bnd): Likewise.
11430 (indirect_thunk_name): Handle return va CX_REG.
11431 (output_indirect_thunk_function): Create alias for
11432 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11433 (ix86_output_indirect_jmp): Remove the bool argument.
11434 (ix86_output_indirect_function_return): New function.
11435 (ix86_split_simple_return_pop_internal): Likewise.
11436 * config/i386/i386.md (*indirect_jump): Don't pass false
11437 to ix86_output_indirect_jmp.
11438 (*tablejump_1): Likewise.
11439 (simple_return_pop_internal): Change it to define_insn_and_split.
11440 Call ix86_split_simple_return_pop_internal to split it for
11441 -mfunction-return=.
11442 (simple_return_indirect_internal): Call
11443 ix86_output_indirect_function_return instead of
11444 ix86_output_indirect_jmp.
11446 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11449 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11450 memset and value initialization afterwards.
11452 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11454 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11456 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11459 * common/config/aarch64/aarch64-common.c
11460 (aarch_option_optimization_table[]): Switch
11461 off fomit-frame-pointer
11463 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11464 Chung-Ju Wu <jasonwucj@gmail.com>
11466 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11468 (store_multiple): Ditto.
11470 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11472 * config.gcc: Add --with-cpu support for nds32 target.
11473 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11474 * config/nds32/nds32.opt: Add -mcpu= option.
11476 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11478 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11479 isel=yes): Warn for these deprecated options.
11481 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11483 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11484 ISA_2_5_MASKS_EMBEDDED.
11486 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11488 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11489 p->max as pointers rather than using iterative_hash_expr.
11491 2018-02-23 Carl Love <cel@us.ibm.com>
11493 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11494 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11496 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11497 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11498 P8V_BUILTIN_VEC_VUNSIGNED2.
11500 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11503 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11504 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11506 * lra-constraints.c (curr_insn_transform): Set up
11507 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11509 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11510 * lra-eliminations.c (spill_pseudos): Ditto.
11511 (process_insn_for_elimination): Ditto.
11512 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11513 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11514 LRA_NON_CLOBBERED_ALT.
11516 2018-02-22 Martin Sebor <msebor@redhat.com>
11518 PR tree-optimization/84480
11519 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11520 to maybe_diag_stxncpy_trunc. Call it.
11521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11522 from gimple_fold_builtin_strcpy. Print inlining stack.
11523 (handle_builtin_stxncpy): Print inlining stack.
11524 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11526 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11529 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11530 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11531 and -fcheck-pointer-bounds are used together.
11532 (indirect_thunk_prefix): New enum.
11533 (indirect_thunk_need_prefix): New function.
11534 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11535 "_nt" instead of "_bnd" for NOTRACK prefix.
11536 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11537 (output_indirect_thunk_function): Likewise.
11539 (ix86_code_end): Update output_indirect_thunk_function calls.
11540 (ix86_output_indirect_branch_via_reg): Replace
11541 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11542 (ix86_output_indirect_branch_via_push): Likewise.
11543 (ix86_output_function_return): Likewise.
11544 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11545 incompatible with -fcf-protection=branch and
11546 -fcheck-pointer-bounds.
11548 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11551 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11552 Change gcc_assert call to output_operand_lossage.
11554 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11556 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11558 2018-02-22 DJ Delorie <dj@redhat.com>
11559 Sebastian Perta <sebastian.perta@renesas.com>
11560 Oleg Endo <olegendo@gcc.gnu.org>
11562 * config/rx/rx.c (rx_rtx_costs): New function.
11563 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11565 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11567 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11569 2018-02-22 Martin Liska <mliska@suse.cz>
11572 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11573 Add "native" as a possible value.
11575 2018-02-22 Martin Liska <mliska@suse.cz>
11578 * config/i386/i386.c (ix86_option_override_internal):
11579 Add "native" as a possible value for -march and -mtune.
11581 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11584 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11585 to all type variants.
11587 PR tree-optimization/84503
11588 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11589 width as info->bitpos + info->bitsize - start.
11590 (merged_store_group::merge_overlapping): Simplify width computation.
11591 (check_no_overlap): New function.
11592 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11593 start + width and last_order of the group, fail if check_no_overlap
11595 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11596 to group if check_no_overlap fails.
11598 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11600 * config/rs6000/altivec.md: Delete contraint arguments to
11601 define_expand, define_split, and define_peephole2, and in
11602 define_insn_and_split if always unused.
11603 * config/rs6000/darwin.md: Ditto.
11604 * config/rs6000/dfp.md: Ditto.
11605 * config/rs6000/rs6000.md: Ditto.
11606 * config/rs6000/sync.md: Ditto.
11607 * config/rs6000/vector.md: Ditto.
11608 * config/rs6000/vsx.md: Ditto.
11610 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11612 * config/rs6000/altivec.md: Write output control strings as braced
11613 blocks instead of double-quoted strings.
11614 * config/rs6000/darwin.md: Ditto.
11615 * config/rs6000/rs6000.md: Ditto.
11616 * config/rs6000/vector.md: Ditto.
11617 * config/rs6000/vsx.md: Ditto.
11619 2018-02-21 Jason Merrill <jason@redhat.com>
11621 PR c++/84314 - ICE with templates and fastcall attribute.
11622 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11624 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11626 * ipa-cp.c (determine_versionability): Fix comment typos.
11628 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11631 * ipa-cp.c (determine_versionability): Do not version functions caling
11634 2018-02-21 Martin Liska <mliska@suse.cz>
11637 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11638 Add "native" as a possible value.
11639 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11640 the macro when native cpu detection is available.
11642 2018-02-21 Martin Liska <mliska@suse.cz>
11645 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11646 Add "native" as a possible value.
11647 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11648 when native cpu detection is available.
11650 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11651 Martin Sebor <msebor@redhat.com>
11653 PR tree-optimization/84478
11654 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11656 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11657 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11658 support which is conservatively correct, for 2 only stay conservative
11659 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11660 argument to the 2 argument get_range_strlen, adjust 6 arg
11661 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11663 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11664 (gimple_fold_builtin_strlen): Pass true as last argument to
11667 2018-02-20 Martin Sebor <msebor@redhat.com>
11669 PR middle-end/84095
11670 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11671 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11672 (builtin_memref::builtin_memref): Factor out parts into
11673 set_base_and_offset and call it.
11675 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11677 PR middle-end/84406
11678 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11679 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11680 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11681 search at the associated MODE_INT.
11683 2018-02-20 Jeff Law <law@redhat.com>
11685 PR middle-end/82123
11686 PR tree-optimization/81592
11687 PR middle-end/79257
11688 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11689 for range data rather than using global data.
11690 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11691 range data rather than using global data.
11692 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11693 pass it to children as needed.
11694 (struct directive::fmtresult): Similarly.
11695 (struct directive::set_width): Similarly.
11696 (struct directive::set_precision): Similarly.
11697 (format_integer, format_directive, parse_directive): Similarly.
11698 (format_none): Accept unnamed vr_values parameter.
11699 (format_percent, format_floating, format_character): Similarly.
11700 (format_string, format_plain): Similarly.
11701 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11702 the EVRP range analyzer for range data rather than using global data.
11703 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11704 gimple-ssa-evrp-analyze.h
11705 (class sprintf_dom_walker): Add after_dom_children member function.
11706 Add evrp_range_analyzer member.
11707 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11708 range analyzer as needed.
11709 (sprintf_dom_walker::after_dom_children): New member function.
11710 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11712 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11713 (evrp_range_analyzer::pop_to_marker): Likewise.
11715 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11717 PR tree-optimization/84419
11718 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11719 with the required type if its current type is compatible but
11722 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11724 PR middle-end/82004
11725 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11726 after vectorization.
11728 2018-02-20 Martin Liska <mliska@suse.cz>
11731 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11732 possible values if we don't have a hint.
11734 2018-02-20 Martin Liska <mliska@suse.cz>
11738 * final.c (shorten_branches): Build align_tab array with one
11740 * opts.c (finish_options): Add alignment option limit check.
11741 (MAX_CODE_ALIGN): Likewise.
11742 (MAX_CODE_ALIGN_VALUE): Likewise.
11743 * doc/invoke.texi: Document maximum allowed option value for
11744 all -falign-* options.
11746 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11749 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11750 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11751 * var-tracking.c (emit_note_insn_var_location): Remove all references
11752 to NOTE_INSN_CALL_ARG_LOCATION.
11753 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11754 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11755 Use copy_rtx_if_shared.
11756 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11757 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11758 (dwarf2out_var_location): Remove handling of
11759 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11761 * final.c (final_scan_insn): Remove all references to
11762 NOTE_INSN_CALL_ARG_LOCATION.
11763 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11764 before dumping final insns.
11765 * except.c (emit_note_eh_region_end): Remove all references to
11766 NOTE_INSN_CALL_ARG_LOCATION.
11767 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11768 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11769 * config/arc/arc.c (hwloop_optimize): Likewise.
11770 * config/arm/arm.c (create_fix_barrier): Likewise.
11771 * config/s390/s390.c (s390_chunkify_start): Likewise.
11772 * config/sh/sh.c (find_barrier): Likewise.
11773 * config/i386/i386.c (rest_of_insert_endbranch,
11774 ix86_seh_fixup_eh_fallthru): Likewise.
11775 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11776 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11777 * config/frv/frv.c (frv_function_prologue): Likewise.
11778 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11780 (note_outside_basic_block_p): Remove all references to
11781 NOTE_INSN_CALL_ARG_LOCATION.
11782 * gengtype.c (adjust_field_rtx_def): Likewise.
11783 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11785 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11786 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11789 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11792 PR tree-optimization/84452
11793 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11794 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11797 2018-02-19 Martin Liska <mliska@suse.cz>
11800 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11802 2018-02-19 Martin Liska <mliska@suse.cz>
11803 Richard Sandiford <richard.sandiford@linaro.org>
11805 PR tree-optimization/82491
11806 * gimple-fold.c (get_base_constructor): Make earlier bail out
11809 2018-02-19 Carl Love <cel@us.ibm.com>
11811 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11812 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11814 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11815 P8V_BUILTIN_VEC_NEG.
11817 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11819 * config/rl78/rl78.md (movdf): New define expand.
11821 2018-02-19 Martin Liska <mliska@suse.cz>
11824 * doc/invoke.texi: Fix typo.
11825 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11827 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11829 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11830 handle rs6000_single_float and rs6000_double_float specially for
11833 2018-02-16 Jeff Law <law@redhat.com>
11835 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11836 (pop_regs): Use it.
11838 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11841 * ipa-inline.c (inline_small_functions): Fix a typo.
11843 2018-02-16 Nathan Sidwell <nathan@acm.org>
11845 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11847 2018-02-16 Carl Love <cel@us.ibm.com>
11849 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11850 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11851 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11852 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11853 expansion to P8V_BUILTIN_VEC_FLOAT2.
11855 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11857 PR rtl-optimization/70023
11858 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11859 src_regno into account.
11861 2018-02-16 Carl Love <cel@us.ibm.com>
11863 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11864 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11865 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11866 * config/rs6000/rs6000.c: Remove case statements for
11867 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11868 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11869 and P9V_BUILTIN_VEC_VINSERT4B.
11870 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11871 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11872 * config/rs6000/vsx.md:
11873 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11876 2018-02-16 Carl Love <cel@us.ibm.com>
11878 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11880 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11882 * config/rs6000/rs6000-c.c: Add the definitions for
11883 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11884 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11885 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11886 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11887 definition for insert4b and define insn *insert3b_internal.
11888 * doc/extend.texi: Add documentation for vec_extract4b.
11890 2018-02-16 Nathan Sidwell <nathan@acm.org>
11892 * doc/extend.texi (Backwards Compatibility): Mention friend
11893 injection. Note for-scope is deprecated.
11894 * doc/invoke.texi (-ffriend-injection): Deprecate.
11896 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11898 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11899 that moved to I2, also allow destinations that are a paradoxical
11900 subreg (instead of a normal reg).
11902 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11905 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11908 2018-02-16 Richard Biener <rguenther@suse.de>
11910 PR tree-optimization/84037
11911 PR tree-optimization/84016
11913 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11914 Adjust vec_construct for the fact we need additional higher latency
11915 128bit inserts for AVX256 and AVX512 vector builds.
11916 (ix86_add_stmt_cost): Scale vector construction cost for
11919 2018-02-16 Richard Biener <rguenther@suse.de>
11921 PR tree-optimization/84417
11922 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11923 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11924 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11926 2018-02-16 Martin Liska <mliska@suse.cz>
11929 * internal-fn.def (ASAN_CHECK): Set proper flags.
11930 (ASAN_MARK): Likewise.
11932 2018-02-16 Julia Koval <julia.koval@intel.com>
11934 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11935 from PTA_CANNONLAKE.
11937 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11940 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11941 Use ++iter rather than iter++ for std::list iterators.
11942 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11943 defer deleting them until all nodes in the forest are processed. Do
11944 free even leaf nodes. Change to_process into auto_vec.
11947 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11948 * vec.h (vec_default_construct): Use memset instead of placement new
11949 if BROKEN_VALUE_INITIALIZATION is defined.
11950 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11951 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11954 PR rtl-optimization/83723
11955 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11956 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11957 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11959 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11961 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11963 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11965 PR rtl-optimization/81443
11966 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11967 from inner REGs to paradoxical SUBREGs.
11969 2018-02-16 Richard Biener <rguenther@suse.de>
11971 PR tree-optimization/84399
11972 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11973 For operands we can analyze at their definition make sure we can
11974 analyze them at each use as well.
11976 2018-02-16 Richard Biener <rguenther@suse.de>
11978 PR tree-optimization/84190
11979 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11980 volatile accesses if the decl isn't volatile.
11982 2018-02-15 Jason Merrill <jason@redhat.com>
11984 PR c++/84314 - ICE with templates and fastcall attribute.
11985 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11986 TYPE_CANONICAL on an existing type.
11988 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11990 PR tree-optimization/84383
11991 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11992 dstoff nor call operand_equal_p if dstbase is NULL.
11994 PR tree-optimization/84334
11995 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11996 also a CONSTANT_CLASS_P, punt.
11998 2018-02-14 Jim Wilson <jimw@sifive.com>
12000 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12001 first SMALL_OPERAND check. New local min_second_step. Move assert
12002 to where locals are set. Add TARGET_RVC support.
12003 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12005 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12007 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12009 2018-02-14 Martin Sebor <msebor@redhat.com>
12011 PR tree-optimization/83698
12012 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12013 arrays constrain the offset range to their bounds.
12014 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12015 (builtin_access::overlap): Avoid setting the size of overlap if it's
12017 (maybe_diag_overlap): Also consider arrays when deciding what values
12018 of offsets to include in diagnostics.
12020 2018-02-14 Martin Sebor <msebor@redhat.com>
12023 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12024 that correspond to the kind of a declaration.
12026 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12029 * config/pa/pa.md: Load address of PIC label using the linkage table
12030 if the label is nonlocal.
12032 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12034 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12035 warning message if user requests -maltivec=be.
12036 * doc/invoke.texi: Document deprecation of -maltivec=be.
12038 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12041 * config/rs6000/rs6000-c.c: Update definitions for
12042 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12043 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12045 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12048 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12049 add _get_ssp intrinsics. Remove argument from
12050 __builtin_ia32_rdssp[d|q].
12051 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12052 * config/i386/i386-builtin.def: Remove argument from
12053 __builtin_ia32_rdssp[d|q].
12054 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12055 ix86_expand_special_args_builtin for _rdssp[d|q].
12056 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12057 Clear register before usage.
12058 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12059 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12061 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12063 PR tree-optimization/84357
12064 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12065 operand 1 of an ARRAY_REF too.
12067 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12070 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12071 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12073 (set_of_reg): New struct.
12074 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12075 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12076 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12078 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12079 Split into bitclr, bitset, bitinvert patterns if appropriate.
12080 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12081 use rx_fuse_in_memory_bitop.
12082 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12083 to named insn, correct maximum insn length.
12085 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12088 * machmode.def: Define a complex mode for PARTIAL_INT.
12089 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12091 * doc/rtl.texi: Document CSPImode.
12092 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12094 (msp430_hard_regno_nregs_with_padding): Likewise.
12096 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12099 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12101 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12103 PR rtl-optimization/84169
12104 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12105 we generated a parallel as new i3 and we split that to new i2 and i3
12106 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12107 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12108 those to i2, not i1. Partially rewrite this scan code.
12110 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12113 * stor-layout.c (place_field): For variable length fields, adjust
12114 offset_align afterwards not just based on the field's alignment,
12115 but also on the size.
12117 PR middle-end/84309
12118 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12119 of exps and logs in the use_exp2 case.
12121 2018-02-13 Jeff Law <law@redhat.com>
12123 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12124 entry for "vector".
12126 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12129 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12133 * common.opt (gas-loc-support, gas-locview-support): New.
12134 (ginline-points, ginternal-reset-location-views): New.
12135 * doc/invoke.texi: Document them. Use @itemx where intended.
12136 (gvariable-location-views): Adjust.
12137 * target.def (reset_location_view): New.
12138 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12139 (TARGET_RESET_LOCATION_VIEW): New.
12140 * doc/tm.texi: Rebuilt.
12141 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12142 (dwarf2out_default_as_locview_support): New.
12143 (output_asm_line_debug_info): Use option variables.
12144 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12145 (output_loc_list): Likewise.
12146 (add_high_low_attributes): Check option variables.
12147 Don't output entry view attribute in strict mode.
12148 (gen_inlined_subroutine_die): Check option variables.
12149 (dwarf2out_inline_entry): Likewise.
12150 (init_sections_and_labels): Likewise.
12151 (dwarf2out_early_finish): Likewise.
12152 (maybe_reset_location_view): New, from...
12153 (dwarf2out_var_location): ... here. Call it.
12154 * debug.h (dwarf2out_default_as_loc_support): Declare.
12155 (dwarf2out_default_as_locview_support): Declare.
12156 * hooks.c (hook_int_rtx_insn_0): New.
12157 * hooks.h (hook_int_rtx_insn_0): Declare.
12158 * toplev.c (process_options): Take -gas-loc-support and
12159 -gas-locview-support from dwarf2out. Enable
12160 -gvariable-location-views by default only with locview
12161 assembler support. Enable -ginternal-reset-location-views by
12162 default only if the target defines the corresponding hook.
12163 Enable -ginline-points by default if location views are
12164 enabled; force it disabled if statement frontiers are
12166 * tree-inline.c (expand_call_inline): Check option variables.
12167 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12169 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12171 PR tree-optimization/84321
12172 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12173 handling. Also check whether the anti-range contains any values
12174 that satisfy the mask; switch to a VR_RANGE if not.
12176 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12179 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12181 2018-02-13 Martin Jambor <mjambor@suse.cz>
12184 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12185 of call statements, also set location of a load to a temporary.
12187 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12189 * config/rl78/rl78.c (add_vector_labels): New function.
12190 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12191 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12192 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12193 which checks that no arguments are passed.
12194 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12195 * doc/extend.texi: Documentation for the new attribute.
12197 2018-02-13 Andreas Schwab <schwab@suse.de>
12199 * config/riscv/linux.h (CPP_SPEC): Define.
12201 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12204 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12205 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12206 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12207 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12208 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12209 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12210 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12211 temporarily for AES and PCLMUL builtins.
12213 PR tree-optimization/84339
12214 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12215 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12218 PR middle-end/84309
12219 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12220 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12221 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12223 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12225 * omp-simd-clone.h: New file.
12226 * omp-simd-clone.c: Include omp-simd-clone.h.
12227 (expand_simd_clones): No longer static.
12228 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12229 cgraph.h and omp-simd-clone.h.
12230 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12231 (vect_recog_widen_shift_pattern): Formatting fix.
12232 (vect_pattern_recog_1): Don't check optab for calls.
12235 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12236 operands[2] into a REG before using gen_lowpart on it.
12238 2018-02-12 Jeff Law <law@redhat.com>
12241 * config/sh/sh.c (find_barrier): Consider a sibling call
12244 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12245 successfully back substituting a reg.
12247 2018-02-12 Richard Biener <rguenther@suse.de>
12249 PR tree-optimization/84037
12250 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12251 parameter, move visited init to caller.
12252 (vect_slp_analyze_operations): Separate cost from validity
12253 check, initialize visited once for all instances.
12254 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12256 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12258 (vect_model_promotion_demotion_cost): Likewise.
12259 (vectorizable_bswap): Guard cost modeling with !slp_node
12260 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12262 (vectorizable_call): Likewise.
12263 (vectorizable_conversion): Likewise.
12264 (vectorizable_assignment): Likewise.
12265 (vectorizable_shift): Likewise.
12266 (vectorizable_operation): Likewise.
12267 (vectorizable_store): Likewise.
12268 (vectorizable_load): Likewise.
12269 (vectorizable_condition): Likewise.
12270 (vectorizable_comparison): Likewise.
12272 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12275 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12276 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12277 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12279 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12281 PR middle-end/83665
12282 * params.def (inline-min-speedup): Increase from 8 to 15.
12283 (max-inline-insns-auto): Decrease from 40 to 30.
12284 * ipa-split.c (consider_split): Add some buffer for function to
12285 be considered inlining candidate.
12286 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12289 2018-02-12 Richard Biener <rguenther@suse.de>
12291 PR tree-optimization/84037
12292 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12293 matched stmts if we cannot swap the non-matched ones.
12295 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12297 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12298 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12299 _mm_maskz_scalef_round_ss): New intrinsics.
12300 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12301 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12302 __builtin_ia32_scalefss_round): Remove.
12303 (__builtin_ia32_scalefsd_mask_round,
12304 __builtin_ia32_scalefss_mask_round): New intrinsics.
12305 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12306 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12307 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12308 "<round_constraint>")): Changed to ...
12309 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12310 "<round_scalar_constraint>")): ... this.
12311 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12312 %0, %1, %2<round_op3>}"): Changed to ...
12313 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12314 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12315 %2<round_scalar_mask_op3>}"): ... this.
12316 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12318 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12320 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12321 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12322 (_mm_maskz_sqrt_round_ss): New intrinsics.
12323 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12324 (__builtin_ia32_sqrtsd_mask_round)
12325 (__builtin_ia32_sqrtss_mask_round): New builtins.
12326 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12327 (__builtin_ia32_sqrtss_round): Remove.
12328 (__builtin_ia32_sqrtsd_mask_round)
12329 (__builtin_ia32_sqrtss_mask_round): New builtins.
12330 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12331 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12332 ((match_operand:VF_128 1 "vector_operand"
12333 "xBm,<round_constraint>")): Changed to ...
12334 ((match_operand:VF_128 1 "vector_operand"
12335 "xBm,<round_scalar_constraint>")): ... this.
12336 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12337 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12338 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12339 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12340 %<iptr>1<round_scalar_mask_op3>}): ... this.
12341 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12342 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12344 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12347 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12348 Cast vec_cmpeq result to correct type.
12349 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12350 Cast vec_cmpgt result to correct type.
12352 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12354 * final.c (final_scan_insn_1): Renamed from...
12355 (final_scan_insn): ... this. New wrapper, to recover
12356 seen from the outermost call in recursive ones.
12357 * config/sparc/sparc.c (output_return): Drop seen from call.
12358 (output_sibcall): Likewise.
12359 * config/visium/visium.c (output_branch): Likewise.
12361 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12363 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12366 2018-02-10 Alan Modra <amodra@gmail.com>
12369 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12370 Specify LR as an input.
12372 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12375 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12376 remove_member_access_dummy_vars): New functions.
12377 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12378 lower_omp_1, execute_lower_omp): Use them.
12380 PR rtl-optimization/84308
12381 * shrink-wrap.c (spread_components): Release todo vector.
12383 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12385 PR rtl-optimization/57193
12386 * ira-color.c (struct allocno_color_data): Add member
12387 conflict_allocno_hard_prefs.
12388 (update_conflict_allocno_hard_prefs): New.
12389 (bucket_allocno_compare_func): Add a preference based on
12390 conflict_allocno_hard_prefs.
12391 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12392 (color_allocnos): Remove a dead code. Initiate
12393 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12395 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12398 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12399 constraint from =wa to wa. Avoid a subreg on the output operand,
12400 instead use a pseudo and subreg it in a move.
12401 (p9_xxbrd_<mode>): Changed to ...
12402 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12403 (p9_xxbrd_v2df): New expander.
12404 (p9_xxbrw_<mode>): Changed to ...
12405 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12406 (p9_xxbrw_v4sf): New expander.
12408 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12410 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12412 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12415 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12416 multiply in 32-bit mode.
12417 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12418 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12421 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12423 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12424 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12425 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12426 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12428 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12431 * dwarf2out.c (is_trivial_indirect_ref): New function.
12432 (dwarf2out_late_global_decl): Do not generate a location
12433 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12434 and that are not defined in the current unit.
12436 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12438 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12439 instead of a libcall for UNORDERED.
12441 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12444 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12445 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12447 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12450 * config/s390/s390.c (s390_set_current_function): Invoke
12451 s390_indirect_branch_settings also if fndecl didn't change.
12453 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12455 * config/rs6000/rs6000.md (blockage): Set length to zero.
12457 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12459 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12461 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12464 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12465 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12469 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12470 PARALLEL incoming that failed vt_get_decl_and_offset check.
12472 PR middle-end/84237
12473 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12474 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12476 (get_variable_section): For decls in named .bss* sections pass true as
12477 second argument to bss_initializer_p.
12479 2018-02-09 Marek Polacek <polacek@redhat.com>
12480 Jakub Jelinek <jakub@redhat.com>
12483 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12484 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12485 Sync some changes from cxx_fold_indirect_ref.
12487 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12489 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12491 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12492 (BLOCK_INLINE_ENTRY_LABEL): New.
12493 (dwarf2out_var_location): Disregard inline entry markers.
12494 (inline_entry_data): New struct.
12495 (inline_entry_data_hasher): New hashtable type.
12496 (inline_entry_data_hasher::hash): New.
12497 (inline_entry_data_hasher::equal): New.
12498 (inline_entry_data_table): New variable.
12499 (add_high_low_attributes): Add DW_AT_entry_pc and
12500 DW_AT_GNU_entry_view attributes if a pending entry is found
12501 in inline_entry_data_table. Add old entry_pc attribute only
12502 if debug nonbinding markers are disabled.
12503 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12504 markers are enabled.
12505 (block_within_block_p, dwarf2out_inline_entry): New.
12506 (dwarf2out_finish): Check that no entries remained in
12507 inline_entry_data_table.
12508 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12509 (final_scan_insn, notice_source_line): Likewise.
12510 (rest_of_clean_state): Skip inline entry markers.
12511 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12513 * gimple.c (gimple_build_debug_inline_entry): New.
12514 * gimple.h (enum gimple_debug_subcode): Add
12515 GIMPLE_DEBUG_INLINE_ENTRY.
12516 (gimple_build_debug_inline_entry): Declare.
12517 (gimple_debug_inline_entry_p): New.
12518 (gimple_debug_nonbind_marker_p): Adjust.
12519 * insn-notes.def (INLINE_ENTRY): New.
12520 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12521 inline entry marker notes.
12522 (print_insn): Likewise.
12523 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12524 (INSN_DEBUG_MARKER_KIND): Likewise.
12525 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12526 * tree-inline.c (expand_call_inline): Build and insert
12527 debug_inline_entry stmt.
12528 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12529 inline entry blocks early, if nonbind markers are enabled.
12530 (dump_scope_block): Dump fragment info.
12531 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12532 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12533 (gimple_build_debug_inline_entry): New.
12534 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12535 Enable/disable inline entry points too.
12536 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12537 (DEBUG_INSN): Describe inline entry markers.
12539 * common.opt (gvariable-location-views): New.
12540 (gvariable-location-views=incompat5): New.
12541 * config.in: Rebuilt.
12542 * configure: Rebuilt.
12543 * configure.ac: Test assembler for view support.
12544 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12545 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12546 * dwarf2out.c (var_loc_view): New typedef.
12547 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12548 (dwarf2out_locviews_in_attribute): New.
12549 (dwarf2out_locviews_in_loclist): New.
12550 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12551 (enum dw_line_info_opcode): Add LI_adv_address.
12552 (struct dw_line_info_table): Add view.
12553 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12554 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12555 (zero_view_p): New variable.
12556 (ZERO_VIEW_P): New macro.
12557 (output_asm_line_debug_info): New.
12558 (struct var_loc_node): Add view.
12559 (add_AT_view_list, AT_loc_list): New.
12560 (add_var_loc_to_decl): Add view param. Test it against last.
12561 (new_loc_list): Add view params. Record them.
12562 (AT_loc_list_ptr): Handle loc and view lists.
12563 (view_list_to_loc_list_val_node): New.
12564 (print_dw_val): Handle dw_val_class_view_list.
12565 (size_of_die): Likewise.
12566 (value_format): Likewise.
12567 (loc_list_has_views): New.
12568 (gen_llsym): Set vl_symbol too.
12569 (maybe_gen_llsym, skip_loc_list_entry): New.
12570 (dwarf2out_maybe_output_loclist_view_pair): New.
12571 (output_loc_list): Output view list or entries too.
12572 (output_view_list_offset): New.
12573 (output_die): Handle dw_val_class_view_list.
12574 (output_dwarf_version): New.
12575 (output_compilation_unit_header): Use it.
12576 (output_skeleton_debug_sections): Likewise.
12577 (output_rnglists, output_line_info): Likewise.
12578 (output_pubnames, output_aranges): Update version comments.
12579 (output_one_line_info_table): Output view numbers in asm comments.
12580 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12581 Call maybe_gen_llsym.
12582 (loc_list_from_tree_1): Adjust.
12583 (add_AT_location_description): Create view list attribute if
12584 needed, check it's absent otherwise.
12585 (convert_cfa_to_fb_loc_list): Adjust.
12586 (maybe_emit_file): Call output_asm_line_debug_info for test.
12587 (dwarf2out_var_location): Reset views as needed. Precompute
12588 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12589 attribute. Set view.
12590 (new_line_info_table): Reset next view.
12591 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12592 (dwarf2out_source_line): Likewise. Output view resets and labels to
12593 the assembler, or select appropriate line info opcodes.
12594 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12595 (optimize_string_length): Catch it. Adjust.
12596 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12597 dw_val_class_view_list, and remove it if no longer needed.
12598 (hash_loc_list): Hash view numbers.
12599 (loc_list_hasher::equal): Compare them.
12600 (optimize_location_lists): Check whether a view list symbol is
12601 needed, and whether the locview attribute is present, and
12602 whether they match. Remove the locview attribute if no longer
12604 (index_location_lists): Call skip_loc_list_entry for test.
12605 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12606 Use output_dwarf_version.
12607 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12608 (struct dw_val_node): Add val_view_list.
12609 * final.c (SEEN_NEXT_VIEW): New.
12610 (set_next_view_needed): New.
12611 (clear_next_view_needed): New.
12612 (maybe_output_next_view): New.
12613 (final_start_function): Rename to...
12614 (final_start_function_1): ... this. Take pointer to FIRST,
12615 add SEEN parameter. Emit param bindings in the initial view.
12616 (final_start_function): Reintroduce SEEN-less interface.
12617 (final): Rename to...
12618 (final_1): ... this. Take SEEN parameter. Output final pending
12619 next view at the end.
12620 (final): Reintroduce seen-less interface.
12621 (final_scan_insn): Output pending next view before switching
12622 sections or ending a block. Mark the next view as needed when
12623 outputting variable locations. Notify debug backend of section
12624 changes, and of location view changes.
12625 (rest_of_handle_final): Adjust.
12626 * toplev.c (process_options): Autodetect value for debug variable
12627 location views option. Warn on incompat5 without -gdwarf-5.
12628 * doc/invoke.texi (gvariable-location-views): New.
12629 (gvariable-location-views=incompat5): New.
12630 (gno-variable-location-views): New.
12632 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12634 PR tree-optimization/84136
12635 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12636 that the result of find_edge is non-NULL.
12638 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12641 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12642 storing integer register in SImode. Fix cost of 256 and 512
12643 byte aligned SSE register store.
12645 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12647 * config/i386/i386.c (ix86_multiplication_cost): Fix
12648 multiplication cost for TARGET_AVX512DQ.
12650 2018-02-08 Marek Polacek <polacek@redhat.com>
12652 PR tree-optimization/84238
12653 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12656 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12658 PR tree-optimization/84265
12659 * tree-vect-stmts.c (vectorizable_store): Don't treat
12660 VMAT_CONTIGUOUS accesses as grouped.
12661 (vectorizable_load): Likewise.
12663 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12665 PR tree-optimization/81635
12666 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12667 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12668 (test_round_for_mask): New functions.
12669 (wide_int_cc_tests): Call test_round_for_mask.
12670 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12671 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12672 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12673 range returned by get_range_info.
12675 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12678 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12679 * symtab.c: Include builtins.h
12680 (symtab_node::output_to_lto_symbol_table_p): Move here
12681 from lto-streamer-out.c:output_symbol_p.
12682 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12683 (output_symbol_p): Move all logic to symtab.c
12684 (produce_symtab): Update.
12686 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12688 * config/s390/s390-opts.h (enum indirect_branch): Define.
12689 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12690 (s390_indirect_branch_via_thunk)
12691 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12692 (enum s390_indirect_branch_type): Define.
12693 * config/s390/s390.c (struct s390_frame_layout, struct
12694 machine_function): Remove.
12695 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12696 (indirect_branch_table_label_no, indirect_branch_table_name):
12698 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12699 (enum s390_indirect_branch_option): Define.
12700 (s390_return_addr_from_memory): New function.
12701 (s390_handle_string_attribute): New function.
12702 (s390_attribute_table): Add new attribute handler.
12703 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12704 (s390_indirect_branch_via_thunk): New function.
12705 (s390_indirect_branch_via_inline_thunk): New function.
12706 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12707 sibling call optimization for non z10 compiles.
12708 (s390_emit_call): Force indirect branch target to be a single
12709 register. Add r1 clobber for non-z10 compiles.
12710 (s390_emit_epilogue): Emit return jump via return_use expander.
12711 (s390_reorg): Handle JUMP_INSNs as execute targets.
12712 (s390_option_override_internal): Perform validity checks for the
12713 new command line options.
12714 (s390_indirect_branch_attrvalue): New function.
12715 (s390_indirect_branch_settings): New function.
12716 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12717 (s390_output_indirect_thunk_function): New function.
12718 (s390_code_end): Implement target hook.
12719 (s390_case_values_threshold): Implement target hook.
12720 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12722 * config/s390/s390.h (struct s390_frame_layout)
12723 (struct machine_function): Move here from s390.c.
12724 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12725 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12726 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12727 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12728 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12729 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12730 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12731 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12732 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12733 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12734 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12735 (mnemonic attribute): Add values which aren't recognized
12737 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12738 pattern for branch conversion. Fix mnemonic attribute.
12739 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12740 indirect branch via thunk if requested.
12741 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12742 ("*indirect_jump"): Disable for branch conversion using out of
12744 ("indirect_jump_via_thunk<mode>_z10")
12745 ("indirect_jump_via_thunk<mode>")
12746 ("indirect_jump_via_inlinethunk<mode>_z10")
12747 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12748 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12749 ("casesi_jump_via_inlinethunk<mode>_z10")
12750 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12751 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12752 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12753 ("*indirect2_jump"): Disable for branch conversion.
12754 ("casesi_jump"): Turn into expander and expand patterns for branch
12756 ("return_use"): New expander.
12757 ("*return"): Emit return via thunk and rename it to ...
12758 ("*return<mode>"): ... this one.
12759 * config/s390/s390.opt: Add new options and and enum for the
12762 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12764 * lra-constraints.c (match_reload): Unconditionally use
12765 gen_lowpart_SUBREG, rather than selecting between that
12766 and equivalent gen_rtx_SUBREG code.
12768 2018-02-08 Richard Biener <rguenther@suse.de>
12770 PR tree-optimization/84233
12771 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12772 changed flag instead of boguously re-using phi_inserted.
12774 2018-02-08 Martin Jambor <mjambor@suse.cz>
12776 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12777 static local variables.
12779 2018-02-08 Richard Biener <rguenther@suse.de>
12781 PR tree-optimization/84278
12782 * tree-vect-stmts.c (vectorizable_store): When looking for
12783 smaller vector types to perform grouped strided loads/stores
12784 make sure the mode is supported by the target.
12785 (vectorizable_load): Likewise.
12787 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12789 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12790 Increase LDP/STP opportunities by adding adjacent callee-saves.
12792 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12794 PR rtl-optimization/84068
12795 PR rtl-optimization/83459
12796 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12798 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12800 PR tree-optimization/84224
12801 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12802 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12803 non-zero arguments.
12805 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12808 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12809 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12810 count, remove one USE.
12812 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12814 * doc/install.texi (Configuration): Document the
12815 --with-long-double-format={ibm,ieee} PowerPC configuration
12819 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12820 Convert from define_expand to be define_insn_and_split. Rework
12821 float/double/_Float128 conversions to QI/HI/SImode to work with
12822 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12823 conversions to QI/HImode types did a store and then a load to
12824 truncate the value. For conversions to VSX registers, don't split
12825 the insn, instead emit the code directly. Use the code iterator
12826 any_fix to combine signed and unsigned conversions.
12827 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12828 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12829 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12830 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12831 (fix_<mode>di2_hw): Likewise.
12832 (fixuns_<mode>di2_hw): Likewise.
12833 (fix_<mode>si2_hw): Likewise.
12834 (fixuns_<mode>si2_hw): Likewise.
12835 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12836 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12837 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12838 fix<uns>_trunc<SFDF:mode>si2_p8.
12839 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12840 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12841 (fix<uns>_<mode>_mem): Likewise.
12842 (fctiw<u>z_<mode>_mem): Likewise.
12843 (fix<uns>_<mode>_mem): Likewise.
12844 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12845 the register allocator from doing a direct move to the GPRs to do
12846 a store, and instead use the ISA 3.0 store byte/half-word from
12847 vector register instruction. For IEEE 128-bit floating point,
12848 also optimize stores of 32-bit ints.
12849 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12851 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12853 * genextract.c (push_pathstr_operand): New function to support
12855 (walk_rtx): Call push_pathstr_operand.
12856 (print_path): Support [a-zA-Z].
12858 2018-02-07 Richard Biener <rguenther@suse.de>
12860 PR tree-optimization/84037
12861 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12862 (cse_and_gimplify_to_preheader): Declare.
12863 (vect_get_place_in_interleaving_chain): Likewise.
12864 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12866 (_loop_vec_info::~_loop_vec_info): Delete it.
12867 (cse_and_gimplify_to_preheader): New function.
12868 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12869 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12870 (vectorizable_load): Likewise. For grouped stores always base
12871 the IV on the first element.
12872 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12873 condition before gimplifying.
12875 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12877 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12878 *DIV_EXPR and *MOD_EXPR.
12880 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12883 * config/i386/i386.c (ix86_option_override_internal): Mask out
12884 the CF_SET bit when checking -fcf-protection.
12886 2018-02-07 Tom de Vries <tom@codesourcery.com>
12889 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12892 2018-02-07 Richard Biener <rguenther@suse.de>
12894 PR tree-optimization/84204
12895 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12898 PR tree-optimization/84205
12899 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12900 special-case isl_ast_op_zdiv_r.
12902 PR tree-optimization/84223
12903 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12904 Only add conditions from within the region.
12905 (gather_bbs::after_dom_children): Adjust.
12907 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12910 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12911 * config/avr/avr.md: Only post-reload split REG-REG moves if
12912 either register is GENERAL_REG_P.
12914 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12916 PR tree-optimization/84235
12917 * tree-ssa-scopedtables.c
12918 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12919 if the subtraction is performed in floating point type where NaNs are
12920 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12921 build 1. Formatting fix.
12923 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12926 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12927 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12928 and skip it regardless of bb boundaries. Use CALL_P macro,
12929 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12930 unnecessarily, formatting fix.
12932 2018-02-06 Michael Collison <michael.collison@arm.com>
12934 * config/arm/thumb2.md:
12935 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12936 (*thumb_mov_notscc): Ditto.
12938 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12941 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12942 unsigned_fix, not "s".
12944 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12946 * configure.ac (gcc_fn_eh_frame_ro): New function.
12947 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12948 correct .eh_frame permissions.
12949 * configure: Regenerate.
12951 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12953 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12954 irrelevant options.
12956 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12958 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12959 Display warning message for -mno-speculate-indirect-jumps.
12961 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12963 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12965 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12967 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12969 PR tree-optimization/84225
12970 * tree-eh.c (find_trapping_overflow): Only call
12971 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12973 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12976 * config/i386/i386.c: Reimplement the check of possible options
12977 -mibt/-mshstk conbination. Change error messages.
12978 * doc/invoke.texi: Fix a typo: remove extra '='.
12980 2018-02-06 Marek Polacek <polacek@redhat.com>
12982 PR tree-optimization/84228
12983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12985 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12988 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12989 emitted arch directives.
12990 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12991 __ARM_FEATURE_COPROC before changing architectures.
12993 2018-02-06 Richard Biener <rguenther@suse.de>
12995 * config/i386/i386.c (print_reg): Fix typo.
12996 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12998 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13000 * configure: Regenerate.
13002 2018-02-05 Martin Sebor <msebor@redhat.com>
13004 PR tree-optimization/83369
13005 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13008 2018-02-05 Martin Liska <mliska@suse.cz>
13010 * doc/invoke.texi: Cherry-pick upstream r323995.
13012 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13014 * ira.c (ira_init_register_move_cost): Adjust comment.
13016 2018-02-05 Martin Liska <mliska@suse.cz>
13018 PR gcov-profile/84137
13019 * doc/gcov.texi: Fix typo in documentation.
13021 2018-02-05 Martin Liska <mliska@suse.cz>
13023 PR gcov-profile/83879
13024 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13025 using dlopen functionality.
13027 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13029 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13030 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13031 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13032 _mm_maskz_range_round_ss): New intrinsics.
13033 (__builtin_ia32_rangesd128_round)
13034 (__builtin_ia32_rangess128_round): Remove.
13035 (__builtin_ia32_rangesd128_mask_round,
13036 __builtin_ia32_rangess128_mask_round): New builtins.
13037 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13038 __builtin_ia32_rangess128_round): Remove.
13039 (__builtin_ia32_rangesd128_mask_round,
13040 __builtin_ia32_rangess128_mask_round): New builtins.
13041 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13042 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13043 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13044 "<round_saeonly_constraint>")): Changed to ...
13045 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13046 "<round_saeonly_scalar_constraint>")): ... this.
13047 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13048 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13049 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13050 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13051 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13053 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13055 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13057 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13058 Remove all values except native, 8540 and 8548.
13060 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13062 * config/i386/i386.c (ix86_output_function_return): Pass
13063 INVALID_REGNUM, instead of -1, as invalid register number to
13064 indirect_thunk_name and output_indirect_thunk.
13066 2018-02-02 Julia Koval <julia.koval@intel.com>
13068 * config.gcc: Add -march=icelake.
13069 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13070 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13071 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13072 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13073 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13074 (processor_target_table): Add icelake.
13075 (ix86_option_override_internal): Handle new PTAs.
13076 (get_builtin_code_for_version): Handle icelake.
13077 (M_INTEL_COREI7_ICELAKE): New.
13078 (fold_builtin_cpu): Handle icelake.
13079 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13080 * doc/invoke.texi: Add -march=icelake.
13082 2018-02-02 Julia Koval <julia.koval@intel.com>
13084 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13085 to wide_int_bitmask.
13086 * wide-int-bitmask.h: New.
13088 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13091 * config/i386/i386.md: Replace Pmode with word_mode in
13092 builtin_setjmp_setup and builtin_longjmp to support x32.
13094 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13098 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13100 (rs6000_supported_cpu_names): New static variable.
13101 (linux_cpu_translation_table): Likewise.
13102 (elf_platform) <cpu>: Define new static variable and use it.
13103 Translate kernel AT_PLATFORM name to canonical name if needed.
13104 Error if platform name is unknown.
13106 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13109 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13111 2018-02-01 Jeff Law <law@redhat.com>
13114 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13115 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13116 the scratch if RELEASE_VIA_POP is false.
13117 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13118 If we have to save a temporary register, decrement SIZE appropriately.
13119 Pass new arguments to release_scratch_register_on_entry.
13120 (ix86_adjust_stack_and_probe): Likewise.
13121 (ix86_emit_probe_stack_range): Pass new arguments to
13122 release_scratch_register_on_entry.
13124 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13126 PR rtl-optimization/84157
13127 * combine.c (change_zero_ext): Use REG_P predicate in
13128 front of HARD_REGISTER_P predicate.
13130 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13132 * config/avr/avr.c (avr_option_override): Move disabling of
13133 -fdelete-null-pointer-checks to...
13134 * common/config/avr/avr-common.c (avr_option_optimization_table):
13137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13139 PR tree-optimization/81635
13140 * tree-data-ref.c (split_constant_offset_1): For types that
13141 wrap on overflow, try to use range info to prove that wrapping
13144 2018-02-01 Renlin Li <renlin.li@arm.com>
13147 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13148 TAILCALL_ADDR_REGS.
13149 (aarch64_register_move_cost): Likewise.
13150 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13151 TAILCALL_ADDR_REGS.
13152 (REG_CLASS_NAMES): Likewise.
13153 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13154 TAILCALL_ADDR_REGS. Remove IP registers.
13155 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13157 2018-02-01 Richard Biener <rguenther@suse.de>
13159 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13160 for specifying RPO order and allow NULL for that.
13161 * domwalk.c (dom_walker::dom_walker): Likewise.
13162 (dom_walker::walk): Handle NULL RPO order.
13163 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13165 (rewrite_update_dom_walker): Likewise.
13166 (mark_def_dom_walker): Likewise.
13168 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13170 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13171 (aarch64_maybe_expand_sve_subreg_move): Declare.
13172 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13173 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13175 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13176 that are semantically a reverse operation.
13177 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13178 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13179 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13181 (aarch64_can_change_mode_class): For big-endian, forbid changes
13182 between two SVE modes if they have different element sizes.
13184 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13186 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13187 the TImode handling for big-endian targets.
13189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13191 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13192 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13194 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13195 Remove BSWAP handing for big-endian targets and use the form of
13196 LD1RQ appropariate for the mode.
13198 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13200 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13201 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13202 duplicated element.
13204 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13207 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13208 check for operands that need to go through aarch64_sve_reload_be.
13210 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13212 PR tree-optimization/81661
13213 PR tree-optimization/84117
13214 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13215 * tree-eh.c: Include gimplify.h.
13216 (find_trapping_overflow, replace_trapping_overflow,
13217 rewrite_to_non_trapping_overflow): New functions.
13218 * tree-vect-loop.c: Include tree-eh.h.
13219 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13220 * tree-data-ref.c: Include tree-eh.h.
13221 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13223 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13225 PR rtl-optimization/84123
13226 * combine.c (change_zero_ext): Check if hard register satisfies
13227 can_change_dest_mode before calling gen_lowpart_SUBREG.
13229 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13232 * ira.c (ira_init_register_move_cost): Remove assert.
13234 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13236 PR rtl-optimization/84071
13237 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13238 * doc/tm.texi: Regenerate.
13240 2018-01-31 Richard Biener <rguenther@suse.de>
13242 PR tree-optimization/84132
13243 * tree-data-ref.c (analyze_miv_subscript): Properly
13244 check whether evolution_function_is_affine_multivariate_p
13245 before calling gcd_of_steps_may_divide_p.
13247 2018-01-31 Julia Koval <julia.koval@intel.com>
13250 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13251 * config/i386/i386.md (rdpid_rex64) New.
13252 (rdpid): Make 32bit only.
13254 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13257 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13258 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13260 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13263 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13265 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13267 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13269 PR rtl-optimization/84071
13270 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13271 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13273 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13275 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13276 (arc_attribute_table): Add 'aux' attribute.
13277 (arc_in_small_data_p): Consider aux like variables.
13278 (arc_is_aux_reg_p): New function.
13279 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13280 (arc_get_aux_arg): New function.
13281 (prepare_move_operands): Handle aux-register access.
13282 (arc_handle_aux_attribute): New function.
13283 * doc/extend.texi (ARC Variable attributes): Add subsection.
13285 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13287 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13288 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13289 (arc_attribute_table): Add 'uncached' attribute.
13290 (arc_print_operand): Print '.di' flag for uncached memory
13292 (arc_in_small_data_p): Do not consider for small data the uncached
13294 (arc_is_uncached_mem_p): New function.
13295 * config/arc/predicates.md (compact_store_memory_operand): Check
13296 for uncached memory accesses.
13297 (nonvol_nonimm_operand): Likewise.
13298 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13300 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13303 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13304 falign-loops=): Add Optimization flag.
13306 2018-01-30 Jeff Law <law@redhat.com>
13309 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13310 INT_REGISTERS_SAVED. Check it prior to calling
13311 get_scratch_register_on_entry.
13312 (ix86_adjust_stack_and_probe): Similarly.
13313 (ix86_emit_probe_stack_range): Similarly.
13314 (ix86_expand_prologue): Corresponding changes.
13316 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13319 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13320 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13322 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13325 * lra-constraints.c (curr_insn_transform): Process AND in the
13328 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13330 PR rtl-optimization/83986
13331 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13332 dependence against last_pending_memory_flush in addition to
13333 pending_jump_insns.
13335 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13337 PR tree-optimization/81611
13338 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13341 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13344 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13347 2018-01-30 Richard Biener <rguenther@suse.de>
13348 Jakub Jelinek <jakub@redhat.com>
13350 PR tree-optimization/84111
13351 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13352 inner loops added during recursion, as they don't have up-to-date
13355 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13358 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13359 (can_inline_edge_by_limits_p): ... here.
13360 (can_early_inline_edge_p, check_callers,
13361 update_caller_keys, update_callee_keys, recursive_inlining,
13362 add_new_edges_to_heap, speculation_useful_p,
13363 inline_small_functions,
13364 inline_small_functions, flatten_function,
13365 inline_to_all_callers_1): Update.
13367 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13369 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13372 2018-01-30 Richard Biener <rguenther@suse.de>
13374 PR tree-optimization/83008
13375 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13376 invariant and constant vector uses in stmts when they need
13377 more than one stmt.
13379 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13382 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13383 * configure: Regenerate.
13385 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13387 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13389 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13390 Use gen_rtx_REG rather than gen_lowpart.
13392 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13394 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13395 rather than 0 when creating partial subregs.
13397 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13399 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13402 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13405 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13406 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13407 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13408 flags. This restores the settings used before the 2017-07-24.
13409 Turning off pre increment/decrement/modify allows IVOPTS to
13410 optimize DF/SF loops where the index is an int.
13412 2018-01-29 Richard Biener <rguenther@suse.de>
13413 Kelvin Nilsen <kelvin@gcc.gnu.org>
13416 * tree-vect-stmts.c (vectorizable_call): Don't call
13417 targetm.vectorize_builtin_md_vectorized_function if callee is
13420 2018-01-22 Carl Love <cel@us.ibm.com>
13422 * doc/extend.tex: Fix typo in second arg in
13423 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13425 2018-01-29 Richard Biener <rguenther@suse.de>
13427 PR tree-optimization/84086
13428 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13429 (flush_ssaname_freelist): When SSA names were released reset
13430 the SCEV hash table.
13432 2018-01-29 Richard Biener <rguenther@suse.de>
13434 PR tree-optimization/84057
13435 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13436 removed paths when removing edges.
13438 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13440 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13441 -mfunction-return=@var{choice}.
13443 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13445 PR diagnostic/84034
13446 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13447 Handle CR like TAB.
13448 (layout::print_source_line): Likewise.
13449 (test_get_line_width_without_trailing_whitespace): Add test cases.
13451 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13453 PR middle-end/84040
13454 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13457 2018-01-26 Jim Wilson <jimw@sifive.com>
13459 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13461 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13464 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13466 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13467 and CMP + SUB-immediate -> SUBS.
13469 2018-01-26 Martin Sebor <msebor@redhat.com>
13471 PR tree-optimization/83896
13472 * tree-ssa-strlen.c (get_string_len): Rename...
13473 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13474 Avoid assuming length is constant.
13475 (handle_char_store): Use HOST_WIDE_INT for string length.
13477 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13480 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13481 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13483 2018-01-26 Richard Biener <rguenther@suse.de>
13485 PR rtl-optimization/84003
13486 * dse.c (record_store): Only record redundant stores when
13487 the earlier store aliases at least all accesses the later one does.
13489 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13491 PR rtl-optimization/83985
13492 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13493 REG_CFA_RESTORE insns.
13494 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13495 REG_CFA_RESTORE insns here.
13498 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13499 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13501 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13503 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13504 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13505 (arc_init): Likewise.
13506 (arc_override_options): Likewise.
13507 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13509 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13511 * config/arc/arc.h (TARGET_DBNZ): Define.
13512 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13513 properly set the tune attribute.
13514 (dbnz): Use TARGET_DBNZ guard.
13515 * config/arc/arc.opt (mtune): Add core3 option.
13517 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13519 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13520 recognize new pic like addresses.
13521 (arc_delegitimize_address): Clean up.
13523 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13525 * config/arc/arc-arches.def: Option mrf16 valid for all
13527 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13528 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13529 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13530 * config/arc/arc-tables.opt: Regenerate.
13531 * config/arc/arc.c (arc_conditional_register_usage): Handle
13532 reduced register file case.
13533 (arc_file_start): Set must have build attributes.
13534 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13535 mrf16 option value.
13536 * config/arc/arc.opt (mrf16): Add new option.
13537 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13538 * config/arc/genmultilib.awk: Handle new mrf16 option.
13539 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13540 * config/arc/t-multilib: Regenerate.
13541 * doc/invoke.texi (ARC Options): Document mrf16 option.
13543 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13545 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13546 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13547 (arc_attribute_table): Add 'secure_call' attribute.
13548 (arc_print_operand): Print secure call operand.
13549 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13551 (arc_is_secure_call_p): New function. * config/arc/arc.md
13552 (call_i): Add support for sjli instruction.
13553 (call_value_i): Likewise.
13554 * config/arc/constraints.md (Csc): New constraint.
13556 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13557 John Eric Martin <John.Martin@emmicro-us.com>
13559 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13560 * config/arc/arc.c (_arc_jli_section): New struct.
13561 (arc_jli_section): New type.
13562 (rc_jli_sections): New static variable.
13563 (arc_handle_jli_attribute): New function.
13564 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13565 (arc_file_end): New function.
13566 (TARGET_ASM_FILE_END): Define.
13567 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13568 (arc_add_jli_section): New function.
13569 (jli_call_scan): Likewise.
13570 (arc_reorg): Call jli_call_scan.
13571 (arc_output_addsi): Remove 'S' from printing asm operand.
13572 (arc_is_jli_call_p): New function.
13573 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13575 (movhi_insn): Likewise.
13576 (movsi_insn): Likewise.
13577 (movsi_set_cc_insn): Likewise.
13578 (loadqi_update): Likewise.
13579 (load_zeroextendqisi_update): Likewise.
13580 (load_signextendqisi_update): Likewise.
13581 (loadhi_update): Likewise.
13582 (load_zeroextendhisi_update): Likewise.
13583 (load_signextendhisi_update): Likewise.
13584 (loadsi_update): Likewise.
13585 (loadsf_update): Likewise.
13586 (movsicc_insn): Likewise.
13587 (bset_insn): Likewise.
13588 (bxor_insn): Likewise.
13589 (bclr_insn): Likewise.
13590 (bmsk_insn): Likewise.
13591 (bicsi3_insn): Likewise.
13592 (cmpsi_cc_c_insn): Likewise.
13593 (movsi_ne): Likewise.
13594 (movsi_cond_exec): Likewise.
13595 (clrsbsi2): Likewise.
13596 (norm_f): Likewise.
13606 (call_i): Remove 'S' asm letter, add jli instruction.
13607 (call_value_i): Likewise.
13608 * config/arc/arc.op (mjli-always): New option.
13609 * config/arc/constraints.md (Cji): New constraint.
13610 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13612 (subsf3_fpx): Likewise.
13613 (mulsf3_fpx): Likewise.
13614 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13616 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13617 function attrbutes.
13618 * doc/invoke.texi (ARC): Document mjli-always option.
13620 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13622 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13623 avoid addition with 0 and use incw and decw where possible.
13625 2018-01-26 Richard Biener <rguenther@suse.de>
13627 PR tree-optimization/81082
13628 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13629 association if it requires casting to unsigned.
13630 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13631 from fold_plusminus_mult_expr to catch important cases late when
13632 range info is available.
13634 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13636 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13637 * configure.ac (hidden_linkonce): New test.
13638 * configure: Regenerate.
13639 * config.in: Regenerate.
13641 2018-01-26 Julia Koval <julia.koval@intel.com>
13643 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13644 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13645 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13646 _mm_mask_bitshuffle_epi64_mask): Fix type.
13647 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13648 USI_FTYPE_V4DI_V4DI_USI): Remove.
13649 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13650 __builtin_ia32_vpshufbitqmb256_mask,
13651 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13652 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13653 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13655 2018-01-26 Alan Modra <amodra@gmail.com>
13658 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13659 UNSPEC_VBPERMQ. Sort other unspecs.
13661 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13663 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13665 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13667 PR middle-end/83055
13668 * predict.c (drop_profile): Do not push/pop cfun; update also
13670 (handle_missing_profiles): Fix logic looking for zero profiles.
13672 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13674 PR middle-end/83977
13675 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13676 on functions with #pragma omp declare simd or functions with simd
13678 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13679 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13680 Remove trailing \n from warning_at calls.
13682 2018-01-25 Tom de Vries <tom@codesourcery.com>
13685 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13686 for neutered workers.
13688 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13691 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13692 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13694 2018-01-24 Jeff Law <law@redhat.com>
13697 * i386.c (get_probe_interval): Move to earlier point.
13698 (ix86_compute_frame_layout): If -fstack-clash-protection and
13699 the frame is larger than the probe interval, then use pushes
13700 to save registers rather than reg->mem moves.
13701 (ix86_expand_prologue): Remove conditional for int_registers_saved
13704 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13707 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13708 min/max for never referenced object.
13710 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13712 PR middle-end/83977
13713 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13715 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13716 attributes from DECL_ATTRIBUTES (decl) without affecting
13717 DECL_ATTRIBUTES (current_function_decl).
13718 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13719 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13721 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13723 PR tree-optimization/83979
13724 * fold-const.c (fold_comparison): Use constant_boolean_node
13725 instead of boolean_{true,false}_node.
13727 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13729 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13732 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13734 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13735 Simplify the clause that sets the length attribute.
13736 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13737 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13738 clause that sets the length attribute.
13739 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13741 2018-01-24 Tom de Vries <tom@codesourcery.com>
13744 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13745 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13746 Add strict parameter.
13747 (prevent_branch_around_nothing): Insert dummy insn between branch to
13748 label and label with no ptx insn inbetween.
13749 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13751 2018-01-24 Tom de Vries <tom@codesourcery.com>
13754 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13755 for neutered threads in warp.
13756 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13758 2018-01-24 Richard Biener <rguenther@suse.de>
13760 PR tree-optimization/83176
13761 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13764 2018-01-24 Richard Biener <rguenther@suse.de>
13766 PR tree-optimization/82819
13767 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13768 code generating pluses that are no-ops in the target precision.
13770 2018-01-24 Richard Biener <rguenther@suse.de>
13772 PR middle-end/84000
13773 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13775 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13777 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13778 to merge probabilities.
13779 * predict.c (probably_never_executed): Also mark as cold functions
13780 with global 0 profile and guessed local profile.
13781 * profile-count.c (profile_probability::combine_with_count): New
13783 * profile-count.h (profile_probability::operator*,
13784 profile_probability::operator*=, profile_probability::operator/,
13785 profile_probability::operator/=): Reduce precision to adjusted
13786 and set value to guessed on contradictory divisions.
13787 (profile_probability::combine_with_freq): Remove.
13788 (profile_probability::combine_wiht_count): Declare.
13789 (profile_count::force_nonzero):: Set to adjusted.
13790 (profile_count::probability_in):: Set quality to adjusted.
13791 * tree-ssa-tail-merge.c (replace_block_by): Use
13792 combine_with_count.
13794 2018-01-23 Andrew Waterman <andrew@sifive.com>
13795 Jim Wilson <jimw@sifive.com>
13797 * config/riscv/riscv.c (riscv_stack_boundary): New.
13798 (riscv_option_override): Set riscv_stack_boundary. Handle
13799 riscv_preferred_stack_boundary_arg.
13800 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13801 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13802 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13803 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13804 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13805 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13807 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13810 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13811 of struct ix86_frame.
13812 (ix86_expand_epilogue): Likewise. Add a local variable for
13813 the reg_save_offset field in struct ix86_frame.
13815 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13817 PR tree-optimization/82604
13818 * tree-loop-distribution.c (enum partition_kind): New enum item
13819 PKIND_PARTIAL_MEMSET.
13820 (partition_builtin_p): Support above new enum item.
13821 (generate_code_for_partition): Ditto.
13822 (compute_access_range): Differentiate cases that equality can be
13823 proven at all loops, the innermost loops or no loops.
13824 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13825 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13826 (finalize_partitions, distribute_loop): Don't fuse partition of
13827 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13828 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13829 parloop is enabled.
13831 2018-01-23 Martin Liska <mliska@suse.cz>
13833 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13834 order to ignore the predictor.
13835 (PRED_POLYMORPHIC_CALL): Likewise.
13836 (PRED_RECURSIVE_CALL): Likewise.
13838 2018-01-23 Martin Liska <mliska@suse.cz>
13840 * tree-profile.c (tree_profiling): Print function header to
13841 aware reader which function we are working on.
13842 * value-prof.c (gimple_find_values_to_profile): Do not print
13843 not interesting value histograms.
13845 2018-01-23 Martin Liska <mliska@suse.cz>
13847 * profile-count.h (enum profile_quality): Add
13848 profile_uninitialized as the first value. Do not number values
13849 as they are zero based.
13850 (profile_count::verify): Update sanity check.
13851 (profile_probability::verify): Likewise.
13853 2018-01-23 Nathan Sidwell <nathan@acm.org>
13855 * doc/invoke.texi (ffor-scope): Deprecate.
13857 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13859 PR tree-optimization/83510
13860 * domwalk.c (set_all_edges_as_executable): New function.
13861 (dom_walker::dom_walker): Convert bool param
13862 "skip_unreachable_blocks" to enum reachability. Move setup of
13863 edge flags to set_all_edges_as_executable and only do it when
13864 reachability is REACHABLE_BLOCKS.
13865 * domwalk.h (enum dom_walker::reachability): New enum.
13866 (dom_walker::dom_walker): Convert bool param
13867 "skip_unreachable_blocks" to enum reachability.
13868 (set_all_edges_as_executable): New decl.
13869 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13870 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13872 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13873 but converting true to REACHABLE_BLOCKS.
13874 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13876 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13877 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13878 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13880 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13881 if check_all_array_refs will be called.
13883 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13885 * tree.c (selftest::test_location_wrappers): Add more test
13888 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13890 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13891 (selftest::test_bit_in_range): Likewise.
13893 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13896 * doc/sourcebuild.texi (vect_float): Say that the selector
13897 only describes the situation when -funsafe-math-optimizations is on.
13898 (vect_float_strict): Document.
13900 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13902 PR tree-optimization/83965
13903 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13904 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13905 instead of checking only for a reduction.
13906 (vect_recog_widen_sum_pattern): Likewise.
13908 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13910 * predict.c (probably_never_executed): Only use precise profile info.
13911 (compute_function_frequency): Skip after inlining hack since we now
13912 have quality checking.
13914 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13916 * profile-count.h (profile_probability::very_unlikely,
13917 profile_probability::unlikely, profile_probability::even): Set
13918 precision to guessed.
13920 2018-01-23 Richard Biener <rguenther@suse.de>
13922 PR tree-optimization/83963
13923 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13924 Properly terminate dominator walk when crossing the exit edge not
13925 when visiting its source block.
13927 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13930 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13931 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13933 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13935 PR tree-optimization/83957
13936 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13937 semicolon after for body surrounded by braces.
13939 PR tree-optimization/83081
13940 * profile-count.h (profile_probability::split): New method.
13941 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13942 Use profile_probability::split.
13943 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13944 when splitting a single conditional jump into 2.
13946 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13948 PR tree-optimization/69452
13949 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13952 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13954 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13955 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13956 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13960 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13962 * config/rl78/rl78.md (movdi): New define_expand.
13963 * config/rl78/rl78.c (rl78_split_movdi): New function.
13965 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13968 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13970 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13971 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13972 128-bit to produce an UNSPEC move to get the double word with the
13973 signbit and then a shift directly to do signbit.
13974 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13975 implementation with a new version that just does either a direct
13976 move or a regular move. Move memory interface to separate insns.
13977 Move insns so they are next to the expander.
13978 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13979 with signbit move. Split big and little endian case.
13980 (signbit<mode>2_dm_mem_le): Likewise.
13981 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13982 (signbit<mode>2_dm2): Likewise.
13984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13986 * config/rl78/rl78.md (anddi3): New define_expand.
13988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13990 * config/rl78/rl78.md (umindi3): New define_expand.
13992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13994 * config/rl78/rl78.md (smindi3): New define_expand.
13996 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13998 * config/rl78/rl78.md (smaxdi3): New define_expand.
14000 2018-01-22 Carl Love <cel@us.ibm.com>
14002 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14003 LVX_V1TI): Add macro expansion.
14004 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14005 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14006 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14007 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14008 Change check to determine if the instruction is a byte reversing
14009 entry. Fix typo in comment.
14010 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14011 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14012 Add def_builtin calls for new builtins.
14013 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14014 Add define_insn expansion.
14016 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14018 * config/rl78/rl78.md (umaxdi3): New define_expand.
14020 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14022 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14023 for non-QImode registers.
14025 2018-01-22 Richard Biener <rguenther@suse.de>
14027 PR tree-optimization/83963
14028 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14029 including the loop exit block.
14030 (scop_detection::merge_sese): Likewise.
14031 (scop_detection::add_scop): Do it here instead.
14033 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14035 * doc/sourcebuild.texi (arm_softfloat): Document.
14037 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14040 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14041 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14042 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14044 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14045 David Edelsohn <dje.gcc@gmail.com>
14048 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14049 Change "crset eq" to "crset 2".
14050 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14051 (*call_indirect_aix<mode>_nospec): Likewise.
14052 (*call_value_indirect_aix<mode>_nospec): Likewise.
14053 (*call_indirect_elfv2<mode>_nospec): Likewise.
14054 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14055 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14056 change assembly output from . to $.
14057 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14058 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14059 (*tablejump<mode>_internal1_nospec): Likewise.
14061 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14064 * config/sh/sh_optimize_sett_clrt.cc:
14065 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14067 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14069 PR tree-optimization/83940
14070 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14071 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14072 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14073 use it to pass back the definition type.
14074 (vect_check_store_rhs): Likewise rhs_dt_out.
14075 (vect_build_gather_load_calls): Add a mask_dt argument and use
14076 it instead of a call to vect_is_simple_use.
14077 (vectorizable_store): Update calls to vect_check_load_store_mask
14078 and vect_check_store_rhs. Use the dt returned by the latter instead
14079 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14080 instead of calls to vect_is_simple_use. Pass the scalar rather
14081 than the vector operand to vect_is_simple_use when handling
14082 second and subsequent copies of an rhs value.
14083 (vectorizable_load): Update calls to vect_check_load_store_mask
14084 and vect_build_gather_load_calls. Use the cached mask_dt and
14085 gs_info.offset_dt instead of calls to vect_is_simple_use.
14087 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14089 PR middle-end/83945
14090 * tree-emutls.c: Include gimplify.h.
14091 (lower_emutls_2): New function.
14092 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14093 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14094 it before further processing.
14097 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14098 UINTVAL (trueop1) instead of INTVAL (op1).
14100 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14104 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14105 INCOMING_FRAME_SP_OFFSET if not defined.
14106 (scan_trace): Add ENTRY argument. If true and
14107 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14108 emit a note to adjust the CFA offset.
14109 (create_cfi_notes): Adjust scan_trace callers.
14110 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14111 INCOMING_FRAME_SP_OFFSET in the CIE.
14112 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14113 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14115 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14116 * doc/tm.texi: Regenerated.
14118 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14120 PR rtl-optimization/83147
14121 * lra-constraints.c (remove_inheritance_pseudos): Use
14122 lra_substitute_pseudo_within_insn.
14124 2018-01-19 Tom de Vries <tom@codesourcery.com>
14125 Cesar Philippidis <cesar@codesourcery.com>
14128 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14130 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14133 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14134 spaces for function labels.
14136 2018-01-19 Martin Liska <mliska@suse.cz>
14138 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14139 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14140 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14141 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14142 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14143 (PRED_CONST_RETURN): Change from 69 to 65.
14144 (PRED_NULL_RETURN): Change from 91 to 71.
14145 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14146 (PRED_LOOP_GUARD): Change from 66 to 73.
14148 2018-01-19 Martin Liska <mliska@suse.cz>
14150 * predict.c (predict_insn_def): Add new assert.
14151 (struct branch_predictor): Change type to signed integer.
14152 (test_prediction_value_range): Amend test to cover
14153 PROB_UNINITIALIZED.
14154 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14155 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14156 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14157 (PRED_LOOP_IV_COMPARE): Likewise.
14158 * predict.h (PROB_UNINITIALIZED): Define new constant.
14160 2018-01-19 Martin Liska <mliska@suse.cz>
14162 * predict.c (dump_prediction): Add new format for
14163 analyze_brprob.py script which is enabled with -details
14165 * profile-count.h (precise_p): New function.
14167 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14169 PR tree-optimization/83922
14170 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14171 there are no statements that need masking.
14172 (vect_active_double_reduction_p): New function.
14173 (vect_analyze_loop_operations): Use it when handling phis that
14174 are not in the loop header.
14176 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14178 PR tree-optimization/83914
14179 * tree-vect-loop.c (vectorizable_induction): Don't convert
14180 init_expr or apply the peeling adjustment for inductions
14181 that are nested within the vectorized loop.
14183 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14185 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14188 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14192 * function.h (gimplify_parameters): Add gimple_seq * argument.
14193 * function.c: Include gimple.h and options.h.
14194 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14195 for the added local temporaries if needed.
14196 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14197 if there are any parameter cleanups, wrap whole body into a
14198 try/finally with the cleanups.
14200 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14203 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14204 Use GET_MODE_CLASS for scalar floating point.
14206 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14210 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14211 Fix call of call_cgraph_insertion_hooks.
14213 2018-01-18 Martin Sebor <msebor@redhat.com>
14215 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14217 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14220 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14223 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14226 * common.opt: (-ffile-prefix-map): New option.
14227 * opts.c (common_handle_option): Defer it.
14228 * opts-global.c (handle_common_deferred_options): Handle it.
14229 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14230 * file-prefix-map.h: New file.
14231 (remap_debug_filename, add_debug_prefix_map): ...here.
14232 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14233 * final.c (debug_prefix_map, add_debug_prefix_map
14234 remap_debug_filename): Move to...
14235 * file-prefix-map.c: New file.
14236 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14237 generalize, get rid of alloca(), use strrchr() instead of strchr().
14238 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14239 Implement in terms of add_prefix_map().
14240 (remap_macro_filename, remap_debug_filename): Implement in term of
14242 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14243 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14244 * dbxout.c: Include file-prefix-map.h.
14245 * varasm.c: Likewise.
14246 * vmsdbgout.c: Likewise.
14247 * xcoffout.c: Likewise.
14248 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14249 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14250 * doc/invoke.texi (-ffile-prefix-map): Document.
14251 (-fdebug-prefix-map): Update description.
14253 2018-01-18 Martin Liska <mliska@suse.cz>
14255 * config/i386/i386.c (indirect_thunk_name): Document that also
14257 (output_indirect_thunk): Document why both instructions
14258 (pause and lfence) are generated.
14260 2018-01-18 Richard Biener <rguenther@suse.de>
14262 PR tree-optimization/83887
14263 * graphite-scop-detection.c
14264 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14265 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14266 (scop_detection::merge_sese): Re-implement with a flood-fill
14267 algorithm that properly finds a SESE region if it exists.
14269 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14272 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14273 pointer_diff optimizations use view_convert instead of convert.
14275 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14277 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14278 Generate different code for -mno-speculate-indirect-jumps.
14279 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14280 (*call_indirect_aix<mode>): Disable for
14281 -mno-speculate-indirect-jumps.
14282 (*call_indirect_aix<mode>_nospec): New define_insn.
14283 (*call_value_indirect_aix<mode>): Disable for
14284 -mno-speculate-indirect-jumps.
14285 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14286 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14287 -mno-speculate-indirect-jumps.
14288 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14290 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14292 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14293 long double type, set the flags for noting the default long double
14294 type, even if we don't pass or return a long double type.
14296 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14299 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14302 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14304 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14305 support for merge[hl].
14306 (fold_mergehl_helper): New helper function.
14307 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14308 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14309 (altivec_vmrglw_direct): Add xxmrglw insn.
14311 2018-01-17 Andrew Waterman <andrew@sifive.com>
14313 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14314 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14316 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14319 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14320 call the lto_location_cache before reading the
14321 DECL_SOURCE_LOCATION of the types.
14323 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14324 Richard Sandiford <richard.sandiford@linaro.org>
14326 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14327 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14328 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14329 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14330 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14332 * config/aarch64/constraints.md (aarch64_movti_operand):
14334 * config/aarch64/predicates.md (Uti): Add new constraint.
14336 2018-01-17 Carl Love <cel@us.ibm.com>
14338 * config/rs6000/vsx.md (define_expand xl_len_r,
14339 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14340 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14342 (define_expand, define_insn): Move the shift left from the
14343 define_insn to the define_expand for lxvl and stxvl instructions.
14344 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14345 and XL_LEN_R definitions to PURE.
14347 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14349 * config/i386/i386.c (indirect_thunk_name): Declare regno
14350 as unsigned int. Compare regno with INVALID_REGNUM.
14351 (output_indirect_thunk): Ditto.
14352 (output_indirect_thunk_function): Ditto.
14353 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14354 in the call to output_indirect_thunk_function.
14356 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14358 PR middle-end/83884
14359 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14360 rather than the size of inner_type to determine the stack slot size
14361 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14363 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14366 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14369 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14371 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14372 endian Linux systems to optionally enable multilibs for selecting
14373 the long double type if the user configured an explicit type.
14374 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14375 have no long double multilibs if not defined.
14376 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14377 warn if the user used -mabi={ieee,ibm}longdouble and we built
14378 multilibs for long double.
14379 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14380 appropriate multilib option.
14381 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14383 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14384 for building long double multilibs.
14385 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14387 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14389 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14392 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14394 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14397 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14400 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14403 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14405 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14406 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14408 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14410 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14411 different rtl trees depending on TARGET_64BIT.
14412 (rs6000_gen_lvx): Likewise.
14414 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14416 * config/visium/visium.md (nop): Tweak comment.
14417 (hazard_nop): Likewise.
14419 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14421 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14422 -mspeculate-indirect-jumps.
14423 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14424 for -mno-speculate-indirect-jumps.
14425 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14426 (*call_value_indirect_elfv2<mode>): Disable for
14427 -mno-speculate-indirect-jumps.
14428 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14429 (indirect_jump): Emit different RTL for
14430 -mno-speculate-indirect-jumps.
14431 (*indirect_jump<mode>): Disable for
14432 -mno-speculate-indirect-jumps.
14433 (*indirect_jump<mode>_nospec): New define_insn.
14434 (tablejump): Emit different RTL for
14435 -mno-speculate-indirect-jumps.
14436 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14437 (tablejumpsi_nospec): New define_expand.
14438 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14439 (tablejumpdi_nospec): New define_expand.
14440 (*tablejump<mode>_internal1): Disable for
14441 -mno-speculate-indirect-jumps.
14442 (*tablejump<mode>_internal1_nospec): New define_insn.
14443 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14446 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14448 * caller-save.c (insert_save): Drop unnecessary parameter. All
14451 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14452 Richard Biener <rguenth@suse.de>
14455 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14456 return early, inline manually is_gimple_sizepos. Make sure if we
14457 call gimplify_expr we don't end up with a gimple constant.
14458 * tree.c (variably_modified_type_p): Don't return true for
14459 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14460 * gimplify.h (is_gimple_sizepos): Remove.
14462 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14464 PR tree-optimization/83857
14465 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14466 vectorizable_live_operation for pure SLP statements.
14467 (vectorizable_live_operation): Handle PHIs.
14469 2018-01-16 Richard Biener <rguenther@suse.de>
14471 PR tree-optimization/83867
14472 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14473 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14475 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14478 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14479 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14480 If off is not INTEGER_CST, issue a may not be aligned warning
14481 rather than isn't aligned. Use isn%'t rather than isn't.
14482 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14484 <case MULT_EXPR>: Improve the case when bottom and one of the
14485 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14486 operand, in that case check if the other operand is multiple of
14487 bottom divided by the INTEGER_CST operand.
14489 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14492 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14493 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14494 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14495 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14496 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14497 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14498 (pa_function_arg_size): New function.
14500 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14502 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14503 in a separate statement.
14505 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14507 PR tree-optimization/83847
14508 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14509 group gathers and scatters.
14511 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14513 PR rtl-optimization/86620
14514 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14516 PR rtl-optimization/83213
14517 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14518 to last if both are JUMP_INSNs.
14520 PR tree-optimization/83843
14521 * gimple-ssa-store-merging.c
14522 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14523 store_immediate_info for bswap/nop orig_stores.
14525 2018-01-15 Andrew Waterman <andrew@sifive.com>
14527 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14529 <UDIV>: Increase cost if !TARGET_DIV.
14531 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14533 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14534 (define_attr "cr_logical_3op"): New.
14535 (cceq_ior_compare): Adjust.
14536 (cceq_ior_compare_complement): Adjust.
14537 (*cceq_rev_compare): Adjust.
14538 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14539 (is_cracked_insn): Adjust.
14540 (insn_must_be_first_in_group): Adjust.
14541 * config/rs6000/40x.md: Adjust.
14542 * config/rs6000/440.md: Adjust.
14543 * config/rs6000/476.md: Adjust.
14544 * config/rs6000/601.md: Adjust.
14545 * config/rs6000/603.md: Adjust.
14546 * config/rs6000/6xx.md: Adjust.
14547 * config/rs6000/7450.md: Adjust.
14548 * config/rs6000/7xx.md: Adjust.
14549 * config/rs6000/8540.md: Adjust.
14550 * config/rs6000/cell.md: Adjust.
14551 * config/rs6000/e300c2c3.md: Adjust.
14552 * config/rs6000/e500mc.md: Adjust.
14553 * config/rs6000/e500mc64.md: Adjust.
14554 * config/rs6000/e5500.md: Adjust.
14555 * config/rs6000/e6500.md: Adjust.
14556 * config/rs6000/mpc.md: Adjust.
14557 * config/rs6000/power4.md: Adjust.
14558 * config/rs6000/power5.md: Adjust.
14559 * config/rs6000/power6.md: Adjust.
14560 * config/rs6000/power7.md: Adjust.
14561 * config/rs6000/power8.md: Adjust.
14562 * config/rs6000/power9.md: Adjust.
14563 * config/rs6000/rs64.md: Adjust.
14564 * config/rs6000/titan.md: Adjust.
14566 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14568 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14569 ix86_indirect_branch_register logic.
14571 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14573 * config/i386/constraints.md (Bs): Update
14574 ix86_indirect_branch_register check. Don't check
14575 ix86_indirect_branch_register with GOT_memory_operand.
14577 * config/i386/predicates.md (GOT_memory_operand): Don't check
14578 ix86_indirect_branch_register here.
14579 (GOT32_symbol_operand): Likewise.
14581 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14583 * config/i386/predicates.md (constant_call_address_operand):
14584 Rewrite ix86_indirect_branch_register logic.
14585 (sibcall_insn_operand): Likewise.
14587 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14589 * config/i386/constraints.md (Bs): Replace
14590 ix86_indirect_branch_thunk_register with
14591 ix86_indirect_branch_register.
14593 * config/i386/i386.md (indirect_jump): Likewise.
14594 (tablejump): Likewise.
14595 (*sibcall_memory): Likewise.
14596 (*sibcall_value_memory): Likewise.
14597 Peepholes of indirect call and jump via memory: Likewise.
14598 * config/i386/i386.opt: Likewise.
14599 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14600 (GOT_memory_operand): Likewise.
14601 (call_insn_operand): Likewise.
14602 (sibcall_insn_operand): Likewise.
14603 (GOT32_symbol_operand): Likewise.
14605 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14607 PR middle-end/83837
14608 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14609 type rather than type addr's type points to.
14610 (expand_omp_atomic_mutex): Likewise.
14611 (expand_omp_atomic): Likewise.
14613 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14616 * config/i386/i386.c (output_indirect_thunk_function): Use
14617 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14618 for __x86_return_thunk.
14620 2018-01-15 Richard Biener <rguenther@suse.de>
14622 PR middle-end/83850
14623 * expmed.c (extract_bit_field_1): Fix typo.
14625 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14628 * config/arm/iterators.md (VF): New mode iterator.
14629 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14630 Remove integer-related logic from pattern.
14631 (neon_vabd<mode>_3): Likewise.
14633 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14635 PR middle-end/82694
14636 * common.opt (fstrict-overflow): No longer an alias.
14637 (fwrapv-pointer): New option.
14638 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14639 also for pointer types based on flag_wrapv_pointer.
14640 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14641 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14642 opts->x_flag_wrapv got set.
14643 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14644 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14645 POINTER_TYPE_OVERFLOW_UNDEFINED.
14646 * match.pd: Likewise in address comparison pattern.
14647 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14649 2018-01-15 Richard Biener <rguenther@suse.de>
14652 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14653 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14654 Reset type names to their identifier if their TYPE_DECL doesn't
14655 have linkage (and thus is used for ODR and devirt).
14656 (save_debug_info_for_decl): Remove.
14657 (save_debug_info_for_type): Likewise.
14658 (add_tree_to_fld_list): Adjust.
14659 * tree-pretty-print.c (dump_generic_node): Make dumping of
14660 type names more robust.
14662 2018-01-15 Richard Biener <rguenther@suse.de>
14664 * BASE-VER: Bump to 8.0.1.
14666 2018-01-14 Martin Sebor <msebor@redhat.com>
14669 * builtins.c (check_access): Avoid warning when the no-warning bit
14672 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14674 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14675 * ira-color (allocno_hard_regs_compare): Likewise.
14677 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14680 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14681 Use .pushsection/.popsection.
14683 2018-01-14 Martin Sebor <msebor@redhat.com>
14686 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14688 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14690 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14691 entry from extra_headers.
14692 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14693 extra_headers, make the list bitwise identical to the i?86-*-* one.
14695 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14697 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14698 -mcmodel=large with -mindirect-branch=thunk,
14699 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14700 -mfunction-return=thunk-extern.
14701 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14702 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14703 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14705 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14707 * config/i386/i386.c (print_reg): Print the name of the full
14708 integer register without '%'.
14709 (ix86_print_operand): Handle 'V'.
14710 * doc/extend.texi: Document 'V' modifier.
14712 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14714 * config/i386/constraints.md (Bs): Disallow memory operand for
14715 -mindirect-branch-register.
14717 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14718 (GOT_memory_operand): Likewise.
14719 (call_insn_operand): Likewise.
14720 (sibcall_insn_operand): Likewise.
14721 (GOT32_symbol_operand): Likewise.
14722 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14723 for -mindirect-branch-register.
14724 (tablejump): Likewise.
14725 (*sibcall_memory): Likewise.
14726 (*sibcall_value_memory): Likewise.
14727 Disallow peepholes of indirect call and jump via memory for
14728 -mindirect-branch-register.
14729 (*call_pop): Replace m with Bw.
14730 (*call_value_pop): Likewise.
14731 (*sibcall_pop_memory): Replace m with Bs.
14732 * config/i386/i386.opt (mindirect-branch-register): New option.
14733 * doc/invoke.texi: Document -mindirect-branch-register option.
14735 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14737 * config/i386/i386-protos.h (ix86_output_function_return): New.
14738 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14739 set function_return_type.
14740 (indirect_thunk_name): Add ret_p to indicate thunk for function
14742 (output_indirect_thunk_function): Pass false to
14743 indirect_thunk_name.
14744 (ix86_output_indirect_branch_via_reg): Likewise.
14745 (ix86_output_indirect_branch_via_push): Likewise.
14746 (output_indirect_thunk_function): Create alias for function
14747 return thunk if regno < 0.
14748 (ix86_output_function_return): New function.
14749 (ix86_handle_fndecl_attribute): Handle function_return.
14750 (ix86_attribute_table): Add function_return.
14751 * config/i386/i386.h (machine_function): Add
14752 function_return_type.
14753 * config/i386/i386.md (simple_return_internal): Use
14754 ix86_output_function_return.
14755 (simple_return_internal_long): Likewise.
14756 * config/i386/i386.opt (mfunction-return=): New option.
14757 (indirect_branch): Mention -mfunction-return=.
14758 * doc/extend.texi: Document function_return function attribute.
14759 * doc/invoke.texi: Document -mfunction-return= option.
14761 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14763 * config/i386/i386-opts.h (indirect_branch): New.
14764 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14765 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14766 with local indirect jump when converting indirect call and jump.
14767 (ix86_set_indirect_branch_type): New.
14768 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14769 (indirectlabelno): New.
14770 (indirect_thunk_needed): Likewise.
14771 (indirect_thunk_bnd_needed): Likewise.
14772 (indirect_thunks_used): Likewise.
14773 (indirect_thunks_bnd_used): Likewise.
14774 (INDIRECT_LABEL): Likewise.
14775 (indirect_thunk_name): Likewise.
14776 (output_indirect_thunk): Likewise.
14777 (output_indirect_thunk_function): Likewise.
14778 (ix86_output_indirect_branch_via_reg): Likewise.
14779 (ix86_output_indirect_branch_via_push): Likewise.
14780 (ix86_output_indirect_branch): Likewise.
14781 (ix86_output_indirect_jmp): Likewise.
14782 (ix86_code_end): Call output_indirect_thunk_function if needed.
14783 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14785 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14786 (ix86_attribute_table): Add indirect_branch.
14787 * config/i386/i386.h (machine_function): Add indirect_branch_type
14788 and has_local_indirect_jump.
14789 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14791 (tablejump): Likewise.
14792 (*indirect_jump): Use ix86_output_indirect_jmp.
14793 (*tablejump_1): Likewise.
14794 (simple_return_indirect_internal): Likewise.
14795 * config/i386/i386.opt (mindirect-branch=): New option.
14796 (indirect_branch): New.
14799 (thunk-inline): Likewise.
14800 (thunk-extern): Likewise.
14801 * doc/extend.texi: Document indirect_branch function attribute.
14802 * doc/invoke.texi: Document -mindirect-branch= option.
14804 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14807 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14809 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14811 * ipa-inline.c (want_inline_small_function_p): Return false if
14812 inlining has already failed with CIF_FINAL_ERROR.
14813 (update_caller_keys): Call want_inline_small_function_p before
14815 (update_callee_keys): Likewise.
14817 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14819 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14821 (rs6000_quadword_masked_address_p): Likewise.
14822 (quad_aligned_load_p): Likewise.
14823 (quad_aligned_store_p): Likewise.
14824 (const_load_sequence_p): Add comment to describe the outer-most loop.
14825 (mimic_memory_attributes_and_flags): New function.
14826 (rs6000_gen_stvx): Likewise.
14827 (replace_swapped_aligned_store): Likewise.
14828 (rs6000_gen_lvx): Likewise.
14829 (replace_swapped_aligned_load): Likewise.
14830 (replace_swapped_load_constant): Capitalize argument name in
14831 comment describing this function.
14832 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14833 and stores that access quad-word aligned addresses and replace
14834 with stvx or lvx instructions when appropriate.
14835 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14836 New function prototype.
14837 (rs6000_quadword_masked_address_p): Likewise.
14838 (rs6000_gen_lvx): Likewise.
14839 (rs6000_gen_stvx): Likewise.
14840 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14841 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14842 when memory address is aligned.
14843 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14844 this split to select lvx instruction when memory address is aligned.
14845 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14846 instruction when memory address is aligned.
14847 (*vsx_le_perm_load_v16qi): Likewise.
14848 (four unnamed splitters): Modify to select the stvx instruction
14849 when memory is aligned.
14851 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14853 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14854 which appears in the queue multiple times.
14856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14857 Alan Hayward <alan.hayward@arm.com>
14858 David Sherwood <david.sherwood@arm.com>
14860 * tree-vectorizer.h (vec_lower_bound): New structure.
14861 (_loop_vec_info): Add check_nonzero and lower_bounds.
14862 (LOOP_VINFO_CHECK_NONZERO): New macro.
14863 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14864 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14865 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14866 fields. Make seg_len the distance travelled, not including the
14868 (dr_direction_indicator): Declare.
14869 (dr_zero_step_indicator): Likewise.
14870 (dr_known_forward_stride_p): Likewise.
14871 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14873 (runtime_alias_check_p): Allow runtime alias checks with
14875 (operator ==): Compare access_size and align.
14876 (prune_runtime_alias_test_list): Rework for new distinction between
14877 the access_size and seg_len.
14878 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14880 (get_segment_min_max): New function.
14881 (create_intersect_range_checks): Use it.
14882 (dr_step_indicator): New function.
14883 (dr_direction_indicator): Likewise.
14884 (dr_zero_step_indicator): Likewise.
14885 (dr_known_forward_stride_p): Likewise.
14886 * tree-loop-distribution.c (data_ref_segment_size): Return
14887 DR_STEP * (niters - 1).
14888 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14890 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14891 (vect_preserves_scalar_order_p): New function, split out from...
14892 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14893 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14894 (vect_vfa_access_size): New function.
14895 (vect_vfa_align): Likewise.
14896 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14897 (dump_lower_bound): New function.
14898 (vect_check_lower_bound): Likewise.
14899 (vect_small_gap_p): Likewise.
14900 (vectorizable_with_step_bound_p): Likewise.
14901 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14902 depencies if the vectorization factor is 1. Convert the checks
14903 for nonzero steps into checks on the bounds of DR_STEP. Try using
14904 a bunds check for variable steps if the minimum required step is
14905 relatively small. Update calls to the dr_with_seg_len
14906 constructor and to vect_compile_time_alias.
14907 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14909 (vect_loop_versioning): Call it.
14910 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14912 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14915 Alan Hayward <alan.hayward@arm.com>
14916 David Sherwood <david.sherwood@arm.com>
14918 * doc/sourcebuild.texi (vect_scatter_store): Document.
14919 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14921 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14923 * genopinit.c (main): Add supports_vec_scatter_store and
14924 supports_vec_scatter_store_cached to target_optabs.
14925 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14926 IFN_MASK_SCATTER_STORE.
14927 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14929 * internal-fn.h (internal_store_fn_p): Declare.
14930 (internal_fn_stored_value_index): Likewise.
14931 * internal-fn.c (scatter_store_direct): New macro.
14932 (expand_scatter_store_optab_fn): New function.
14933 (direct_scatter_store_optab_supported_p): New macro.
14934 (internal_store_fn_p): New function.
14935 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14936 IFN_MASK_SCATTER_STORE.
14937 (internal_fn_mask_index): Likewise.
14938 (internal_fn_stored_value_index): New function.
14939 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14940 for scatter stores.
14941 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14942 * optabs-query.c (supports_vec_scatter_store_p): New function.
14943 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14944 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14945 true for scatter stores.
14946 (vect_gather_scatter_fn_p): Handle scatter stores too.
14947 (vect_check_gather_scatter): Consider using scatter stores if
14948 supports_vec_scatter_store_p.
14949 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14950 scatter stores too.
14951 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14952 internal_fn_stored_value_index.
14953 (check_load_store_masking): Handle scatter stores too.
14954 (vect_get_store_rhs): Make public.
14955 (vectorizable_call): Use internal_store_fn_p.
14956 (vectorizable_store): Handle scatter store internal functions.
14957 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14958 when deciding whether the end of the group has been reached.
14959 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14960 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14961 (mask_scatter_store<mode>): New insns.
14963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14964 Alan Hayward <alan.hayward@arm.com>
14965 David Sherwood <david.sherwood@arm.com>
14967 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14968 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14969 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14971 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14972 Use vect_truncate_gather_scatter_offset if we can't treat the
14973 operation as a normal gather load or scatter store.
14974 (get_group_load_store_type): Take the gather_scatter_info
14975 as argument. Try using a gather load or scatter store for
14976 single-element groups.
14977 (get_load_store_type): Update calls to get_group_load_store_type
14978 and vect_use_strided_gather_scatters_p.
14980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14981 Alan Hayward <alan.hayward@arm.com>
14982 David Sherwood <david.sherwood@arm.com>
14984 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14985 optional tree argument.
14986 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14988 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14989 but continue to use the current value as a fallback.
14990 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14991 to compare the updates.
14992 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14993 (get_load_store_type): Use it when handling a strided access.
14994 (vect_get_strided_load_store_ops): New function.
14995 (vect_get_data_ptr_increment): Likewise.
14996 (vectorizable_load): Handle strided gather loads. Always pass
14997 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15000 Alan Hayward <alan.hayward@arm.com>
15001 David Sherwood <david.sherwood@arm.com>
15003 * doc/md.texi (gather_load@var{m}): Document.
15004 (mask_gather_load@var{m}): Likewise.
15005 * genopinit.c (main): Add supports_vec_gather_load and
15006 supports_vec_gather_load_cached to target_optabs.
15007 * optabs-tree.c (init_tree_optimization_optabs): Use
15008 ggc_cleared_alloc to allocate target_optabs.
15009 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15010 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15012 * internal-fn.h (internal_load_fn_p): Declare.
15013 (internal_gather_scatter_fn_p): Likewise.
15014 (internal_fn_mask_index): Likewise.
15015 (internal_gather_scatter_fn_supported_p): Likewise.
15016 * internal-fn.c (gather_load_direct): New macro.
15017 (expand_gather_load_optab_fn): New function.
15018 (direct_gather_load_optab_supported_p): New macro.
15019 (direct_internal_fn_optab): New function.
15020 (internal_load_fn_p): Likewise.
15021 (internal_gather_scatter_fn_p): Likewise.
15022 (internal_fn_mask_index): Likewise.
15023 (internal_gather_scatter_fn_supported_p): Likewise.
15024 * optabs-query.c (supports_at_least_one_mode_p): New function.
15025 (supports_vec_gather_load_p): Likewise.
15026 * optabs-query.h (supports_vec_gather_load_p): Declare.
15027 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15028 and memory_type field.
15029 (NUM_PATTERNS): Bump to 15.
15030 * tree-vect-data-refs.c: Include internal-fn.h.
15031 (vect_gather_scatter_fn_p): New function.
15032 (vect_describe_gather_scatter_call): Likewise.
15033 (vect_check_gather_scatter): Try using internal functions for
15034 gather loads. Recognize existing calls to a gather load function.
15035 (vect_analyze_data_refs): Consider using gather loads if
15036 supports_vec_gather_load_p.
15037 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15038 (vect_get_gather_scatter_offset_type): Likewise.
15039 (vect_convert_mask_for_vectype): Likewise.
15040 (vect_add_conversion_to_patterm): Likewise.
15041 (vect_try_gather_scatter_pattern): Likewise.
15042 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15043 (vect_vect_recog_func_ptrs): Add it.
15044 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15045 internal_fn_mask_index and internal_gather_scatter_fn_p.
15046 (check_load_store_masking): Take the gather_scatter_info as an
15047 argument and handle gather loads.
15048 (vect_get_gather_scatter_ops): New function.
15049 (vectorizable_call): Check internal_load_fn_p.
15050 (vectorizable_load): Likewise. Handle gather load internal
15052 (vectorizable_store): Update call to check_load_store_masking.
15053 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15054 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15055 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15056 (aarch64_gather_scale_operand_d): New predicates.
15057 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15058 (mask_gather_load<mode>): New insns.
15060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15061 Alan Hayward <alan.hayward@arm.com>
15062 David Sherwood <david.sherwood@arm.com>
15064 * optabs.def (fold_left_plus_optab): New optab.
15065 * doc/md.texi (fold_left_plus_@var{m}): Document.
15066 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15067 * internal-fn.c (fold_left_direct): Define.
15068 (expand_fold_left_optab_fn): Likewise.
15069 (direct_fold_left_optab_supported_p): Likewise.
15070 * fold-const-call.c (fold_const_fold_left): New function.
15071 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15072 * tree-parloops.c (valid_reduction_p): New function.
15073 (gather_scalar_reductions): Use it.
15074 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15075 (vect_finish_replace_stmt): Declare.
15076 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15077 (needs_fold_left_reduction_p): New function, split out from...
15078 (vect_is_simple_reduction): ...here. Accept reductions that
15079 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15080 (vect_force_simple_reduction): Also store the reduction type in
15081 the assignment's STMT_VINFO_REDUC_TYPE.
15082 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15083 (merge_with_identity): New function.
15084 (vect_expand_fold_left): Likewise.
15085 (vectorize_fold_left_reduction): Likewise.
15086 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15087 scalar phi in place for it. Check for target support and reject
15088 cases that would reassociate the operation. Defer the transform
15089 phase to vectorize_fold_left_reduction.
15090 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15091 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15092 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15096 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15097 call to ifc_temp_var.
15099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15100 Alan Hayward <alan.hayward@arm.com>
15101 David Sherwood <david.sherwood@arm.com>
15103 * target.def (legitimize_address_displacement): Take the original
15104 offset as a poly_int.
15105 * targhooks.h (default_legitimize_address_displacement): Update
15107 * targhooks.c (default_legitimize_address_displacement): Likewise.
15108 * doc/tm.texi: Regenerate.
15109 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15110 as an argument, moving assert of ad->disp == ad->disp_term to...
15111 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15112 Try calling targetm.legitimize_address_displacement before expanding
15113 the address rather than afterwards, and adjust for the new interface.
15114 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15115 Match the new hook interface. Handle SVE addresses.
15116 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15117 new hook interface.
15119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15121 * Makefile.in (OBJS): Add early-remat.o.
15122 * target.def (select_early_remat_modes): New hook.
15123 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15124 * doc/tm.texi: Regenerate.
15125 * targhooks.h (default_select_early_remat_modes): Declare.
15126 * targhooks.c (default_select_early_remat_modes): New function.
15127 * timevar.def (TV_EARLY_REMAT): New timevar.
15128 * passes.def (pass_early_remat): New pass.
15129 * tree-pass.h (make_pass_early_remat): Declare.
15130 * early-remat.c: New file.
15131 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15133 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15136 Alan Hayward <alan.hayward@arm.com>
15137 David Sherwood <david.sherwood@arm.com>
15139 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15140 vfm1 with a bound_epilog parameter.
15141 (vect_do_peeling): Update calls accordingly, and move the prologue
15142 call earlier in the function. Treat the base bound_epilog as 0 for
15143 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15144 this base when peeling for gaps.
15145 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15146 with fully-masked loops.
15147 (vect_estimate_min_profitable_iters): Handle the single peeled
15148 iteration in that case.
15150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15151 Alan Hayward <alan.hayward@arm.com>
15152 David Sherwood <david.sherwood@arm.com>
15154 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15155 single-element interleaving even if the size is not a power of 2.
15156 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15157 accesses for single-element interleaving if the group size is
15160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15161 Alan Hayward <alan.hayward@arm.com>
15162 David Sherwood <david.sherwood@arm.com>
15164 * doc/md.texi (fold_extract_last_@var{m}): Document.
15165 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15166 * optabs.def (fold_extract_last_optab): New optab.
15167 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15168 * internal-fn.c (fold_extract_direct): New macro.
15169 (expand_fold_extract_optab_fn): Likewise.
15170 (direct_fold_extract_optab_supported_p): Likewise.
15171 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15172 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15173 EXTRACT_LAST_REDUCTION.
15174 (get_initial_def_for_reduction): Do not create an initial vector
15175 for EXTRACT_LAST_REDUCTION reductions.
15176 (vectorizable_reduction): Leave the scalar phi in place for
15177 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15178 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15179 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15180 transform phase to vectorizable_condition.
15181 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15183 (vect_finish_stmt_generation): ...here.
15184 (vect_finish_replace_stmt): New function.
15185 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15186 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15188 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15191 Alan Hayward <alan.hayward@arm.com>
15192 David Sherwood <david.sherwood@arm.com>
15194 * doc/md.texi (extract_last_@var{m}): Document.
15195 * optabs.def (extract_last_optab): New optab.
15196 * internal-fn.def (EXTRACT_LAST): New internal function.
15197 * internal-fn.c (cond_unary_direct): New macro.
15198 (expand_cond_unary_optab_fn): Likewise.
15199 (direct_cond_unary_optab_supported_p): Likewise.
15200 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15201 loops using EXTRACT_LAST.
15202 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15203 (extract_last_<mode>): ...this optab.
15204 (vec_extract<mode><Vel>): Update accordingly.
15206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15207 Alan Hayward <alan.hayward@arm.com>
15208 David Sherwood <david.sherwood@arm.com>
15210 * target.def (empty_mask_is_expensive): New hook.
15211 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15212 * doc/tm.texi: Regenerate.
15213 * targhooks.h (default_empty_mask_is_expensive): Declare.
15214 * targhooks.c (default_empty_mask_is_expensive): New function.
15215 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15216 if the target says that empty masks are expensive.
15217 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15219 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15222 Alan Hayward <alan.hayward@arm.com>
15223 David Sherwood <david.sherwood@arm.com>
15225 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15226 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15227 (vect_use_loop_mask_for_alignment_p): New function.
15228 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15229 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15230 niters_skip argument. Make sure that the first niters_skip elements
15231 of the first iteration are inactive.
15232 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15233 Update call to vect_set_loop_masks_directly.
15234 (get_misalign_in_elems): New function, split out from...
15235 (vect_gen_prolog_loop_niters): ...here.
15236 (vect_update_init_of_dr): Take a code argument that specifies whether
15237 the adjustment should be added or subtracted.
15238 (vect_update_init_of_drs): Likewise.
15239 (vect_prepare_for_masked_peels): New function.
15240 (vect_do_peeling): Skip prologue peeling if we're using a mask
15241 instead. Update call to vect_update_inits_of_drs.
15242 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15244 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15245 alignment. Do not include the number of peeled iterations in
15246 the minimum threshold in that case.
15247 (vectorizable_induction): Adjust the start value down by
15248 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15249 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15250 Take the number of skipped iterations into account when calculating
15252 * tree-vect-stmts.c (vect_gen_while_not): New function.
15254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15255 Alan Hayward <alan.hayward@arm.com>
15256 David Sherwood <david.sherwood@arm.com>
15258 * doc/sourcebuild.texi (vect_fully_masked): Document.
15259 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15260 default value to 0.
15261 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15263 (vect_analyze_loop_2): ...here. Don't check the vectorization
15264 factor against the number of loop iterations if the loop is
15267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15268 Alan Hayward <alan.hayward@arm.com>
15269 David Sherwood <david.sherwood@arm.com>
15271 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15272 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15273 (dump_groups): Update accordingly.
15274 (iv_use::mem_type): New member variable.
15275 (address_p): New function.
15276 (record_use): Add a mem_type argument and initialize the new
15278 (record_group_use): Add a mem_type argument. Use address_p.
15279 Remove obsolete null checks of base_object. Update call to record_use.
15280 (find_interesting_uses_op): Update call to record_group_use.
15281 (find_interesting_uses_cond): Likewise.
15282 (find_interesting_uses_address): Likewise.
15283 (get_mem_type_for_internal_fn): New function.
15284 (find_address_like_use): Likewise.
15285 (find_interesting_uses_stmt): Try find_address_like_use before
15286 calling find_interesting_uses_op.
15287 (addr_offset_valid_p): Use the iv mem_type field as the type
15288 of the addressed memory.
15289 (add_autoinc_candidates): Likewise.
15290 (get_address_cost): Likewise.
15291 (split_small_address_groups_p): Use address_p.
15292 (split_address_groups): Likewise.
15293 (add_iv_candidate_for_use): Likewise.
15294 (autoinc_possible_for_pair): Likewise.
15295 (rewrite_groups): Likewise.
15296 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15297 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15298 (get_alias_ptr_type_for_ptr_address): New function.
15299 (rewrite_use_address): Rewrite address uses in calls that were
15300 identified by find_address_like_use.
15302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15303 Alan Hayward <alan.hayward@arm.com>
15304 David Sherwood <david.sherwood@arm.com>
15306 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15308 * gimple-expr.h (is_gimple_addressable: Likewise.
15309 * gimple-expr.c (is_gimple_address): Likewise.
15310 * internal-fn.c (expand_call_mem_ref): New function.
15311 (expand_mask_load_optab_fn): Use it.
15312 (expand_mask_store_optab_fn): Likewise.
15314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15315 Alan Hayward <alan.hayward@arm.com>
15316 David Sherwood <david.sherwood@arm.com>
15318 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15319 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15320 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15321 (cond_umax@var{mode}): Document.
15322 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15323 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15324 (cond_umin_optab, cond_umax_optab): New optabs.
15325 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15326 (COND_IOR, COND_XOR): New internal functions.
15327 * internal-fn.h (get_conditional_internal_fn): Declare.
15328 * internal-fn.c (cond_binary_direct): New macro.
15329 (expand_cond_binary_optab_fn): Likewise.
15330 (direct_cond_binary_optab_supported_p): Likewise.
15331 (get_conditional_internal_fn): New function.
15332 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15333 Cope with reduction statements that are vectorized as calls rather
15335 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15336 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15337 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15338 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15339 (UNSPEC_COND_EOR): New unspecs.
15340 (optab): Add mappings for them.
15341 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15342 (sve_int_op, sve_fp_op): New int attributes.
15344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15345 Alan Hayward <alan.hayward@arm.com>
15346 David Sherwood <david.sherwood@arm.com>
15348 * optabs.def (while_ult_optab): New optab.
15349 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15350 * internal-fn.def (WHILE_ULT): New internal function.
15351 * internal-fn.h (direct_internal_fn_supported_p): New override
15352 that takes two types as argument.
15353 * internal-fn.c (while_direct): New macro.
15354 (expand_while_optab_fn): New function.
15355 (convert_optab_supported_p): Likewise.
15356 (direct_while_optab_supported_p): New macro.
15357 * wide-int.h (wi::udiv_ceil): New function.
15358 * tree-vectorizer.h (rgroup_masks): New structure.
15359 (vec_loop_masks): New typedef.
15360 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15361 and fully_masked_p.
15362 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15363 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15364 (vect_max_vf): New function.
15365 (slpeel_make_loop_iterate_ntimes): Delete.
15366 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15367 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15368 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15369 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15370 internal-fn.h, stor-layout.h and optabs-query.h.
15371 (vect_set_loop_mask): New function.
15372 (add_preheader_seq): Likewise.
15373 (add_header_seq): Likewise.
15374 (interleave_supported_p): Likewise.
15375 (vect_maybe_permute_loop_masks): Likewise.
15376 (vect_set_loop_masks_directly): Likewise.
15377 (vect_set_loop_condition_masked): Likewise.
15378 (vect_set_loop_condition_unmasked): New function, split out from
15379 slpeel_make_loop_iterate_ntimes.
15380 (slpeel_make_loop_iterate_ntimes): Rename to..
15381 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15382 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15383 (vect_do_peeling): Update call accordingly.
15384 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15386 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15387 mask_compare_type, can_fully_mask_p and fully_masked_p.
15388 (release_vec_loop_masks): New function.
15389 (_loop_vec_info): Use it to free the loop masks.
15390 (can_produce_all_loop_masks_p): New function.
15391 (vect_get_max_nscalars_per_iter): Likewise.
15392 (vect_verify_full_masking): Likewise.
15393 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15394 retries, and free the mask rgroups before retrying. Check loop-wide
15395 reasons for disallowing fully-masked loops. Make the final decision
15396 about whether use a fully-masked loop or not.
15397 (vect_estimate_min_profitable_iters): Do not assume that peeling
15398 for the number of iterations will be needed for fully-masked loops.
15399 (vectorizable_reduction): Disable fully-masked loops.
15400 (vectorizable_live_operation): Likewise.
15401 (vect_halve_mask_nunits): New function.
15402 (vect_double_mask_nunits): Likewise.
15403 (vect_record_loop_mask): Likewise.
15404 (vect_get_loop_mask): Likewise.
15405 (vect_transform_loop): Handle the case in which the final loop
15406 iteration might handle a partial vector. Call vect_set_loop_condition
15407 instead of slpeel_make_loop_iterate_ntimes.
15408 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15409 (check_load_store_masking): New function.
15410 (prepare_load_store_mask): Likewise.
15411 (vectorizable_store): Handle fully-masked loops.
15412 (vectorizable_load): Likewise.
15413 (supportable_widening_operation): Use vect_halve_mask_nunits for
15415 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15416 (vect_gen_while): New function.
15417 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15418 (aarch64_uqdec<mode>): New insn.
15420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15421 Alan Hayward <alan.hayward@arm.com>
15422 David Sherwood <david.sherwood@arm.com>
15424 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15425 (reduc_xor_scal_optab): New optabs.
15426 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15427 (reduc_xor_scal_@var{m}): Document.
15428 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15429 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15430 internal functions.
15431 * fold-const-call.c (fold_const_call): Handle them.
15432 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15433 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15434 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15435 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15436 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15437 (UNSPEC_XORV): New unspecs.
15438 (optab): Add entries for them.
15439 (BITWISEV): New int iterator.
15440 (bit_reduc_op): New int attributes.
15442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15443 Alan Hayward <alan.hayward@arm.com>
15444 David Sherwood <david.sherwood@arm.com>
15446 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15447 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15448 * optabs.def (vec_shl_insert_optab): New optab.
15449 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15450 (duplicate_and_interleave): Likewise.
15451 * tree-vect-loop.c: Include internal-fn.h.
15452 (neutral_op_for_slp_reduction): New function, split out from
15453 get_initial_defs_for_reduction.
15454 (get_initial_def_for_reduction): Handle option 2 for variable-length
15455 vectors by loading the neutral value into a vector and then shifting
15456 the initial value into element 0.
15457 (get_initial_defs_for_reduction): Replace the code argument with
15458 the neutral value calculated by neutral_op_for_slp_reduction.
15459 Use gimple_build_vector for constant-length vectors.
15460 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15461 but the first group_size elements have a neutral value.
15462 Use duplicate_and_interleave otherwise.
15463 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15464 Update call to get_initial_defs_for_reduction. Handle SLP
15465 reductions for variable-length vectors by creating one vector
15466 result for each scalar result, with the elements associated
15467 with other scalar results stubbed out with the neutral value.
15468 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15469 Require IFN_VEC_SHL_INSERT for double reductions on
15470 variable-length vectors, or SLP reductions that have
15471 a neutral value. Require can_duplicate_and_interleave_p
15472 support for variable-length unchained SLP reductions if there
15473 is no neutral value, such as for MIN/MAX reductions. Also require
15474 the number of vector elements to be a multiple of the number of
15475 SLP statements when doing variable-length unchained SLP reductions.
15476 Update call to vect_create_epilog_for_reduction.
15477 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15478 and remove initial values.
15479 (duplicate_and_interleave): Make public.
15480 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15481 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15484 Alan Hayward <alan.hayward@arm.com>
15485 David Sherwood <david.sherwood@arm.com>
15487 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15488 (can_duplicate_and_interleave_p): New function.
15489 (vect_get_and_check_slp_defs): Take the vector of statements
15490 rather than just the current one. Remove excess parentheses.
15491 Restriction rejectinon of vect_constant_def and vect_external_def
15492 for variable-length vectors to boolean types, or types for which
15493 can_duplicate_and_interleave_p is false.
15494 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15495 (duplicate_and_interleave): New function.
15496 (vect_get_constant_vectors): Use gimple_build_vector for
15497 constant-length vectors and suitable variable-length constant
15498 vectors. Use duplicate_and_interleave for other variable-length
15499 vectors. Don't defer the update when inserting new statements.
15501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15502 Alan Hayward <alan.hayward@arm.com>
15503 David Sherwood <david.sherwood@arm.com>
15505 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15506 min_profitable_iters doesn't go negative.
15508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15509 Alan Hayward <alan.hayward@arm.com>
15510 David Sherwood <david.sherwood@arm.com>
15512 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15513 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15514 * optabs.def (vec_mask_load_lanes_optab): New optab.
15515 (vec_mask_store_lanes_optab): Likewise.
15516 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15517 (MASK_STORE_LANES): Likewise.
15518 * internal-fn.c (mask_load_lanes_direct): New macro.
15519 (mask_store_lanes_direct): Likewise.
15520 (expand_mask_load_optab_fn): Handle masked operations.
15521 (expand_mask_load_lanes_optab_fn): New macro.
15522 (expand_mask_store_optab_fn): Handle masked operations.
15523 (expand_mask_store_lanes_optab_fn): New macro.
15524 (direct_mask_load_lanes_optab_supported_p): Likewise.
15525 (direct_mask_store_lanes_optab_supported_p): Likewise.
15526 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15528 (vect_load_lanes_supported): Likewise.
15529 * tree-vect-data-refs.c (strip_conversion): New function.
15530 (can_group_stmts_p): Likewise.
15531 (vect_analyze_data_ref_accesses): Use it instead of checking
15532 for a pair of assignments.
15533 (vect_store_lanes_supported): Take a masked_p parameter.
15534 (vect_load_lanes_supported): Likewise.
15535 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15536 vect_store_lanes_supported and vect_load_lanes_supported.
15537 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15538 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15539 parameter. Don't allow gaps for masked accesses.
15540 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15541 and vect_load_lanes_supported.
15542 (get_load_store_type): Take a masked_p parameter and update
15543 call to get_group_load_store_type.
15544 (vectorizable_store): Update call to get_load_store_type.
15545 Handle IFN_MASK_STORE_LANES.
15546 (vectorizable_load): Update call to get_load_store_type.
15547 Handle IFN_MASK_LOAD_LANES.
15549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15550 Alan Hayward <alan.hayward@arm.com>
15551 David Sherwood <david.sherwood@arm.com>
15553 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15555 * config/aarch64/aarch64-protos.h
15556 (aarch64_sve_struct_memory_operand_p): Declare.
15557 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15558 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15559 (VPRED, vpred): Handle SVE structure modes.
15560 * config/aarch64/constraints.md (Utx): New constraint.
15561 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15562 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15563 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15564 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15565 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15566 structure modes. Split into pieces after RA.
15567 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15568 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15570 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15571 SVE structure modes.
15572 (aarch64_classify_address): Likewise.
15573 (sizetochar): Move earlier in file.
15574 (aarch64_print_operand): Handle SVE register lists.
15575 (aarch64_array_mode): New function.
15576 (aarch64_sve_struct_memory_operand_p): Likewise.
15577 (TARGET_ARRAY_MODE): Redefine.
15579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15580 Alan Hayward <alan.hayward@arm.com>
15581 David Sherwood <david.sherwood@arm.com>
15583 * target.def (array_mode): New target hook.
15584 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15585 * doc/tm.texi: Regenerate.
15586 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15587 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15588 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15589 targetm.array_mode.
15590 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15594 Alan Hayward <alan.hayward@arm.com>
15595 David Sherwood <david.sherwood@arm.com>
15597 * fold-const.c (fold_binary_loc): Check the argument types
15598 rather than the result type when testing for a vector operation.
15600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15602 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15603 * doc/tm.texi: Regenerate.
15605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15606 Alan Hayward <alan.hayward@arm.com>
15607 David Sherwood <david.sherwood@arm.com>
15609 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15610 (sve): Document new AArch64 extension.
15611 * doc/md.texi (w): Extend the description of the AArch64
15612 constraint to include SVE vectors.
15613 (Upl, Upa): Document new AArch64 predicate constraints.
15614 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15616 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15617 (msve-vector-bits=): New option.
15618 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15619 SVE when these are disabled.
15620 (sve): New extension.
15621 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15622 modes. Adjust their number of units based on aarch64_sve_vg.
15623 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15624 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15625 aarch64_addr_query_type.
15626 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15627 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15628 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15629 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15630 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15631 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15632 (aarch64_simd_imm_zero_p): Delete.
15633 (aarch64_check_zero_based_sve_index_immediate): Declare.
15634 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15635 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15636 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15637 (aarch64_sve_float_mul_immediate_p): Likewise.
15638 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15639 rather than an rtx.
15640 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15641 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15642 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15643 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15644 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15645 (aarch64_regmode_natural_size): Likewise.
15646 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15647 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15649 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15650 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15651 for VG and the SVE predicate registers.
15652 (V_ALIASES): Add a "z"-prefixed alias.
15653 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15654 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15655 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15656 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15657 (REG_CLASS_NAMES): Add entries for them.
15658 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15659 and the predicate registers.
15660 (aarch64_sve_vg): Declare.
15661 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15662 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15663 (REGMODE_NATURAL_SIZE): Define.
15664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15666 * config/aarch64/aarch64.c: Include cfgrtl.h.
15667 (simd_immediate_info): Add a constructor for series vectors,
15668 and an associated step field.
15669 (aarch64_sve_vg): New variable.
15670 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15671 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15672 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15673 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15674 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15675 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15676 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15677 (aarch64_get_mask_mode): New functions.
15678 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15679 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15680 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15681 predicate modes and predicate registers. Explicitly restrict
15682 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15683 to store a vector mode if it is recognized by
15684 aarch64_classify_vector_mode.
15685 (aarch64_regmode_natural_size): New function.
15686 (aarch64_hard_regno_caller_save_mode): Return the original mode
15688 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15689 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15690 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15691 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15693 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15694 does not overlap dest if the function is frame-related. Handle
15696 (aarch64_split_add_offset): New function.
15697 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15698 them aarch64_add_offset.
15699 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15700 and update call to aarch64_sub_sp.
15701 (aarch64_add_cfa_expression): New function.
15702 (aarch64_expand_prologue): Pass extra temporary registers to the
15703 functions above. Handle the case in which we need to emit new
15704 DW_CFA_expressions for registers that were originally saved
15705 relative to the stack pointer, but now have to be expressed
15706 relative to the frame pointer.
15707 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15709 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15710 IP0 and IP1 values for SVE frames.
15711 (aarch64_expand_vec_series): New function.
15712 (aarch64_expand_sve_widened_duplicate): Likewise.
15713 (aarch64_expand_sve_const_vector): Likewise.
15714 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15715 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15716 into the register, rather than emitting a SET directly.
15717 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15718 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15719 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15720 (offset_9bit_signed_scaled_p): New functions.
15721 (aarch64_replicate_bitmask_imm): New function.
15722 (aarch64_bitmask_imm): Use it.
15723 (aarch64_cannot_force_const_mem): Reject expressions involving
15724 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15725 (aarch64_classify_index): Handle SVE indices, by requiring
15726 a plain register index with a scale that matches the element size.
15727 (aarch64_classify_address): Handle SVE addresses. Assert that
15728 the mode of the address is VOIDmode or an integer mode.
15729 Update call to aarch64_classify_symbol.
15730 (aarch64_classify_symbolic_expression): Update call to
15731 aarch64_classify_symbol.
15732 (aarch64_const_vec_all_in_range_p): New function.
15733 (aarch64_print_vector_float_operand): Likewise.
15734 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15735 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15736 and the FP immediates 1.0 and 0.5.
15737 (aarch64_print_address_internal): Handle SVE addresses.
15738 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15739 (aarch64_regno_regclass): Handle predicate registers.
15740 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15742 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15743 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15744 (aarch64_convert_sve_vector_bits): New function.
15745 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15746 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15747 rather than an rtx.
15748 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15749 Handle SVE vector and predicate modes. Accept VL-based constants
15750 that need only one temporary register, and VL offsets that require
15751 no temporary registers.
15752 (aarch64_conditional_register_usage): Mark the predicate registers
15753 as fixed if SVE isn't available.
15754 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15755 Return true for SVE vector and predicate modes.
15756 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15757 rather than an unsigned int. Handle SVE modes.
15758 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15760 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15762 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15763 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15764 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15765 (aarch64_sve_float_mul_immediate_p): New functions.
15766 (aarch64_sve_valid_immediate): New function.
15767 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15768 Explicitly reject structure modes. Check for INDEX constants.
15769 Handle PTRUE and PFALSE constants.
15770 (aarch64_check_zero_based_sve_index_immediate): New function.
15771 (aarch64_simd_imm_zero_p): Delete.
15772 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15773 vector modes. Accept constants in the range of CNT[BHWD].
15774 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15775 ask for an Advanced SIMD mode.
15776 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15777 (aarch64_simd_vector_alignment): Handle SVE predicates.
15778 (aarch64_vectorize_preferred_vector_alignment): New function.
15779 (aarch64_simd_vector_alignment_reachable): Use it instead of
15781 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15782 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15784 (MAX_VECT_LEN): Delete.
15785 (expand_vec_perm_d): Add a vec_flags field.
15786 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15787 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15788 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15790 (aarch64_evpc_rev): Rename to...
15791 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15792 (aarch64_evpc_rev_global): New function.
15793 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15794 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15796 (aarch64_evpc_sve_tbl): New function.
15797 (aarch64_expand_vec_perm_const_1): Update after rename of
15798 aarch64_evpc_rev. Handle SVE permutes too, trying
15799 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15800 than aarch64_evpc_tbl.
15801 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15802 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15803 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15804 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15805 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15806 (aarch64_expand_sve_vcond): New functions.
15807 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15808 of aarch64_vector_mode_p.
15809 (aarch64_dwarf_poly_indeterminate_value): New function.
15810 (aarch64_compute_pressure_classes): Likewise.
15811 (aarch64_can_change_mode_class): Likewise.
15812 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15813 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15814 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15815 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15816 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15817 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15818 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15819 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15821 (Dn, Dl, Dr): Accept const as well as const_vector.
15822 (Dz): Likewise. Compare against CONST0_RTX.
15823 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15824 of "vector" where appropriate.
15825 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15826 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15827 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15828 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15829 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15830 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15831 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15832 (v_int_equiv): Extend to SVE modes.
15833 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15835 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15836 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15837 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15838 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15839 (SVE_COND_FP_CMP): New int iterators.
15840 (perm_hilo): Handle the new unpack unspecs.
15841 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15843 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15844 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15845 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15846 (aarch64_equality_operator, aarch64_constant_vector_operand)
15847 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15848 (aarch64_sve_nonimmediate_operand): Likewise.
15849 (aarch64_sve_general_operand): Likewise.
15850 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15851 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15852 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15853 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15854 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15855 (aarch64_sve_float_arith_immediate): Likewise.
15856 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15857 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15858 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15859 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15860 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15861 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15862 (aarch64_sve_float_arith_operand): Likewise.
15863 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15864 (aarch64_sve_float_mul_operand): Likewise.
15865 (aarch64_sve_vec_perm_operand): Likewise.
15866 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15867 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15868 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15869 as well as const_vector.
15870 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15871 in file. Use CONST0_RTX and CONSTM1_RTX.
15872 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15873 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15874 Use aarch64_simd_imm_zero.
15875 * config/aarch64/aarch64-sve.md: New file.
15876 * config/aarch64/aarch64.md: Include it.
15877 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15878 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15879 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15880 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15881 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15882 (sve): New attribute.
15883 (enabled): Disable instructions with the sve attribute unless
15885 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15886 aarch64_expand_mov_immediate.
15887 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15888 CNT[BHSD] immediates.
15889 (movti): Split CONST_POLY_INT moves into two halves.
15890 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15891 Split additions that need a temporary here if the destination
15892 is the stack pointer.
15893 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15894 (*add<mode>3_poly_1): New instruction.
15895 (set_clobber_cc): New expander.
15897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15899 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15900 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15901 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15902 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15903 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15904 Change innermode from fixed_mode_size to machine_mode.
15905 (simplify_subreg): Update call accordingly. Handle a constant-sized
15906 subreg of a variable-length CONST_VECTOR.
15908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15909 Alan Hayward <alan.hayward@arm.com>
15910 David Sherwood <david.sherwood@arm.com>
15912 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15913 (add_offset_to_base): New function, split out from...
15914 (create_mem_ref): ...here. When handling a scale other than 1,
15915 check first whether the address is valid without the offset.
15916 Add it into the base if so, leaving the index and scale as-is.
15918 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15921 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15922 fold_for_warn before checking if arg2 is INTEGER_CST.
15924 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15926 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15927 (store_multiple_operation): Delete.
15928 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15929 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15930 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15931 guarded by TARGET_STRING.
15932 (rs6000_output_load_multiple): Delete.
15933 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15934 OPTION_MASK_STRING / TARGET_STRING handling.
15935 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15936 (const rs6000_opt_masks) <"string">: Change mask to 0.
15937 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15938 (MASK_STRING): Delete.
15939 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15941 (load_multiple): Delete.
15948 (store_multiple): Delete.
15955 (movmemsi_8reg): Delete.
15956 (corresponding unnamed define_insn): Delete.
15957 (movmemsi_6reg): Delete.
15958 (corresponding unnamed define_insn): Delete.
15959 (movmemsi_4reg): Delete.
15960 (corresponding unnamed define_insn): Delete.
15961 (movmemsi_2reg): Delete.
15962 (corresponding unnamed define_insn): Delete.
15963 (movmemsi_1reg): Delete.
15964 (corresponding unnamed define_insn): Delete.
15965 * config/rs6000/rs6000.opt (mno-string): New.
15966 (mstring): Replace by deprecation warning stub.
15967 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15969 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15971 * regrename.c (regrename_do_replace): If replacing the same
15972 reg multiple times, try to reuse last created gen_raw_REG.
15975 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15976 main to workaround a bug in GDB.
15978 2018-01-12 Tom de Vries <tom@codesourcery.com>
15981 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15983 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15985 PR rtl-optimization/80481
15986 * ira-color.c (get_cap_member): New function.
15987 (allocnos_conflict_by_live_ranges_p): Use it.
15988 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15989 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15991 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15994 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15995 (*saddl_se_1): Ditto.
15996 (*ssubsi_1): Ditto.
15997 (*ssubl_se_1): Ditto.
15999 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16001 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16002 rather than wi::to_widest for DR_INITs.
16003 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16004 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16005 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16007 (vect_analyze_group_access_1): Note that here.
16009 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16011 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16012 polynomial type sizes.
16014 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16016 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16017 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16018 (gimple_add_tmp_var): Likewise.
16020 2018-01-12 Martin Liska <mliska@suse.cz>
16022 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16023 (gimple_alloc_sizes): Likewise.
16024 (dump_gimple_statistics): Use PRIu64 in printf format.
16025 * gimple.h: Change uint64_t to int.
16027 2018-01-12 Martin Liska <mliska@suse.cz>
16029 * tree-core.h: Use uint64_t instead of int.
16030 * tree.c (tree_node_counts): Likewise.
16031 (tree_node_sizes): Likewise.
16032 (dump_tree_statistics): Use PRIu64 in printf format.
16034 2018-01-12 Martin Liska <mliska@suse.cz>
16036 * Makefile.in: As qsort_chk is implemented in vec.c, add
16037 vec.o to linkage of gencfn-macros.
16038 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16039 passing the info to record_node_allocation_statistics.
16040 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16042 * ggc-common.c (struct ggc_usage): Add operator== and use
16043 it in operator< and compare function.
16044 * mem-stats.h (struct mem_usage): Likewise.
16045 * vec.c (struct vec_usage): Remove operator< and compare
16046 function. Can be simply inherited.
16048 2018-01-12 Martin Jambor <mjambor@suse.cz>
16051 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16052 * tree-ssa-math-opts.c: Include domwalk.h.
16053 (convert_mult_to_fma_1): New function.
16054 (fma_transformation_info): New type.
16055 (fma_deferring_state): Likewise.
16056 (cancel_fma_deferring): New function.
16057 (result_of_phi): Likewise.
16058 (last_fma_candidate_feeds_initial_phi): Likewise.
16059 (convert_mult_to_fma): Added deferring logic, split actual
16060 transformation to convert_mult_to_fma_1.
16061 (math_opts_dom_walker): New type.
16062 (math_opts_dom_walker::after_dom_children): New method, body moved
16063 here from pass_optimize_widening_mul::execute, added deferring logic
16065 (pass_optimize_widening_mul::execute): Moved most of code to
16066 math_opts_dom_walker::after_dom_children.
16067 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16068 * config/i386/i386.c (ix86_option_override_internal): Added
16069 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16071 2018-01-12 Richard Biener <rguenther@suse.de>
16074 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16075 inline instance vars.
16077 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16080 * config/rx/rx.c (rx_is_restricted_memory_address):
16081 Handle SUBREG case.
16083 2018-01-12 Richard Biener <rguenther@suse.de>
16085 PR tree-optimization/80846
16086 * target.def (split_reduction): New target hook.
16087 * targhooks.c (default_split_reduction): New function.
16088 * targhooks.h (default_split_reduction): Declare.
16089 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16090 target requests first reduce vectors by combining low and high
16092 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16093 (get_vectype_for_scalar_type_and_size): Export.
16094 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16095 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16096 * doc/tm.texi: Regenerate.
16097 * config/i386/i386.c (ix86_split_reduction): Implement
16098 TARGET_VECTORIZE_SPLIT_REDUCTION.
16100 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16103 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16104 in PIC mode except for TARGET_VXWORKS_RTP.
16105 * config/sparc/sparc.c: Include cfgrtl.h.
16106 (TARGET_INIT_PIC_REG): Define.
16107 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16108 (sparc_pic_register_p): New predicate.
16109 (sparc_legitimate_address_p): Use it.
16110 (sparc_legitimize_pic_address): Likewise.
16111 (sparc_delegitimize_address): Likewise.
16112 (sparc_mode_dependent_address_p): Likewise.
16113 (gen_load_pcrel_sym): Remove 4th parameter.
16114 (load_got_register): Adjust call to above. Remove obsolete stuff.
16115 (sparc_expand_prologue): Do not call load_got_register here.
16116 (sparc_flat_expand_prologue): Likewise.
16117 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16118 (sparc_use_pseudo_pic_reg): New function.
16119 (sparc_init_pic_reg): Likewise.
16120 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16121 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16123 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16125 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16126 Add item for branch_cost.
16128 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16130 PR rtl-optimization/83565
16131 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16132 not extend the result to a larger mode for rotate operations.
16133 (num_sign_bit_copies1): Likewise.
16135 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16138 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16140 Use values-Xc.o for -pedantic.
16141 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16143 2018-01-12 Martin Liska <mliska@suse.cz>
16146 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16148 (possible_polymorphic_call_targets): Use it.
16149 (ipa_devirt): Likewise.
16151 2018-01-12 Martin Liska <mliska@suse.cz>
16153 * profile-count.h (enum profile_quality): Use 0 as invalid
16154 enum value of profile_quality.
16156 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16158 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16159 -mext-string options.
16161 2018-01-12 Richard Biener <rguenther@suse.de>
16163 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16164 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16165 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16167 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16169 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16171 * configure.ac (--with-long-double-format): Add support for the
16172 configuration option to change the default long double format on
16174 * config.gcc (powerpc*-linux*-*): Likewise.
16175 * configure: Regenerate.
16176 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16177 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16178 used without modification.
16180 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16182 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16183 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16184 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16185 MISC_BUILTIN_SPEC_BARRIER.
16186 (rs6000_init_builtins): Likewise.
16187 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16189 (speculation_barrier): New define_insn.
16190 * doc/extend.texi: Document __builtin_speculation_barrier.
16192 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16195 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16196 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16197 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16199 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16200 integral modes instead of "ss" and "sd".
16201 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16202 vectors with 32-bit and 64-bit elements.
16203 (vecdupssescalarmodesuffix): New mode attribute.
16204 (vec_dup<mode>): Use it.
16206 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16209 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16210 frame if argument is passed on stack.
16212 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16215 * ree.c (combine_reaching_defs): Optimize also
16216 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16217 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16219 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16221 PR middle-end/83189
16222 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16224 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16226 PR middle-end/83718
16227 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16228 after they are computed.
16230 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16232 PR tree-optimization/83695
16233 * gimple-loop-linterchange.cc
16234 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16235 reset cached scev information after interchange.
16236 (pass_linterchange::execute): Remove call to scev_reset_htab.
16238 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16240 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16241 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16242 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16243 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16244 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16245 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16246 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16247 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16248 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16249 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16250 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16251 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16252 (V_lane_reg): Likewise.
16253 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16255 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16256 (vfmal_lane_low<mode>_intrinsic,
16257 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16258 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16259 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16260 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16261 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16262 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16264 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16266 * config/arm/arm-cpus.in (fp16fml): New feature.
16267 (ALL_SIMD): Add fp16fml.
16268 (armv8.2-a): Add fp16fml as an option.
16269 (armv8.3-a): Likewise.
16270 (armv8.4-a): Add fp16fml as part of fp16.
16271 * config/arm/arm.h (TARGET_FP16FML): Define.
16272 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16274 * config/arm/arm-modes.def (V2HF): Define.
16275 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16276 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16277 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16278 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16279 vfmsl_low, vfmsl_high): New set of builtins.
16280 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16281 (vfml_op): New code attribute.
16282 (VFMLHALVES): New int iterator.
16283 (VFML, VFMLSEL): New mode attributes.
16284 (V_reg): Define mapping for V2HF.
16285 (V_hi, V_lo): New mode attributes.
16286 (VF_constraint): Likewise.
16287 (vfml_half, vfml_half_selector): New int attributes.
16288 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16290 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16291 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16293 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16294 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16295 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16296 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16298 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16299 Document new effective target and option set.
16301 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16303 * config/arm/arm-cpus.in (armv8_4): New feature.
16304 (ARMv8_4a): New fgroup.
16305 (armv8.4-a): New arch.
16306 * config/arm/arm-tables.opt: Regenerate.
16307 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16308 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16309 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16310 Add matching rules for -march=armv8.4-a and extensions.
16311 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16313 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16316 * config/rx/rx.md (BW): New mode attribute.
16317 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16319 2018-01-11 Richard Biener <rguenther@suse.de>
16321 PR tree-optimization/83435
16322 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16323 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16324 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16326 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16327 Alan Hayward <alan.hayward@arm.com>
16328 David Sherwood <david.sherwood@arm.com>
16330 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16332 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16333 (aarch64_print_address_internal): Use it to check for a zero offset.
16335 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16336 Alan Hayward <alan.hayward@arm.com>
16337 David Sherwood <david.sherwood@arm.com>
16339 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16340 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16341 Return a poly_int64 rather than a HOST_WIDE_INT.
16342 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16343 rather than a HOST_WIDE_INT.
16344 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16345 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16346 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16347 final_offset from HOST_WIDE_INT to poly_int64.
16348 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16349 to_constant when getting the number of units in an Advanced SIMD
16351 (aarch64_builtin_vectorized_function): Check for a constant number
16353 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16355 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16356 attribute instead of GET_MODE_NUNITS.
16357 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16358 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16359 GET_MODE_SIZE for fixed-size registers.
16360 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16361 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16362 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16363 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16364 (aarch64_print_operand, aarch64_print_address_internal)
16365 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16366 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16367 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16368 Handle polynomial GET_MODE_SIZE.
16369 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16370 wider than SImode without modification.
16371 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16372 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16373 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16374 passing and returning SVE modes.
16375 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16376 rather than GEN_INT.
16377 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16378 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16379 (aarch64_allocate_and_probe_stack_space): Likewise.
16380 (aarch64_layout_frame): Cope with polynomial offsets.
16381 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16382 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16383 polynomial offsets.
16384 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16385 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16386 poly_int64 rather than a HOST_WIDE_INT.
16387 (aarch64_get_separate_components, aarch64_process_components)
16388 (aarch64_expand_prologue, aarch64_expand_epilogue)
16389 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16390 (aarch64_anchor_offset): New function, split out from...
16391 (aarch64_legitimize_address): ...here.
16392 (aarch64_builtin_vectorization_cost): Handle polynomial
16393 TYPE_VECTOR_SUBPARTS.
16394 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16396 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16397 number of elements from the PARALLEL rather than the mode.
16398 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16399 rather than GET_MODE_BITSIZE.
16400 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16401 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16402 (aarch64_expand_vec_perm_const_1): Handle polynomial
16403 d->perm.length () and d->perm elements.
16404 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16405 Apply to_constant to d->perm elements.
16406 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16407 polynomial CONST_VECTOR_NUNITS.
16408 (aarch64_move_pointer): Take amount as a poly_int64 rather
16410 (aarch64_progress_pointer): Avoid temporary variable.
16411 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16412 the mode attribute instead of GET_MODE.
16414 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16415 Alan Hayward <alan.hayward@arm.com>
16416 David Sherwood <david.sherwood@arm.com>
16418 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16419 x exists before using it.
16420 (aarch64_add_constant_internal): Rename to...
16421 (aarch64_add_offset_1): ...this. Replace regnum with separate
16422 src and dest rtxes. Handle the case in which they're different,
16423 including when the offset is zero. Replace scratchreg with an rtx.
16424 Use 2 additions if there is no spare register into which we can
16425 move a 16-bit constant.
16426 (aarch64_add_constant): Delete.
16427 (aarch64_add_offset): Replace reg with separate src and dest
16428 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16429 Use aarch64_add_offset_1.
16430 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16431 an rtx rather than an int. Take the delta as a poly_int64
16432 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16433 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16434 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16435 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16436 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16437 and aarch64_add_sp.
16438 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16439 aarch64_add_constant.
16441 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16443 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16444 Use scalar_float_mode.
16446 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16448 * config/aarch64/aarch64-simd.md
16449 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16450 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16451 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16452 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16453 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16454 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16455 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16456 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16457 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16458 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16460 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16463 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16464 targ_options->x_arm_arch_string is non NULL.
16466 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16468 * config/aarch64/aarch64.h
16469 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16471 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16474 * expmed.c (emit_store_flag_force): Swap if const op0
16475 and change VOIDmode to mode of op0.
16477 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16479 PR rtl-optimization/83761
16480 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16481 than bytes to mode_for_size.
16483 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16485 PR middle-end/83189
16486 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16487 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16490 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16492 PR middle-end/83575
16493 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16494 when in layout mode.
16495 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16496 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16499 2018-01-10 Michael Collison <michael.collison@arm.com>
16501 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16502 * config/aarch64/aarch64-option-extension.def: Add
16503 AARCH64_OPT_EXTENSION of 'fp16fml'.
16504 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16505 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16506 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16507 * config/aarch64/constraints.md (Ui7): New constraint.
16508 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16509 (VFMLA_SEL_W): Ditto.
16512 (VFMLA16_LOW): New int iterator.
16513 (VFMLA16_HIGH): Ditto.
16514 (UNSPEC_FMLAL): New unspec.
16515 (UNSPEC_FMLSL): Ditto.
16516 (UNSPEC_FMLAL2): Ditto.
16517 (UNSPEC_FMLSL2): Ditto.
16518 (f16mac): New code attribute.
16519 * config/aarch64/aarch64-simd-builtins.def
16520 (aarch64_fmlal_lowv2sf): Ditto.
16521 (aarch64_fmlsl_lowv2sf): Ditto.
16522 (aarch64_fmlalq_lowv4sf): Ditto.
16523 (aarch64_fmlslq_lowv4sf): Ditto.
16524 (aarch64_fmlal_highv2sf): Ditto.
16525 (aarch64_fmlsl_highv2sf): Ditto.
16526 (aarch64_fmlalq_highv4sf): Ditto.
16527 (aarch64_fmlslq_highv4sf): Ditto.
16528 (aarch64_fmlal_lane_lowv2sf): Ditto.
16529 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16530 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16531 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16532 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16533 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16534 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16535 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16536 (aarch64_fmlal_lane_highv2sf): Ditto.
16537 (aarch64_fmlsl_lane_highv2sf): Ditto.
16538 (aarch64_fmlal_laneq_highv2sf): Ditto.
16539 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16540 (aarch64_fmlalq_lane_highv4sf): Ditto.
16541 (aarch64_fmlsl_lane_highv4sf): Ditto.
16542 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16543 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16544 * config/aarch64/aarch64-simd.md:
16545 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16546 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16547 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16548 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16549 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16550 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16551 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16552 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16553 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16554 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16555 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16556 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16557 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16558 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16559 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16560 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16561 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16562 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16563 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16564 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16565 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16566 (vfmlsl_low_u32): Ditto.
16567 (vfmlalq_low_u32): Ditto.
16568 (vfmlslq_low_u32): Ditto.
16569 (vfmlal_high_u32): Ditto.
16570 (vfmlsl_high_u32): Ditto.
16571 (vfmlalq_high_u32): Ditto.
16572 (vfmlslq_high_u32): Ditto.
16573 (vfmlal_lane_low_u32): Ditto.
16574 (vfmlsl_lane_low_u32): Ditto.
16575 (vfmlal_laneq_low_u32): Ditto.
16576 (vfmlsl_laneq_low_u32): Ditto.
16577 (vfmlalq_lane_low_u32): Ditto.
16578 (vfmlslq_lane_low_u32): Ditto.
16579 (vfmlalq_laneq_low_u32): Ditto.
16580 (vfmlslq_laneq_low_u32): Ditto.
16581 (vfmlal_lane_high_u32): Ditto.
16582 (vfmlsl_lane_high_u32): Ditto.
16583 (vfmlal_laneq_high_u32): Ditto.
16584 (vfmlsl_laneq_high_u32): Ditto.
16585 (vfmlalq_lane_high_u32): Ditto.
16586 (vfmlslq_lane_high_u32): Ditto.
16587 (vfmlalq_laneq_high_u32): Ditto.
16588 (vfmlslq_laneq_high_u32): Ditto.
16589 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16590 (AARCH64_FL_FOR_ARCH8_4): New.
16591 (AARCH64_ISA_F16FML): New ISA flag.
16592 (TARGET_F16FML): New feature flag for fp16fml.
16593 (doc/invoke.texi): Document new fp16fml option.
16595 2018-01-10 Michael Collison <michael.collison@arm.com>
16597 * config/aarch64/aarch64-builtins.c:
16598 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16599 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16600 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16601 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16602 (AARCH64_ISA_SHA3): New ISA flag.
16603 (TARGET_SHA3): New feature flag for sha3.
16604 * config/aarch64/iterators.md (sha512_op): New int attribute.
16605 (CRYPTO_SHA512): New int iterator.
16606 (UNSPEC_SHA512H): New unspec.
16607 (UNSPEC_SHA512H2): Ditto.
16608 (UNSPEC_SHA512SU0): Ditto.
16609 (UNSPEC_SHA512SU1): Ditto.
16610 * config/aarch64/aarch64-simd-builtins.def
16611 (aarch64_crypto_sha512hqv2di): New builtin.
16612 (aarch64_crypto_sha512h2qv2di): Ditto.
16613 (aarch64_crypto_sha512su0qv2di): Ditto.
16614 (aarch64_crypto_sha512su1qv2di): Ditto.
16615 (aarch64_eor3qv8hi): Ditto.
16616 (aarch64_rax1qv2di): Ditto.
16617 (aarch64_xarqv2di): Ditto.
16618 (aarch64_bcaxqv8hi): Ditto.
16619 * config/aarch64/aarch64-simd.md:
16620 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16621 (aarch64_crypto_sha512su0qv2di): Ditto.
16622 (aarch64_crypto_sha512su1qv2di): Ditto.
16623 (aarch64_eor3qv8hi): Ditto.
16624 (aarch64_rax1qv2di): Ditto.
16625 (aarch64_xarqv2di): Ditto.
16626 (aarch64_bcaxqv8hi): Ditto.
16627 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16628 (vsha512h2q_u64): Ditto.
16629 (vsha512su0q_u64): Ditto.
16630 (vsha512su1q_u64): Ditto.
16631 (veor3q_u16): Ditto.
16632 (vrax1q_u64): Ditto.
16633 (vxarq_u64): Ditto.
16634 (vbcaxq_u16): Ditto.
16635 * config/arm/types.md (crypto_sha512): New type attribute.
16636 (crypto_sha3): Ditto.
16637 (doc/invoke.texi): Document new sha3 option.
16639 2018-01-10 Michael Collison <michael.collison@arm.com>
16641 * config/aarch64/aarch64-builtins.c:
16642 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16643 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16644 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16645 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16646 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16647 (AARCH64_ISA_SM4): New ISA flag.
16648 (TARGET_SM4): New feature flag for sm4.
16649 * config/aarch64/aarch64-simd-builtins.def
16650 (aarch64_sm3ss1qv4si): Ditto.
16651 (aarch64_sm3tt1aq4si): Ditto.
16652 (aarch64_sm3tt1bq4si): Ditto.
16653 (aarch64_sm3tt2aq4si): Ditto.
16654 (aarch64_sm3tt2bq4si): Ditto.
16655 (aarch64_sm3partw1qv4si): Ditto.
16656 (aarch64_sm3partw2qv4si): Ditto.
16657 (aarch64_sm4eqv4si): Ditto.
16658 (aarch64_sm4ekeyqv4si): Ditto.
16659 * config/aarch64/aarch64-simd.md:
16660 (aarch64_sm3ss1qv4si): Ditto.
16661 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16662 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16663 (aarch64_sm4eqv4si): Ditto.
16664 (aarch64_sm4ekeyqv4si): Ditto.
16665 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16666 (sm3part_op): Ditto.
16667 (CRYPTO_SM3TT): Ditto.
16668 (CRYPTO_SM3PART): Ditto.
16669 (UNSPEC_SM3SS1): New unspec.
16670 (UNSPEC_SM3TT1A): Ditto.
16671 (UNSPEC_SM3TT1B): Ditto.
16672 (UNSPEC_SM3TT2A): Ditto.
16673 (UNSPEC_SM3TT2B): Ditto.
16674 (UNSPEC_SM3PARTW1): Ditto.
16675 (UNSPEC_SM3PARTW2): Ditto.
16676 (UNSPEC_SM4E): Ditto.
16677 (UNSPEC_SM4EKEY): Ditto.
16678 * config/aarch64/constraints.md (Ui2): New constraint.
16679 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16680 * config/arm/types.md (crypto_sm3): New type attribute.
16681 (crypto_sm4): Ditto.
16682 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16683 (vsm3tt1aq_u32): Ditto.
16684 (vsm3tt1bq_u32): Ditto.
16685 (vsm3tt2aq_u32): Ditto.
16686 (vsm3tt2bq_u32): Ditto.
16687 (vsm3partw1q_u32): Ditto.
16688 (vsm3partw2q_u32): Ditto.
16689 (vsm4eq_u32): Ditto.
16690 (vsm4ekeyq_u32): Ditto.
16691 (doc/invoke.texi): Document new sm4 option.
16693 2018-01-10 Michael Collison <michael.collison@arm.com>
16695 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16696 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16697 (AARCH64_FL_FOR_ARCH8_4): New.
16698 (AARCH64_FL_V8_4): New flag.
16699 (doc/invoke.texi): Document new armv8.4-a option.
16701 2018-01-10 Michael Collison <michael.collison@arm.com>
16703 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16704 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16705 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16706 * config/aarch64/aarch64-option-extension.def: Add
16707 AARCH64_OPT_EXTENSION of 'sha2'.
16708 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16709 (crypto): Disable sha2 and aes if crypto disabled.
16710 (crypto): Enable aes and sha2 if enabled.
16711 (simd): Disable sha2 and aes if simd disabled.
16712 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16714 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16715 (TARGET_SHA2): New feature flag for sha2.
16716 (TARGET_AES): New feature flag for aes.
16717 * config/aarch64/aarch64-simd.md:
16718 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16719 conditional on TARGET_AES.
16720 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16721 (aarch64_crypto_sha1hsi): Make pattern conditional
16723 (aarch64_crypto_sha1hv4si): Ditto.
16724 (aarch64_be_crypto_sha1hv4si): Ditto.
16725 (aarch64_crypto_sha1su1v4si): Ditto.
16726 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16727 (aarch64_crypto_sha1su0v4si): Ditto.
16728 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16729 (aarch64_crypto_sha256su0v4si): Ditto.
16730 (aarch64_crypto_sha256su1v4si): Ditto.
16731 (doc/invoke.texi): Document new aes and sha2 options.
16733 2018-01-10 Martin Sebor <msebor@redhat.com>
16735 PR tree-optimization/83781
16736 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16739 2018-01-11 Martin Sebor <msebor@gmail.com>
16740 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16742 PR tree-optimization/83501
16743 PR tree-optimization/81703
16745 * tree-ssa-strlen.c (get_string_cst): Rename...
16746 (get_string_len): ...to this. Handle global constants.
16747 (handle_char_store): Adjust.
16749 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16750 Jim Wilson <jimw@sifive.com>
16752 * config/riscv/riscv-protos.h (riscv_output_return): New.
16753 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16754 (riscv_attribute_table, riscv_output_return),
16755 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16756 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16757 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16759 (riscv_expand_prologue): Add early return for naked function.
16760 (riscv_expand_epilogue): Likewise.
16761 (riscv_function_ok_for_sibcall): Return false for naked function.
16762 (riscv_set_current_function): New.
16763 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16764 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16765 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16766 * doc/extend.texi (RISC-V Function Attributes): New.
16768 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16770 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16771 check for 128-bit long double before checking TCmode.
16772 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16773 128-bit long doubles before checking TFmode or TCmode.
16774 (FLOAT128_IBM_P): Likewise.
16776 2018-01-10 Martin Sebor <msebor@redhat.com>
16778 PR tree-optimization/83671
16779 * builtins.c (c_strlen): Unconditionally return zero for the empty
16781 Use -Warray-bounds for warnings.
16782 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16783 for non-constant array indices with COMPONENT_REF, arrays of
16784 arrays, and pointers to arrays.
16785 (gimple_fold_builtin_strlen): Determine and set length range for
16786 non-constant character arrays.
16788 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16790 PR middle-end/81897
16791 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16794 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16796 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16798 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16801 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16802 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16803 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16804 indexed_or_indirect_operand predicate.
16805 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16806 (*vsx_le_perm_load_v8hi): Likewise.
16807 (*vsx_le_perm_load_v16qi): Likewise.
16808 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16809 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16810 (*vsx_le_perm_store_v8hi): Likewise.
16811 (*vsx_le_perm_store_v16qi): Likewise.
16812 (eight unnamed splitters): Likewise.
16814 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16816 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16817 * config/rs6000/emmintrin.h: Likewise.
16818 * config/rs6000/mmintrin.h: Likewise.
16819 * config/rs6000/xmmintrin.h: Likewise.
16821 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16824 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16826 * tree.c (tree_nop_conversion): Return true for location wrapper
16828 (maybe_wrap_with_location): New function.
16829 (selftest::check_strip_nops): New function.
16830 (selftest::test_location_wrappers): New function.
16831 (selftest::tree_c_tests): Call it.
16832 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16833 (maybe_wrap_with_location): New decl.
16834 (EXPR_LOCATION_WRAPPER_P): New macro.
16835 (location_wrapper_p): New inline function.
16836 (tree_strip_any_location_wrapper): New inline function.
16838 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16841 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16842 stack_realign_offset for the largest alignment of stack slot
16844 (ix86_find_max_used_stack_alignment): New function.
16845 (ix86_finalize_stack_frame_flags): Use it. Set
16846 max_used_stack_alignment if we don't realign stack.
16847 * config/i386/i386.h (machine_function): Add
16848 max_used_stack_alignment.
16850 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16852 * config/arm/arm.opt (-mbranch-cost): New option.
16853 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16856 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16859 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16860 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16862 2018-01-10 Richard Biener <rguenther@suse.de>
16865 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16866 early out so it also covers the case where we have a non-NULL
16869 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16871 PR tree-optimization/83753
16872 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16873 for non-strided grouped accesses if the number of elements is 1.
16875 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16878 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16879 * i386.h (TARGET_USE_GATHER): Define.
16880 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16882 2018-01-10 Martin Liska <mliska@suse.cz>
16885 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16886 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16888 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16889 CLEANUP_NO_PARTITIONING is not set.
16891 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16893 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16894 for vectors, as a partial revert of r254296.
16895 * rtl.h (const_vec_p): Delete.
16896 (const_vec_duplicate_p): Don't test for vector CONSTs.
16897 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16898 * expmed.c (make_tree): Likewise.
16901 * common.md (E, F): Use CONSTANT_P instead of checking for
16903 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16904 checking for CONST_VECTOR.
16906 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16908 PR middle-end/83575
16909 * predict.c (force_edge_cold): Handle in more sane way edges
16910 with no prediction.
16912 2018-01-09 Carl Love <cel@us.ibm.com>
16914 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16916 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16917 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16918 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16919 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16920 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16921 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16922 * config/rs6000/rs6000-protos.h: Add extern defition for
16923 rs6000_generate_float2_double_code.
16924 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16926 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16927 (float2_v2df): Add define_expand.
16929 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16932 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16933 op_mode in the force_to_mode call.
16935 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16937 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16938 instead of checking each element individually.
16939 (aarch64_evpc_uzp): Likewise.
16940 (aarch64_evpc_zip): Likewise.
16941 (aarch64_evpc_ext): Likewise.
16942 (aarch64_evpc_rev): Likewise.
16943 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16944 instead of checking each element individually. Return true without
16946 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16947 whether all selected elements come from the same input, instead of
16948 checking each element individually. Remove calls to gen_rtx_REG,
16949 start_sequence and end_sequence and instead assert that no rtl is
16952 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16954 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16955 order of HIGH and CONST checks.
16957 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16959 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16960 if the destination isn't an SSA_NAME.
16962 2018-01-09 Richard Biener <rguenther@suse.de>
16964 PR tree-optimization/83668
16965 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16967 (canonicalize_loop_form): ... here, renamed from ...
16968 (canonicalize_loop_closed_ssa_form): ... this and amended to
16969 swap successor edges for loop exit blocks to make us use
16970 the RPO order we need for initial schedule generation.
16972 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16974 PR tree-optimization/64811
16975 * match.pd: When optimizing comparisons with Inf, avoid
16976 introducing or losing exceptions from comparisons with NaN.
16978 2018-01-09 Martin Liska <mliska@suse.cz>
16981 * asan.c (shadow_mem_size): Add gcc_assert.
16983 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16985 Don't save registers in main().
16988 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16989 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16990 * config/avr/avr.c (avr_set_current_function): Don't error if
16991 naked, OS_task or OS_main are specified at the same time.
16992 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16994 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16996 * common/config/avr/avr-common.c (avr_option_optimization_table):
16997 Switch on -mmain-is-OS_task for optimizing compilations.
16999 2018-01-09 Richard Biener <rguenther@suse.de>
17001 PR tree-optimization/83572
17002 * graphite.c: Include cfganal.h.
17003 (graphite_transform_loops): Connect infinite loops to exit
17004 and remove fake edges at the end.
17006 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17008 * ipa-inline.c (edge_badness): Revert accidental checkin.
17010 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17013 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17014 symbols; not inline clones.
17016 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17019 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17020 hard registers. Formatting fixes.
17022 PR preprocessor/83722
17023 * gcc.c (try_generate_repro): Pass
17024 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17025 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17028 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17029 Kito Cheng <kito.cheng@gmail.com>
17031 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17032 (riscv_leaf_function_p): Delete.
17033 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17035 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17037 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17039 (do_ifelse): New function.
17040 (do_isel): New function.
17041 (do_sub3): New function.
17042 (do_add3): New function.
17043 (do_load_mask_compare): New function.
17044 (do_overlap_load_compare): New function.
17045 (expand_compare_loop): New function.
17046 (expand_block_compare): Call expand_compare_loop() when appropriate.
17047 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17048 option description.
17049 (-mblock-compare-inline-loop-limit): New option.
17051 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17054 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17055 Reverse order of second and third operands in first alternative.
17056 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17057 of first and second elements in UNSPEC_VPERMR vector.
17058 (altivec_expand_vec_perm_le): Likewise.
17060 2018-01-08 Jeff Law <law@redhat.com>
17062 PR rtl-optimizatin/81308
17063 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17064 (process_switch): If group_case_labels makes a change, then set
17066 (pass_convert_switch::execute): If a switch is converted, then
17067 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17069 PR rtl-optimization/81308
17070 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17073 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17075 PR target/83663 - Revert r255946
17076 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17077 generation for cases where splatting a value is not useful.
17078 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17079 across a vec_duplicate and a paradoxical subreg forming a vector
17080 mode to a vec_concat.
17082 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17084 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17085 -march=armv8.3-a variants.
17086 * config/arm/t-multilib: Likewise.
17087 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17089 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17091 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17093 (cceq_ior_compare_complement): Give it a name so I can use it, and
17094 change boolean_or_operator predicate to boolean_operator so it can
17095 be used to generate a crand.
17096 (eqne): New code iterator.
17097 (bd/bd_neg): New code_attrs.
17098 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17099 a single define_insn.
17100 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17101 decrement (bdnzt/bdnzf/bdzt/bdzf).
17102 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17103 with the new names of the branch decrement patterns, and added the
17104 names of the branch decrement conditional patterns.
17106 2018-01-08 Richard Biener <rguenther@suse.de>
17108 PR tree-optimization/83563
17109 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17112 2018-01-08 Richard Biener <rguenther@suse.de>
17114 PR middle-end/83713
17115 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17117 2018-01-08 Richard Biener <rguenther@suse.de>
17119 PR tree-optimization/83685
17120 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17121 references to abnormals.
17123 2018-01-08 Richard Biener <rguenther@suse.de>
17126 * dwarf2out.c (output_indirect_strings): Handle empty
17127 skeleton_debug_str_hash.
17128 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17130 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17132 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17133 (emit_store_direct): Likewise.
17134 (arc_trampoline_adjust_address): Likewise.
17135 (arc_asm_trampoline_template): New function.
17136 (arc_initialize_trampoline): Use asm_trampoline_template.
17137 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17138 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17139 * config/arc/arc.md (flush_icache): Delete pattern.
17141 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17143 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17144 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17147 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17150 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17151 by not USED_FOR_TARGET.
17152 (make_pass_resolve_sw_modes): Likewise.
17154 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17156 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17159 2018-01-08 Richard Biener <rguenther@suse.de>
17161 PR middle-end/83580
17162 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17164 2018-01-08 Richard Biener <rguenther@suse.de>
17166 PR middle-end/83517
17167 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17169 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17171 PR middle-end/81897
17172 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17173 basic blocks with a small number of successors.
17174 (convert_control_dep_chain_into_preds): Improve handling of
17176 (dump_predicates): Split apart into...
17177 (dump_pred_chain): ...here...
17178 (dump_pred_info): ...and here.
17179 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17180 (can_chain_union_be_invalidated_p): Improve check for invalidation
17182 (uninit_uses_cannot_happen): Avoid unnecessary if
17183 convert_control_dep_chain_into_preds yielded nothing.
17185 2018-01-06 Martin Sebor <msebor@redhat.com>
17187 PR tree-optimization/83640
17188 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17189 subtracting negative offset from size.
17190 (builtin_access::overlap): Adjust offset bounds of the access to fall
17191 within the size of the object if possible.
17193 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17195 PR rtl-optimization/83699
17196 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17197 extract_bit_field_as_subreg to cases in which the extracted
17198 value is also a vector.
17200 * lra-constraints.c (process_alt_operands): Test for the equivalence
17201 substitutions when detecting a possible reload cycle.
17203 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17206 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17207 by default if flag_selective_schedling{,2}. Formatting fixes.
17209 PR rtl-optimization/83682
17210 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17211 if it has non-VECTOR_MODE element mode.
17212 (vec_duplicate_p): Likewise.
17214 PR middle-end/83694
17215 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17216 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17218 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17221 * config/i386/i386-builtin.def
17222 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17223 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17224 Require also OPTION_MASK_ISA_AVX512F in addition to
17225 OPTION_MASK_ISA_GFNI.
17226 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17227 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17228 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17229 to OPTION_MASK_ISA_GFNI.
17230 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17231 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17232 OPTION_MASK_ISA_AVX512BW.
17233 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17234 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17235 addition to OPTION_MASK_ISA_GFNI.
17236 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17237 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17238 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17239 to OPTION_MASK_ISA_GFNI.
17240 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17241 a requirement for all ISAs rather than any of them with a few
17243 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17245 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17246 bitmasks to be enabled with 3 exceptions, instead of requiring any
17247 enabled ISA with lots of exceptions.
17248 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17249 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17250 Change avx512bw in isa attribute to avx512f.
17251 * config/i386/sgxintrin.h: Add license boilerplate.
17252 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17253 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17254 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17255 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17257 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17258 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17259 temporarily sse2 rather than sse if not enabled already.
17262 * config/i386/sse.md (VI248_VLBW): Rename to ...
17263 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17264 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17265 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17266 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17267 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17268 mode iterator instead of VI248_VLBW.
17270 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17272 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17273 (record_modified): Skip clobbers; add debug output.
17274 (param_change_prob): Use sreal frequencies.
17276 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17278 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17279 punt for user-aligned variables.
17281 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17283 * tree-chrec.c (chrec_contains_symbols): Return true for
17286 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17289 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17290 of (x|y) == x for BICS pattern.
17292 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17294 PR tree-optimization/83605
17295 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17296 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17299 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17301 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17302 * config/epiphany/rtems.h: New file.
17304 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17305 Uros Bizjak <ubizjak@gmail.com>
17308 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17309 QIreg_operand instead of register_operand predicate.
17310 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17311 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17312 comments instead of -fmitigate[-_]rop.
17314 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17317 * cgraphunit.c (symbol_table::compile): Switch to text_section
17318 before calling assembly_start debug hook.
17319 * run-rtl-passes.c (run_rtl_passes): Likewise.
17322 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17324 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17325 range_int_cst_p rather than !symbolic_range_p before calling
17326 extract_range_from_multiplicative_op_1.
17328 2018-01-04 Jeff Law <law@redhat.com>
17330 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17331 redundant test in assertion.
17333 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17335 * doc/rtl.texi: Document machine_mode wrapper classes.
17337 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17339 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17340 using tree_to_uhwi.
17342 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17344 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17345 the VEC_PERM_EXPR fold to fail.
17347 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17350 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17351 to switched_sections.
17353 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17356 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17357 test for d.testing.
17359 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17362 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17363 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17365 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17368 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17369 is BLKmode and bitpos not zero or mode change is needed.
17371 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17374 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17377 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17380 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17381 instead of MULT rtx. Update all corresponding splitters.
17382 (*saddl_se): Ditto.
17383 (*ssub<modesuffix>): Ditto.
17384 (*ssubl_se): Ditto.
17385 (*cmp_sadd_di): Update split patterns.
17386 (*cmp_sadd_si): Ditto.
17387 (*cmp_sadd_sidi): Ditto.
17388 (*cmp_ssub_di): Ditto.
17389 (*cmp_ssub_si): Ditto.
17390 (*cmp_ssub_sidi): Ditto.
17391 * config/alpha/predicates.md (const23_operand): New predicate.
17392 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17393 Look for ASHIFT, not MULT inner operand.
17394 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17396 2018-01-04 Martin Liska <mliska@suse.cz>
17398 PR gcov-profile/83669
17399 * gcov.c (output_intermediate_file): Add version to intermediate
17401 * doc/gcov.texi: Document new field 'version' in intermediate
17402 file format. Fix location of '-k' option of gcov command.
17404 2018-01-04 Martin Liska <mliska@suse.cz>
17407 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17409 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17411 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17413 2018-01-03 Martin Sebor <msebor@redhat.com>
17415 PR tree-optimization/83655
17416 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17417 checking calls with invalid arguments.
17419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17421 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17422 (vectorizable_mask_load_store): Delete.
17423 (vectorizable_call): Return false for masked loads and stores.
17424 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17425 instead of gimple_assign_rhs1.
17426 (vectorizable_load): Handle IFN_MASK_LOAD.
17427 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17431 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17433 (vectorizable_mask_load_store): ...here.
17434 (vectorizable_load): ...and here.
17436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17438 * tree-vect-stmts.c (vect_build_all_ones_mask)
17439 (vect_build_zero_merge_argument): New functions, split out from...
17440 (vectorizable_load): ...here.
17442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17444 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17446 (vectorizable_mask_load_store): ...here.
17447 (vectorizable_store): ...and here.
17449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17451 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17453 (vectorizable_mask_load_store): ...here.
17455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17457 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17458 (vect_model_store_cost): Take a vec_load_store_type instead of a
17460 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17461 (vect_model_store_cost): Take a vec_load_store_type instead of a
17463 (vectorizable_mask_load_store): Update accordingly.
17464 (vectorizable_store): Likewise.
17465 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17469 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17470 IFN_MASK_LOAD calls here rather than...
17471 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17474 Alan Hayward <alan.hayward@arm.com>
17475 David Sherwood <david.sherwood@arm.com>
17477 * expmed.c (extract_bit_field_1): For vector extracts,
17478 fall back to extract_bit_field_as_subreg if vec_extract
17481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17482 Alan Hayward <alan.hayward@arm.com>
17483 David Sherwood <david.sherwood@arm.com>
17485 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17486 they are variable or constant sized.
17487 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17488 slots for constant-sized data.
17490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17491 Alan Hayward <alan.hayward@arm.com>
17492 David Sherwood <david.sherwood@arm.com>
17494 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17495 handling COND_EXPRs with boolean comparisons, try to find a better
17496 basis for the mask type than the boolean itself.
17498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17500 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17501 is calculated and how it can be overridden.
17502 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17503 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17505 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17509 Alan Hayward <alan.hayward@arm.com>
17510 David Sherwood <david.sherwood@arm.com>
17512 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17513 Remove the mode argument.
17514 (aarch64_simd_valid_immediate): Remove the mode and inverse
17516 * config/aarch64/iterators.md (bitsize): New iterator.
17517 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17518 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17519 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17520 aarch64_simd_valid_immediate.
17521 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17522 (aarch64_reg_or_bic_imm): Likewise.
17523 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17524 with an insn_type enum and msl with a modifier_type enum.
17525 Replace element_width with a scalar_mode. Change the shift
17526 to unsigned int. Add constructors for scalar_float_mode and
17527 scalar_int_mode elements.
17528 (aarch64_vect_float_const_representable_p): Delete.
17529 (aarch64_can_const_movi_rtx_p)
17530 (aarch64_simd_scalar_immediate_valid_for_move)
17531 (aarch64_simd_make_constant): Update call to
17532 aarch64_simd_valid_immediate.
17533 (aarch64_advsimd_valid_immediate_hs): New function.
17534 (aarch64_advsimd_valid_immediate): Likewise.
17535 (aarch64_simd_valid_immediate): Remove mode and inverse
17536 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17537 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17538 and aarch64_float_const_representable_p on the result.
17539 (aarch64_output_simd_mov_immediate): Remove mode argument.
17540 Update call to aarch64_simd_valid_immediate and use of
17541 simd_immediate_info.
17542 (aarch64_output_scalar_simd_mov_immediate): Update call
17545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17546 Alan Hayward <alan.hayward@arm.com>
17547 David Sherwood <david.sherwood@arm.com>
17549 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17550 (mode_nunits): Likewise CONST_MODE_NUNITS.
17551 * machmode.def (ADJUST_NUNITS): Document.
17552 * genmodes.c (mode_data::need_nunits_adj): New field.
17553 (blank_mode): Update accordingly.
17554 (adj_nunits): New variable.
17555 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17557 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17558 listed in adj_nunits.
17559 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17560 listed in adj_nunits. Don't emit case statements for such modes.
17561 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17562 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17563 nothing if adj_nunits is nonnull.
17564 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17565 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17566 (emit_mode_fbit): Update use of print_maybe_const_decl.
17567 (emit_move_size): Likewise. Treat the array as non-const
17569 (emit_mode_adjustments): Handle adj_nunits.
17571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17573 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17574 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17575 (VECTOR_MODES): Use it.
17576 (make_vector_modes): Take the prefix as an argument.
17578 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17579 Alan Hayward <alan.hayward@arm.com>
17580 David Sherwood <david.sherwood@arm.com>
17582 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17583 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17584 for MODE_VECTOR_BOOL.
17585 * machmode.def (VECTOR_BOOL_MODE): Document.
17586 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17587 (make_vector_bool_mode): New function.
17588 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17590 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17591 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17593 * stor-layout.c (int_mode_for_mode): Likewise.
17594 * tree.c (build_vector_type_for_mode): Likewise.
17595 * varasm.c (output_constant_pool_2): Likewise.
17596 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17597 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17598 for MODE_VECTOR_BOOL.
17599 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17600 of mode class checks.
17601 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17602 instead of a list of mode class checks.
17603 (expand_vector_scalar_condition): Likewise.
17604 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17607 Alan Hayward <alan.hayward@arm.com>
17608 David Sherwood <david.sherwood@arm.com>
17610 * machmode.h (mode_size): Change from unsigned short to
17612 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17613 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17614 or if measurement_type is not polynomial.
17615 (fixed_size_mode::includes_p): Check for constant-sized modes.
17616 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17617 return a poly_uint16 rather than an unsigned short.
17618 (emit_mode_size): Change the type of mode_size from unsigned short
17619 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17620 (emit_mode_adjustments): Cope with polynomial vector sizes.
17621 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17623 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17625 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17626 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17627 * caller-save.c (setup_save_areas): Likewise.
17628 (replace_reg_with_saved_mem): Likewise.
17629 * calls.c (emit_library_call_value_1): Likewise.
17630 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17631 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17632 (gen_lowpart_for_combine): Likewise.
17633 * convert.c (convert_to_integer_1): Likewise.
17634 * cse.c (equiv_constant, cse_insn): Likewise.
17635 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17636 (cselib_subst_to_values): Likewise.
17637 * dce.c (word_dce_process_block): Likewise.
17638 * df-problems.c (df_word_lr_mark_ref): Likewise.
17639 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17640 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17641 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17642 (rtl_for_decl_location): Likewise.
17643 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17644 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17645 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17646 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17647 (expand_expr_real_1): Likewise.
17648 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17649 (pad_below): Likewise.
17650 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17651 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17652 * ira.c (get_subreg_tracking_sizes): Likewise.
17653 * ira-build.c (ira_create_allocno_objects): Likewise.
17654 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17655 (ira_sort_regnos_for_alter_reg): Likewise.
17656 * ira-costs.c (record_operand_costs): Likewise.
17657 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17658 (resolve_simple_move): Likewise.
17659 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17660 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17661 (lra_constraints): Likewise.
17662 (CONST_POOL_OK_P): Reject variable-sized modes.
17663 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17664 (add_pseudo_to_slot, lra_spill): Likewise.
17665 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17666 * optabs-query.c (get_best_extraction_insn): Likewise.
17667 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17668 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17669 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17670 * recog.c (offsettable_address_addr_space_p): Likewise.
17671 * regcprop.c (maybe_mode_change): Likewise.
17672 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17673 * regrename.c (build_def_use): Likewise.
17674 * regstat.c (dump_reg_info): Likewise.
17675 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17676 (find_reloads, find_reloads_subreg_address): Likewise.
17677 * reload1.c (eliminate_regs_1): Likewise.
17678 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17679 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17680 (simplify_binary_operation_1, simplify_subreg): Likewise.
17681 * targhooks.c (default_function_arg_padding): Likewise.
17682 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17683 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17684 (verify_gimple_assign_ternary): Likewise.
17685 * tree-inline.c (estimate_move_cost): Likewise.
17686 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17687 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17688 (get_address_cost_ainc): Likewise.
17689 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17690 (vect_supportable_dr_alignment): Likewise.
17691 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17692 (vectorizable_reduction): Likewise.
17693 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17694 (vectorizable_operation, vectorizable_load): Likewise.
17695 * tree.c (build_same_sized_truth_vector_type): Likewise.
17696 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17697 * var-tracking.c (emit_note_insn_var_location): Likewise.
17698 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17699 (ADDR_VEC_ALIGN): Likewise.
17701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17702 Alan Hayward <alan.hayward@arm.com>
17703 David Sherwood <david.sherwood@arm.com>
17705 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17707 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17708 or if measurement_type is polynomial.
17709 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17710 * combine.c (make_extraction): Likewise.
17711 * dse.c (find_shift_sequence): Likewise.
17712 * dwarf2out.c (mem_loc_descriptor): Likewise.
17713 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17714 (extract_bit_field, extract_low_bits): Likewise.
17715 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17716 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17717 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17718 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17719 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17720 * reload.c (find_reloads): Likewise.
17721 * reload1.c (alter_reg): Likewise.
17722 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17723 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17724 * tree-if-conv.c (predicate_mem_writes): Likewise.
17725 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17726 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17727 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17728 * valtrack.c (dead_debug_insert_temp): Likewise.
17729 * varasm.c (mergeable_constant_section): Likewise.
17730 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17733 Alan Hayward <alan.hayward@arm.com>
17734 David Sherwood <david.sherwood@arm.com>
17736 * expr.c (expand_assignment): Cope with polynomial mode sizes
17737 when assigning to a CONCAT.
17739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17740 Alan Hayward <alan.hayward@arm.com>
17741 David Sherwood <david.sherwood@arm.com>
17743 * machmode.h (mode_precision): Change from unsigned short to
17745 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17747 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17748 or if measurement_type is not polynomial.
17749 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17750 in which the mode is already known to be a scalar_int_mode.
17751 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17752 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17754 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17755 for GET_MODE_PRECISION.
17756 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17757 for GET_MODE_PRECISION.
17758 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17760 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17761 (expand_field_assignment, make_extraction): Likewise.
17762 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17763 (get_last_value): Likewise.
17764 * convert.c (convert_to_integer_1): Likewise.
17765 * cse.c (cse_insn): Likewise.
17766 * expr.c (expand_expr_real_1): Likewise.
17767 * lra-constraints.c (simplify_operand_subreg): Likewise.
17768 * optabs-query.c (can_atomic_load_p): Likewise.
17769 * optabs.c (expand_atomic_load): Likewise.
17770 (expand_atomic_store): Likewise.
17771 * ree.c (combine_reaching_defs): Likewise.
17772 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17773 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17774 * tree.h (type_has_mode_precision_p): Likewise.
17775 * ubsan.c (instrument_si_overflow): Likewise.
17777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17778 Alan Hayward <alan.hayward@arm.com>
17779 David Sherwood <david.sherwood@arm.com>
17781 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17782 polynomial numbers of units.
17783 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17784 (valid_vector_subparts_p): New function.
17785 (build_vector_type): Remove temporary shim and take the number
17786 of units as a poly_uint64 rather than an int.
17787 (build_opaque_vector_type): Take the number of units as a
17788 poly_uint64 rather than an int.
17789 * tree.c (build_vector_from_ctor): Handle polynomial
17790 TYPE_VECTOR_SUBPARTS.
17791 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17792 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17793 (build_vector_from_val): If the number of units is variable,
17794 use build_vec_duplicate_cst for constant operands and
17795 VEC_DUPLICATE_EXPR otherwise.
17796 (make_vector_type): Remove temporary is_constant ().
17797 (build_vector_type, build_opaque_vector_type): Take the number of
17798 units as a poly_uint64 rather than an int.
17799 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17801 * cfgexpand.c (expand_debug_expr): Likewise.
17802 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17803 (store_constructor, expand_expr_real_1): Likewise.
17804 (const_scalar_mask_from_tree): Likewise.
17805 * fold-const-call.c (fold_const_reduction): Likewise.
17806 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17807 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17808 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17809 (fold_relational_const): Likewise.
17810 (native_interpret_vector): Likewise. Change the size from an
17811 int to an unsigned int.
17812 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17813 TYPE_VECTOR_SUBPARTS.
17814 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17815 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17816 duplicating a non-constant operand into a variable-length vector.
17817 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17818 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17819 * ipa-icf.c (sem_variable::equals): Likewise.
17820 * match.pd: Likewise.
17821 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17822 * print-tree.c (print_node): Likewise.
17823 * stor-layout.c (layout_type): Likewise.
17824 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17825 * tree-cfg.c (verify_gimple_comparison): Likewise.
17826 (verify_gimple_assign_binary): Likewise.
17827 (verify_gimple_assign_ternary): Likewise.
17828 (verify_gimple_assign_single): Likewise.
17829 * tree-pretty-print.c (dump_generic_node): Likewise.
17830 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17831 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17832 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17833 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17834 (vect_shift_permute_load_chain): Likewise.
17835 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17836 (expand_vector_condition, optimize_vector_constructor): Likewise.
17837 (lower_vec_perm, get_compute_type): Likewise.
17838 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17839 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17840 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17841 (vect_recog_mask_conversion_pattern): Likewise.
17842 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17843 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17844 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17845 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17846 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17847 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17848 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17849 (supportable_widening_operation): Likewise.
17850 (supportable_narrowing_operation): Likewise.
17851 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17853 * varasm.c (output_constant): Likewise.
17855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17856 Alan Hayward <alan.hayward@arm.com>
17857 David Sherwood <david.sherwood@arm.com>
17859 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17860 so that both the length == 3 and length != 3 cases set up their
17861 own permute vectors. Add comments explaining why we know the
17862 number of elements is constant.
17863 (vect_permute_load_chain): Likewise.
17865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17866 Alan Hayward <alan.hayward@arm.com>
17867 David Sherwood <david.sherwood@arm.com>
17869 * machmode.h (mode_nunits): Change from unsigned char to
17871 (ONLY_FIXED_SIZE_MODES): New macro.
17872 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17873 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17874 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17876 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17877 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17878 or if measurement_type is not polynomial.
17879 * genmodes.c (ZERO_COEFFS): New macro.
17880 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17882 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17883 Use ZERO_COEFFS when emitting initializers.
17884 * data-streamer.h (bp_pack_poly_value): New function.
17885 (bp_unpack_poly_value): Likewise.
17886 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17887 for GET_MODE_NUNITS.
17888 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17889 for GET_MODE_NUNITS.
17890 * tree.c (make_vector_type): Remove temporary shim and make
17891 the real function take the number of units as a poly_uint64
17892 rather than an int.
17893 (build_vector_type_for_mode): Handle polynomial nunits.
17894 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17895 * emit-rtl.c (const_vec_series_p_1): Likewise.
17896 (gen_rtx_CONST_VECTOR): Likewise.
17897 * fold-const.c (test_vec_duplicate_folding): Likewise.
17898 * genrecog.c (validate_pattern): Likewise.
17899 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17900 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17901 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17902 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17903 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17904 * rtlanal.c (subreg_get_info): Likewise.
17905 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17906 (vect_grouped_load_supported): Likewise.
17907 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17908 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17909 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17910 (simplify_const_unary_operation, simplify_binary_operation_1)
17911 (simplify_const_binary_operation, simplify_ternary_operation)
17912 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17913 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17914 instead of CONST_VECTOR_NUNITS.
17915 * varasm.c (output_constant_pool_2): Likewise.
17916 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17917 explicit-encoded elements in the XVEC for variable-length vectors.
17919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17921 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17924 Alan Hayward <alan.hayward@arm.com>
17925 David Sherwood <david.sherwood@arm.com>
17927 * coretypes.h (fixed_size_mode): Declare.
17928 (fixed_size_mode_pod): New typedef.
17929 * builtins.h (target_builtins::x_apply_args_mode)
17930 (target_builtins::x_apply_result_mode): Change type to
17931 fixed_size_mode_pod.
17932 * builtins.c (apply_args_size, apply_result_size, result_vector)
17933 (expand_builtin_apply_args_1, expand_builtin_apply)
17934 (expand_builtin_return): Update accordingly.
17936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17938 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17939 * cselib.c (cselib_hash_rtx): Likewise.
17940 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17941 CONST_VECTOR encoding.
17943 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17944 Jeff Law <law@redhat.com>
17947 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17948 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17949 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17950 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17953 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17954 explicitly probe *sp in a noreturn function if there were any callee
17955 register saves or frame pointer is needed.
17957 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17960 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17961 BLKmode for ternary, binary or unary expressions.
17964 * var-tracking.c (delete_vta_debug_insn): New inline function.
17965 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17966 insns from get_insns () to NULL instead of each bb separately.
17967 Use delete_vta_debug_insn. No longer static.
17968 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17969 delete_vta_debug_insns callers.
17970 * rtl.h (delete_vta_debug_insns): Declare.
17971 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17972 instead of variable_tracking_main.
17974 2018-01-03 Martin Sebor <msebor@redhat.com>
17976 PR tree-optimization/83603
17977 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17978 arguments past the endof the argument list in functions declared
17979 without a prototype.
17980 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17981 Avoid checking when arguments are null.
17983 2018-01-03 Martin Sebor <msebor@redhat.com>
17986 * doc/extend.texi (attribute const): Fix a typo.
17987 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17988 issuing -Wsuggest-attribute for void functions.
17990 2018-01-03 Martin Sebor <msebor@redhat.com>
17992 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17993 offset_int::from instead of wide_int::to_shwi.
17994 (maybe_diag_overlap): Remove assertion.
17995 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17996 * gimple-ssa-sprintf.c (format_directive): Same.
17997 (parse_directive): Same.
17998 (sprintf_dom_walker::compute_format_length): Same.
17999 (try_substitute_return_value): Same.
18001 2018-01-03 Jeff Law <law@redhat.com>
18003 PR middle-end/83654
18004 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18005 non-constant residual for zero at runtime and avoid probing in
18006 that case. Reorganize code for trailing problem to mirror handling
18009 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18011 PR tree-optimization/83501
18012 * tree-ssa-strlen.c (get_string_cst): New.
18013 (handle_char_store): Call get_string_cst.
18015 2018-01-03 Martin Liska <mliska@suse.cz>
18017 PR tree-optimization/83593
18018 * tree-ssa-strlen.c: Include tree-cfg.h.
18019 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18020 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18021 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18023 (strlen_dom_walker::before_dom_children): Call
18024 gimple_purge_dead_eh_edges. Dump tranformation with details
18026 (strlen_dom_walker::before_dom_children): Update call by adding
18027 new argument cleanup_eh.
18028 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18030 2018-01-03 Martin Liska <mliska@suse.cz>
18033 * cif-code.def (VARIADIC_THUNK): New enum value.
18034 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18037 2018-01-03 Jan Beulich <jbeulich@suse.com>
18039 * sse.md (mov<mode>_internal): Tighten condition for when to use
18040 vmovdqu<ssescalarsize> for TI and OI modes.
18042 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18044 Update copyright years.
18046 2018-01-03 Martin Liska <mliska@suse.cz>
18049 * ipa-visibility.c (function_and_variable_visibility): Skip
18050 functions with noipa attribure.
18052 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18054 * gcc.c (process_command): Update copyright notice dates.
18055 * gcov-dump.c (print_version): Ditto.
18056 * gcov.c (print_version): Ditto.
18057 * gcov-tool.c (print_version): Ditto.
18058 * gengtype.c (create_file): Ditto.
18059 * doc/cpp.texi: Bump @copying's copyright year.
18060 * doc/cppinternals.texi: Ditto.
18061 * doc/gcc.texi: Ditto.
18062 * doc/gccint.texi: Ditto.
18063 * doc/gcov.texi: Ditto.
18064 * doc/install.texi: Ditto.
18065 * doc/invoke.texi: Ditto.
18067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18069 * vector-builder.h (vector_builder::m_full_nelts): Change from
18070 unsigned int to poly_uint64.
18071 (vector_builder::full_nelts): Update prototype accordingly.
18072 (vector_builder::new_vector): Likewise.
18073 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18074 (vector_builder::operator ==): Likewise.
18075 (vector_builder::finalize): Likewise.
18076 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18077 Take the number of elements as a poly_uint64 rather than an
18079 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18080 from unsigned int to poly_uint64.
18081 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18082 (vec_perm_indices::new_vector): Likewise.
18083 (vec_perm_indices::length): Likewise.
18084 (vec_perm_indices::nelts_per_input): Likewise.
18085 (vec_perm_indices::input_nelts): Likewise.
18086 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18087 number of elements per input as a poly_uint64 rather than an
18088 unsigned int. Use the original encoding for variable-length
18089 vectors, rather than clamping each individual element.
18090 For the second and subsequent elements in each pattern,
18091 clamp the step and base before clamping their sum.
18092 (vec_perm_indices::series_p): Handle polynomial element counts.
18093 (vec_perm_indices::all_in_range_p): Likewise.
18094 (vec_perm_indices_to_tree): Likewise.
18095 (vec_perm_indices_to_rtx): Likewise.
18096 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18097 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18098 (tree_vector_builder::new_binary_operation): Handle polynomial
18099 element counts. Return false if we need to know the number
18100 of elements at compile time.
18101 * fold-const.c (fold_vec_perm): Punt if the number of elements
18102 isn't known at compile time.
18104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18106 * vec-perm-indices.h (vec_perm_builder): Change element type
18107 from HOST_WIDE_INT to poly_int64.
18108 (vec_perm_indices::element_type): Update accordingly.
18109 (vec_perm_indices::clamp): Handle polynomial element_types.
18110 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18111 (vec_perm_indices::all_in_range_p): Likewise.
18112 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18114 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18115 polynomial vec_perm_indices element types.
18116 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18117 * fold-const.c (fold_vec_perm): Likewise.
18118 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18119 * tree-vect-generic.c (lower_vec_perm): Likewise.
18120 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18121 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18122 element type to HOST_WIDE_INT.
18124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18125 Alan Hayward <alan.hayward@arm.com>
18126 David Sherwood <david.sherwood@arm.com>
18128 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18129 rather than an int. Use plus_constant.
18130 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18131 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18134 Alan Hayward <alan.hayward@arm.com>
18135 David Sherwood <david.sherwood@arm.com>
18137 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18138 a HOST_WIDE_INT to a poly_int64.
18140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18141 Alan Hayward <alan.hayward@arm.com>
18142 David Sherwood <david.sherwood@arm.com>
18144 * calls.c (load_register_parameters): Cope with polynomial
18145 mode sizes. Require a constant size for BLKmode parameters
18146 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18147 forces a parameter to be padded at the lsb end in order to
18148 fill a complete number of words, require the parameter size
18149 to be ordered wrt UNITS_PER_WORD.
18151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18152 Alan Hayward <alan.hayward@arm.com>
18153 David Sherwood <david.sherwood@arm.com>
18155 * reload1.c (spill_stack_slot_width): Change element type
18156 from unsigned int to poly_uint64_pod.
18157 (alter_reg): Treat mode sizes as polynomial.
18159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18160 Alan Hayward <alan.hayward@arm.com>
18161 David Sherwood <david.sherwood@arm.com>
18163 * reload.c (complex_word_subreg_p): New function.
18164 (reload_inner_reg_of_subreg, push_reload): Use it.
18166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18167 Alan Hayward <alan.hayward@arm.com>
18168 David Sherwood <david.sherwood@arm.com>
18170 * lra-constraints.c (process_alt_operands): Reject matched
18171 operands whose sizes aren't ordered.
18172 (match_reload): Refer to this check here.
18174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18175 Alan Hayward <alan.hayward@arm.com>
18176 David Sherwood <david.sherwood@arm.com>
18178 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18179 that the mode size is in the set {1, 2, 4, 8, 16}.
18181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18182 Alan Hayward <alan.hayward@arm.com>
18183 David Sherwood <david.sherwood@arm.com>
18185 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18186 Use plus_constant instead of gen_rtx_PLUS.
18188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18189 Alan Hayward <alan.hayward@arm.com>
18190 David Sherwood <david.sherwood@arm.com>
18192 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18193 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18194 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18195 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18196 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18197 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18198 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18199 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18200 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18201 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18203 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18204 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18205 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18206 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18207 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18208 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18209 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18210 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18211 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18212 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18214 * expr.c (emit_move_resolve_push): Treat the input and result
18215 of PUSH_ROUNDING as a poly_int64.
18216 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18217 (emit_push_insn): Likewise.
18218 * lra-eliminations.c (mark_not_eliminable): Likewise.
18219 * recog.c (push_operand): Likewise.
18220 * reload1.c (elimination_effects): Likewise.
18221 * rtlanal.c (nonzero_bits1): Likewise.
18222 * calls.c (store_one_arg): Likewise. Require the padding to be
18223 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18230 Use plus_constant instead of gen_rtx_PLUS.
18232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18233 Alan Hayward <alan.hayward@arm.com>
18234 David Sherwood <david.sherwood@arm.com>
18236 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18237 rather than an int.
18239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18240 Alan Hayward <alan.hayward@arm.com>
18241 David Sherwood <david.sherwood@arm.com>
18243 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18244 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18245 via stack temporaries. Treat the mode size as polynomial too.
18247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18248 Alan Hayward <alan.hayward@arm.com>
18249 David Sherwood <david.sherwood@arm.com>
18251 * expr.c (expand_expr_real_2): When handling conversions involving
18252 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18253 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18254 as a poly_uint64 too.
18256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18257 Alan Hayward <alan.hayward@arm.com>
18258 David Sherwood <david.sherwood@arm.com>
18260 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18263 Alan Hayward <alan.hayward@arm.com>
18264 David Sherwood <david.sherwood@arm.com>
18266 * combine.c (can_change_dest_mode): Handle polynomial
18267 REGMODE_NATURAL_SIZE.
18268 * expmed.c (store_bit_field_1): Likewise.
18269 * expr.c (store_constructor): Likewise.
18270 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18271 and polynomial REGMODE_NATURAL_SIZE.
18272 (gen_lowpart_common): Likewise.
18273 * reginfo.c (record_subregs_of_mode): Likewise.
18274 * rtlanal.c (read_modify_subreg_p): Likewise.
18276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18277 Alan Hayward <alan.hayward@arm.com>
18278 David Sherwood <david.sherwood@arm.com>
18280 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18281 numbers of elements.
18283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18284 Alan Hayward <alan.hayward@arm.com>
18285 David Sherwood <david.sherwood@arm.com>
18287 * match.pd: Cope with polynomial numbers of vector elements.
18289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18290 Alan Hayward <alan.hayward@arm.com>
18291 David Sherwood <david.sherwood@arm.com>
18293 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18294 in a POINTER_PLUS_EXPR.
18296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18297 Alan Hayward <alan.hayward@arm.com>
18298 David Sherwood <david.sherwood@arm.com>
18300 * omp-simd-clone.c (simd_clone_subparts): New function.
18301 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18302 (ipa_simd_modify_function_body): Likewise.
18304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18305 Alan Hayward <alan.hayward@arm.com>
18306 David Sherwood <david.sherwood@arm.com>
18308 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18309 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18310 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18311 (expand_vector_condition, vector_element): Likewise.
18312 (subparts_gt): New function.
18313 (get_compute_type): Use subparts_gt.
18314 (count_type_subparts): Delete.
18315 (expand_vector_operations_1): Use subparts_gt instead of
18316 count_type_subparts.
18318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18319 Alan Hayward <alan.hayward@arm.com>
18320 David Sherwood <david.sherwood@arm.com>
18322 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18323 (vect_compile_time_alias): ...this new function. Do the calculation
18324 on poly_ints rather than trees.
18325 (vect_prune_runtime_alias_test_list): Update call accordingly.
18327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18328 Alan Hayward <alan.hayward@arm.com>
18329 David Sherwood <david.sherwood@arm.com>
18331 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18333 (vect_schedule_slp_instance): Likewise.
18335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18336 Alan Hayward <alan.hayward@arm.com>
18337 David Sherwood <david.sherwood@arm.com>
18339 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18340 constant and extern definitions for variable-length vectors.
18341 (vect_get_constant_vectors): Note that the number of units
18342 is known to be constant.
18344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18345 Alan Hayward <alan.hayward@arm.com>
18346 David Sherwood <david.sherwood@arm.com>
18348 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18349 of units as polynomial. Choose between WIDE and NARROW based
18352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18353 Alan Hayward <alan.hayward@arm.com>
18354 David Sherwood <david.sherwood@arm.com>
18356 * tree-vect-stmts.c (simd_clone_subparts): New function.
18357 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18360 Alan Hayward <alan.hayward@arm.com>
18361 David Sherwood <david.sherwood@arm.com>
18363 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18364 vectors as polynomial. Use build_index_vector for
18365 IFN_GOMP_SIMD_LANE.
18367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18368 Alan Hayward <alan.hayward@arm.com>
18369 David Sherwood <david.sherwood@arm.com>
18371 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18372 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18373 for variable-length vectors.
18374 (vectorizable_mask_load_store): Treat the number of units as
18375 polynomial, asserting that it is constant if the condition has
18376 already been enforced.
18377 (vectorizable_store, vectorizable_load): Likewise.
18379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18380 Alan Hayward <alan.hayward@arm.com>
18381 David Sherwood <david.sherwood@arm.com>
18383 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18384 of units as polynomial. Punt if we can't tell at compile time
18385 which vector contains the final result.
18387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18388 Alan Hayward <alan.hayward@arm.com>
18389 David Sherwood <david.sherwood@arm.com>
18391 * tree-vect-loop.c (vectorizable_induction): Treat the number
18392 of units as polynomial. Punt on SLP inductions. Use an integer
18393 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18394 cast of such a series for variable-length floating-point
18397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18398 Alan Hayward <alan.hayward@arm.com>
18399 David Sherwood <david.sherwood@arm.com>
18401 * tree.h (build_index_vector): Declare.
18402 * tree.c (build_index_vector): New function.
18403 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18404 of units as polynomial, forcibly converting it to a constant if
18405 vectorizable_reduction has already enforced the condition.
18406 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18407 to create a {1,2,3,...} vector.
18408 (vectorizable_reduction): Treat the number of units as polynomial.
18409 Choose vectype_in based on the largest scalar element size rather
18410 than the smallest number of units. Enforce the restrictions
18413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18414 Alan Hayward <alan.hayward@arm.com>
18415 David Sherwood <david.sherwood@arm.com>
18417 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18418 number of units as polynomial.
18420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18421 Alan Hayward <alan.hayward@arm.com>
18422 David Sherwood <david.sherwood@arm.com>
18424 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18425 * target.def (autovectorize_vector_sizes): Return the vector sizes
18426 by pointer, using vector_sizes rather than a bitmask.
18427 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18428 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18429 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18431 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18432 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18433 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18434 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18435 * omp-general.c (omp_max_vf): Likewise.
18436 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18437 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18438 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18439 * tree-vect-slp.c (vect_slp_bb): Likewise.
18440 * doc/tm.texi: Regenerate.
18441 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18443 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18444 the vector size as a poly_uint64 rather than an unsigned int.
18445 (current_vector_size): Change from an unsigned int to a poly_uint64.
18446 (get_vectype_for_scalar_type): Update accordingly.
18447 * tree.h (build_truth_vector_type): Take the size and number of
18448 units as a poly_uint64 rather than an unsigned int.
18449 (build_vector_type): Add a temporary overload that takes
18450 the number of units as a poly_uint64 rather than an unsigned int.
18451 * tree.c (make_vector_type): Likewise.
18452 (build_truth_vector_type): Take the number of units as a poly_uint64
18453 rather than an unsigned int.
18455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18456 Alan Hayward <alan.hayward@arm.com>
18457 David Sherwood <david.sherwood@arm.com>
18459 * target.def (get_mask_mode): Take the number of units and length
18460 as poly_uint64s rather than unsigned ints.
18461 * targhooks.h (default_get_mask_mode): Update accordingly.
18462 * targhooks.c (default_get_mask_mode): Likewise.
18463 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18464 * doc/tm.texi: Regenerate.
18466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18467 Alan Hayward <alan.hayward@arm.com>
18468 David Sherwood <david.sherwood@arm.com>
18470 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18471 * omp-general.c (omp_max_vf): Likewise.
18472 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18473 (expand_omp_simd): Handle polynomial safelen.
18474 * omp-low.c (omplow_simd_context): Add a default constructor.
18475 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18476 (lower_rec_simd_input_clauses): Update accordingly.
18477 (lower_rec_input_clauses): Likewise.
18479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18480 Alan Hayward <alan.hayward@arm.com>
18481 David Sherwood <david.sherwood@arm.com>
18483 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18484 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18485 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18486 (vect_analyze_slp_cost): Likewise.
18487 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18488 (vect_model_load_cost): Likewise.
18490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18491 Alan Hayward <alan.hayward@arm.com>
18492 David Sherwood <david.sherwood@arm.com>
18494 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18495 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18496 from an unsigned int * to a poly_uint64_pod *.
18497 (calculate_unrolling_factor): New function.
18498 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18501 Alan Hayward <alan.hayward@arm.com>
18502 David Sherwood <david.sherwood@arm.com>
18504 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18505 from an unsigned int to a poly_uint64.
18506 (_loop_vec_info::slp_unrolling_factor): Likewise.
18507 (_loop_vec_info::vectorization_factor): Change from an int
18509 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18510 (vect_get_num_vectors): New function.
18511 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18512 (vect_get_num_copies): Use vect_get_num_vectors.
18513 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18514 to an unsigned int *.
18515 (vect_analyze_data_refs): Change min_vf from an int * to a
18517 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18518 than an unsigned HOST_WIDE_INT.
18519 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18520 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18521 to an unsigned int *.
18522 (vect_analyze_data_ref_dependences): Likewise.
18523 (vect_compute_data_ref_alignment): Handle polynomial vf.
18524 (vect_enhance_data_refs_alignment): Likewise.
18525 (vect_prune_runtime_alias_test_list): Likewise.
18526 (vect_shift_permute_load_chain): Likewise.
18527 (vect_supportable_dr_alignment): Likewise.
18528 (dependence_distance_ge_vf): Take the vectorization factor as a
18529 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18530 (vect_analyze_data_refs): Change min_vf from an int * to a
18532 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18533 vfm1 as a poly_uint64 rather than an int. Make the same change
18534 for the returned bound_scalar.
18535 (vect_gen_vector_loop_niters): Handle polynomial vf.
18536 (vect_do_peeling): Likewise. Update call to
18537 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18538 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18540 * tree-vect-loop.c (vect_determine_vectorization_factor)
18541 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18542 (vect_get_known_peeling_cost): Likewise.
18543 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18544 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18545 (vect_transform_loop): Likewise. Use the lowest possible VF when
18546 updating the upper bounds of the loop.
18547 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18548 rather than an int.
18549 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18550 polynomial unroll factors.
18551 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18552 (vect_make_slp_decision): Likewise.
18553 (vect_supported_load_permutation_p): Likewise, and polynomial
18555 (vect_analyze_slp_cost): Handle polynomial vf.
18556 (vect_slp_analyze_node_operations): Likewise.
18557 (vect_slp_analyze_bb_1): Likewise.
18558 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18559 than an unsigned HOST_WIDE_INT.
18560 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18561 (vectorizable_load): Handle polynomial vf.
18562 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18564 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18567 Alan Hayward <alan.hayward@arm.com>
18568 David Sherwood <david.sherwood@arm.com>
18570 * match.pd: Handle bit operations involving three constants
18571 and try to fold one pair.
18573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18575 * tree-vect-loop-manip.c: Include gimple-fold.h.
18576 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18577 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18578 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18579 Add a path that uses a step of VF instead of 1, but disable it
18581 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18582 and niters_no_overflow parameters. Update calls to
18583 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18584 Create a new SSA name if the latter choses to use a ste other
18585 than zero, and return it via niters_vector_mult_vf_var.
18586 * tree-vect-loop.c (vect_transform_loop): Update calls to
18587 vect_do_peeling, vect_gen_vector_loop_niters and
18588 slpeel_make_loop_iterate_ntimes.
18589 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18590 (vect_gen_vector_loop_niters): Update declarations after above changes.
18592 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18594 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18595 128-bit round to integer instructions.
18596 (ceil<mode>2): Likewise.
18597 (btrunc<mode>2): Likewise.
18598 (round<mode>2): Likewise.
18600 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18602 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18603 unaligned VSX load/store on P8/P9.
18604 (expand_block_clear): Allow the use of unaligned VSX
18605 load/store on P8/P9.
18607 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18609 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18611 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18612 swap associated with both a load and a store.
18614 2018-01-02 Andrew Waterman <andrew@sifive.com>
18616 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18617 * config/riscv/riscv.md (clear_cache): Use it.
18619 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18621 * web.c: Remove out-of-date comment.
18623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18625 * expr.c (fixup_args_size_notes): Check that any existing
18626 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18627 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18628 (emit_single_push_insn): ...here.
18630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18632 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18633 (const_vector_encoded_nelts): New function.
18634 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18635 (const_vector_int_elt, const_vector_elt): Declare.
18636 * emit-rtl.c (const_vector_int_elt_1): New function.
18637 (const_vector_elt): Likewise.
18638 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18639 of CONST_VECTOR_ELT.
18641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18643 * expr.c: Include rtx-vector-builder.h.
18644 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18645 directly on the tree encoding.
18646 (const_vector_from_tree): Likewise.
18647 * optabs.c: Include rtx-vector-builder.h.
18648 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18649 sequence of "u" values.
18650 * vec-perm-indices.c: Include rtx-vector-builder.h.
18651 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18652 directly on the vec_perm_indices encoding.
18654 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18656 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18657 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18658 * rtx-vector-builder.h: New file.
18659 * rtx-vector-builder.c: Likewise.
18660 * rtl.h (rtx_def::u2): Add a const_vector field.
18661 (CONST_VECTOR_NPATTERNS): New macro.
18662 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18663 (CONST_VECTOR_DUPLICATE_P): Likewise.
18664 (CONST_VECTOR_STEPPED_P): Likewise.
18665 (CONST_VECTOR_ENCODED_ELT): Likewise.
18666 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18667 (unwrap_const_vec_duplicate): Likewise.
18668 (const_vec_series_p): Check for a non-duplicated vector encoding.
18669 Say that the function only returns true for integer vectors.
18670 * emit-rtl.c: Include rtx-vector-builder.h.
18671 (gen_const_vec_duplicate_1): Delete.
18672 (gen_const_vector): Call gen_const_vec_duplicate instead of
18673 gen_const_vec_duplicate_1.
18674 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18675 (gen_const_vec_duplicate): Use rtx_vector_builder.
18676 (gen_const_vec_series): Likewise.
18677 (gen_rtx_CONST_VECTOR): Likewise.
18678 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18679 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18680 Build a new vector rather than modifying a CONST_VECTOR in-place.
18681 (handle_special_swappables): Update call accordingly.
18682 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18683 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18684 Build a new vector rather than modifying a CONST_VECTOR in-place.
18685 (handle_special_swappables): Update call accordingly.
18687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18689 * simplify-rtx.c (simplify_const_binary_operation): Use
18690 CONST_VECTOR_ELT instead of XVECEXP.
18692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18694 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18695 the selector elements to be different from the data elements
18696 if the selector is a VECTOR_CST.
18697 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18698 ssizetype for the selector.
18700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18702 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18703 before testing each element individually.
18704 * tree-vect-generic.c (lower_vec_perm): Likewise.
18706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18708 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18709 * selftest-run-tests.c (selftest::run_tests): Call it.
18710 * vector-builder.h (vector_builder::operator ==): New function.
18711 (vector_builder::operator !=): Likewise.
18712 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18713 (vec_perm_indices::all_from_input_p): New function.
18714 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18715 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18716 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18717 instead of reading the VECTOR_CST directly. Detect whether both
18718 vector inputs are the same before constructing the vec_perm_indices,
18719 and update the number of inputs argument accordingly. Use the
18720 utility functions added above. Only construct sel2 if we need to.
18722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18724 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18725 the broadcast of the low byte.
18726 (expand_mult_highpart): Use an explicit encoding for the permutes.
18727 * optabs-query.c (can_mult_highpart_p): Likewise.
18728 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18729 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18730 (vectorizable_bswap): Likewise.
18731 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18732 explicit encoding for the power-of-2 permutes.
18733 (vect_permute_store_chain): Likewise.
18734 (vect_grouped_load_supported): Likewise.
18735 (vect_permute_load_chain): Likewise.
18737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18739 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18740 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18741 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18742 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18743 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18744 (vect_gen_perm_mask_any): Likewise.
18746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18748 * int-vector-builder.h: New file.
18749 * vec-perm-indices.h: Include int-vector-builder.h.
18750 (vec_perm_indices): Redefine as an int_vector_builder.
18751 (auto_vec_perm_indices): Delete.
18752 (vec_perm_builder): Redefine as a stand-alone class.
18753 (vec_perm_indices::vec_perm_indices): New function.
18754 (vec_perm_indices::clamp): Likewise.
18755 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18756 (vec_perm_indices::new_vector): New function.
18757 (vec_perm_indices::new_expanded_vector): Update for new
18758 vec_perm_indices class.
18759 (vec_perm_indices::rotate_inputs): New function.
18760 (vec_perm_indices::all_in_range_p): Operate directly on the
18761 encoded form, without computing elided elements.
18762 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18763 encoding. Update for new vec_perm_indices class.
18764 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18765 the given vec_perm_builder.
18766 (expand_vec_perm_var): Update vec_perm_builder constructor.
18767 (expand_mult_highpart): Use vec_perm_builder instead of
18768 auto_vec_perm_indices.
18769 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18770 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18771 or double series encoding as appropriate.
18772 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18773 vec_perm_indices instead of auto_vec_perm_indices.
18774 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18775 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18776 (vect_permute_store_chain): Likewise.
18777 (vect_grouped_load_supported): Likewise.
18778 (vect_permute_load_chain): Likewise.
18779 (vect_shift_permute_load_chain): Likewise.
18780 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18781 (vect_transform_slp_perm_load): Likewise.
18782 (vect_schedule_slp_instance): Likewise.
18783 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18784 (vectorizable_mask_load_store): Likewise.
18785 (vectorizable_bswap): Likewise.
18786 (vectorizable_store): Likewise.
18787 (vectorizable_load): Likewise.
18788 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18789 vec_perm_indices instead of auto_vec_perm_indices. Use
18790 tree_to_vec_perm_builder to read the vector from a tree.
18791 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18792 vec_perm_builder instead of a vec_perm_indices.
18793 (have_whole_vector_shift): Use vec_perm_builder and
18794 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18795 truncation to calc_vec_perm_mask_for_shift.
18796 (vect_create_epilog_for_reduction): Likewise.
18797 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18798 from auto_vec_perm_indices to vec_perm_indices.
18799 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18800 instead of changing individual elements.
18801 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18802 the vector in d.perm.
18803 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18804 from auto_vec_perm_indices to vec_perm_indices.
18805 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18806 instead of changing individual elements.
18807 (arm_vectorize_vec_perm_const): Use new_vector to install
18808 the vector in d.perm.
18809 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18810 Update vec_perm_builder constructor.
18811 (rs6000_expand_interleave): Likewise.
18812 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18813 (rs6000_expand_interleave): Likewise.
18815 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18817 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18818 to qimode could truncate the indices.
18819 * optabs.c (expand_vec_perm_var): Likewise.
18821 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18823 * Makefile.in (OBJS): Add vec-perm-indices.o.
18824 * vec-perm-indices.h: New file.
18825 * vec-perm-indices.c: Likewise.
18826 * target.h (vec_perm_indices): Replace with a forward class
18828 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18829 * optabs.h: Include vec-perm-indices.h.
18830 (expand_vec_perm): Delete.
18831 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18832 (expand_vec_perm_const): Declare.
18833 * target.def (vec_perm_const_ok): Replace with...
18834 (vec_perm_const): ...this new hook.
18835 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18836 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18837 * doc/tm.texi: Regenerate.
18838 * optabs.def (vec_perm_const): Delete.
18839 * doc/md.texi (vec_perm_const): Likewise.
18840 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18841 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18842 expand_vec_perm for constant permutation vectors. Assert that
18843 the mode of variable permutation vectors is the integer equivalent
18844 of the mode that is being permuted.
18845 * optabs-query.h (selector_fits_mode_p): Declare.
18846 * optabs-query.c: Include vec-perm-indices.h.
18847 (selector_fits_mode_p): New function.
18848 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18849 is defined, instead of checking whether the vec_perm_const_optab
18850 exists. Use targetm.vectorize.vec_perm_const instead of
18851 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18852 fit in the vector mode before using a variable permute.
18853 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18854 vec_perm_indices instead of an rtx.
18855 (expand_vec_perm): Replace with...
18856 (expand_vec_perm_const): ...this new function. Take the selector
18857 as a vec_perm_indices rather than an rtx. Also take the mode of
18858 the selector. Update call to shift_amt_for_vec_perm_mask.
18859 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18860 Use vec_perm_indices::new_expanded_vector to expand the original
18861 selector into bytes. Check whether the indices fit in the vector
18862 mode before using a variable permute.
18863 (expand_vec_perm_var): Make global.
18864 (expand_mult_highpart): Use expand_vec_perm_const.
18865 * fold-const.c: Includes vec-perm-indices.h.
18866 * tree-ssa-forwprop.c: Likewise.
18867 * tree-vect-data-refs.c: Likewise.
18868 * tree-vect-generic.c: Likewise.
18869 * tree-vect-loop.c: Likewise.
18870 * tree-vect-slp.c: Likewise.
18871 * tree-vect-stmts.c: Likewise.
18872 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18874 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18875 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18876 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18877 (aarch64_vectorize_vec_perm_const): ...this new function.
18878 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18879 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18880 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18881 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18882 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18883 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18884 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18886 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18887 check for NEON modes.
18888 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18889 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18890 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18891 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18893 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18894 the old VEC_PERM_CONST conditions.
18895 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18896 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18897 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18898 (ia64_vectorize_vec_perm_const_ok): Merge into...
18899 (ia64_vectorize_vec_perm_const): ...this new function.
18900 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18901 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18902 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18903 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18904 * config/mips/mips.c (mips_expand_vec_perm_const)
18905 (mips_vectorize_vec_perm_const_ok): Merge into...
18906 (mips_vectorize_vec_perm_const): ...this new function.
18907 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18908 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18909 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18910 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18911 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18912 (rs6000_expand_vec_perm_const): Delete.
18913 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18915 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18916 (altivec_expand_vec_perm_const_le): Take each operand individually.
18917 Operate on constant selectors rather than rtxes.
18918 (altivec_expand_vec_perm_const): Likewise. Update call to
18919 altivec_expand_vec_perm_const_le.
18920 (rs6000_expand_vec_perm_const): Delete.
18921 (rs6000_vectorize_vec_perm_const_ok): Delete.
18922 (rs6000_vectorize_vec_perm_const): New function.
18923 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18924 an element count and rtx array.
18925 (rs6000_expand_extract_even): Update call accordingly.
18926 (rs6000_expand_interleave): Likewise.
18927 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18928 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18929 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18930 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18931 (rs6000_expand_vec_perm_const): Delete.
18932 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18933 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18934 (altivec_expand_vec_perm_const_le): Take each operand individually.
18935 Operate on constant selectors rather than rtxes.
18936 (altivec_expand_vec_perm_const): Likewise. Update call to
18937 altivec_expand_vec_perm_const_le.
18938 (rs6000_expand_vec_perm_const): Delete.
18939 (rs6000_vectorize_vec_perm_const_ok): Delete.
18940 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18941 reference to the SPE evmerge intructions.
18942 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18943 an element count and rtx array.
18944 (rs6000_expand_extract_even): Update call accordingly.
18945 (rs6000_expand_interleave): Likewise.
18946 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18947 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18949 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18951 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18953 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18954 vector mode and that that mode matches the mode of the data
18956 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18957 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18958 directly using expand_vec_perm_1 when forcing selectors into
18960 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18964 * optabs-query.h (can_vec_perm_p): Delete.
18965 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18966 * optabs-query.c (can_vec_perm_p): Split into...
18967 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18968 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18969 particular selector is valid.
18970 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18971 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18972 (vect_grouped_load_supported): Likewise.
18973 (vect_shift_permute_load_chain): Likewise.
18974 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18975 (vect_transform_slp_perm_load): Likewise.
18976 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18977 (vectorizable_bswap): Likewise.
18978 (vect_gen_perm_mask_checked): Likewise.
18979 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18980 implementations of variable permutation vectors into account
18981 when deciding which selector to use.
18982 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18983 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18984 with a false third argument.
18985 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18986 to test whether the constant selector is valid and can_vec_perm_var_p
18987 to test whether a variable selector is valid.
18989 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18991 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18992 * optabs-query.c (can_vec_perm_p): Likewise.
18993 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18994 instead of vec_perm_indices.
18995 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18996 (vect_gen_perm_mask_checked): Likewise,
18997 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18998 (vect_gen_perm_mask_checked): Likewise,
19000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19002 * optabs-query.h (qimode_for_vec_perm): Declare.
19003 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19004 (qimode_for_vec_perm): ...this new function.
19005 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19007 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19009 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19010 does not have a conditional at the top.
19012 2018-01-02 Richard Biener <rguenther@suse.de>
19014 * ipa-inline.c (big_speedup_p): Fix expression.
19016 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19019 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19022 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19026 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19027 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19028 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19029 cond_taken_branch_cost 3->4.
19031 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19033 PR tree-optimization/83581
19034 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19035 TODO_cleanup_cfg if any changes have been made.
19037 PR middle-end/83608
19038 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19039 convert_modes if target mode has the right side, but different mode
19042 PR middle-end/83609
19043 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19044 last argument when extracting from CONCAT. If either from_real or
19045 from_imag is NULL, use expansion through memory. If result is not
19046 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19047 the parts directly to inner mode, if even that fails, use expansion
19050 PR middle-end/83623
19051 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19052 check for bswap in mode rather than HImode and use that in expand_unop
19055 Copyright (C) 2018 Free Software Foundation, Inc.
19057 Copying and distribution of this file, with or without modification,
19058 are permitted in any medium without royalty provided the copyright
19059 notice and this notice are preserved.