1 2018-07-05 James Clarke <jrtc27@jrtc27.com>
3 * configure: Regenerated.
5 2018-07-05 Carl Love <cel@us.ibm.com>
7 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
10 VSX_BUILTIN_DOUBLEL_V4SF.
12 2018-07-05 Martin Sebor <msebor@redhat.com>
15 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
16 than its domain to compute its the upper bound of a char array.
18 2018-07-05 Nathan Sidwell <nathan@acm.org>
20 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
21 * doc/cpp.texi: Update comment.
22 * doc/tm.texi: Rebuilt.
23 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
24 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
25 * doc/extend.texi (Backwards Compatibility): Clarify it is system
26 headers affected by extern "C".
27 * system.h: Poison NO_IMPLICIT_EXTERN_C.
28 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
29 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
30 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
31 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
32 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
33 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
34 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
35 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
36 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
37 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
38 config/sparc/sp64-elf.h, config/spu/spu.h,
39 config/stormy16/stormy16.h, config/v850/v850.h,
40 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
41 define NO_IMPLICIT_EXTERN_C.
42 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
44 2018-07-05 Tamar Christina <tamar.christina@arm.com>
47 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
48 instead of GET_MODE_SIZE when comparing Units.
50 2018-07-05 Tamar Christina <tamar.christina@arm.com>
53 * rtlanal.c (set_noop_p): Constrain on mode change,
54 include hard-reg-set.h
56 2018-07-05 Tamar Christina <tamar.christina@arm.com>
58 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
60 2018-07-05 Jakub Jelinek <jakub@redhat.com>
63 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
66 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
68 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
70 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
73 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
75 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
77 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
80 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
82 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
84 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
87 2018-07-04 Martin Liska <mliska@suse.cz>
89 * tree-switch-conversion.c: Define
90 max_ratio_for_speed and max_ratio_for_size constants.
92 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
93 Martin Liska <mliska@suse.cz>
98 * common.opt: Rename align options with 'str_' prefix.
99 * common/config/i386/i386-common.c (set_malign_value): New
101 (ix86_handle_option): Use it to set -falign-* options/
102 * config/aarch64/aarch64-protos.h (struct tune_params): Change
103 type from int to string.
104 * config/aarch64/aarch64.c: Update default values from int
106 * config/alpha/alpha.c (alpha_override_options_after_change):
108 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
109 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
110 max skip conditionally.
111 * config/i386/freebsd.h (SUBALIGN_LOG): New.
112 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
113 max skip conditionally.
114 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
115 max skip conditionally.
116 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
117 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
118 max skip conditionally.
119 * config/i386/i386.c (struct ptt): Change type from int to
121 (ix86_default_align): Set default values.
122 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
123 max skip conditionally.
124 * config/i386/iamcu.h (SUBALIGN_LOG): New.
125 (ASM_OUTPUT_MAX_SKIP_ALIGN):
126 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
127 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
128 max skip conditionally.
129 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
130 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
131 * config/i386/x86-64.h (SUBALIGN_LOG): New.
132 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
133 max skip conditionally.
134 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
135 * config/ia64/ia64.c (ia64_option_override): Set default values
136 for alignment options.
137 * config/m68k/m68k.c: Handle new str_align_* options.
138 * config/mips/mips.c (mips_set_compression_mode): Change
140 (mips_option_override): Set default values for options.
141 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
143 * config/rs6000/rs6000.c (rs6000_option_override_internal):
145 * config/rx/rx.c (rx_option_override): Likewise.
146 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
147 (LABEL_ALIGN): Use align_labels_log.
148 (LOOP_ALIGN): Use align_loops_align.
149 * config/s390/s390.c (s390_asm_output_function_label): Use new
151 * config/sh/sh.c (sh_override_options_after_change):
152 Change type of constants.
153 * config/spu/spu.c (spu_sched_init): Likewise.
154 * config/sparc/sparc.c (sparc_option_override): Set default
156 * config/visium/visium.c (visium_option_override): Likewise.
157 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
158 emit p2align format with last argument if it's not needed.
159 * doc/invoke.texi: Document extended format of -falign-*.
160 * final.c: Use align_labels alignment.
161 * flags.h (struct target_flag_state): Change type to use
163 (struct align_flags_tuple): New.
164 (struct align_flags): Likewise.
165 (align_loops_log): Redefine macro to use new types.
166 (align_loops_max_skip): Redefine macro to use new types.
167 (align_jumps_log): Redefine macro to use new types.
168 (align_jumps_max_skip): Redefine macro to use new types.
169 (align_labels_log): Redefine macro to use new types.
170 (align_labels_max_skip): Redefine macro to use new types.
171 (align_functions_log): Redefine macro to use new types.
172 (align_loops): Redefine macro to use new types.
173 (align_jumps): Redefine macro to use new types.
174 (align_labels): Redefine macro to use new types.
175 (align_functions): Redefine macro to use new types.
176 (align_functions_max_skip): Redefine macro to use new types.
177 (align_loops_value): New macro.
178 (align_jumps_value): New macro.
179 (align_labels_value): New macro.
180 (align_functions_value): New macro.
181 * function.c (invoke_set_current_function_hook): Propagate
182 alignment values from flags to global variables default in
184 * ipa-icf.c (sem_function::equals_wpa): Use
185 cl_optimization_option_eq instead of memcmp.
186 * lto-streamer.h (cl_optimization_stream_out): Support streaming
188 (cl_optimization_stream_in): Likewise.
189 * optc-save-gen.awk: Support strings in cl_optimization.
190 * opth-gen.awk: Likewise.
191 * opts.c (finish_options): Remove error checking of invalid
193 (MAX_CODE_ALIGN): Remove.
194 (MAX_CODE_ALIGN_VALUE): Likewise.
195 (parse_and_check_align_values): New function.
196 (check_alignment_argument): Likewise.
197 (common_handle_option): Use check_alignment_argument.
198 * opts.h (parse_and_check_align_values): Declare.
199 * toplev.c (init_alignments): Remove.
200 (read_log_maxskip): New.
201 (parse_N_M): Likewise.
202 (parse_alignment_opts): Likewise.
203 (backend_init_target): Remove usage of init_alignments.
204 * toplev.h (parse_alignment_opts): Declare.
205 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
207 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
208 * tree.c (cl_option_hasher::equal): New.
209 * varasm.c: Use new global macros.
211 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
213 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
214 Use a simpler align directive also if MAXSKIP = ALIGN-1.
215 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
216 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
217 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
218 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
219 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
220 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
221 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
222 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
223 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
224 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
225 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
227 2018-07-04 Martin Liska <mliska@suse.cz>
228 Jonathan Wakely <jwakely@redhat.com>
230 * coverage.c: Use correct type.
231 * doc/invoke.texi: Language correction.
233 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
236 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
237 ENDBRANCH for non-tail call which may return via indirect branch.
238 * doc/extend.texi: Document indirect_return attribute.
240 2018-07-03 Martin Sebor <msebor@redhat.com>
242 PR tree-optimization/86274
243 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
245 (format_floating): Correct handling of infinities and NaNs.
247 2018-07-03 Martin Sebor <msebor@redhat.com>
249 * print-tree.c (print_real_cst): New function.
250 (print_node_brief): Call it.
253 2018-07-03 Jeff Law <law@redhat.com>
255 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
256 into a single pattern.
258 * config/h8300/h8300.md (ors code_iterator): New.
259 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
260 a single pattern and single splitter.
261 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
262 (iorqi3_1, xorqi3_1): Likewise.
263 (iorqi3, xorqi3 expanders): Similarly.
265 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
266 (movmd_internal) into a single pattern using the P mode iterator.
267 (movmd splitters): Similarly.
268 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
269 (movsd splitters): Similarly.
271 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
272 ADDB, ADDW and ADDL into a single ADD attribute which selects the
273 right table based on the size of the operand.
274 * config/h8300/h8300.md (length_table): Corresponding changes. All
275 references to "addb", "addw" and "addl" changed to "add".
276 (btst patterns): Merge two variants into a single pattern.
277 (tstqi, tsthi): Likewise.
278 (addhi3_incdec, addsi3_incdec): Likewise.
279 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
280 (mulhi3, mulsi3): Likewise.
281 (udivhi3, udivsi3): Likewise.
282 (divhi3, divsi3): Likewise.
283 (andorqi3, andorhi3, andorsi3): Likewise.
285 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
288 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
289 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
291 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
293 PR tree-optimization/85694
294 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
295 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
297 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
298 (<u>avg<mode>3_ceil): New patterns.
300 2018-07-03 David Malcolm <dmalcolm@redhat.com>
302 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
303 scan-tree-dump directive.
304 * gcc.dg/vect/slp-perm-2.c: Likewise.
305 * gcc.dg/vect/slp-perm-3.c: Likewise.
306 * gcc.dg/vect/slp-perm-5.c: Likewise.
307 * gcc.dg/vect/slp-perm-6.c: Likewise.
308 * gcc.dg/vect/slp-perm-7.c: Likewise.
309 * gcc.dg/vect/slp-perm-8.c: Likewise.
311 2018-07-03 Marek Polacek <polacek@redhat.com>
314 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
316 2018-07-03 Richard Biener <rguenther@suse.de>
319 * tree-ssa-structalias.c (find_func_clobbers): Properly
320 handle indirect calls.
322 2018-07-03 Jeff Law <law@redhat.com>
324 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
325 (shifts): New code iterator.
326 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
327 expander. Fix HImode handling on H8/SX.
328 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
329 (subqi3, subhi3, subsi3 expanders): Likewise.
330 (andqi3, andhi3, andsi3 expanders): Likewise.
331 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
332 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
333 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
334 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
335 (zero_extendqihi2, zero_extendqisi2): Likewise.
336 (extendqihi2, extendqisi2): Likewise.
337 (rotlqi3, rotlhi3, rotlsi3): Likewise.
338 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
339 (rotlqi3_1, rotlhi3_1): Likewise.
340 (logicalhi3_sn, logicalsi3_sn): Likewise.
341 (logicalhi3, logicalsi3): Likewise.
343 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
345 * tree-vect-patterns.c (vect_recog_rotate_pattern)
346 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
347 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
348 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
349 type to append_pattern_def_seq instead of creating a stmt_vec_info
351 (build_mask_conversion): Likewise. Remove vinfo argument.
352 (vect_add_conversion_to_patterm): Likewise, renaming to...
353 (vect_add_conversion_to_pattern): ...this.
354 (vect_recog_mask_conversion_pattern): Update call to
355 build_mask_conversion. Pass the vector type to
356 append_pattern_def_seq here too.
357 (vect_recog_gather_scatter_pattern): Update call to
358 vect_add_conversion_to_pattern.
360 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
362 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
363 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
364 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
365 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
366 STMT_VINFO_PATTERN_DEF_SEQ to null here.
367 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
368 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
369 append_pattern_def_seq instead of new_pattern_def_seq.
370 (vect_recog_divmod_pattern): Do both of the above.
371 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
374 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
376 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
377 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
378 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
379 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
380 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
381 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
382 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
383 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
384 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
385 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
386 parameter with a single stmt_vec_info.
387 (vect_recog_func_ptr): Likewise.
388 (vect_recog_gather_scatter_pattern): Likewise, folding in...
389 (vect_try_gather_scatter_pattern): ...this.
390 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
391 the stmt_vec_info of the statement to be matched. Don't clear
392 STMT_VINFO_RELATED_STMT.
393 (vect_pattern_recog): Update call accordingly.
395 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
397 PR tree-optimization/85694
398 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
399 (uavgM3_ceil): Document new optabs.
400 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
401 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
403 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
404 (savg_ceil_optab): New optabs.
405 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
406 (vect_vect_recog_func_ptrs): Add it.
407 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
408 constant directly from the associated lhs.
410 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
412 * tree-vect-patterns.c (vect_split_statement): New function.
413 (vect_convert_input): Use it to try to split an existing cast.
415 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
417 * poly-int.h (print_hex): New function.
418 * dumpfile.h (dump_dec, dump_hex): Declare.
419 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
420 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
421 min_input_precision, operation_precision and operation_sign.
422 * tree-vect-patterns.c (vect_get_range_info): New function.
423 (vect_same_loop_or_bb_p, vect_single_imm_use)
424 (vect_operation_fits_smaller_type): Delete.
425 (vect_look_through_possible_promotion): Add an optional
426 single_use_p parameter.
427 (vect_recog_over_widening_pattern): Rewrite to use new
428 stmt_vec_info infomration. Handle one operation at a time.
429 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
430 (vect_truncatable_operation_p, vect_set_operation_type)
431 (vect_set_min_input_precision): New functions.
432 (vect_determine_min_output_precision_1): Likewise.
433 (vect_determine_min_output_precision): Likewise.
434 (vect_determine_precisions_from_range): Likewise.
435 (vect_determine_precisions_from_users): Likewise.
436 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
437 (vect_vect_recog_func_ptrs): Put over_widening first.
439 (vect_pattern_recog): Call vect_determine_precisions.
441 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
443 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
444 statements that have been replaced by further pattern statements.
445 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
447 2018-07-03 Richard Biener <rguenther@suse.de>
449 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
450 always set *dt. Dump vectype in vectype overload.
451 * dumpfile.h (dump_gimple_expr): New function.
452 (dump_gimple_expr_loc): Likewise.
453 * dumpfile.c (dump_gimple_expr): New function.
454 (dump_gimple_expr_loc): Likewise.
456 2018-07-02 Jeff Law <law@redhat.com>
458 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
459 the H8/300, H8/300H and H8/S variants into a single pattern.
460 (movhi_h8300, movqi_h8300hs): Similarly.
461 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
462 (QHI mode iterator): New.
464 * config/h8300/h8300.md: Remove trailing whitespace.
466 2018-07-02 Jim Wilson <jimw@sifive.com>
468 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
469 instead of emit_insn for interrupt returns.
470 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
471 (riscv_sret, riscv_uret): Likewise.
473 2018-07-02 David Malcolm <dmalcolm@redhat.com>
475 * pretty-print.c (selftest::test_pp_format): Move save and restore
476 of quotes to class auto_fix_quotes, and add an instance.
477 * selftest.c: Include "intl.h".
478 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
479 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
480 * selftest.h (selftest::auto_fix_quotes): New class.
482 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
484 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
485 (aarch64_sve_prepare_conditional_op): Remove.
486 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
487 Allow aarch64_simd_reg_or_zero as select operand; remove
488 the aarch64_sve_prepare_conditional_op call.
489 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
490 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
491 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
492 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
493 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
494 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
495 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
496 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
497 and a splitters to match all of the *_any patterns.
498 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
500 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
501 (SVE_COND_FP_BINARY_REV): Remove.
502 (sve_int_op_rev, sve_fp_op_rev): New.
503 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
504 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
505 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
506 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
507 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
508 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
509 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
510 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
511 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
513 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
514 Remove match_dup 1 from the inner unspec.
515 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
517 * config/aarch64/aarch64.md (movprfx): New attr.
518 (length): Default movprfx to 8.
519 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
520 (*madd<SVE_I>, *msub<SVE_I): Likewise.
521 (*<su>mul<SVE_I>3_highpart): Likewise.
522 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
523 (*v<ASHIFT><SVE_I>3): Likewise.
524 (*<su><MAXMIN><SVE_I>3): Likewise.
525 (*<su><MAXMIN><SVE_F>3): Likewise.
526 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
527 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
528 (*div<SVE_F>4): Likewise.
530 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
532 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
535 2018-07-02 Richard Biener <rguenther@suse.de>
537 PR tree-optimization/86363
538 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
539 memset argument refers to a non-variable address.
541 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
543 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
544 {PLUS,MINUS}_EXPR code to...
545 (adjust_symbolic_bound): ...here,
546 (combine_bound): ...here,
547 (set_value_range_with_overflow): ...and here.
549 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
551 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
553 (extract_range_from_abs_expr): ...here.
555 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
557 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
558 -fno-omit-frame-pointer when not optimizing.
560 2018-07-02 Martin Liska <mliska@suse.cz>
563 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
564 (propagate_nothrow): Likewise.
566 2018-07-02 Martin Liska <mliska@suse.cz>
569 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
571 2018-07-02 David Malcolm <dmalcolm@redhat.com>
573 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
574 function in r262149, changing "loc" param from source_location to
575 const dump_location_t &.
576 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
577 declaration, as above.
579 2018-07-01 Paul Koning <ni1d@arrl.net>
581 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
582 -munit-asm, -mgnu-asm, -mdec-asm.
583 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
584 (pdp11_output_labelref): New.
585 (pdp11_output_def): New.
586 (pdp11_output_addr_vec_elt): New.
587 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
588 %# and %@ format codes.
589 (pdp11_option_override): New.
590 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
591 (pdp11_output_ident): New.
592 (pdp11_asm_named_section): New.
593 (pdp11_asm_init_sections): New.
594 (pdp11_file_start): New.
595 (pdp11_file_end): New.
596 (output_ascii): Use .ascii/.asciz for -mdec-asm.
597 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
598 %o, like %c but octal.
599 (pdp11_option_override): New.
600 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
602 (DATA_SECTION_ASM_OP): Ditto.
603 (READONLY_DATA_SECTION_ASM_OP): New.
604 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
605 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
606 (ASM_OUTPUT_LABELREF): Ditto.
607 (ASM_OUTPUT_DEF): Ditto.
608 (ASM_OUTPUT_EXTERNAL): New.
609 (ASM_OUTPUT_SOURCE_FILENAME): New.
610 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
611 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
612 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
613 %# and %@ format codes.
614 * config/pdp11/pdp11.opt (mgnu-asm): New.
615 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
616 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
617 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
619 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
621 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
622 dereferencing path[] beyond its length.
624 (debug_all_paths): New.
625 (rewire_first_differing_edge): New.
626 (adjust_paths_after_duplication): New.
627 (duplicate_thread_path): Call adjust_paths_after_duplication.
629 (thread_through_all_blocks): Add new argument to
630 duplicate_thread_path.
632 2018-06-30 Jim Wilson <jimw@sifive.com>
634 * config/riscv/predicates.md (p2m1_shift_operand): New.
635 (high_mask_shift_operand): New.
636 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
637 pattern using p2m1_shift_operand.
638 (lshsi3_zero_extend_3+2): New combiner pattern using
639 high_mask_shift_operand.
641 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
643 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
645 (vect_recog_rotate_pattern): ...here.
646 (vect_convert_input): Try to insert casts of invariants in the
648 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
649 preheader to be empty.
651 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
653 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
654 vector type. If given, install it in the new statement's
656 (vect_element_precision): New function.
657 (vect_unpromoted_value): New struct.
658 (vect_unpromoted_value::vect_unpromoted_value): New function.
659 (vect_unpromoted_value::set_op): Likewise.
660 (vect_look_through_possible_promotion): Likewise.
661 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
662 (vect_widened_op_tree, vect_convert_input): Likewise.
663 (vect_convert_inputs, vect_convert_output): Likewise.
664 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
665 to handle the optional cast of the multiplication result and
666 vect_widened_op_tree to detect the widened multiplication itself.
667 Do not require the input and output of promotion casts to have
668 the same sign, but base the signedness of the operation on the
669 input rather than the result. If the pattern includes two
670 promotions, check that those promotions have the same sign.
671 Do not restrict the MULT_EXPR handling to a double-width result;
672 handle quadruple-width results and wider. Use vect_convert_inputs
673 to convert the inputs to the common type.
674 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
675 to handle the optional cast of the ABS result. Also allow a sign
676 change or a sign extension between the ABS and MINUS.
677 Use vect_widened_op_tree to detect the widened subtraction and use
678 vect_convert_inputs to convert the inputs to the common type.
679 (vect_handle_widen_op_by_const): Delete.
680 (vect_recog_widen_op_pattern): New function.
681 (vect_recog_widen_mult_pattern): Use it.
682 (vect_recog_widen_shift_pattern): Likewise.
683 (vect_recog_widen_sum_pattern): Use
684 vect_look_through_possible_promotion to handle the promoted
687 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
689 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
690 the containing gimple_seq *.
691 * gimple-iterator.h (gsi_for_stmt): Declare it.
692 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
693 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
694 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
695 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
696 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
698 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
700 (vect_mark_pattern_stmts): ...here. Handle cases in which the
701 statement being replaced is part of an existing pattern
702 definition sequence, inserting the new pattern statements before
704 (vect_pattern_recog_1): Don't return a bool. If the statement
705 is already part of a pattern, instead apply pattern matching
706 to the pattern definition statements. Don't clear the
707 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
708 (vect_pattern_recog): Don't break after the first match;
709 continue processing the pattern definition statements instead.
710 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
712 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
714 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
715 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
716 (vect_recog_widen_sum_pattern): Use it.
718 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
720 * tree-vect-loop.c (vectorizable_reduction): Assert that the
721 phi is not a pattern statement and has not been replaced by
723 * tree-vect-patterns.c (type_conversion_p): Don't check
724 STMT_VINFO_IN_PATTERN_P.
725 (vect_recog_vector_vector_shift_pattern): Likewise.
726 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
727 the pattern statement rather than the original statement; check
728 directly for a WIDEN_MULT_EXPR here.
729 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
730 vect_is_simple_use to return the pattern statement rather
731 than the original statement; use is_pattern_stmt_p to check
732 for such a pattern statement.
733 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
734 to return the pattern statement rather than the original statement;
735 don't do the same transformation here.
736 (vect_is_simple_use): If the defining statement has been replaced
737 by a pattern statement, return the pattern statement instead.
738 Remove the corresponding (local) transformation from the vectype
741 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
743 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
744 end and default to null.
745 * tree-vect-loop.c (vect_create_epilog_for_reduction)
746 (vectorizable_reduction): Update calls accordingly, dropping the
747 gimple ** argument if the passed-back statement isn't needed.
748 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
749 (vect_recog_rotate_pattern): Likewise.
750 (vect_recog_mask_conversion_pattern): Likewise.
751 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
752 (vect_mask_constant_operand_p): Likewise.
753 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
754 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
755 (get_group_load_store_type, get_load_store_type): Likewise.
756 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
757 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
758 (vectorizable_conversion, vectorizable_assignment): Likewise.
759 (vectorizable_shift, vectorizable_operation): Likewise.
760 (vectorizable_store, vect_is_simple_cond): Likewise.
761 (vectorizable_condition, vectorizable_comparison): Likewise.
762 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
763 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
764 and move it to the end. Cope with null def_stmt_outs.
766 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
768 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
770 2018-06-29 Jeff Law <law@redhat.com>
772 * config/v850/v850.c (v850_legitimate_address_p): Handle large
773 displacements for TARGET_V850E2V3 and newer.
774 (TARGET_LRA_P): Remove. Defaults to LRA now.
775 * config/v850/v850.md (sign23byte_load): Remove.
776 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
777 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
779 2018-06-29 Martin Liska <mliska@suse.cz>
782 * coverage.c (coverage_init): Mangle full path name.
783 * doc/invoke.texi: Document the change.
784 * gcov-io.c (mangle_path): New.
785 * gcov-io.h (mangle_path): Likewise.
786 * gcov.c (mangle_name): Use mangle_path for path mangling.
788 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
790 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
791 if starting source register is not even.
793 2018-06-29 Martin Liska <mliska@suse.cz>
795 PR tree-optimization/86263
796 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
797 Make edge redirection.
799 2018-06-29 David Malcolm <dmalcolm@redhat.com>
801 * dumpfile.c (dump_loc): Add indentation based on scope depth.
802 (dump_scope_depth): New variable.
803 (get_dump_scope_depth): New function.
804 (dump_begin_scope): New function.
805 (dump_end_scope): New function.
806 * dumpfile.h (get_dump_scope_depth): New declaration.
807 (dump_begin_scope): New declaration.
808 (dump_end_scope): New declaration.
809 (class auto_dump_scope): New class.
810 (AUTO_DUMP_SCOPE): New macro.
811 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
814 2018-06-29 Richard Biener <rguenther@suse.de>
816 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
817 compute_all_dependences succeeds.
818 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
819 exceed --param loop-max-datarefs-for-datadeps.
821 2018-06-29 Jakub Jelinek <jakub@redhat.com>
823 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
825 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
828 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
829 alternative 0 in preferred_for_speed attribute.
831 2018-06-28 Paul Koning <ni1d@arrl.net>
833 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
834 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
835 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
836 * config/pdp11/pdp11.md: Correct "length" attribute calculation
837 for shift insn patterns.
839 2018-06-28 David Malcolm <dmalcolm@redhat.com>
841 * cgraph.c (cgraph_node::get_body): Replace assignments to
842 "dump_file" with calls to set_dump_file.
843 * dumpfile.c (alt_dump_file): Make static, and group with...
844 (alt_flags): ...this definition.
845 (dumps_are_enabled): New variable.
846 (refresh_dumps_are_enabled): New function.
847 (set_dump_file): New function.
848 (set_alt_dump_file): New function.
849 (gcc::dump_manager::dump_start): Replace assignments to
850 "dump_file" and "alt_dump_file" with calls to set_dump_file and
852 (gcc::dump_manager::dump_finish): Likewise.
853 * dumpfile.h (alt_dump_file): Delete decl.
854 (dumps_are_enabled): New variable decl.
855 (set_dump_file): New function decl.
856 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
858 * tree-nested.c (lower_nested_functions): Replace assignments to
859 "dump_file" with calls to set_dump_file.
861 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
863 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
864 goto_locus of each outgoing edge of each basic block.
866 2018-06-28 Richard Biener <rguenther@suse.de>
868 * dwarf2out.c (decl_scope_table): Remove.
869 (push_decl_scope): Likewise.
870 (pop_decl_scope): Likewise.
871 (gen_type_die_for_member): Do not call push/pop_decl_scope.
872 (gen_struct_or_union_type_die): Likewise.
873 (gen_tagged_type_die): Likewise.
874 (dwarf2out_init): Do not initialize decl_scope_table.
875 (dwarf2out_c_finalize): Do not free it.
877 2018-06-28 Richard Biener <rguenther@suse.de>
879 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
880 deciding whether to not re-use a DIE.
882 2018-06-28 Richard Biener <rguenther@suse.de>
884 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
885 DW_AT_abstract_origin attribute.
887 2018-06-28 Martin Liska <mliska@suse.cz>
889 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
890 Use newly introduced constants.
891 * tree-switch-conversion.h (struct jump_table_cluster):
892 Define max_ratio_for_size and max_ratio_for_speed.
894 2018-06-28 Martin Liska <mliska@suse.cz>
896 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
897 Add new checking assert to catch invalid state.
898 (jump_table_cluster::can_be_handled): Handle single case
900 (jump_table_cluster::is_beneficial): Bail out for such case.
901 (bit_test_cluster::find_bit_tests):
902 Add new checking assert to catch invalid state.
903 (bit_test_cluster::can_be_handled): Handle single case
905 (bit_test_cluster::is_beneficial): Bail out for such case.
906 (switch_decision_tree::analyze_switch_statement):
909 2018-06-28 Martin Liska <mliska@suse.cz>
911 * common.opt: Introduce -completion option.
912 * gcc.c (driver_handle_option): Handle it.
913 (driver::main): Print completions if completion
915 * opt-suggestions.c (option_proposer::get_completions):
917 (option_proposer::suggest_completion): Likewise.
918 (option_proposer::find_param_completions): Likewise.
919 (verify_autocompletions): Likewise.
920 (test_completion_valid_options): Likewise.
921 (test_completion_valid_params): Likewise.
922 (in_completion_p): Likewise.
923 (empty_completion_p): Likewise.
924 (test_completion_partial_match): Likewise.
925 (test_completion_garbage): Likewise.
926 (opt_proposer_c_tests): Likewise.
927 * opt-suggestions.h: Declare new functions.
928 * opts.c (common_handle_option): Handle OPT__completion_.
929 * selftest-run-tests.c (selftest::run_tests): Add
930 opt_proposer_c_tests.
931 * selftest.c (assert_str_startswith): New.
932 * selftest.h (assert_str_startswith): Likewise.
933 (opt_proposer_c_tests): New.
934 (ASSERT_STR_STARTSWITH): Likewise.
936 2018-06-28 Martin Liska <mliska@suse.cz>
938 * Makefile.in: Add opt-suggestions.o.
939 * gcc-main.c: Include opt-suggestions.h.
940 * gcc.c (driver::driver): Likewise.
941 (driver::~driver): Remove m_option_suggestions.
942 (driver::build_option_suggestions): Moved to option_proposer.
943 (driver::suggest_option): Likewise.
944 (driver::handle_unrecognized_options): Use option_proposer.
945 * gcc.h (class driver): Add new memver m_option_proposer.
946 * opt-suggestions.c: New file.
947 * opt-suggestions.h: New file.
949 2018-06-28 Martin Liska <mliska@suse.cz>
951 * vec.h (class auto_string_vec): New (moved from auto_argvec).
952 (auto_string_vec::~auto_string_vec): Likewise.
954 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
956 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
957 prevent_decl_creation_for_types fields up and add reset_location field.
958 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
959 statement if id->reset_location is true.
960 (copy_edges_for_bb): Do not set goto_locus on the new edges if
961 id->reset_location is true.
962 (copy_phis_for_bb): Force input_location on the arguments if
963 id->reset_location is true.
964 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
965 is set on the function to be inlined.
967 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
969 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
971 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
973 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
975 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
976 hard registers for the clobbered pseudo.
978 2018-06-27 Paul Koning <ni1d@arrl.net>
980 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
981 mutually exclusive options.
982 * config/pdp11/constraints.md (h): New constraint.
983 (O): Update definition to match shift code generation.
985 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
987 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
989 (output_jump): Change arguments.
990 (pdp11_fixed_cc_regs): New function.
991 (pdp11_cc_mode): Ditto.
992 (pdp11_expand_shift): Ditto.
993 (pdp11_assemble_shift): Ditto.
994 (pdp11_small_shift): Ditto.
995 (pdp11_branch_cost): Remove.
996 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
998 (pdp11_register_move_cost): Update for CC registers.
999 (pdp11_rtx_costs): Add case for LSHIFTRT.
1000 (pdp11_output_jump): Add CCNZ mode conditional branches.
1001 (notice_update_cc_on_set): Remove.
1002 (pdp11_cc_mode): New function.
1003 (simple_memory_operand): Correct pre/post decrement case.
1004 (no_side_effect_operand): New function.
1005 (pdp11_regno_reg_class): Add CC_REGS class.
1006 (pdp11_fixed_cc_regs): New function.
1007 (pdp11_small_shift): New function.
1008 (pdp11_expand_shift): New function to expand shift insns.
1009 (pdp11_assemble_shift): New function to output shifts.
1010 (pdp11_branch_cost): Remove.
1011 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1012 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1013 (WCHAR_TYPE): Ditto.
1014 (PTRDIFF_TYPE): Ditto.
1015 (ADJUST_INSN_LENGTH): New macro.
1016 (FIXED_REGISTERS): Add CC registers.
1017 (CALL_USED_REGISTERS): Ditto.
1019 (REG_CLASS_NAMES): Ditto.
1020 (REG_CLASS_CONTENTS): Ditto.
1021 (SELECT_CC_MODE): Use new function.
1022 (TARGET_FLAGS_REGNUM): New macro.
1023 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1024 (cc0_reg_rtx): Remove.
1025 (CC_STATUS_MDEP): Remove.
1026 (CC_STATUS_MDEFP_INIT): Remove.
1027 (CC_IN_FPU): Remove.
1028 (NOTICE_UPDATE_CC): Remove.
1029 (REGISTER_NAMES): Add CC registers.
1030 (BRANCH_COST): Change to constant 1.
1031 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1033 * config/pdp11/pdp11.opt (mbcopy): Remove.
1034 (mbcopy-builtin): Remove.
1035 (mbranch-cheap): Remove.
1036 (mbranch-expensive): Remove.
1037 * config/pdp11/predicates.md (expand_shift_operand): Update to
1038 match shift code generation.
1039 (ccnz_operator): New predicate.
1040 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1041 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1042 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1044 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1045 description of O constraint.
1047 2018-06-27 Jeff Law <law@redhat.com>
1048 Austin Law <austinklaw@gmail.com>
1050 * config/v850/v850.md (addsi3_set_flags): New pattern.
1051 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1052 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1053 (zero_extendhisi2_v850_set_flags): Likewise.
1054 (zero_extendqisi2_v850_set_flags): Likewise.
1055 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1056 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1057 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1059 * config/v850/v850-protos.h (notice_update_cc): Remove.
1060 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1061 (v850_print_operand): Handle 'D' and "d".
1062 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1063 Add handling of arithmetic/logical operations compared against zero.
1064 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1065 Do not look at v850_compare_op, instead get mode from last argument.
1066 (v850_gen_compare): Remove
1067 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1068 after reload for prologue insns.
1069 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1071 (construct_save_jarl): Likewise.
1072 (TARGET_FLAGS_REGNUM): Define.
1073 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1074 (NOTICE_UPDATE_CC): Remove.
1075 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1076 than cc0. Conditionalize on reload_completed.
1077 (cmpsi_insn, setfcc_insn): Likewise.
1078 (tst1 splitter): Turn into define_and_split which sets the flags
1080 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1081 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1082 (cstoresf4, cstoredf4): Clobber the flags.
1083 (cmpsi, cmpsf, cmpdf): Remove expanders.
1084 (setf_insn): Remove pattern.
1085 (addsi3): Turn into define_and_split which clobbers the flags after
1086 reload and a suitable pattern (addsi3_clobber_flags) for use after
1088 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1089 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1090 (ashrsi3, ashrsi3_v850e2): Likewise.
1091 (bins): Clobber the flags.
1092 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1093 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1094 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1095 (call_value_internal_short, call_value_internal_long): Likewise.
1096 (callt_save_interrupt, callt_return_interrupt): Likewise.
1097 (save_interrupt, return_interrupt): Likewise.
1098 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1099 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1100 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1101 (All FP comparisons): Only allow after reload has completed.
1103 (divh, divhu): Tweak output template.
1104 (branch_z_normal, branch_z_invert): Remove
1105 (branch_nz_normal, branch_nz_invert): Likewise.
1106 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1108 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1109 * config/v850/v850.c (notice_update_cc): Remove.
1110 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1111 (CC_NO_CARRY): Likewise.
1112 (NOTICE_UPDATE_CC): Define to nothing.
1113 * config/v850/v850.md: Remove block comment on cc0 handling
1114 Remove "cc" attribute from all patterns. Remove cc_status handling
1115 from all patterns. Minor formatting fixes.
1117 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1119 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1120 (cortex-a76.cortex-a55): Likewise.
1121 * config/aarch64/aarch64-tune.md: Regenerate.
1122 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1123 cortex-a76.cortex-a55.
1125 2018-06-27 Jeff Law <law@redhat.com>
1127 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1128 (MULTILIB_DIRNAMES): Similarly.
1130 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1132 * gimple.h (gimple_return_retbnd): Delete.
1133 (gimple_return_set_retbnd): Likewise.
1134 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1135 gimple_return_set_retbnd.
1136 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1137 gimple_return_retbnd and adjust.
1138 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1139 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1140 Explicitly return NULL in a couple more cases. Move assertion
1141 on debug statements and remove unreachable code.
1142 (reset_debug_binding): Do not test id->retbnd.
1143 (expand_call_inline): Do not set it.
1145 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1147 * configure.ac: Add --disable-gcov option.
1148 * configure: Regenerate.
1149 * Makefile.in: Honour @enable_gcov@.
1150 * doc/install.texi: Document --disable-gcov.
1152 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1154 * config/arm/arm-cpus.in (cortex-a76): New entry.
1155 (cortex-a76.cortex-a55): Likewise.
1156 * config/arm/arm-tables.opt: Regenerate.
1157 * config/arm/arm-tune.md: Likewise.
1158 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1159 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1160 cortex-a76.cortex-a55.
1162 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1165 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1167 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1169 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1171 (EPILOGUE_USES): Likewise.
1173 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1175 * tree-inline.c (remap_location): New function extracted from...
1176 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1177 (copy_phis_for_bb): ...here. Call remap_location.
1178 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1180 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1182 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1183 unaligned vsx for 16B memset.
1185 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1188 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1189 ieee128_float_type_node to long_double_type_node unless
1190 TARGET_LONG_DOUBLE_128 is set.
1192 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1194 * cfgloop.c (get_loop_location): Convert return type from
1195 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1196 by implicit construction from rtx_insn *, and using
1197 dump_user_location_t::from_function_decl for the fallback case.
1198 * cfgloop.h (get_loop_location): Convert return type from
1199 location_t to dump_user_location_t.
1200 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1201 dump_printf_loc to pass in a dump_location_t rather than a
1202 location_t, via the gimple stmt.
1203 * coverage.c (get_coverage_counts): Update calls to
1204 dump_printf_loc to pass in dump_location_t rather than a
1206 * doc/optinfo.texi (Dump types): Convert example of
1207 dump_printf_loc from taking "locus" to taking "insn". Update
1208 description of the "_loc" calls to cover dump_location_t.
1209 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1211 (dump_user_location_t::dump_user_location_t): New constructors,
1212 from gimple *stmt and rtx_insn *.
1213 (dump_user_location_t::from_function_decl): New function.
1214 (dump_loc): Make static.
1215 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1216 const dump_location_t &.
1217 (dump_generic_expr_loc): Delete.
1218 (dump_printf_loc): Convert param "loc" from location_t to
1219 const dump_location_t &.
1220 (selftest::test_impl_location): New function.
1221 (selftest::dumpfile_c_tests): New function.
1222 * dumpfile.h: Include "profile-count.h".
1223 (class dump_user_location_t): New class.
1224 (struct dump_impl_location_t): New struct.
1225 (class dump_location_t): New class.
1226 (dump_printf_loc): Convert 2nd param from source_location to
1227 const dump_location_t &.
1228 (dump_generic_expr_loc): Delete.
1229 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1230 const dump_location_t &.
1231 * gimple-fold.c (fold_gimple_assign): Update call to
1232 dump_printf_loc to pass in a dump_location_t rather than a
1233 location_t, via the gimple stmt.
1234 (gimple_fold_call): Likewise.
1235 * gimple-loop-interchange.cc
1236 (loop_cand::analyze_iloop_reduction_var): Update for change to
1237 check_reduction_path.
1238 (tree_loop_interchange::interchange): Update for change to
1240 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1241 change in return-type of find_loop_location.
1242 (graphite_regenerate_ast_isl): Likewise.
1243 * graphite-optimize-isl.c (optimize_isl): Likewise.
1244 * graphite.c (graphite_transform_loops): Likewise.
1245 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1246 pass in a dump_location_t rather than a location_t, via the
1248 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1249 * ipa.c (walk_polymorphic_call_targets): Likewise.
1250 * loop-unroll.c (report_unroll): Convert "locus" param from
1251 location_t to dump_location_t.
1252 (decide_unrolling): Update for change to get_loop_location's
1254 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1255 location_t to dump_user_location_t.
1256 (grid_find_single_omp_among_assignments_1): Updates calls to
1257 dump_printf_loc to pass in a dump_location_t rather than a
1258 location_t, via the gimple stmt.
1259 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1260 location_t to dump_location_t. Updates calls to dump_printf_loc
1261 to pass in a dump_location_t rather than a location_t, via the
1263 (grid_inner_loop_gridifiable_p): Likewise.
1264 (grid_dist_follows_simple_pattern): Likewise.
1265 (grid_gfor_follows_tiling_pattern): Likewise.
1266 (grid_target_follows_gridifiable_pattern): Likewise.
1267 (grid_attempt_target_gridification): Convert initialization
1268 of local "grid" from memset to zero-initialization; FIXME: does
1269 this require C++11? Update call to dump_printf_loc to pass in a
1270 optinfo_location rather than a location_t, via the gimple stmt.
1271 * profile.c (read_profile_edge_counts): Updates call to
1272 dump_printf_loc to pass in a dump_location_t rather than a
1274 (compute_branch_probabilities): Likewise.
1275 * selftest-run-tests.c (selftest::run_tests): Call
1277 * selftest.h (dumpfile_c_tests): New decl.
1278 * tree-loop-distribution.c (pass_loop_distribution::execute):
1279 Update for change in return type of find_loop_location.
1280 * tree-parloops.c (parallelize_loops): Likewise.
1281 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1282 "locus" from location_t to dump_user_location_t.
1283 (canonicalize_loop_induction_variables): Likewise.
1284 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1285 for change in return type of find_loop_location.
1286 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1287 to dump_printf_loc to pass in a dump_location_t rather than a
1288 location_t, via the stmt.
1289 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1291 * tree-vect-loop-manip.c (find_loop_location): Convert return
1292 type from source_location to dump_user_location_t.
1293 (vect_do_peeling): Update for above change.
1294 (vect_loop_versioning): Update for change in type of
1296 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1297 from location_t to dump_user_location_t.
1298 (vect_estimate_min_profitable_iters): Update for change in type
1300 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1301 location_t to dump_location_t.
1302 (vect_slp_bb): Update for change in type of vect_location.
1303 * tree-vectorizer.c (vect_location): Convert from source_location
1304 to dump_user_location_t.
1305 (try_vectorize_loop_1): Update for change in vect_location's type.
1306 (vectorize_loops): Likewise.
1307 (increase_alignment): Likewise.
1308 * tree-vectorizer.h (vect_location): Convert from source_location
1309 to dump_user_location_t.
1310 (find_loop_location): Convert return type from source_location to
1311 dump_user_location_t.
1312 (check_reduction_path): Convert 1st param from location_t to
1313 dump_user_location_t.
1314 * value-prof.c (check_counter): Update call to dump_printf_loc to
1315 pass in a dump_user_location_t rather than a location_t; update
1316 call to error_at for change in type of "locus".
1317 (check_ic_target): Update call to dump_printf_loc to
1318 pass in a dump_user_location_t rather than a location_t, via the
1321 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1323 * config/s390/s390.h (enum processor_flags): Do not use
1324 default tune parameter when -march was specified.
1326 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1329 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1330 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1333 2018-06-26 Richard Biener <rguenther@suse.de>
1335 PR tree-optimization/86287
1337 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1338 (vect_analyze_loop): Initialize n_stmts.
1340 2018-06-26 Richard Biener <rguenther@suse.de>
1343 * fold-const.c (fold_convertible_p): Pointer extension
1346 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1349 * dwarf2out.c (loc_list_has_views): Adjust comments.
1350 (dw_loc_list): Split single cross-partition range with
1353 2018-06-25 Jeff Law <law@redhat.com>
1355 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1356 on -mbig-switch by default.
1358 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1360 (const_float_0_operand): Remove unused predicate.
1361 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1362 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1363 (recipsf2): New expander. Original pattern now called
1365 (recipdf2, recipdf2_insn): Similarly.
1366 (rsqrtsf2, rsqrtsf2_insn): Similarly
1367 (rsqrtdf2, rsqrtdf2_insn): Similarly
1369 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1371 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1372 Simplify logic for FreeBSD (twice).
1374 2018-06-25 Martin Sebor <msebor@redhat.com>
1376 PR tree-optimization/86204
1377 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1378 a strnlen result if it's less than the length of the string.
1380 2018-06-25 Martin Sebor <msebor@redhat.com>
1382 PR tree-optimization/85700
1383 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1384 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1385 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1387 2018-06-25 Martin Sebor <msebor@redhat.com>
1389 * doc/extend.texi (Zero-length arrays): Update and clarify.
1391 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1393 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1394 added IEEE/IBM long double multilib support on PowerPC little
1395 endian Linux systems.
1396 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1397 (MULTILIB_DEFAULTS): Likewise.
1398 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1400 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1401 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1402 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1404 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1407 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1408 (REORDER_45): Likewise.
1410 2018-06-25 Jeff Law <law@redhat.com>
1412 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1413 dividend to 32 bits. Adjust length.
1414 (udivmodhi4): Cleanup output template. Fix length.
1416 2018-06-25 Carl Love <cel@us.ibm.com>
1418 * config/rs6000/vsx.md: Change word selector to prefered location.
1420 2018-06-25 Richard Biener <rguenther@suse.de>
1422 PR tree-optimization/86304
1423 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1424 epilogue-if-converted loops as well.
1426 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1428 * lto-section-out.c (lto_begin_section): Do not print section
1429 name for noaddr and unnumbered dumps.
1431 2018-06-25 Richard Biener <rguenther@suse.de>
1433 * tree-vectorizer.h (struct vec_info_shared): New structure
1434 with parts split out from struct vec_info and loop_nest from
1435 struct _loop_vec_info.
1436 (struct vec_info): Adjust accordingly.
1437 (struct _loop_vec_info): Likewise.
1438 (LOOP_VINFO_LOOP_NEST): Adjust.
1439 (LOOP_VINFO_DATAREFS): Likewise.
1440 (LOOP_VINFO_DDRS): Likewise.
1441 (struct _bb_vec_info): Likewise.
1442 (BB_VINFO_DATAREFS): Likewise.
1443 (BB_VINFO_DDRS): Likewise.
1444 (struct _stmt_vec_info): Add dr_aux member.
1445 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1446 (DR_MISALIGNMENT_UNINITIALIZED): New.
1447 (set_dr_misalignment): Adjust.
1448 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1449 (vect_analyze_loop): Adjust prototype.
1450 (vect_analyze_loop_form): Likewise.
1451 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1452 Compute dependences lazily.
1453 (vect_record_base_alignments): Use shared datarefs/ddrs.
1454 (vect_verify_datarefs_alignment): Likewise.
1455 (vect_analyze_data_refs_alignment): Likewise.
1456 (vect_analyze_data_ref_accesses): Likewise.
1457 (vect_analyze_data_refs): Likewise.
1458 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1459 constructor parameter for shared part.
1460 (vect_analyze_loop_form): Pass in shared part and adjust.
1461 (vect_analyze_loop_2): Pass in storage for the number of
1462 stmts. Move loop nest finding to the caller. Compute
1464 (vect_analyze_loop): Pass in shared part.
1465 (vect_transform_loop): Verify shared datarefs are unchanged.
1466 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1467 constructor parameter for shared part.
1468 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1469 (vect_slp_bb): Verify shared datarefs are unchanged before
1471 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1473 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1474 DR_MISALIGNMENT_UNINITIALIZED.
1475 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1476 parameter for shared part.
1477 (vec_info::~vec_info): Adjust.
1478 (vec_info_shared::vec_info_shared): New.
1479 (vec_info_shared::~vec_info_shared): Likewise.
1480 (vec_info_shared::save_datarefs): Likewise.
1481 (vec_info_shared::check_datarefs): Likewise.
1482 (try_vectorize_loop_1): Construct shared part live for analyses
1483 of a single loop for multiple vector sizes.
1484 * tree-parloops.c (gather_scalar_reductions): Adjust.
1486 2018-06-25 Richard Biener <rguenther@suse.de>
1488 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1489 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1490 (vect_analyze_data_refs): Remove similar code from here and
1491 simplify accordingly.
1493 2018-06-25 Richard Biener <rguenther@suse.de>
1495 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1496 for reverse storage order accesses rather than asserting
1497 they cannot happen here.
1499 2018-06-25 Tom de Vries <tdevries@suse.de>
1502 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1503 Use data16 instead of .byte for insn prefix.
1505 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1508 * parser.c (make_char_string_pack): Pass this literal chars
1509 through cpp_interpret_string.
1510 (cp_parser_userdef_numeric_literal): Check the result of
1511 make_char_string_pack.
1513 2018-06-24 Maya Rashish <coypu@sdf.org>
1515 * ginclude/stddef.h: Simplify conditions around avoiding
1516 re-definition of __size_t.
1518 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1520 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1521 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1523 2018-06-22 Maya Rashish <coypu@sdf.org>
1525 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1526 made to OpenBSD and NetBSD through the years.
1528 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1530 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1531 behavior of vec_pack (vector double, vector double) to match
1532 behavior of vec_float2 (vector double, vector double).
1534 2018-06-22 Olivier Hainque <hainque@adacore.com>
1536 * gimplify.c (gimplify_function_tree): Prevent creation
1537 of a trampoline for the address of the current function
1538 passed to entry/exit instrumentation hooks.
1540 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1543 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1546 2018-06-22 Martin Liska <mliska@suse.cz>
1548 PR tree-optimization/86263
1549 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1550 Bail out if is_enabled is false.
1551 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1553 (jump_table_cluster::is_enabled): New function.
1555 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1557 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1558 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1559 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1560 (lto_input_ts_binfo_tree_pointers): Likewise.
1561 * tree-streamer-out.c (streamer_write_tree_bitfields,
1562 write_ts_binfo_tree_pointers): Likewise.
1563 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1565 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1567 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1569 2018-06-22 Martin Liska <mliska@suse.cz>
1571 * symbol-summary.h (get): Make it pure and inline move
1572 functionality from ::get function.
1573 (get): Remove and inline into ::get and ::get_create.
1574 (get_create): Move code from ::get function.
1576 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1579 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1580 -x assembler-with-cpp.
1582 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1584 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1585 _FILE_OFFSET_BITS=64 for C++.
1587 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1589 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1590 conversion insn that shows up when pr85657-3.c is compiled using
1591 IEEE 128-bit long double.
1592 (neg<mode>2_internal): Use the correct mode to check whether the
1593 mode is IBM extended.
1594 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1595 multiply and divide external functions from being created more
1598 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1600 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1602 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1603 the edge can be forwarded.
1604 (cfg_layout_merge_blocks): Likewise.
1606 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1608 * except.c (finish_eh_generation): Commit edge insertions only after
1609 the EH edges have been redirected from post-landing to landing pads.
1611 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1613 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1614 create_tmp_var_for to create the FRAME decl.
1615 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1617 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1619 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1620 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1621 debug statement when resetting its value.
1622 (expand_call_inline): Copy the locus of the call onto the assignment
1623 of the return value, if any. Use local variable in more cases.
1625 2018-06-21 Martin Liska <mliska@suse.cz>
1627 * ipa-pure-const.c (propagate_nothrow): Use
1628 funct_state_summaries->get.
1629 (dump_malloc_lattice): Likewise.
1630 (propagate_malloc): Likewise.
1632 2018-06-21 Richard Biener <rguenther@suse.de>
1634 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1635 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1636 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1638 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1639 BLOCK_ABSTRACT_ORIGIN unconditionally.
1641 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1643 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1645 * ipa-reference.c (ipa_reference_c_finalize): Delete
1646 ipa_ref_opt_sum_summaries and set it to NULL.
1648 2018-06-21 Tom de Vries <tdevries@suse.de>
1650 PR tree-optimization/85859
1651 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1652 test with comment from bb_no_side_effects_p.
1654 2018-06-21 Richard Biener <rguenther@suse.de>
1656 PR tree-optimization/86232
1657 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1658 max for constant niter.
1660 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1662 * config/aarch64/aarch64-simd.md
1663 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1665 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1667 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1668 Make opernads of the unspec commutative.
1670 2018-06-21 Richard Biener <rguenther@suse.de>
1672 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1673 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1674 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1675 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1677 (vect_record_base_alignments): Do not record base alignment
1678 for gathers or scatters.
1679 (vect_compute_data_ref_alignment): Drop return value that is always
1680 true. Bail out early for gathers or scatters.
1681 (vect_enhance_data_refs_alignment): Bail out early for gathers
1683 (vect_find_same_alignment_drs): Likewise.
1684 (vect_analyze_data_refs_alignment): Remove dead code.
1685 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1686 (vect_analyze_data_refs): For possible gathers or scatters do
1687 not create an alternate DR, just check their possible validity
1688 and mark them. Adjust DECL_NONALIASED handling to not rely
1690 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1691 update inits of gathers or scatters.
1692 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1693 Also copy gather/scatter flag to pattern vinfo.
1695 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1697 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1698 behavior of vec_packsu (vector unsigned long long, vector unsigned
1699 long long) to match behavior of vec_packs with same signature.
1701 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1702 Thomas Schwinge <thomas@codesourcery.com>
1703 Cesar Philippidis <cesar@codesourcery.com>
1705 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1706 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1707 (gimplify_adjust_omp_clauses): Likewise.
1708 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1709 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1710 (gimplify_omp_target_update): Update handling of acc update and
1712 * omp-low.c (install_var_field): Remove unused parameter
1713 base_pointers_restrict.
1714 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1715 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1717 (omp_target_base_pointers_restrict_p): Delete.
1718 (scan_omp_target): Update call to scan_sharing_clauses.
1719 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1721 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1722 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1723 (convert_local_omp_clauses): Likewise.
1724 * tree-pretty-print.c (dump_omp_clause): Likewise.
1725 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1727 (omp_clause_code_name): Likewise.
1729 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1732 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1735 PR tree-optimization/86231
1736 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1737 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1739 2018-06-20 Tom de Vries <tdevries@suse.de>
1741 PR tree-optimization/86097
1742 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1743 iv type if signedness of iv type is not the same as that of *nit.
1745 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1747 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1748 EDGE_EH edges, verify they are all EDGE_EH.
1750 2018-06-20 Maya Rashish <coypu@sdf.org>
1752 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1754 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1756 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1757 * config/aarch64/aarch64.c (xgene1_tunings): Add
1758 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1759 (aarch64_mode_valid_for_sched_fusion_p):
1760 Allow 16-byte modes.
1761 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1762 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1764 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1766 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1767 * config/aarch64/iterators.md (VQ2): New mode iterator.
1769 2018-06-20 Martin Liska <mliska@suse.cz>
1771 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1772 Change default ratio from 10 to 8.
1774 2018-06-20 Martin Liska <mliska@suse.cz>
1776 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1778 (bit_test_cluster::find_bit_tests): Likewise.
1779 (switch_decision_tree::analyze_switch_statement): Find clusters.
1780 * tree-switch-conversion.h (struct jump_table_cluster): Document
1783 2018-06-20 Martin Liska <mliska@suse.cz>
1785 * tree-switch-conversion.c (switch_conversion::collect):
1786 Record m_uniq property.
1787 (switch_conversion::expand): Bail out for special conditions.
1788 (group_cluster::~group_cluster): New.
1789 (group_cluster::group_cluster): Likewise.
1790 (group_cluster::dump): Likewise.
1791 (jump_table_cluster::emit): New.
1792 (switch_decision_tree::fix_phi_operands_for_edges): New.
1793 (struct case_node): Remove struct.
1794 (jump_table_cluster::can_be_handled): New.
1795 (case_values_threshold): Moved to header.
1796 (reset_out_edges_aux): Likewise.
1797 (jump_table_cluster::is_beneficial): New.
1798 (bit_test_cluster::can_be_handled): Likewise.
1799 (add_case_node): Remove.
1800 (bit_test_cluster::is_beneficial): New.
1801 (case_bit_test::cmp): New.
1802 (bit_test_cluster::emit): New.
1803 (expand_switch_as_decision_tree_p): Remove.
1804 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1805 (fix_phi_operands_for_edge): Likewise.
1806 (switch_decision_tree::analyze_switch_statement): New.
1807 (compute_cases_per_edge): Move ...
1808 (switch_decision_tree::compute_cases_per_edge): ... here.
1809 (try_switch_expansion): Likewise.
1810 (switch_decision_tree::try_switch_expansion): Likewise.
1811 (record_phi_operand_mapping): Likewise.
1812 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1813 (emit_case_decision_tree): Likewise.
1814 (switch_decision_tree::emit): Likewise.
1815 (balance_case_nodes): Likewise.
1816 (switch_decision_tree::balance_case_nodes): Likewise.
1817 (dump_case_nodes): Likewise.
1818 (switch_decision_tree::dump_case_nodes): Likewise.
1819 (emit_jump): Likewise.
1820 (switch_decision_tree::emit_jump): Likewise.
1821 (emit_cmp_and_jump_insns): Likewise.
1822 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1823 (emit_case_nodes): Likewise.
1824 (switch_decision_tree::emit_case_nodes): Likewise.
1825 (conditional_probability): Remove.
1826 * tree-switch-conversion.h (enum cluster_type): New.
1828 (struct cluster): Likewise.
1829 (cluster::cluster): Likewise.
1830 (struct simple_cluster): Likewise.
1831 (simple_cluster::simple_cluster): Likewise.
1832 (struct group_cluster): Likewise.
1833 (struct jump_table_cluster): Likewise.
1834 (struct bit_test_cluster): Likewise.
1835 (struct min_cluster_item): Likewise.
1836 (struct case_tree_node): Likewise.
1837 (case_tree_node::case_tree_node): Likewise.
1838 (jump_table_cluster::case_values_threshold): Likewise.
1839 (struct case_bit_test): Likewise.
1840 (struct switch_decision_tree): Likewise.
1841 (struct switch_conversion): Likewise.
1842 (switch_decision_tree::reset_out_edges_aux): Likewise.
1844 2018-06-20 Martin Liska <mliska@suse.cz>
1846 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1847 (hoist_edge_and_branch_if_true): Likewise.
1848 (expand_switch_using_bit_tests_p): Likewise.
1849 (struct case_bit_test): Likewise.
1850 (case_bit_test_cmp): Likewise.
1851 (emit_case_bit_tests): Likewise.
1852 (switch_conversion::switch_conversion): New class.
1853 (struct switch_conv_info): Remove old struct.
1854 (collect_switch_conv_info): More to ...
1855 (switch_conversion::collect): ... this.
1856 (check_range): Likewise.
1857 (switch_conversion::check_range): Likewise.
1858 (check_all_empty_except_final): Likewise.
1859 (switch_conversion::check_all_empty_except_final): Likewise.
1860 (check_final_bb): Likewise.
1861 (switch_conversion::check_final_bb): Likewise.
1862 (create_temp_arrays): Likewise.
1863 (switch_conversion::create_temp_arrays): Likewise.
1864 (free_temp_arrays): Likewise.
1865 (gather_default_values): Likewise.
1866 (switch_conversion::gather_default_values): Likewise.
1867 (build_constructors): Likewise.
1868 (switch_conversion::build_constructors): Likewise.
1869 (constructor_contains_same_values_p): Likewise.
1870 (switch_conversion::contains_same_values_p): Likewise.
1871 (array_value_type): Likewise.
1872 (switch_conversion::array_value_type): Likewise.
1873 (build_one_array): Likewise.
1874 (switch_conversion::build_one_array): Likewise.
1875 (build_arrays): Likewise.
1876 (switch_conversion::build_arrays): Likewise.
1877 (gen_def_assigns): Likewise.
1878 (switch_conversion::gen_def_assigns): Likewise.
1879 (prune_bbs): Likewise.
1880 (switch_conversion::prune_bbs): Likewise.
1881 (fix_phi_nodes): Likewise.
1882 (switch_conversion::fix_phi_nodes): Likewise.
1883 (gen_inbound_check): Likewise.
1884 (switch_conversion::gen_inbound_check): Likewise.
1885 (process_switch): Use the newly created class.
1886 (switch_conversion::expand): New.
1887 (switch_conversion::~switch_conversion): New.
1888 * tree-switch-conversion.h: New file.
1890 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1892 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1893 tree-vect-patterns.c.
1894 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1895 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1896 (vect_recog_sad_pattern): Likewise.
1897 (vect_recog_widen_sum_pattern): Likewise.
1898 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1899 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1900 (vect_recog_rotate_pattern): Likewise.
1901 (vect_recog_mult_pattern): Likewise.
1902 (vect_recog_vector_vector_shift_pattern): Likewise.
1903 (vect_recog_divmod_pattern): Likewise.
1904 (vect_recog_mixed_size_cond_pattern): Likewise.
1905 (vect_recog_bool_pattern): Likewise.
1906 (vect_recog_mask_conversion_pattern): Likewise.
1907 (vect_try_gather_scatter_pattern): Likewise.
1908 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1909 (vect_recog_over_widening_pattern): Likewise.
1910 (vect_recog_gather_scatter_pattern): Likewise.
1911 (vect_recog_func_ptr): Move from tree-vectorizer.h
1912 (vect_vect_recog_func_ptrs): Move further down the file.
1913 (vect_recog_func): Likewise. Remove the third argument.
1914 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1915 (vect_pattern_recog_1): Expect the pattern function to do any
1916 necessary target tests. Also expect it to provide a vector type.
1917 Remove the type_in handling.
1919 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1921 * tree-vect-patterns.c (vect_pattern_detected): New function.
1922 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1923 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1924 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1925 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1926 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1927 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1928 (vect_recog_mask_conversion_pattern)
1929 (vect_try_gather_scatter_pattern): Likewise.
1931 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1933 * tree-vect-patterns.c (vect_get_internal_def): New function.
1934 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1935 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1936 (search_type_for_mask_1): Use it.
1938 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1940 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1941 redundant WIDEN_SUM_EXPR handling.
1942 (vect_recog_sad_pattern): Likewise.
1944 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1946 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1947 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1948 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1949 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1950 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1952 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1954 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1955 use the stmt_vec_info of the original bb statement for the
1956 new zero assignment, even if the call is part of a pattern.
1958 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1960 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1961 that the sequence is attached to the original statement rather
1962 than the pattern statement.
1963 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1964 PATTERN_DEF_SEQ from the original statement rather than
1965 the main pattern statement.
1966 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1967 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1968 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1970 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1972 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1973 definition statements before the early exit for statements that aren't
1975 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1977 (vect_transform_loop): ...here. Process pattern definition
1978 statements without first checking whether the main pattern
1979 statement is live or relevant.
1981 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1983 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1984 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1986 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1988 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1989 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1990 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1991 (expand_block_compare): Change select_block_compare_mode call.
1992 (expand_strncmp_align_check): Use new functions, fix comment.
1993 (emit_final_str_compare_gpr): New function.
1994 (expand_strn_compare): Refactor and clean up code.
1995 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1997 2018-06-19 Tony Reix <tony.reix@atos.com>
1998 Damien Bergamini <damien.bergamini@atos.com>
1999 David Edelsohn <dje.gcc@gmail.com>
2001 * collect2.c (static_obj): New variable.
2002 (static_libs): New variable.
2003 (is_in_list): Uncomment declaration.
2004 (main): Track AIX libraries linked statically.
2005 (is_in_list): Uncomment definition.
2006 (scan_prog_file): Don't add AIX shared libraries initializer
2007 to constructor list if linking statically.
2009 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2011 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2013 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2015 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2017 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2020 2018-06-19 Martin Liska <mliska@suse.cz>
2022 * config/i386/i386.c (ix86_can_inline_p): Do not use
2023 ipa_fn_summaries::get_create.
2024 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2026 (devirtualization_time_bonus): Likewise.
2027 (ipcp_propagate_stage): Likewise.
2028 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2029 (edge_set_predicate): Likewise.
2030 (evaluate_conditions_for_known_args): Likewise.
2031 (evaluate_properties_for_edge): Likewise.
2032 (ipa_call_summary::reset): Tranform to ...
2033 (ipa_call_summary::~ipa_call_summary): ... this.
2034 (ipa_fn_summary::reset): Transform to ...
2035 (ipa_fn_summary::~ipa_fn_summary): ... this.
2036 (ipa_fn_summary_t::remove): Rename to ...
2037 (ipa_fn_summary_t::remove_callees): ... this.
2038 (ipa_fn_summary_t::duplicate): Use placement new
2039 instead of memory copy.
2040 (ipa_call_summary_t::duplicate): Likewise.
2041 (ipa_call_summary_t::remove): Remove.
2042 (dump_ipa_call_summary): Change get_create to get.
2043 (ipa_dump_fn_summary): Dump only when summary exists.
2044 (analyze_function_body): Use symbol_summary::get instead
2046 (compute_fn_summary): Likewise.
2047 (estimate_edge_devirt_benefit): Likewise.
2048 (estimate_edge_size_and_time): Likewise.
2049 (inline_update_callee_summaries): Likewise.
2050 (remap_edge_change_prob): Likewise.
2051 (remap_edge_summaries): Likewise.
2052 (ipa_merge_fn_summary_after_inlining): Likewise.
2053 (write_ipa_call_summary): Likewise.
2054 (ipa_fn_summary_write): Likewise.
2055 (ipa_free_fn_summary): Likewise.
2056 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2057 (struct ipa_call_summary): Likewise.
2058 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2060 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2061 (estimate_size_after_inlining): Likewise.
2062 (estimate_growth): Likewise.
2063 (growth_likely_positive): Likewise.
2064 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2065 (inline_call): Likewise.
2066 * ipa-inline.c (caller_growth_limits): Likewise.
2067 (can_inline_edge_p): Likewise.
2068 (can_inline_edge_by_limits_p): Likewise.
2069 (compute_uninlined_call_time): Likewise.
2070 (compute_inlined_call_time): Likewise.
2071 (want_inline_small_function_p): Likewise.
2072 (edge_badness): Likewise.
2073 (update_caller_keys): Likewise.
2074 (update_callee_keys): Likewise.
2075 (inline_small_functions): Likewise.
2076 (inline_to_all_callers_1): Likewise.
2077 (dump_overall_stats): Likewise.
2078 (early_inline_small_functions): Likewise.
2079 (early_inliner): Likewise.
2080 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2081 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2082 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2083 * ipa-split.c (execute_split_functions): Likewise.
2084 * symbol-summary.h: Likewise.
2085 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2087 2018-06-19 Richard Biener <rguenther@suse.de>
2089 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2090 (vectorize_loops): ... here. Fix dbgcnt handling.
2091 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2093 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2096 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2097 ieee128 argument takes up only one (vector) register, not two (floating
2100 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2102 * gimplify.c (gimplify_init_constructor): Really never clear for an
2103 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2105 2018-06-19 Richard Biener <rguenther@suse.de>
2107 PR tree-optimization/86179
2108 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2109 after failed recognition.
2111 2018-06-18 Martin Sebor <msebor@redhat.com>
2114 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2115 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2116 Handle integer subtraction.
2117 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2118 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2120 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2122 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2123 param from rtx to rtx_insn *.
2124 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2126 (frv_ifcvt_modify_insn): Likwise.
2127 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2128 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2129 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2130 as_a <rtx_insn *> cast to local "unprotected_region" once
2131 it's been established that it's not NULL or pc_rtx.
2132 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2133 param "sethi" from rtx to rtx_insn *.
2134 (nds32_group_float_insns): Likewise for param "insn".
2135 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2137 (vax_output_int_subtract): Likewise.
2138 * config/vax/vax.c (vax_output_int_add): Likewise for param
2140 (vax_output_int_subtract): Likewise.
2141 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2142 (emit_pattern_after): Likewise for param "after".
2143 (emit_insn_after): Likewise.
2144 (emit_jump_insn_after): Likewise.
2145 (emit_call_insn_after): Likewise.
2146 (emit_debug_insn_after): Likewise.
2147 (emit_pattern_before): Likewise for param "before".
2148 (emit_insn_before): Likewise.
2149 (emit_jump_insn_before): Likewise.
2150 * final.c (get_insn_template): Likewise for param "insn", removing
2152 * output.h (get_insn_template): Likewise for 2nd param.
2153 * rtl.h (emit_insn_before): Likewise.
2154 (emit_jump_insn_before): Likewise.
2155 (emit_debug_insn_before_noloc): Likewise.
2156 (emit_insn_after): Likewise.
2157 (emit_jump_insn_after): Likewise.
2158 (emit_call_insn_after): Likewise.
2159 (emit_debug_insn_after): Likewise.
2160 (set_insn_deleted): Likewise for param.
2162 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2165 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2166 floating point modes, so that IFmode is numerically greater than
2167 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2168 to declare the ordering. This prevents IFmode from being
2169 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2170 machine. Include rs6000-modes.h to share the fractional values
2171 between genmodes* and the rest of the compiler.
2175 * config/rs6000/rs6000-modes.h: New file.
2176 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2177 meaning of rs6000_long_double_size so that 126..128 selects an
2178 appropriate 128-bit floating point type.
2179 (rs6000_option_override_internal): Likewise.
2180 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2181 (TARGET_LONG_DOUBLE_128): Change the meaning of
2182 rs6000_long_double_size so that 126..128 selects an appropriate
2183 128-bit floating point type.
2184 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2185 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2186 source and destination to match the standard usage.
2187 (truncifkf2): Likewise.
2188 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2189 ISA 2.07 to use an explicit clobber, instead of passing in a
2191 (copysign<mode>3_soft): Likewise.
2193 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2195 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2196 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2197 (vect_slp_analyze_instance_dependence): Likewise.
2198 (vect_enhance_data_refs_alignment): Likewise.
2199 (vect_analyze_data_refs_alignment): Likewise.
2200 (vect_slp_analyze_and_verify_instance_alignment
2201 (vect_analyze_data_ref_accesses): Likewise.
2202 (vect_prune_runtime_alias_test_list): Likewise.
2203 (vect_analyze_data_refs): Likewise.
2204 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2205 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2206 (vect_analyze_scalar_cycles_1): Likewise.
2207 (vect_get_loop_niters): Likewise.
2208 (vect_analyze_loop_form_1): Likewise.
2209 (vect_update_vf_for_slp): Likewise.
2210 (vect_analyze_loop_operations): Likewise.
2211 (vect_analyze_loop): Likewise.
2212 (vectorizable_induction): Likewise.
2213 (vect_transform_loop): Likewise.
2214 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2215 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2216 (vect_make_slp_decision): Likewise.
2217 (vect_detect_hybrid_slp): Likewise.
2218 (vect_slp_analyze_operations): Likewise.
2219 (vect_slp_bb): Likewise.
2220 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2221 (vectorizable_bswap): Likewise.
2222 (vectorizable_call): Likewise.
2223 (vectorizable_simd_clone_call): Likewise.
2224 (vectorizable_conversion): Likewise.
2225 (vectorizable_assignment): Likewise.
2226 (vectorizable_shift): Likewise.
2227 (vectorizable_operation): Likewise.
2228 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2230 2018-06-18 Martin Sebor <msebor@redhat.com>
2232 PR tree-optimization/81384
2233 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2234 * builtins.c (expand_builtin_strnlen): New function.
2235 (expand_builtin): Call it.
2236 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2237 * builtins.def (BUILT_IN_STRNLEN): New.
2238 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2239 Warn for bounds in excess of maximum object size.
2240 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2241 single-value ranges. Handle strnlen.
2242 (handle_builtin_strlen): Handle strnlen.
2243 (strlen_check_and_optimize_stmt): Same.
2244 * doc/extend.texi (Other Builtins): Document strnlen.
2246 2018-06-18 Maya Rashish <coypu@sdf.org>
2248 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2249 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2250 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2252 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2254 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2256 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2258 * tree.c (escaped_string::escape): Replace cast to char * by
2259 const_cast<char *> (unescaped).
2261 2018-06-18 Nick Clifton <nickc@redhat.com>
2264 * tree.c (escaped_string): New class. Converts an unescaped
2265 string into its escaped equivalent.
2266 (warn_deprecated_use): Use the new class to convert the
2267 deprecation message, if present.
2268 (test_escaped_strings): New self test.
2269 (test_c_tests): Add test_escaped_strings.
2270 * doc/extend.texi (deprecated): Add a note that the
2271 deprecation message is affected by the -fmessage-length
2272 option, and that control characters will be escaped.
2273 (#pragma GCC error): Document this pragma.
2274 (#pragma GCC warning): Likewise.
2275 * doc/invoke.texi (-fmessage-length): Document this option's
2276 effect on the #warning and #error preprocessor directives and
2277 the deprecated attribute.
2279 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2281 * tree.c (decl_value_expr_lookup): Revert latest change.
2282 (decl_value_expr_insert): Likewise.
2284 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2286 * gimplify.c (nonlocal_vlas): Delete.
2287 (nonlocal_vla_vars): Likewise.
2288 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2290 (gimplify_body): Do not create and destroy nonlocal_vlas.
2291 * tree-nested.c: Include diagnostic.h.
2292 (use_pointer_in_frame): Tweak.
2293 (lookup_field_for_decl): Add assertion and declare the transformation.
2294 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2295 internal error when the reference is in a wrong context. Do not
2296 create a debug decl by default.
2297 (note_nonlocal_block_vlas): Delete.
2298 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2299 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2300 create a debug decl by default.
2301 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2302 call is in a wrong context.
2303 (fixup_vla_decls): New function.
2304 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2305 debug variables were created.
2306 * tree.c (decl_value_expr_lookup): Add checking assertion.
2307 (decl_value_expr_insert): Likewise.
2309 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2312 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2313 * tree-scalar-evolution.c (interpret_expr): Likewise.
2314 (expression_expensive_p): Likewise.
2315 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2316 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2317 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2318 (ssa_defined_by_minus_one_stmt_p): New.
2320 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2323 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2324 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2325 * dojump.c (do_jump): Likewise.
2326 * expr.c (expand_expr_real_2): Check operand type's sign.
2327 * fold-const.c (const_unop): Handle ABSU_EXPR.
2328 (fold_abs_const): Likewise.
2329 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2330 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2331 (strip_sign_op_1): Likesise.
2332 * match.pd: Add new pattern to generate ABSU_EXPR.
2333 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2334 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2335 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2336 * tree-inline.c (estimate_operator_cost): Likewise.
2337 * tree-pretty-print.c (dump_generic_node): Likewise.
2338 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2339 * tree.def (ABSU_EXPR): New.
2341 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2344 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2345 documented as preserved for backward compatibility only.
2346 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2348 PR rtl-optimization/86108
2349 * bb-reorder.c (create_forwarder_block): Renamed to ...
2350 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2351 jump from new landing pad to the second part.
2352 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2355 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2358 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2359 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2360 Only call store_expr for halves if the mode is the same.
2363 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2364 Fix up comment formatting.
2366 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2368 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2369 (typed_splay_tree::closure,
2370 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2371 (typed_splay_tree::typed_splay_tree,
2372 typed_splay_tree::operator =): Declared private.
2373 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2374 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2375 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2376 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2377 typed_splay_tree::splay_tree_splay,
2378 typed_splay_tree::splay_tree_foreach_helper,
2379 typed_splay_tree::splay_tree_insert,
2380 typed_splay_tree::splay_tree_remove,
2381 typed_splay_tree::splay_tree_lookup,
2382 typed_splay_tree::splay_tree_predecessor,
2383 typed_splay_tree::splay_tree_successor,
2384 typed_splay_tree::splay_tree_min,
2385 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2386 (typed_splay_tree::root, typed_splay_tree::comp,
2387 typed_splay_tree::delete_key,
2388 typed_splay_tree::delete_value): New data members.
2389 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2390 typed_splay_tree::remove.
2392 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2394 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2395 -mginv and -mno-ginv to the assembler.
2396 * config/mips/mips.opt (-mcrc): New option.
2398 * doc/invoke.text (-mcrc): Document.
2401 2018-06-15 Nick Clifton <nickc@redhat.com>
2404 * tree.c (escaped_string): New class. Converts an unescaped
2405 string into its escaped equivalent.
2406 (warn_deprecated_use): Use the new class to convert the
2407 deprecation message, if present.
2408 (test_escaped_strings): New self test.
2409 (test_c_tests): Add test_escaped_strings.
2410 * doc/extend.texi (deprecated): Add a note that the
2411 deprecation message is affected by the -fmessage-length
2412 option, and that control characters will be escaped.
2413 (#pragma GCC error): Document this pragma.
2414 (#pragma GCC warning): Likewise.
2415 * doc/invoke.texi (-fmessage-length): Document this option's
2416 effect on the #warning and #error preprocessor directives and
2417 the deprecated attribute.
2419 2018-06-15 Richard Biener <rguenther@suse.de>
2421 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2422 here, also noting vector size used.
2423 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2424 size used in MSG_OPTIMIZED_LOCATIONS dump.
2425 (pass_slp_vectorize::execute): Adjust.
2427 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2430 * config/arc/arc.c (arc_return_address_register): Fix
2433 2018-06-15 Richard Biener <rguenther@suse.de>
2436 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2437 leave useless conversion stripping to force_gimple_operand_gsi.
2438 (gimplify_build2): Likewise.
2439 (gimplify_build1): Likewise.
2441 2018-06-15 Richard Biener <rguenther@suse.de>
2444 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2445 before adjusting their block.
2447 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2449 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2450 multilibs for *-*-rtems*.
2451 * config/riscv/t-rtems: New file.
2453 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2456 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2457 unsigned_type_for returns NULL.
2460 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2461 subregs of multi-word pseudos unless the float mode has word size.
2463 2018-06-14 Richard Biener <rguenther@suse.de>
2466 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2467 duplicate and harmful type_hash_canon.
2468 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2470 2018-06-14 Richard Biener <rguenther@suse.de>
2473 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2476 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2478 * config/rtems.h (STDINT_LONG32): Define.
2480 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2481 Prachi Godbole <prachi.godbole@imgtec.com>
2483 * config/mips/mips-cpus.def: Define P6600.
2484 * config/mips/mips-tables.opt: Regenerate.
2485 * config/mips/mips.c (mips_ucbranch_type): New enum.
2486 (mips_rtx_cost_data): Add support for P6600.
2487 (mips_issue_rate): Likewise.
2488 (mips_multipass_dfa_lookahead): Likewise.
2489 (mips_avoid_hazard): Likewise.
2490 (mips_reorg_process_insns): Likewise.
2491 (mips_classify_branch_p6600): New function.
2492 * config/mips/mips.h (TUNE_P6600): New define.
2493 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2494 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2495 * config/mips/mips.md: Include p6600.md.
2496 (processor): Add p6600.
2497 * config/mips/p6600.md: New file.
2498 * doc/invoke.texi: Add p6600 to supported architectures.
2500 2018-06-13 Martin Sebor <msebor@redhat.com>
2502 PR tree-optimization/86114
2503 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2505 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2507 2018-06-13 Richard Biener <rguenther@suse.de>
2509 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2510 Properly set vector type of the intermediate stmt.
2511 * tree-vect-stmts.c (vectorizable_operation): The destination
2512 var always has vectype_out type.
2514 2018-06-13 Jeff Law <law@redhat.com>
2516 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2517 integer 0 for argument to print_rtl_with_bb.
2518 (rl78_reorg): Likewise.
2520 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2522 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2523 from rtx to rtx_insn *.
2524 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2526 (add_sched_insns_for_speculation): Likewise for local "target",
2527 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2528 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2529 from rtx_insn ** to rtx_code_label **.
2530 (reorg_emit_nops): Likewise.
2531 (c6x_reorg): Likewise for local "call_labels".
2532 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2534 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2535 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2536 the loops over LABEL_REFS.
2537 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2539 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2540 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2541 (split_branches): Strengthen local "olabel" from rtx to
2542 rtx_insn *, adding a safe_as_a cast.
2543 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2545 (add_insn_after): Likewise for first two params.
2546 (add_insn_before): Likewise.
2547 (remove_insn): Likewise for param.
2548 (emit_pattern_before_noloc): Likewise for second and third params.
2549 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2550 (emit_call_insn_before_noloc): Likewise.
2551 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2553 (emit_barrier_before): Likewise.
2554 (emit_label_before): Strengthen "label" param from "rtx" to
2555 "rtx_code_label *". Strengthen "before" param from "rtx" to
2557 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2559 (emit_pattern_after_noloc): Likewise.
2560 (emit_insn_after_noloc): Likewise.
2561 (emit_jump_insn_after_noloc): Likewise.
2562 (emit_call_insn_after_noloc): Likewise.
2563 (emit_debug_insn_after_noloc): Likewise.
2564 (emit_barrier_after): Likewise.
2565 (emit_label_after): Likewise for both params.
2566 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2567 "loc" param from "int" to "location_t".
2568 (emit_insn_after_setloc): Likewise.
2569 (emit_jump_insn_after_setloc): Likewise.
2570 (emit_call_insn_after_setloc): Likewise.
2571 (emit_debug_insn_after_setloc): Likewise.
2572 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2573 "loc" param from "int" to "location_t".
2574 (emit_pattern_before): Convert NULL_RTX to NULL.
2575 (emit_insn_before_setloc): Convert "loc" param from "int" to
2577 (emit_jump_insn_before_setloc): Likewise.
2578 (emit_call_insn_before_setloc): Likewise.
2579 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2580 rtx_insn *. Convert "loc" param from "int" to "location_t".
2581 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2582 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2583 Convert 3rd param from "int" to "location_t".
2584 (emit_barrier_before, emit_barrier_after, next_real_insn):
2585 Strengthen param from rtx to rtx_insn *.
2586 (emit_label_before): Strengthen 1st param from "rtx" to
2587 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2589 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2590 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2591 Strengthen 2nd param from "rtx" to "rtx_insn *".
2592 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2593 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2594 Likewise. Convert 3rd param from "int" to "location_t".
2595 (emit_label_after): Strengthen 1st param from "rtx" to
2597 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2599 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2600 from "rtx" to "rtx_insn *".
2602 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2604 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2605 bodies streamed in with -Q.
2606 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2607 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2608 * gimple-streamer-out.c: Include gimple-pretty-print.h
2609 (output_bb): Dump stmts streamed.
2610 * lto-section-out.c: Include print-tree.h
2611 (lto_begin_section): Dump sections created.
2612 (lto_output_decl_index): Dump decl encoded.
2613 * lto-streamer-out.c: Include print-tree.h
2614 (create_output_block): Dump output block created.
2615 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2616 (output_function): Dump function output.
2617 (output_constructor): Dump constructor streamed.
2618 (write_global_stream): Output indexes encoded.
2619 (produce_asm_for_decls): Dump streams encoded.
2620 * lto-streamer.c (streamer_dump_file): New global var.
2621 * lto-streamer.h (streamer_dump_file): Declare.
2622 * passes.c (ipa_write_summaries): Initialize streamer dump.
2623 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2626 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2629 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2630 offsets for register save directives. Emit a second batch of save
2631 directives, if need be, when the function accesses prior frames.
2633 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2635 * config/arc/fpu.md (fmasf4): Force operand to register.
2636 (fnmasf4): Likewise.
2638 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2640 * config/arc/arc-protos.h (arc_pad_return): Remove.
2641 * config/arc/arc.c (machine_function): Remove force_short_suffix
2643 (arc_print_operand): Adjust printing of '&'.
2644 (arc_verify_short): Remove conditional printing of short suffix.
2645 (arc_final_prescan_insn): Remove reference to size_reason.
2646 (pad_return): New function.
2647 (arc_reorg): Call pad_return.
2648 (arc_pad_return): Remove.
2649 (arc_init_machine_status): Remove reference to force_short_suffix.
2650 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2651 (attr length): When attribute iscompact is true force to 2
2652 regardless; in the case of maybe check if we want to force the
2653 instruction to have 4 bytes length.
2654 (nopv): Change it to generate 4 byte long nop as well.
2655 (blockage): New pattern.
2656 (simple_return): Remove call to arc_pad_return.
2657 (p_return_i): Likewise.
2659 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2661 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2663 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2665 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2668 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2670 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2671 for ARC700 and ARCv2.
2673 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2676 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2677 operands[2] instead of operands[1].
2680 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2682 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2683 case, check whether the outer register overlaps an unallocatable
2684 register, not just whether it fits the required class.
2686 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2688 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2690 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2691 (memrefs_conflict_p): Likewise.
2692 (init_alias_analysis): Likewise.
2693 * cfgexpand.c (expand_debug_expr): Likewise.
2694 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2695 * cse.c (fold_rtx): Likewise.
2696 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2697 * expr.c (emit_block_move_hints): Likewise.
2698 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2699 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2700 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2701 (emit_group_store): Likewise.
2702 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2703 to read the PRE/POST_MODIFY increment.
2704 * calls.c (store_one_arg): Use strip_offset.
2705 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2707 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2709 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2710 (simplify_binary_operation_1): Extend CONST_INT handling to
2712 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2713 than a HOST_WIDE_INT.
2714 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2716 (adjust_mems, add_stores): Update accodingly.
2717 (vt_canonicalize_addr): Track polynomial offsets.
2718 (emit_note_insn_var_location): Likewise.
2719 (vt_add_function_parameter): Likewise.
2720 (vt_initialize): Likewise.
2722 2018-06-12 Jeff Law <law@redhat.com>
2724 * config.gcc (alpha*-*-freebsd*): Remove.
2725 * config/alpha/freebsd.h: Remove.
2727 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2730 * spellcheck-tree.c (levenshtein_distance): Rename to...
2731 (get_edit_distance): ...this, and update for underlying renaming.
2732 * spellcheck-tree.h (levenshtein_distance): Rename to...
2733 (get_edit_distance): ...this.
2734 * spellcheck.c (levenshtein_distance): Rename to...
2735 (get_edit_distance): ...this. Convert from Levenshtein distance
2736 to Damerau-Levenshtein distance by supporting transpositions of
2737 adjacent characters. Rename "v1" to "v_next" and "v0" to
2739 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2740 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2741 update for underlying renaming.
2742 (selftest::levenshtein_distance_unit_test): Rename to...
2743 (selftest::test_get_edit_distance_unit): ...this, and update for
2744 underlying renaming.
2745 (selftest::test_find_closest_string): Add example from PR 69968
2746 where transposition helps
2747 (selftest::test_metric_conditions): Update for renaming.
2748 (selftest::test_metric_conditions): Likewise.
2749 (selftest::spellcheck_c_tests): Likewise.
2750 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2751 (get_edit_distance): ...this.
2752 (best_match::consider): Update for renaming.
2754 2018-06-12 Martin Sebor <msebor@redhat.com>
2756 PR tree-optimization/85259
2757 * builtins.c (compute_objsize): Handle constant offsets.
2758 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2759 true iff a warning has been issued.
2760 * gimple.h (gimple_nonartificial_location): New function.
2761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2762 gimple_nonartificial_location and handle -Wno-system-headers.
2763 (handle_builtin_stxncpy): Same.
2765 2018-06-12 Martin Sebor <msebor@redhat.com>
2768 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2770 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2772 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2773 BUILTIN_VEC_XST entries for pointer to double and long long.
2775 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2778 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2780 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2783 2018-06-12 Martin Liska <mliska@suse.cz>
2785 * doc/options.texi: Document IntegerRange.
2787 2018-06-12 Martin Liska <mliska@suse.cz>
2789 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2790 * opt-functions.awk: Handle Deprecated flag.
2791 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2793 (read_cmdline_option): Report warning for a deprecated option.
2794 * opts.h (struct cl_option): Add new field cl_deprecated.
2795 (CL_ERR_DEPRECATED): New.
2797 2018-06-12 Martin Liska <mliska@suse.cz>
2799 * doc/options.texi: Document Deprecated option flag.
2801 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2803 * config/arc/arc-arch.h (arc_extras): New enum.
2804 (arc_cpu_t):Add field extra.
2805 (arc_cpu_types): Consider the extras.
2806 * config/arc/arc-cpus.def: Add extras info.
2807 * config/arc/arc-opts.h (processor_type): Consider extra field.
2808 * config/arc/arc.c (arc_override_options): Handle extra field.
2810 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2812 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2813 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2815 * config/arc/arc.c (arc_init): Update pic warning.
2816 (irq_range): Update irq range parsing warnings.
2817 (arc_override_options): Update various warning messages.
2818 (arc_handle_aux_attribute): Likewise.
2820 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2822 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2824 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2826 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2827 compiler messages directives.
2829 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2831 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2832 * config/mips/mips-tables.opt: Regenerate.
2833 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2835 * doc/invoke.texi: Document -march=i6500.
2837 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2839 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2840 (i6400_gpmul): Add cpu_unit.
2841 (i6400_gpdiv): Likewise.
2842 (i6400_msa_add_d): Update reservations.
2843 (i6400_msa_int_add) Likewise.
2844 (i6400_msa_short_logic3) Likewise.
2845 (i6400_msa_short_logic2) Likewise.
2846 (i6400_msa_short_logic) Likewise.
2847 (i6400_msa_move) Likewise.
2848 (i6400_msa_cmp) Likewise.
2849 (i6400_msa_short_float2) Likewise.
2850 (i6400_msa_div_d) Likewise.
2851 (i6400_msa_long_logic1) Likewise.
2852 (i6400_msa_long_logic2) Likewise.
2853 (i6400_msa_mult) Likewise.
2854 (i6400_msa_long_float2) Likewise.
2855 (i6400_msa_long_float4) Likewise.
2856 (i6400_msa_long_float5) Likewise.
2857 (i6400_msa_long_float8) Likewise.
2858 (i6400_fpu_fadd): Include frint type.
2859 (i6400_fpu_store): New define_insn_reservation.
2860 (i6400_fpu_load): Likewise.
2861 (i6400_fpu_move): Likewise.
2862 (i6400_fpu_fcmp): Likewise.
2863 (i6400_fpu_fmadd): Likewise.
2864 (i6400_int_mult): Include imul3nc type and update reservation.
2865 (i6400_int_div): Include idiv3 type and update reservation.
2866 (i6400_int_load): Update to check type not move_type.
2867 (i6400_int_store): Likewise.
2868 (i6400_int_prefetch): Set zero latency.
2870 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2872 * gcc.c: Document new %@{...} sequence.
2873 (LINK_COMMAND_SPEC): Use it for the -L switches.
2874 (cpp_unique_options): Use it for the -I switches.
2875 (at_file_argbuf): New global variable.
2876 (in_at_file): Likewise.
2877 (alloc_args): Create at_file_argbuf.
2878 (clear_args): Truncate at_file_argbuf.
2879 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2880 (open_at_file): New function.
2881 (close_at_file): Likewise.
2882 (create_at_file): Delete.
2883 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2886 (validate_switches_from_spec): Deal with %@{...} sequence.
2887 (validate_switches): Likewise.
2888 (driver::finalize): Call clear_args.
2890 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2892 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2894 2018-06-11 Martin Sebor <msebor@redhat.com>
2896 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2897 (Wno-class-memaccess): Add @opindex.
2898 (Wno-templates, Wno-multiple-inheritance): Same.
2899 (Wno-virtual-inheritance, Wno-namespaces): Same.
2900 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2901 (Wno-format-overflow, Wno-format-truncation): Same.
2902 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2903 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2904 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2905 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2906 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2907 (Wno-misspelled-isr): Same.
2909 2018-06-11 Martin Sebor <msebor@redhat.com>
2911 * PR tree-optimization/86083
2912 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2914 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2916 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2918 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2921 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2922 on the correct operand.
2923 (*movdi_internal64): Ditto.
2925 2018-06-11 Martin Liska <mliska@suse.cz>
2927 PR tree-optimization/86089
2928 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2930 2018-06-11 Julia Koval <julia.koval@intel.com>
2932 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2933 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2934 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2936 2018-06-11 Olivier Hainque <hainque@adacore.com>
2938 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2939 for Ada with strict dwarf2.
2941 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2944 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2947 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2949 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2951 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2953 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2956 2018-06-08 Martin Liska <mliska@suse.cz>
2958 * tree-cfg.h (debug_function): Fix argument type to match
2961 2018-06-08 Martin Liska <mliska@suse.cz>
2963 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2964 Remove usage of MPX-related (and removed) fields.
2965 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2967 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2969 * cfg.c (debug): Use TDF_NONE rather than 0.
2970 * cfghooks.c (debug): Likewise.
2971 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2972 (struct dump_option_value_info): Convert to...
2973 (struct kv_pair): ...this template type.
2974 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2976 (optinfo_verbosity_options): Likewise.
2977 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2979 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2980 than int for "optgroup_flags" param.
2981 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2983 (dump_dec): Likewise.
2984 (dump_finish): Use TDF_NONE rather than 0.
2985 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2986 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2987 than 0. Update for change to option_ptr.
2988 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2989 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2990 0. Update for changes to optinfo_verbosity_options and
2992 (opt_info_switch_p): Convert optgroup_flags from int to
2994 (dump_basic_block): Use dump_flags_t rather than int
2995 for "dump_kind" param.
2996 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2997 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2998 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2999 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3000 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3001 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3002 TDF_NONE): Convert from macros to...
3003 (enum dump_flag): ...this new enum.
3004 (dump_flags_t): Update to use enum.
3005 (operator|, operator&, operator~, operator|=, operator&=):
3006 Implement for dump_flags_t.
3007 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3008 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3009 Convert from macros to...
3010 (enum optgroup_flag): ...this new enum.
3011 (optgroup_flags_t): New typedef.
3012 (operator|, operator|=): Implement for optgroup_flags_t.
3013 (struct dump_file_info): Convert field "alt_flags" to
3014 dump_flags_t. Convert field "optgroup_flags" to
3016 (dump_basic_block): Use dump_flags_t rather than int for param.
3017 (dump_generic_expr_loc): Likewise.
3018 (dump_dec): Likewise.
3019 (dump_register): Convert param "optgroup_flags" to
3021 (opt_info_enable_passes): Likewise.
3022 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3024 * gimple-pretty-print.c (debug): Likewise.
3025 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3026 (merged_store_group::apply_stores): Likewise.
3027 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3028 * gimple.c (verify_gimple_pp): Likewise.
3029 * graphite-poly.c (print_pbb_body): Likewise.
3030 * passes.c (pass_manager::register_one_dump_file): Convert
3031 local "optgroup_flags" to optgroup_flags_t.
3032 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3034 (debug_body): Likewise.
3035 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3036 to optgroup_flags_t.
3037 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3039 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3040 (convert_mult_to_fma): Likewise.
3041 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3042 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3043 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3044 "dump_kind" to dump_flags_t.
3046 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3048 * config/rs6000/rs6000.c (min, max): Delete.
3050 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3052 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3053 -mabi=spe and -mabi=no-spe.
3055 2018-06-08 Martin Liska <mliska@suse.cz>
3057 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3058 where we expect an existing summary.
3060 2018-06-08 Martin Liska <mliska@suse.cz>
3062 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3063 * ipa-inline.h (estimate_edge_growth): Likewise.
3065 2018-06-08 Martin Liska <mliska@suse.cz>
3067 * cgraph.c (function_version_hasher::hash): Use
3068 cgraph_node::get_uid ().
3069 (function_version_hasher::equal):
3070 * cgraph.h (cgraph_node::get_uid): New method.
3071 * ipa-inline.c (update_caller_keys): Use
3072 cgraph_node::get_uid ().
3073 (update_callee_keys): Likewise.
3074 * ipa-utils.c (searchc): Likewise.
3075 (ipa_reduced_postorder): Likewise.
3076 * lto-cgraph.c (input_node): Likewise.
3077 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3078 * symbol-summary.h (symtab_insertion): Likewise.
3079 (symtab_removal): Likewise.
3080 (symtab_duplication): Likewise.
3081 * tree-pretty-print.c (dump_function_header): Likewise.
3082 * tree-sra.c (convert_callers_for_node): Likewise.
3084 2018-06-08 Martin Liska <mliska@suse.cz>
3086 * cgraph.c (symbol_table::create_edge): Always assign a new
3088 (symbol_table::free_edge): Do not recycle numbers.
3089 * cgraph.h (cgraph_edge::get): New method.
3090 * symbol-summary.h (symtab_removal): Use it.
3091 (symtab_duplication): Likewise.
3092 (call_summary::hashable_uid): Remove.
3094 2018-06-08 Martin Liska <mliska@suse.cz>
3096 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3097 (initialize_growth_caches): Remove.
3098 (free_growth_caches): Likewise.
3099 (do_estimate_edge_time): Use edge_growth_cache.
3100 (do_estimate_edge_size): Likewise.
3101 (do_estimate_edge_hints): Likewise.
3102 * ipa-inline.c (reset_edge_caches): Likewise.
3103 (recursive_inlining): Likewise.
3104 (inline_small_functions): Likewise.
3105 * ipa-inline.h (initialize_growth_caches): Remove.
3106 (estimate_edge_size): Likewise.
3107 (estimate_edge_time): Likewise.
3108 (estimate_edge_hints): Likewise.
3109 (reset_edge_growth_cache): Likewise.
3110 * symbol-summary.h (call_summary::remove): New method.
3112 2018-06-08 Martin Liska <mliska@suse.cz>
3114 * ipa-cp.c (class edge_clone_summary): New summary.
3115 (grow_edge_clone_vectors): Remove.
3116 (ipcp_edge_duplication_hook): Remove.
3117 (class edge_clone_summary_t): New call_summary class.
3118 (ipcp_edge_removal_hook): Remove.
3119 (edge_clone_summary_t::duplicate): New function.
3120 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3121 (create_specialized_node): Likewise.
3122 (ipcp_driver): Initialize edge_clone_summaries and do not
3125 2018-06-08 Martin Liska <mliska@suse.cz>
3127 * symbol-summary.h (get): New function.
3128 (call_summary::m_initialize_when_cloning): New class member.
3130 2018-06-08 Martin Liska <mliska@suse.cz>
3132 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3133 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3134 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3135 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3136 (remove_cgraph_node_from_order): Use the removed_nodes set.
3137 (do_per_function_toporder): Likwise.
3138 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3139 instead of summary_uid.
3140 (symtab_removal): Likewise.
3141 (symtab_duplication): Likewise.
3143 2018-06-08 Martin Liska <mliska@suse.cz>
3145 * ipa-cp.c (ipcp_store_bits_results): Use
3146 ipcp_transformation_sum.
3147 (ipcp_store_vr_results): Likewise.
3148 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3150 (ipcp_transformation_initialize): ... this.
3151 (ipa_set_node_agg_value_chain):
3152 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3153 (write_ipcp_transformation_info): Likewise.
3154 (read_ipcp_transformation_info): Likewise.
3155 (ipcp_update_bits): Likewise.
3156 (ipcp_update_vr): Likewise.
3157 (ipcp_transform_function): Likewise.
3158 * ipa-prop.h: Rename ipcp_transformation_summary to
3159 ipcp_transformation.
3160 (class ipcp_transformation_t): New function summary.
3161 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3162 (ipa_get_agg_replacements_for_node): Likewise.
3164 2018-06-08 Martin Liska <mliska@suse.cz>
3166 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3168 (class funct_state_summary_t): New function_summary class.
3169 (has_function_state): Remove.
3170 (get_function_state): Likewise.
3171 (set_function_state): Likewise.
3172 (add_new_function): Likewise.
3173 (funct_state_summary_t::insert): New function.
3174 (duplicate_node_data): Remove.
3175 (remove_node_data): Remove.
3176 (funct_state_summary_t::duplicate): New function.
3177 (register_hooks): Create new funct_state_summaries.
3178 (pure_const_generate_summary): Use it.
3179 (pure_const_write_summary): Likewise.
3180 (pure_const_read_summary): Likewise.
3181 (propagate_pure_const): Likewise.
3182 (propagate_nothrow): Likewise.
3183 (dump_malloc_lattice): Likewise.
3184 (propagate_malloc): Likewise.
3185 (execute): Do not register hooks, just remove summary
3187 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3190 2018-06-08 Martin Liska <mliska@suse.cz>
3192 * ipa-reference.c (remove_node_data): Remove.
3193 (duplicate_node_data): Likewise.
3194 (class ipa_ref_var_info_summary_t): New class.
3195 (class ipa_ref_opt_summary_t): Likewise.
3196 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3197 (get_reference_optimization_summary): Use
3198 ipa_ref_opt_sum_summaries.
3199 (set_reference_vars_info): Remove.
3200 (set_reference_optimization_summary): Likewise.
3201 (ipa_init): Create summaries.
3202 (init_function_info): Use function summary.
3203 (ipa_ref_opt_summary_t::duplicate): New function.
3204 (ipa_ref_opt_summary_t::remove): New function.
3205 (get_read_write_all_from_node): Fix GNU coding style.
3206 (propagate): Use function summary.
3207 (write_node_summary_p): Fix GNU coding style.
3208 (stream_out_bitmap): Likewise.
3209 (ipa_reference_read_optimization_summary): Use function summary.
3210 (ipa_reference_c_finalize): Do not release hooks.
3212 2018-06-08 Martin Liska <mliska@suse.cz>
3214 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3215 (analyze_function_body): Extract multiple calls of get_create.
3216 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3217 * ipa-inline.c (recursive_inlining): Use ::get method.
3218 * ipa-inline.h (estimate_edge_growth): Likewise.
3220 2018-06-08 Martin Liska <mliska@suse.cz>
3222 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3224 (hsa_function_summary::hsa_function_summary): Use the new enum
3226 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3227 * hsa-gen.c (hsa_get_host_function): Likewise.
3228 (get_brig_function_name): Likewise.
3229 * ipa-hsa.c (process_hsa_functions): Likewise.
3230 (ipa_hsa_write_summary): Likewise.
3231 * symbol-summary.h (symtab_duplication): Use ::get function/
3232 (get): New function.
3234 2018-06-08 Martin Liska <mliska@suse.cz>
3236 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3238 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3239 (hsa_register_kernel): Likewise.
3240 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3241 * hsa-gen.c (hsa_get_host_function): Likewise.
3242 (get_brig_function_name): Likewise.
3243 (generate_hsa): Likewise.
3244 (pass_gen_hsail::execute): Likewise.
3245 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3246 (devirtualization_time_bonus): Likewise.
3247 (ipcp_propagate_stage): Likewise.
3248 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3249 (edge_set_predicate): Likewise.
3250 (evaluate_conditions_for_known_args): Likewise.
3251 (evaluate_properties_for_edge): Likewise.
3252 (ipa_fn_summary::reset): Likewise.
3253 (ipa_fn_summary_t::duplicate): Likewise.
3254 (dump_ipa_call_summary): Likewise.
3255 (ipa_dump_fn_summary): Likewise.
3256 (analyze_function_body): Likewise.
3257 (compute_fn_summary): Likewise.
3258 (estimate_edge_devirt_benefit): Likewise.
3259 (estimate_edge_size_and_time): Likewise.
3260 (estimate_calls_size_and_time): Likewise.
3261 (estimate_node_size_and_time): Likewise.
3262 (inline_update_callee_summaries): Likewise.
3263 (remap_edge_change_prob): Likewise.
3264 (remap_edge_summaries): Likewise.
3265 (ipa_merge_fn_summary_after_inlining): Likewise.
3266 (ipa_update_overall_fn_summary): Likewise.
3267 (read_ipa_call_summary): Likewise.
3268 (inline_read_section): Likewise.
3269 (write_ipa_call_summary): Likewise.
3270 (ipa_fn_summary_write): Likewise.
3271 (ipa_free_fn_summary): Likewise.
3272 * ipa-hsa.c (process_hsa_functions): Likewise.
3273 (ipa_hsa_write_summary): Likewise.
3274 (ipa_hsa_read_section): Likewise.
3275 * ipa-icf.c (sem_function::merge): Likewise.
3276 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3277 (do_estimate_edge_time): Likewise.
3278 (estimate_size_after_inlining): Likewise.
3279 (estimate_growth): Likewise.
3280 (growth_likely_positive): Likewise.
3281 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3282 (inline_call): Likewise.
3283 * ipa-inline.c (caller_growth_limits): Likewise.
3284 (can_inline_edge_p): Likewise.
3285 (can_inline_edge_by_limits_p): Likewise.
3286 (compute_uninlined_call_time): Likewise.
3287 (compute_inlined_call_time): Likewise.
3288 (want_inline_small_function_p): Likewise.
3289 (edge_badness): Likewise.
3290 (update_caller_keys): Likewise.
3291 (update_callee_keys): Likewise.
3292 (recursive_inlining): Likewise.
3293 (inline_small_functions): Likewise.
3294 (inline_to_all_callers_1): Likewise.
3295 (dump_overall_stats): Likewise.
3296 (early_inline_small_functions): Likewise.
3297 (early_inliner): Likewise.
3298 * ipa-inline.h (estimate_edge_growth): Likewise.
3299 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3300 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3301 * ipa-prop.h (IPA_NODE_REF): Likewise.
3302 (IPA_EDGE_REF): Likewise.
3303 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3304 (propagate_malloc): Likewise.
3305 * ipa-split.c (execute_split_functions): Likewise.
3306 * symbol-summary.h: Rename get to get_create.
3308 (get_create): Likewise.
3309 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3311 2018-06-08 Martin Liska <mliska@suse.cz>
3313 * symbol-summary.h (release): Move definition out of class
3315 (symtab_removal): Likewise.
3316 (symtab_duplication): Likewise.
3318 2018-06-08 Martin Liska <mliska@suse.cz>
3320 * symbol-summary.h (function_summary): Move constructor
3321 implementation out of class declaration.
3322 (release): Likewise.
3323 (symtab_insertion): Likewise.
3324 (symtab_removal): Likewise.
3325 (symtab_duplication): Likewise.
3328 2018-06-08 Martin Liska <mliska@suse.cz>
3330 * Makefile.in: Remove support for MPX (macros, related functions,
3331 fields in cgraph_node, ...).
3332 * builtin-types.def (BT_BND): Likewise.
3333 (BT_FN_BND_CONST_PTR): Likewise.
3334 (BT_FN_CONST_PTR_BND): Likewise.
3335 (BT_FN_VOID_PTR_BND): Likewise.
3336 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3337 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3338 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3339 (expand_builtin_mempcpy_with_bounds): Likewise.
3340 (expand_builtin_memset_with_bounds): Likewise.
3341 (expand_builtin_memset_args): Likewise.
3342 (std_expand_builtin_va_start): Likewise.
3343 (expand_builtin): Likewise.
3344 (expand_builtin_with_bounds): Likewise.
3345 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3346 (DEF_LIB_BUILTIN_CHKP): Likewise.
3347 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3348 (DEF_CHKP_BUILTIN): Likewise.
3349 (BUILT_IN_MEMCPY): Likewise.
3350 (BUILT_IN_MEMMOVE): Likewise.
3351 (BUILT_IN_MEMPCPY): Likewise.
3352 (BUILT_IN_MEMSET): Likewise.
3353 (BUILT_IN_STPCPY): Likewise.
3354 (BUILT_IN_STRCAT): Likewise.
3355 (BUILT_IN_STRCHR): Likewise.
3356 (BUILT_IN_STRCPY): Likewise.
3357 (BUILT_IN_STRLEN): Likewise.
3358 (BUILT_IN_MEMCPY_CHK): Likewise.
3359 (BUILT_IN_MEMMOVE_CHK): Likewise.
3360 (BUILT_IN_MEMPCPY_CHK): Likewise.
3361 (BUILT_IN_MEMSET_CHK): Likewise.
3362 (BUILT_IN_STPCPY_CHK): Likewise.
3363 (BUILT_IN_STRCAT_CHK): Likewise.
3364 (BUILT_IN_STRCPY_CHK): Likewise.
3365 * calls.c (store_bounds): Likewise.
3366 (emit_call_1): Likewise.
3367 (special_function_p): Likewise.
3368 (maybe_warn_nonstring_arg): Likewise.
3369 (initialize_argument_information): Likewise.
3370 (finalize_must_preallocate): Likewise.
3371 (compute_argument_addresses): Likewise.
3372 (expand_call): Likewise.
3373 * cfgexpand.c (expand_call_stmt): Likewise.
3374 (expand_return): Likewise.
3375 (expand_gimple_stmt_1): Likewise.
3376 (pass_expand::execute): Likewise.
3377 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3378 (cgraph_node::remove): Likewise.
3379 (cgraph_node::dump): Likewise.
3380 (cgraph_node::verify_node): Likewise.
3381 * cgraph.h (chkp_function_instrumented_p): Likewise.
3382 (symtab_node::get_alias_target): Likewise.
3383 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3384 (cgraph_local_p): Likewise.
3385 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3386 (cgraph_edge::rebuild_references): Likewise.
3387 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3388 (walk_polymorphic_call_targets): Likewise.
3389 (cgraph_node::expand_thunk): Likewise.
3390 (symbol_table::output_weakrefs): Likewise.
3391 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3392 (ix86_handle_option): Likewise.
3393 * config/i386/constraints.md: Likewise.
3394 * config/i386/i386-builtin-types.def (BND): Likewise.
3398 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3399 (BDESC_FIRST): Likewise.
3401 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3402 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3403 * config/i386/i386.c (enum reg_class): Likewise.
3404 (ix86_target_string): Likewise.
3405 (ix86_option_override_internal): Likewise.
3406 (ix86_conditional_register_usage): Likewise.
3407 (ix86_valid_target_attribute_inner_p): Likewise.
3408 (ix86_set_indirect_branch_type): Likewise.
3409 (ix86_set_current_function): Likewise.
3410 (ix86_function_arg_regno_p): Likewise.
3411 (init_cumulative_args): Likewise.
3412 (ix86_function_arg_advance): Likewise.
3413 (ix86_function_arg): Likewise.
3414 (ix86_pass_by_reference): Likewise.
3415 (ix86_function_value_regno_p): Likewise.
3416 (ix86_function_value_1): Likewise.
3417 (ix86_function_value_bounds): Likewise.
3418 (ix86_return_in_memory): Likewise.
3419 (ix86_setup_incoming_vararg_bounds): Likewise.
3420 (ix86_va_start): Likewise.
3421 (indirect_thunk_need_prefix): Likewise.
3422 (print_reg): Likewise.
3423 (ix86_print_operand): Likewise.
3424 (ix86_expand_call): Likewise.
3425 (ix86_output_function_return): Likewise.
3426 (reg_encoded_number): Likewise.
3427 (BDESC_VERIFYS): Likewise.
3428 (ix86_init_mpx_builtins): Likewise.
3429 (ix86_init_builtins): Likewise.
3430 (ix86_emit_cmove): Likewise.
3431 (ix86_emit_move_max): Likewise.
3432 (ix86_expand_builtin): Likewise.
3433 (ix86_builtin_mpx_function): Likewise.
3434 (ix86_get_arg_address_for_bt): Likewise.
3435 (ix86_load_bounds): Likewise.
3436 (ix86_store_bounds): Likewise.
3437 (ix86_load_returned_bounds): Likewise.
3438 (ix86_store_returned_bounds): Likewise.
3439 (ix86_class_likely_spilled_p): Likewise.
3440 (ix86_hard_regno_mode_ok): Likewise.
3441 (x86_order_regs_for_local_alloc): Likewise.
3442 (ix86_mitigate_rop): Likewise.
3443 (ix86_bnd_prefixed_insn_p): Likewise.
3444 (ix86_mpx_bound_mode): Likewise.
3445 (ix86_make_bounds_constant): Likewise.
3446 (ix86_initialize_bounds): Likewise.
3447 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3448 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3449 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3450 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3451 (TARGET_CHKP_BOUND_MODE): Likewise.
3452 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3453 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3454 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3455 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3456 * config/i386/i386.h (TARGET_MPX): Likewise.
3457 (TARGET_MPX_P): Likewise.
3458 (VALID_BND_REG_MODE): Likewise.
3459 (FIRST_BND_REG): Likewise.
3460 (LAST_BND_REG): Likewise.
3461 (enum reg_class): Likewise.
3462 (BND_REG_P): Likewise.
3463 (BND_REGNO_P): Likewise.
3464 (BNDmode): Likewise.
3465 (ADJUST_INSN_LENGTH): Likewise.
3466 * config/i386/i386.md: Likewise.
3467 * config/i386/i386.opt: Likewise.
3468 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3469 (defined): Likewise.
3470 (LINK_MPX): Likewise.
3471 (MPX_SPEC): Likewise.
3472 (LIBMPX_SPEC): Likewise.
3473 (LIBMPXWRAPPERS_SPEC): Likewise.
3474 (CHKP_SPEC): Likewise.
3475 * config/i386/predicates.md: Likewise.
3476 * dbxout.c (dbxout_type): Likewise.
3477 * doc/extend.texi: Likewise.
3478 * doc/invoke.texi: Likewise.
3479 * doc/md.texi: Likewise.
3480 * doc/tm.texi: Likewise.
3481 * doc/tm.texi.in: Likewise.
3482 * dwarf2out.c (is_base_type): Likewise.
3483 (gen_formal_types_die): Likewise.
3484 (gen_subprogram_die): Likewise.
3485 (gen_type_die_with_usage): Likewise.
3486 (gen_decl_die): Likewise.
3487 (dwarf2out_late_global_decl): Likewise.
3488 * expr.c (expand_assignment): Likewise.
3489 (emit_storent_insn): Likewise.
3490 (store_expr_with_bounds): Likewise.
3491 (store_expr): Likewise.
3492 (expand_expr_real_1): Likewise.
3493 * expr.h (store_expr_with_bounds): Likewise.
3494 * function.c (use_register_for_decl): Likewise.
3495 (struct bounds_parm_data): Likewise.
3496 (assign_parms_augmented_arg_list): Likewise.
3497 (assign_parm_find_entry_rtl): Likewise.
3498 (assign_parm_is_stack_parm): Likewise.
3499 (assign_parm_load_bounds): Likewise.
3500 (assign_bounds): Likewise.
3501 (assign_parms): Likewise.
3502 (expand_function_start): Likewise.
3503 * gcc.c (CHKP_SPEC): Likewise.
3504 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3505 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3506 (wrestrict_dom_walker::check_call): Likewise.
3507 * gimple.c (gimple_build_call_from_tree): Likewise.
3508 * gimple.h (enum gf_mask): Likewise.
3509 (gimple_call_with_bounds_p): Likewise.
3510 (gimple_call_set_with_bounds): Likewise.
3511 * gimplify.c (gimplify_init_constructor): Likewise.
3512 * ipa-cp.c (initialize_node_lattices): Likewise.
3513 (propagate_constants_across_call): Likewise.
3514 (find_more_scalar_values_for_callers_subset): Likewise.
3515 * ipa-hsa.c (process_hsa_functions): Likewise.
3516 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3517 * ipa-icf.c (sem_function::merge): Likewise.
3518 * ipa-inline.c (early_inliner): Likewise.
3519 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3520 (warn_function_cold): Likewise.
3521 (propagate_pure_const): Likewise.
3522 * ipa-ref.h (enum GTY): Likewise.
3523 * ipa-split.c (find_retbnd): Likewise.
3524 (consider_split): Likewise.
3525 (split_function): Likewise.
3526 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3527 * ipa.c (walk_polymorphic_call_targets): Likewise.
3528 (symbol_table::remove_unreachable_nodes): Likewise.
3529 (process_references): Likewise.
3530 (cgraph_build_static_cdtor_1): Likewise.
3531 * lto-cgraph.c (lto_output_node): Likewise.
3532 (output_refs): Likewise.
3533 (compute_ltrans_boundary): Likewise.
3534 (input_overwrite_node): Likewise.
3535 (input_node): Likewise.
3536 (input_cgraph_1): Likewise.
3537 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3538 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3539 (class pass_chkp_instrumentation_passes): Likewise.
3540 (make_pass_chkp_instrumentation_passes): Likewise.
3541 * passes.def: Likewise.
3542 * rtl.h (struct GTY): Likewise.
3543 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3544 * stor-layout.c (layout_type): Likewise.
3545 * symtab.c: Likewise.
3546 * target.def: Likewise.
3547 * targhooks.c (default_chkp_bound_type): Likewise.
3548 (default_chkp_bound_mode): Likewise.
3549 (default_builtin_chkp_function): Likewise.
3550 (default_chkp_function_value_bounds): Likewise.
3551 (default_chkp_make_bounds_constant): Likewise.
3552 (default_chkp_initialize_bounds): Likewise.
3553 * targhooks.h (default_chkp_bound_type): Likewise.
3554 (default_chkp_bound_mode): Likewise.
3555 (default_builtin_chkp_function): Likewise.
3556 (default_chkp_function_value_bounds): Likewise.
3557 (default_chkp_make_bounds_constant): Likewise.
3558 (default_chkp_initialize_bounds): Likewise.
3559 * toplev.c (compile_file): Likewise.
3560 (process_options): Likewise.
3561 * tree-core.h (DEF_BUILTIN): Likewise.
3562 (DEF_BUILTIN_CHKP): Likewise.
3563 * tree-inline.c (declare_return_variable): Likewise.
3564 (remap_gimple_stmt): Likewise.
3565 (copy_bb): Likewise.
3566 (initialize_inlined_parameters): Likewise.
3567 (expand_call_inline): Likewise.
3568 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3569 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3570 (make_pass_ipa_chkp_produce_thunks): Likewise.
3571 (make_pass_chkp): Likewise.
3572 (make_pass_chkp_opt): Likewise.
3573 (make_pass_chkp_instrumentation_passes): Likewise.
3574 * tree-pretty-print.c (dump_generic_node): Likewise.
3575 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3576 * tree-ssa-dce.c (propagate_necessity): Likewise.
3577 (eliminate_unnecessary_stmts): Likewise.
3578 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3579 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3580 * tree-ssa-sccvn.h: Likewise.
3581 * tree-ssa-strlen.c (get_string_length): Likewise.
3582 (valid_builtin_call): Likewise.
3583 (adjust_last_stmt): Likewise.
3584 (handle_builtin_strchr): Likewise.
3585 (handle_builtin_strcpy): Likewise.
3586 (handle_builtin_stxncpy): Likewise.
3587 (handle_builtin_memcpy): Likewise.
3588 (handle_builtin_strcat): Likewise.
3589 (strlen_check_and_optimize_stmt): Likewise.
3590 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3591 * tree-streamer-in.c: Likewise.
3592 * tree-streamer.c (record_common_node): Likewise.
3593 * tree.c (tree_code_size): Likewise.
3594 (wide_int_to_tree_1): Likewise.
3595 (type_contains_placeholder_1): Likewise.
3596 (build_common_tree_nodes): Likewise.
3597 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3598 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3599 (POINTER_BOUNDS_P): Likewise.
3600 (BOUNDED_TYPE_P): Likewise.
3601 (BOUNDED_P): Likewise.
3602 (CALL_WITH_BOUNDS_P): Likewise.
3603 (pointer_bounds_type_node): Likewise.
3604 * value-prof.c (gimple_ic): Likewise.
3605 * var-tracking.c (vt_add_function_parameters): Likewise.
3606 * varasm.c (make_decl_rtl): Likewise.
3607 (assemble_start_function): Likewise.
3608 (output_constant): Likewise.
3609 (maybe_assemble_visibility): Likewise.
3610 * varpool.c (ctor_for_folding): Likewise.
3611 * chkp-builtins.def: Remove.
3612 * ipa-chkp.c: Remove.
3613 * ipa-chkp.h: Remove.
3614 * rtl-chkp.c: Remove.
3615 * rtl-chkp.h: Remove.
3616 * tree-chkp-opt.c: Remove.
3617 * tree-chkp.c: Remove.
3618 * tree-chkp.h: Remove.
3620 2018-06-07 Carl Love <cel@us.ibm.com>
3622 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3623 vextract_fp_from_shortl): Add BE support.
3625 2018-06-07 Paul Koning <ni1d@arrl.net>
3627 * compare-elim.c (try_merge_compare): Don't merge compare if
3628 address contains a side effect.
3629 (try_eliminate_compare): Likewise.
3631 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3633 * config.gcc: Support "tremont".
3634 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3635 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3637 * config/i386/i386.c (m_TREMONT): Define.
3638 (processor_target_table): Add "tremont".
3639 (PTA_TREMONT): Define.
3640 (ix86_lea_outperforms): Add TARGET_TREMONT.
3641 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3642 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3643 and M_INTEL_GOLDMONT_PLUS.
3644 (fold_builtin_cpu): Add "tremont".
3645 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3646 (ix86_option_override_internal): Add "tremont".
3647 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3648 (processor_type): Add PROCESSOR_TREMONT.
3649 * config/i386/x86-tune.def: Add m_TREMONT.
3650 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3652 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3654 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3655 symbol defined for msp430i* devices to be lower case.
3657 gcc/testsuite/gcc.target/msp430/
3658 * msp430i-device-symbol.c: New test.
3659 * msp430f-device-symbol.c: New test.
3660 * msp430.h: New test header file.
3662 2018-06-07 Richard Biener <rguenther@suse.de>
3664 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3665 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3666 Properly wrap signed arithmetic if overflow wraps.
3668 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3670 PR tree-optimization/69615
3671 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3672 of a cast from a same precision integral SSA_NAME in a bb dominated
3673 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3674 cast to utype if rhs2 has already a compatible type.
3676 2018-06-07 Richard Biener <rguenther@suse.de>
3678 PR tree-optimization/85935
3679 * graphite-scop-detection.c (find_params_in_bb): Analyze
3680 condition operands with respect to the correct loop. Assert
3681 the analysis doesn't fail.
3683 2018-06-04 Carl Love <cel@us.ibm.com>
3685 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3686 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3687 as it is slightly cheaper.
3688 (first_match_or_eos_index_<mode>):
3689 Calculate index using natural element order.
3690 (first_match_index_<mode>):
3691 Calculate index using natural element order.
3692 (first_match_or_eos_index_<mode>):
3693 Calculate index using natural order.
3694 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3695 for BE and LE modes.
3696 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3697 P9V_BUILTIN_VCLZLSBB_V16QI.
3698 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3701 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3703 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3704 indentation and line wrap for many prototypes. Add missing
3705 @smallexample directives around block of prototypes for vec_xl and
3708 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3710 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3711 track if we pass or return IEEE 128-bit floating point.
3712 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3713 C++ mangling that is compatible with GCC 8.1.
3714 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3715 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3716 floating point types.
3717 (rs6000_function_arg_advance_1): Likewise.
3718 (rs6000_mangle_type): Optionally generate mangled names that match
3719 what GCC 8.1 generated for IEEE 128-bit floating point types.
3720 (rs6000_globalize_decl_name): If we have an external function that
3721 passes or returns IEEE 128-bit types, generate a weak reference
3722 from the mangled name used in GCC 8.1 to the current mangled
3724 (rs6000_init_builtins): Make __ibm128 use the long double type if
3725 long double is IBM extended double. Make __float128 use the long
3726 double type if long double is IEEE 128-bit.
3729 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3730 macro for __ibm128 built-in functions.
3731 (PACK_IF): Add __ibm128 pack/unpack functions.
3732 (UNPACK_IF): Likewise.
3733 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3734 enable long double built-in functions if long double is IEEE
3735 128-bit floating point.
3736 (rs6000_invalid_builtin): Update long double built-in function
3738 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3739 functions, adjust the built-in function to use the long double
3740 built-in function if __ibm128 and long double are the same type.
3741 * doc/extend.texi (PowerPC builtins): Update documention for
3742 __builtin_{,un}pack_longdouble. Add documentation for
3743 __builtin_{,un}pack_ibm128.
3745 2018-06-06 Jim Wilson <jimw@sifive.com>
3747 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3748 (struct machine_function): New field interrupt_mode.
3749 (riscv_handle_type_attribute): New function. Add forward declaration.
3750 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3751 (riscv_expand_epilogue): Check interrupt_mode field.
3752 (riscv_set_current_function): Check interrupt attribute args and
3753 set interrupt_mode field.
3754 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3755 (riscv_sret, riscv_uret): New.
3756 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3757 new arguments to interrupt attribute.
3759 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3762 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3763 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3765 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3767 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3768 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3769 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3770 several redundant entries.
3772 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3774 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3775 type from "rtx" to "rtx_insn *".
3776 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3777 for local "call_insn", removing cast.
3778 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3780 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3782 PR tree-optimization/86066
3783 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3784 for BIT_INSERT_EXPR stores.
3786 2018-06-06 Richard Biener <rguenther@suse.de>
3788 PR tree-optimization/86062
3789 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3790 component refs ontop
3791 of to be offsetted base.
3793 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3795 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3796 to be static and remove check on interrupt attribute name.
3798 gcc/testsuite/gcc.target/msp430/
3799 * function-attributes-4.c: New test.
3800 * static-interrupts.c: New test.
3802 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3804 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3805 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3807 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3810 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3812 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3813 Remove second argument, change how error is called.
3814 (aarch64_layout_arg): Remove second argument from
3815 aarch64_err_no_fpadvsimd call.
3816 (aarch64_init_cumulative_args): Ditto.
3817 (aarch64_gimplify_va_arg_expr): Ditto.
3818 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3820 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3822 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3823 (*simple_return_indirect_internal<mode>): Rename from
3824 simple_return_indirect_internal. Use W mode iterator.
3825 (rstorssp): New expander.
3826 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3827 (clrssbsy): New expander.
3828 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3830 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3832 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3834 (cmse_check_pointed_object): Likewise.
3836 2018-06-05 Martin Liska <mliska@suse.cz>
3838 PR gcov-profile/47618
3839 * doc/invoke.texi: Document how -fprofile-dir format
3842 2018-06-05 Richard Biener <rguenther@suse.de>
3844 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3845 removal pretend DOM info isn't available so we do not update
3846 it and only remove edges, not dominated blocks. Actually free
3847 DOM info in case we removed something. Remove unreachable blocks.
3848 (mfb_keep_latches): Work with either DOM info or marked backedges.
3849 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3850 first. Mark backedges if DOM info isn't available.
3851 (Re-)compute DOM info after cleanup_control_flow_pre.
3853 2018-06-05 Richard Biener <rguenther@suse.de>
3855 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3856 (locus_discrim_hasher::hash): Adjust.
3857 (locus_discrim_hasher::equal): Likewise.
3858 (next_discriminator_for_locus): Work on line directly.
3859 (same_line_p): Pass in expanded locus1 as well.
3860 (assign_discriminators): Avoid redundant location expansions.
3862 2018-06-05 Richard Biener <rguenther@suse.de>
3864 PR tree-optimization/86046
3865 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3866 if required after clearing TREE_ADDRESSABLE.
3868 2018-06-05 Richard Biener <rguenther@suse.de>
3870 PR tree-optimization/86047
3871 * tree-ssa-loop.c (for_each_index): Glob handling of all
3872 decls and constants and really handle all of them.
3874 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3877 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3878 qualifier_void_pointer and qualifier_const_void_pointer.
3879 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3880 (arm_init_builtins): Handle the above.
3881 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3882 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3883 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3886 2018-06-05 Martin Liska <mliska@suse.cz>
3888 * auto-profile.c (read_autofdo_file): Do not use
3889 gcov_ctr_summary struct.
3890 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3891 * coverage.c (struct counts_entry): Likewise.
3892 (read_counts_file): Read just single summary entry.
3893 (get_coverage_counts): Use gcov_summary struct.
3894 * coverage.h (get_coverage_counts): Likewise.
3895 * gcov-dump.c (dump_working_sets): Likewise.
3896 (tag_summary): Dump just single summary.
3897 * gcov-io.c (gcov_write_summary): Write just histogram
3899 (gcov_read_summary): Read just single summary.
3900 (compute_working_sets): Use gcov_summary struct.
3901 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3902 of GCOV_COUNTERS_SUMMABLE.
3903 (GCOV_COUNTERS_SUMMABLE): Remove.
3904 (GCOV_FIRST_VALUE_COUNTER): Replace with
3905 GCOV_COUNTER_V_INTERVAL.
3906 (struct gcov_ctr_summary): Remove.
3907 (struct gcov_summary): Directly use fields of former
3909 (compute_working_sets): Use gcov_summary struct.
3910 * gcov.c (read_count_file): Do not use ctrs fields.
3911 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3913 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3915 * profile.c: Likewise.
3916 * profile.h: Likewise.
3918 2018-06-05 Martin Liska <mliska@suse.cz>
3920 PR gcov-profile/84846
3921 * gcov.c (output_lines): Print working directory only
3922 in intermediate format.
3924 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3926 * config/s390/s390-builtin-types.def: Add void function type.
3927 * config/s390/s390-builtins.def: Use the function type for the
3930 2018-06-04 Jim Wilson <jimw@sifive.com>
3932 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3934 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3935 and maybe_eh_return. Change regno to unsigned int. Use new args to
3936 handle EH_RETURN_DATA_REGNO registers properly.
3937 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3938 (riscv_expand_epilogue): Update comment. Change argument name and
3939 type. Update code to use new name and type. Pass new args to
3940 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3942 * config/riscv/riscv.md (NORMAL_RETURN): New.
3943 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3944 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3945 (eh_return): Call gen_eh_return_internal and emit barrier.
3946 (eh_return_internal): Call riscv_expand_epilogue.
3948 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3950 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3951 bit_insertion field and declare can_be_merged_into method.
3952 (merged_store_group::can_be_merged_into): New method.
3953 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3954 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3955 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3957 2018-06-04 Richard Biener <rguenther@suse.de>
3959 PR tree-optimization/85955
3960 * builtins.c (fold_builtin_sincos): Convert pointers to
3961 destination to appropriate type before dereferencing.
3963 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3965 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3967 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3969 * expr.c (expand_expr_real_1): Force the operand into memory if
3970 its TYPE_MODE is BLKmode and if there is no integer mode for
3971 the number of bits being extracted.
3973 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3977 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3978 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3980 2018-06-04 Richard Biener <rguenther@suse.de>
3982 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3983 (cleanup_tree_cfg_noloop): ... single caller. Do
3984 start_recording_case_labels later.
3986 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3988 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3989 to _IMMINTRIN_H_INCLUDED.
3990 * config/i386/pconfigintrin.h: Ditto.
3991 * config/i386/waitpkgintrin.h: Ditto.
3992 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3993 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3994 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3995 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3996 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3997 waitpkgintrin.h and cldemoteintrin.h.
3999 2018-06-04 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/86038
4002 * tracer.c (find_best_successor): Check probability for
4003 being initialized, bail out if not.
4005 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4008 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4009 of bits to ignore when comparing architectures.
4011 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4013 PR tree-optimization/69615
4014 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4015 maximum or minimum of the type, try to merge it also as if
4016 range1 is + [-, x - 1] or + [x + 1, -].
4019 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4021 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4023 PR tree-optimization/86034
4024 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4025 the unsigned bitfield type in a bit insertion sequence if it does not
4026 have a larger precision than the bitfield size.
4027 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4029 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4031 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4033 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4035 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4036 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4037 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4038 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4040 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4042 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4043 Disable -fdelete-null-pointer-checks for ELF toolchain.
4045 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4046 Kito Cheng <kito.cheng@gmail.com>
4048 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4049 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4050 (nds32*-*-*): Add glibc and uclibc conditions.
4051 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4052 (TARGET_EXCEPT_UNWIND_INFO): Define.
4053 * config/nds32/elf.h: New file.
4054 * config/nds32/linux.h: New file.
4055 * config/nds32/nds32-elf.opt: New file.
4056 * config/nds32/nds32-linux.opt: New file.
4057 * config/nds32/nds32-fp-as-gp.c
4058 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4059 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4061 (nds32_asm_file_end): Ditto.
4062 (nds32_print_operand): Ditto.
4063 (nds32_insert_attributes): Ditto.
4064 (nds32_init_libfuncs): New function.
4065 (TARGET_HAVE_TLS): Define.
4066 (TARGET_INIT_LIBFUNCS): Define.
4067 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4069 (TARGET_ELF): Apply different mcmodel setting.
4070 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4071 been migrated into elf.h and linux.h files.
4072 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4073 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4074 (mcmodel): The content has been migrated into nds32-elf.opt and
4075 nds32-linux.opt files.
4076 * config/nds32/t-elf: New file.
4077 * config/nds32/t-linux: New file.
4079 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4080 Shiva Chen <shiva0217@gmail.com>
4082 * config/nds32/constants.md (unspec_volatile_element): Add
4083 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4084 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4086 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4087 (make_pass_nds32_fp_as_gp): Declare.
4088 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4090 (nds32_asm_function_end_prologue): Remove unused asm output.
4091 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4092 (nds32_asm_file_start): Output necessary fp_as_gp information.
4093 (nds32_option_override): Adjust register usage.
4094 (nds32_expand_prologue): Consider fp_as_gp situation.
4095 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4096 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4099 (simple_return): Ditto.
4100 (omit_fp_begin): Output special directive for fp_as_gp.
4101 (omit_fp_end): Output special directive for fp_as_gp.
4102 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4103 mforbid-fp-as-gp): New options.
4105 2018-06-01 Mark Wielaard <mark@klomp.org>
4107 * dwarf2out.c (dwarf2out_finish): Remove generation of
4108 DW_AT_loclists_base.
4110 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4112 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4113 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4114 (struct merged_store_group): Add bit_insertion field.
4115 (dump_char_array): Use standard hexadecimal format.
4116 (merged_store_group::merged_store_group): Set bit_insertion to false.
4117 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4118 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4119 also print the mask in the dump file.
4120 (pass_store_merging::gate): Minor tweak.
4121 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4122 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4123 stores with INTEGER_CST stores.
4124 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4125 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4126 and use it throughout. Generate bit insertion sequences if need be.
4127 (pass_store_merging::process_store): Remove redundant condition.
4128 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4130 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4132 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4133 the 128-bit floating point types. Fix function comment.
4135 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4137 * config/aarch64/aarch64-simd.md
4138 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4140 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4143 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4145 PR tree-optimization/85989
4146 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4148 (backprop::intersect_uses): Check it when deciding whether this
4149 is a backedge reference.
4150 (backprop::process_block): Add each phi to m_visited_phis
4151 after visiting it, then clear it at the end.
4153 2018-06-01 Richard Biener <rguenther@suse.de>
4155 * tree-vectorizer.h (vect_dr_stmt): New function.
4156 (vect_get_load_cost): Adjust.
4157 (vect_get_store_cost): Likewise.
4158 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4159 Use vect_dr_stmt instead of DR_SMTT.
4160 (vect_record_base_alignments): Likewise.
4161 (vect_calculate_target_alignment): Likewise.
4162 (vect_compute_data_ref_alignment): Likewise and make static.
4163 (vect_update_misalignment_for_peel): Likewise.
4164 (vect_verify_datarefs_alignment): Likewise.
4165 (vector_alignment_reachable_p): Likewise.
4166 (vect_get_data_access_cost): Likewise. Pass down
4167 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4168 (vect_get_peeling_costs_all_drs): Likewise.
4169 (vect_peeling_hash_get_lowest_cost): Likewise.
4170 (vect_enhance_data_refs_alignment): Likewise.
4171 (vect_find_same_alignment_drs): Likewise.
4172 (vect_analyze_data_refs_alignment): Likewise.
4173 (vect_analyze_group_access_1): Likewise.
4174 (vect_analyze_group_access): Likewise.
4175 (vect_analyze_data_ref_access): Likewise.
4176 (vect_analyze_data_ref_accesses): Likewise.
4177 (vect_vfa_segment_size): Likewise.
4178 (vect_small_gap_p): Likewise.
4179 (vectorizable_with_step_bound_p): Likewise.
4180 (vect_prune_runtime_alias_test_list): Likewise.
4181 (vect_analyze_data_refs): Likewise.
4182 (vect_supportable_dr_alignment): Likewise.
4183 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4184 (vect_gen_prolog_loop_niters): Likewise.
4185 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4186 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4188 (vect_recog_mask_conversion_pattern): Likewise.
4189 (vect_try_gather_scatter_pattern): Likewise.
4190 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4191 to vect_get_store_cost.
4192 (vect_get_store_cost): Get stmt_info instead of DR.
4193 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4194 (vect_get_load_cost): Get stmt_info instead of DR.
4196 2018-06-01 Richard Biener <rguenther@suse.de>
4199 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4200 vars wrapped in MEM_REFs.
4202 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4204 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4205 Fix subreg tests so that we only return a choice between
4206 GENERAL_REGS and FP_REGS if the original classes included both.
4208 2018-06-01 Richard Biener <rguenther@suse.de>
4211 * tree-ssa-structalias.c (get_function_part_constraint):
4212 Handle NULL fi->decl.
4213 (find_func_aliases_for_call): Properly handle indirect
4214 fi from direct call.
4215 (find_func_clobbers): Likewise.
4216 (ipa_pta_execute): Likewise.
4217 (create_variable_info_for): For functions that are ifunc_resolver
4218 resolve to a varinfo that contains the result of the resolver call.
4219 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4222 2018-05-31 Michael Collison <michael.collison@arm.com>
4224 * config/aarch64/aarch64.md:
4225 (*fix_to_zero_extenddfdi2): New pattern.
4226 * gcc.target/aarch64/fix_extend1.c: New testcase.
4228 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4232 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4233 and BUILT_IN_STRNCMP_EQ.
4234 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4235 BUILT_IN_STRNCMP_EQ.
4236 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4237 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4238 (gimple_fold_builtin): Likewise.
4239 * tree-ssa-strlen.c (compute_string_length): New function.
4240 (determine_min_obsize): New function.
4241 (handle_builtin_string_cmp): New function to handle calls to
4242 string compare functions.
4243 (strlen_optimize_stmt): Add handling to builtin string compare
4245 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4246 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4247 * tree.c (build_common_builtin_nodes): Add new defines of
4248 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4250 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4253 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4254 functions with naked attribute.
4256 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4258 * config/i386/sse.md (avx_vec_concat<mode>):
4259 Substitute concat_tg_mode mode attribute with xtg_mode.
4260 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4261 (concat_tg_mode): Remove mode attribute.
4263 2018-05-31 Martin Sebor <msebor@redhat.com>
4266 * calls.c (alloc_max_size): Correct a logic error/typo.
4267 Treat excessive arguments as infinite. Warn for invalid arguments.
4268 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4270 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4273 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4274 and movx for Haswell.
4276 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4277 Cesar Philippidis <cesar@codesourcery.com>
4280 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4281 when emitting error on private/firstprivate reductions.
4282 * omp-low.c (lower_omp_target): Avoid reference-type processing
4283 on pointers for firstprivate clause.
4285 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4287 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4290 * config/aarch64/aarch64-simd.md
4291 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4292 (aarch64_ld1_x3_<mode>): Likewise
4293 (aarch64_st1x2<VALLDIF:mode>): Likewise
4294 (aarch64_st1_x2_<mode>): Likewise
4295 (aarch64_st1x3<VALLDIF:mode>): Likewise
4296 (aarch64_st1_x3_<mode>): Likewise
4297 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4298 (vld1_s8_x3): Likewise.
4299 (vld1_u16_x3): Likewise.
4300 (vld1_s16_x3): Likewise.
4301 (vld1_u32_x3): Likewise.
4302 (vld1_s32_x3): Likewise.
4303 (vld1_u64_x3): Likewise.
4304 (vld1_s64_x3): Likewise.
4305 (vld1_f16_x3): Likewise.
4306 (vld1_f32_x3): Likewise.
4307 (vld1_f64_x3): Likewise.
4308 (vld1_p8_x3): Likewise.
4309 (vld1_p16_x3): Likewise.
4310 (vld1_p64_x3): Likewise.
4311 (vld1q_u8_x3): Likewise.
4312 (vld1q_s8_x3): Likewise.
4313 (vld1q_u16_x3): Likewise.
4314 (vld1q_s16_x3): Likewise.
4315 (vld1q_u32_x3): Likewise.
4316 (vld1q_s32_x3): Likewise.
4317 (vld1q_u64_x3): Likewise.
4318 (vld1q_s64_x3): Likewise.
4319 (vld1q_f16_x3): Likewise.
4320 (vld1q_f32_x3): Likewise.
4321 (vld1q_f64_x3): Likewise.
4322 (vld1q_p8_x3): Likewise.
4323 (vld1q_p16_x3): Likewise.
4324 (vld1q_p64_x3): Likewise.
4325 (vst1_s64_x2): Likewise.
4326 (vst1_u64_x2): Likewise.
4327 (vst1_f64_x2): Likewise.
4328 (vst1_s8_x2): Likewise.
4329 (vst1_p8_x2): Likewise.
4330 (vst1_s16_x2): Likewise.
4331 (vst1_p16_x2): Likewise.
4332 (vst1_s32_x2): Likewise.
4333 (vst1_u8_x2): Likewise.
4334 (vst1_u16_x2): Likewise.
4335 (vst1_u32_x2): Likewise.
4336 (vst1_f16_x2): Likewise.
4337 (vst1_f32_x2): Likewise.
4338 (vst1_p64_x2): Likewise.
4339 (vst1q_s8_x2): Likewise.
4340 (vst1q_p8_x2): Likewise.
4341 (vst1q_s16_x2): Likewise.
4342 (vst1q_p16_x2): Likewise.
4343 (vst1q_s32_x2): Likewise.
4344 (vst1q_s64_x2): Likewise.
4345 (vst1q_u8_x2): Likewise.
4346 (vst1q_u16_x2): Likewise.
4347 (vst1q_u32_x2): Likewise.
4348 (vst1q_u64_x2): Likewise.
4349 (vst1q_f16_x2): Likewise.
4350 (vst1q_f32_x2): Likewise.
4351 (vst1q_f64_x2): Likewise.
4352 (vst1q_p64_x2): Likewise.
4353 (vst1_s64_x3): Likewise.
4354 (vst1_u64_x3): Likewise.
4355 (vst1_f64_x3): Likewise.
4356 (vst1_s8_x3): Likewise.
4357 (vst1_p8_x3): Likewise.
4358 (vst1_s16_x3): Likewise.
4359 (vst1_p16_x3): Likewise.
4360 (vst1_s32_x3): Likewise.
4361 (vst1_u8_x3): Likewise.
4362 (vst1_u16_x3): Likewise.
4363 (vst1_u32_x3): Likewise.
4364 (vst1_f16_x3): Likewise.
4365 (vst1_f32_x3): Likewise.
4366 (vst1_p64_x3): Likewise.
4367 (vst1q_s8_x3): Likewise.
4368 (vst1q_p8_x3): Likewise.
4369 (vst1q_s16_x3): Likewise.
4370 (vst1q_p16_x3): Likewise.
4371 (vst1q_s32_x3): Likewise.
4372 (vst1q_s64_x3): Likewise.
4373 (vst1q_u8_x3): Likewise.
4374 (vst1q_u16_x3): Likewise.
4375 (vst1q_u32_x3): Likewise.
4376 (vst1q_u64_x3): Likewise.
4377 (vst1q_f16_x3): Likewise.
4378 (vst1q_f32_x3): Likewise.
4379 (vst1q_f64_x3): Likewise.
4380 (vst1q_p64_x3): Likewise.
4382 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4384 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4385 user_label_prefix to name.
4387 * tree-core.h: Update comment about the format of NAME string
4388 passed to handler in attribute_spec.
4390 * config/msp430/msp430.md: Remove erroneous subreg expression from
4391 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4392 zero_extend{q,h}isi2.
4394 2018-05-30 Borislav Petkov <bp@suse.de>
4396 * doc/extend.texi: Document some architecture specific
4397 constraints and sort entries.
4399 2018-05-30 Martin Sebor <msebor@redhat.com>
4402 * builtins.c (expand_builtin_stpcpy_1): New function.
4403 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4404 only if the former succeeds.
4406 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4408 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4411 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4413 * doc/invoke.texi (-flinker-output): Document
4415 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4417 * passes.c (ipa_write_summaries): Only modify statements if body
4419 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4420 incrementally linking.
4421 (ipa_passes): Likewise.
4422 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4423 pass down resolution info.
4424 * common.opt (flag_incremental_link): Update info.
4425 * gcc.c (plugin specs): Turn flinker-output=* to
4426 -plugin-opt=-linker-output-known
4427 * toplev.c (compile_file): Also cut compilation when doing incremental
4429 * flag-types. (enum lto_partition_model): Add
4430 LTO_LINKER_OUTPUT_NOLTOREL.
4431 (invoke.texi): Add -flinker-output docs.
4432 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4433 link same way as WPA; do not stream in dead initializers.
4435 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4436 darf2out_register_external_decl): Support incremental link.
4438 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4440 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4442 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4444 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4445 it down to simple_object_copy_lto_debug_sections.
4446 (run_gcc): Determine incremental LTO link time and configure
4447 lto1 into non-wpa mode, disable renaming of debug sections.
4449 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4451 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4452 descriptions of various incorrectly documented functions.
4454 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4457 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4458 address check not strict.
4460 2018-05-30 Richard Biener <rguenther@suse.de>
4462 PR tree-optimization/85964
4463 * tracer.c (better_p): Drop initialized count check, we only
4464 call the function with initialized counts now.
4465 (find_best_successor): Do find a best edge if one
4466 has uninitialized count.
4467 (find_best_predecessor): Likewise. Do BB frequency check only
4468 if count is initialized.
4470 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4472 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4473 (aarch64_ldrstr_offset_compare): New.
4474 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4475 load/store orderings.
4476 (aarch64_gen_adjusted_ldpstp): Likewise.
4478 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4480 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4481 Check for subset of GENERAL_REGS and FP_REGS.
4482 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4485 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4487 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4488 and wi::to_poly_offset. Add the current offset and then check
4489 whether the sum fits, rather than using an unchecked addition of
4490 a checked term. Check for a shwi rather than a uhwi.
4491 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4492 (store_constructor): Use poly_int_tree_p.
4493 (expand_expr_real_1): Likewise.
4494 * function.c (assign_temp): Likewise.
4495 * fold-const.c (const_binop): Use poly_int_tree_p and
4497 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4499 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4500 to_poly_offset for MEM offsets.
4501 * ipa-icf.c (sem_variable::equals): Likewise.
4502 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4503 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4504 wi::to_poly_offset for BIT_FIELD_REF offsets.
4505 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4507 * var-tracking.c (emit_note_insn_var_location): Use
4508 tree_to_poly_uint64.
4510 2018-05-29 Jim Wilson <jimw@sifive.com>
4512 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4514 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4517 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4518 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4520 (sse4_1_round<mode>2): Use nonimmediate_operand
4521 for operand 1 predicate.
4523 2018-05-29 Martin Sebor <msebor@redhat.com>
4524 Richard Biener <rguenther@suse.de>
4527 * calls.c (get_size_range): Call determine_value_range instead
4528 of get_value_range..
4529 * tree-vrp.h (determine_value_range): Declared new function.
4530 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4532 2018-05-29 Richard Biener <rguenther@suse.de>
4534 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4535 sure to use non-pattern stmts for get_earlier_stmt arguments.
4536 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4537 called on pattern stmts.
4538 (get_later_stmt): Likewise.
4540 2018-05-29 Martin Liska <mliska@suse.cz>
4542 PR gcov-profile/85759
4543 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4546 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4548 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4550 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4554 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4555 VEC_PACK_FLOAT_EXPR): New tree codes.
4556 * tree-pretty-print.c (op_code_prio): Handle
4557 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4558 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4559 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4560 * tree-inline.c (estimate_operator_cost): Likewise.
4561 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4562 * fold-const.c (const_binop): Likewise.
4563 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4564 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4565 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4566 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4567 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4568 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4569 * expr.c (expand_expr_real_2): Likewise.
4570 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4571 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4572 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4574 * optabs.c (expand_widen_pattern_expr): For
4575 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4576 sign from result type rather than operand's type.
4577 (expand_binop_directly): For vec_packu_float_optab and
4578 vec_packs_float_optab allow result type to be different from operand's
4580 * optabs-tree.c (optab_for_tree_code): Handle
4581 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4582 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4583 * tree-vect-generic.c (expand_vector_operations_1): Handle
4584 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4585 VEC_PACK_FLOAT_EXPR.
4586 * tree-vect-stmts.c (supportable_widening_operation): Handle
4588 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4589 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4590 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4591 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4592 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4594 (vec_pack<floatprefix>_float_<mode>): New expander.
4595 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4597 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4598 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4599 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4600 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4601 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4603 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4604 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4605 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4606 VEC_PACK_FLOAT_EXPR): Document.
4608 2018-05-29 Richard Biener <rguenther@suse.de>
4610 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4612 (stmt_vec_info_vec): Make pointer.
4613 (init_stmt_vec_info_vec): Remove.
4614 (free_stmt_vec_info_vec): Likewise.
4615 (set_stmt_vec_info_vec): New function.
4616 (free_stmt_vec_infos): Likewise.
4617 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4618 (set_vinfo_for_stmt): Likewise.
4619 (get_earlier_stmt): Likewise.
4620 (get_later_stmt): Likewise.
4621 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4622 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4623 (vec_info::~vec_info): Free stmt_vec_infos.
4624 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4625 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4626 (pass_slp_vectorize::execute): Likewise.
4627 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4628 (free_stmt_vec_info_vec): Likewise.
4629 (set_stmt_vec_info_vec): New function.
4630 (free_stmt_vec_infos): Likewise.
4631 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4632 the global stmt_vec_info_vec.
4633 * tree-parloops.c (gather_scalar_reductions): Use
4634 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4637 2018-05-29 Richard Biener <rguenther@suse.de>
4639 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4641 2018-05-29 Martin Liska <mliska@suse.cz>
4642 David Malcolm <dmalcolm@redhat.com>
4644 * vec.c (test_reverse): New.
4645 (vec_c_tests): Add new test.
4646 * vec.h (vl_ptr>::reverse): New function.
4648 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4650 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4652 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4655 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4657 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4659 2018-05-28 Richard Biener <rguenther@suse.de>
4661 PR tree-optimization/85933
4662 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4663 look at stmts marked as vectorizable.
4665 2018-05-28 Richard Biener <rguenther@suse.de>
4667 PR tree-optimization/85934
4668 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4669 vector boolean check before scalar optimization.
4671 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4673 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4676 2018-05-28 Mark Wielaard <mark@klomp.org>
4678 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4679 if it is an expression containing a minus sign.
4681 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4683 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4685 2018-05-27 Paul Koning <ni1d@arrl.net>
4687 * config/pdp11/pdp11.md (truncsihi2): Remove.
4689 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4690 Chung-Ju Wu <jasonwucj@gmail.com>
4692 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4694 (unaligned_store_dw): Ditto.
4695 * config/nds32/nds32-memory-manipulation.c
4696 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4697 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4698 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4699 (emit_setmem_word_loop): Rename to ...
4700 (emit_setmem_doubleword_loop): ... this.
4701 (nds32_gen_dup_4_byte_to_word_value): New function.
4702 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4703 (nds32_expand_setmem_loop): Refine implementation.
4704 (nds32_expand_setmem_loop_v3m): Ditto.
4705 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4708 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4712 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4714 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4715 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4716 machine->attr_no_prologue_p.
4717 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4718 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4719 (nds32_expand_epilogue): Consider attr_naked_p.
4720 (nds32_expand_epilogue_v3pop): Likewise.
4721 (nds32_can_use_return_insn): Likewise.
4722 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4723 attr_no_prologue_p fields.
4724 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4726 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4729 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4731 * config/i386/sse.md
4732 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4734 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4736 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4738 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4740 (*<floatsuffix>floatv2div2sf2): Rename to ...
4741 (*float<floatunssuffix>v2div2sf2): ... this.
4742 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4743 (float<floatunssuffix>v2div2sf2_mask): ... this.
4744 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4745 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4746 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4748 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4750 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4752 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4754 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4756 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4758 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4759 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4760 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4761 gen_ufix_truncv8dfv8si2.
4762 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4763 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4764 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4765 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4766 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4767 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4768 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4769 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4771 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4775 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4777 2018-05-25 Jim Wilson <jimw@sifive.com>
4779 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4780 * config/riscv/riscv.c (struct machine_function): Add
4781 interrupt_handler_p and attribute_checked_p fields.
4782 (riscv_attribute_table): Add interrupt.
4783 (riscv_interrupt_type_p): New.
4784 (riscv_save_reg_p): Save extra regs for interrupt handler.
4785 (riscv_use_save_libcall): Return false for interrupt handler.
4786 (riscv_first_stack_step): Add forward declaration.
4787 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4788 for interrupt handler with large frame. Use it for saved reg list.
4789 (riscv_expand_prologue): Move flag_stack_usage_info support to
4790 eliminate duplication.
4791 (riscv_expand_epilogue): Generate mret for interrupt handler.
4792 (riscv_epilogue_uses): New.
4793 (riscv_can_use_return_insn): Return false for interrupt handler.
4794 (riscv_function_ok_for_sibcall): Likewise.
4795 (riscv_set_current_function): Add interrupt handler support.
4796 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4797 * config/riscv/riscv.md (UNSPECV_MRET): New.
4799 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4801 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4803 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4805 PR tree-optimization/85712
4806 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4807 this candidate has already been replaced in-situ by a copy.
4809 2018-05-25 Jason Merrill <jason@redhat.com>
4811 PR c++/80485 - inline function non-zero address.
4812 * symtab.c (nonzero_address): Check DECL_COMDAT.
4814 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4817 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4818 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4819 extension of SImode operation. Use const123_operand predicate.
4820 (*saddsi_1): Remove.
4821 (*saddl_se_1): Ditto.
4823 (*ssubl_se_1): Ditto.
4824 * config/alpha/predicates.md (const123_operand): New predicate.
4825 * config/alpha/constraints.md (P): Use IN_RANGE.
4827 2018-05-25 Richard Biener <rguenther@suse.de>
4829 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4831 (ref_maybe_used_by_stmt_p): Likewise.
4832 (stmt_may_clobber_ref_p): Likewise.
4833 (stmt_may_clobber_ref_p_1): Likewise.
4834 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4836 (ref_maybe_used_by_stmt_p): Likewise.
4837 (stmt_may_clobber_ref_p): Likewise.
4838 (stmt_may_clobber_ref_p_1): Likewise.
4839 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4840 the alias oracle to disambiguate DRs with stmts DR analysis
4842 (vect_analyze_data_refs): Do not give up on not analyzable
4843 DRs for BB vectorization. Remove code truncating the dataref
4846 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4849 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4850 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4851 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4853 2018-05-25 Richard Biener <rguenther@suse.de>
4855 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4856 function, combining stmt data ref gathering and fatal analysis
4858 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4859 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4860 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4861 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4862 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4864 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4866 PR tree-optimization/85720
4867 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4868 SCC if all partitions are builtins.
4869 (version_loop_by_alias_check): New parameter. Generate cancelable
4870 runtime alias check if all partitions are builtins.
4871 (distribute_loop): Update call to above function.
4873 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4875 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4876 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4877 (parm_default_def_partition_arg): Ditto.
4878 (set_parm_default_def_partition): Ditto.
4879 (get_parm_default_def_partitions): Ditto and make it static.
4880 (get_undefined_value_partitions): Ditto and make it static.
4881 (remove_ssa_form): Refactor call to init_var_map here.
4882 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4883 computation for loop region.
4884 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4885 (register_default_def): Delete.
4886 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4887 (parm_default_def_partition_arg): Ditto.
4888 (set_parm_default_def_partition): Ditto.
4889 (get_parm_default_def_partitions): Ditto and make it static.
4890 (get_undefined_value_partitions): Ditto and make it static.
4891 (coalesce_with_default, coalesce_with_default): Update comment.
4892 (create_coalesce_list_for_region): New func factored out from
4893 create_outofssa_var_map.
4894 (populate_coalesce_list_for_outofssa): New func factored out from
4895 create_outofssa_var_map and coalesce_ssa_name.
4896 (create_outofssa_var_map): Delete.
4897 (coalesce_ssa_name): Refactor to support live range computation.
4898 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4899 (get_parm_default_def_partitions): Delete.
4900 (get_undefined_value_partitions): Ditto.
4901 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4902 computation for loop region.
4903 (new_tree_live_info, loe_visit_block): Ditto.
4904 (live_worklist, set_var_live_on_entry): Ditto.
4905 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4906 * tree-ssa-live.h (struct _var_map): New fields.
4907 (init_var_map): Change decl.
4908 (region_contains_p): New.
4910 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4912 * tree-ssa-live.h (live_merge_and_clear): Delete.
4914 2018-05-25 Richard Biener <rguenther@suse.de>
4917 * tree-dump.c (dequeue_and_dump): Remove access to removed
4918 operand 2 of a SWITCH_EXPR.
4920 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4922 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4923 multiplication and division.
4924 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4925 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4926 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4927 (cond_udiv_optab, cond_umod_optab): New optabs.
4928 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4929 (IFN_COND_RDIV): New internal functions.
4930 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4931 TRUNC_MOD_EXPR and RDIV_EXPR.
4932 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4933 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4935 (SVE_INT_BINARY): Include mult.
4936 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4937 (optab, sve_int_op): Handle mult.
4938 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4940 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4941 for SVE_INT_BINARY_SD.
4943 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4945 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4946 (optab, sve_int_op): Handle div and udiv.
4947 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4948 for SVE_INT_BINARY_SD.
4949 (*<optab><mode>3): New insn for the same.
4951 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4953 * tree-vect-patterns.c: Include predict.h.
4954 (vect_recog_divmod_pattern): Restrict check for division support
4955 to when optimizing for size.
4957 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4959 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4960 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4961 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4962 (gimple_match_op::set_op): Likewise.
4963 (gimple_resimplify4): Declare.
4964 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4965 (expr::gen_transform): Likewise.
4966 (decision_tree::gen): Generate a simplification routine for 4 operands.
4967 * gimple-match-head.c (gimple_simplify): Add an overload for
4968 4 operands. In the top-level function, handle up to 4 call
4969 arguments and call gimple_resimplify4.
4970 (gimple_resimplify4): New function.
4971 (build_call_internal): Pass a fourth operand.
4972 (maybe_push_to_seq): Likewise.
4973 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4974 Fold VEC_COND_EXPRs of an operation and a default value into
4975 an IFN_COND_* function if possible.
4976 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4978 (SVE_COND_FP_BINARY): Include them.
4979 (optab, sve_fp_op): Handle them.
4980 (SVE_INT_BINARY_REV): New code iterator.
4981 (SVE_COND_FP_BINARY_REV): New int iterator.
4982 (commutative): New int attribute.
4983 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4985 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4987 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4988 (*cond_<optab><mode>): New patterns for reversed operands.
4990 2018-05-25 Richard Biener <rguenther@suse.de>
4992 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4993 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4994 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4995 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4996 * tree-vect-data-refs.c (everywhere): Adjust users.
4997 * tree-vect-loop.c (everywhere): Likewise.
4998 * tree-vect-slp.c (everywhere): Likewise.
4999 * tree-vect-stmts.c (everywhere): Likewise.
5000 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5002 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5004 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5006 (gcc_cv_as_section_exclude): ... this.
5007 Try Solaris as #exclude syntax.
5008 * configure: Regenerate.
5009 * config.in: Regenerate.
5010 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5012 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5013 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5015 * varasm.c (default_elf_asm_named_section): Don't check if
5016 HAVE_GAS_SECTION_EXCLUDE is defined.
5018 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5020 * doc/md.texi: Update the documentation of the cond_* optabs
5021 to mention the new final operand. Fix GET_MODE_NUNITS call.
5022 Describe the scalar case too.
5023 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5024 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5026 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5027 (get_conditional_internal_fn): Update comment.
5028 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5029 accumulator value as a final argument to conditional functions.
5030 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5031 a define_expand and add an "else" operand. Assert for now that
5032 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5033 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5034 (*cond_<optab><mode>): New patterns.
5035 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5036 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5037 (UNSPEC_COND_EOR): Delete.
5038 (optab): Remove associated mappings.
5039 (SVE_INT_BINARY): New code iterator.
5040 (sve_int_op): Remove int attribute and add "minus" to the code
5042 (SVE_COND_INT_OP): Delete.
5043 (SVE_COND_FP_OP): Rename to...
5044 (SVE_COND_FP_BINARY): ...this.
5046 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5048 * optabs.c (can_reuse_operands_p): New function.
5049 (maybe_legitimize_operands): Try to reuse the results for
5052 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5054 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5055 Add {q} suffix to insn mnemonic.
5057 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5059 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5060 (msp430_warn_func_return): New.
5062 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5064 * fold-const.c (tree_nonzero_bits): New function.
5065 * fold-const.h (tree_nonzero_bits): Likewise.
5066 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5067 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5069 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5073 * varasm.c (assemble_alias): Check ifunc_resolver only on
5076 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5079 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5080 when memory input operand is handled.
5082 2018-05-24 Luis Machado <luis.machado@linaro.org>
5084 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5086 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5088 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5090 * match.pd: Delay FMA folds until after vectorization.
5092 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5095 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5096 address check not strict.
5098 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5100 * gimple-match.h (gimple_match_op): New class.
5101 (mprts_hook): Replace parameters with a gimple_match_op *.
5102 (maybe_build_generic_op): Likewise.
5103 (gimple_simplified_result_is_gimple_val): Replace parameters with
5104 a const gimple_match_op *.
5105 (gimple_simplify): Replace code_helper * and tree * parameters with
5106 a gimple_match_op * parameter.
5107 (gimple_resimplify1): Replace code_helper *, tree and tree *
5108 parameters with a gimple_match_op * parameter.
5109 (gimple_resimplify2): Likewise.
5110 (gimple_resimplify3): Likewise.
5111 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5112 parameters with a gimple_match_op * parameter.
5113 * gimple-match-head.c (gimple_simplify): Change prototypes of
5114 auto-generated functions to take a gimple_match_op * instead of
5115 separate code_helper * and tree * parameters. Make the same
5116 change in the top-level overload and update calls to the
5117 gimple_resimplify routines. Update calls to the auto-generated
5118 functions and to maybe_push_res_to_seq in the publicly-facing
5119 operation-specific gimple_simplify overloads.
5120 (gimple_match_op::MAX_NUM_OPS): Define.
5121 (gimple_resimplify1): Replace rcode and ops with a single res_op
5122 parameter. Update call to gimple_simplify.
5123 (gimple_resimplify2): Likewise.
5124 (gimple_resimplify3): Likewise.
5125 (mprts_hook): Replace parameters with a gimple_match_op *.
5126 (maybe_build_generic_op): Likewise.
5127 (build_call_internal): Replace type, nargs and ops with
5128 a gimple_match_op *.
5129 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5130 with a single gimple_match_op *. Update calls to mprts_hook,
5131 build_call_internal and gimple_simplified_result_is_gimple_val.
5132 Factor out code that is common to the tree_code and combined_fn cases.
5133 * genmatch.c (expr::gen_transform): Replace tem_code and
5134 tem_ops with a gimple_match_op called tem_op. Update calls
5135 to the gimple_resimplify functions and maybe_push_res_to_seq.
5136 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5137 res_ops. Update call to the gimple_resimplify functions.
5138 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5139 (decision_tree::gen): Make the functions take a gimple_match_op *
5140 called res_op instead of separate res_code and res_ops parameters.
5141 Update call accordingly.
5142 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5143 and ops with a single res_op parameter. Update calls to
5144 maybe_build_generic_op and maybe_push_res_to_seq.
5145 (fold_stmt_1): Update calls to gimple_simplify and
5146 replace_stmt_with_simplification.
5147 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5148 and gimple_simplified_result_is_gimple_val.
5149 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5151 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5152 with a gimple_match_op *.
5153 (vn_nary_build_or_lookup): Likewise. Update call to
5154 vn_nary_build_or_lookup_1.
5155 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5156 gimple_match_op *. Update calls to the gimple_resimplify routines
5157 and to gimple_simplified_result_is_gimple_val.
5158 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5159 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5160 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5161 (visit_nary_op): Likewise.
5162 (visit_reference_op_load): Likewise.
5164 2018-05-23 Luis Machado <luis.machado@linaro.org>
5166 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5167 modifier for printing the step amount.
5169 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5172 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5175 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5177 * doc/sourcebuild.texi (Endianness): New subsubsection.
5179 2018-05-23 Luis Machado <luis.machado@linaro.org>
5181 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5182 <prefetch_dynamic_strides>: New const bool field.
5183 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5184 prefetch_dynamic_strides.
5185 (exynosm1_prefetch_tune): Likewise.
5186 (thunderxt88_prefetch_tune): Likewise.
5187 (thunderx_prefetch_tune): Likewise.
5188 (thunderx2t99_prefetch_tune): Likewise.
5189 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5191 (aarch64_override_options_internal): Update to set
5192 PARAM_PREFETCH_DYNAMIC_STRIDES.
5193 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5194 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5195 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5196 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5197 prefetch-dynamic-strides setting.
5199 2018-05-23 Luis Machado <luis.machado@linaro.org>
5201 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5202 <minimum_stride>: New const int field.
5203 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5204 minimum_stride field defaulting to -1.
5205 (exynosm1_prefetch_tune): Likewise.
5206 (thunderxt88_prefetch_tune): Likewise.
5207 (thunderx_prefetch_tune): Likewise.
5208 (thunderx2t99_prefetch_tune): Likewise.
5209 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5210 <default_opt_level>: Set to 3.
5211 (aarch64_override_options_internal): Update to set
5212 PARAM_PREFETCH_MINIMUM_STRIDE.
5213 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5214 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5215 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5216 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5217 stride is constant and is below the minimum stride threshold.
5219 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5221 * config/arm/arm-cpus.in (mode26): Delete.
5222 (armv4): Delete mode26 reference.
5223 * config/arm/arm.c (arm_configure_build_target): Delete use of
5226 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5228 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5230 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5231 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5233 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5234 (floatunsdidf2): Ditto.
5236 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5238 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5239 (fixuns_trunc<mode>si2_avx512f): Ditto.
5240 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5241 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5242 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5244 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5246 PR rtl-optimization/79985
5247 * df-scan.c (df_insn_refs_collect): Remove special case for
5248 global registers and asm statements.
5250 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5252 * extend.texi (Global Register Variables): Rewrite the bullet list.
5253 Note that the register is available for allocation. Note that access
5254 via inline asm must use constraints. Add note about async-signal
5255 handlers. Remove paragraph about automagic register selection.
5257 2018-05-23 Richard Biener <rguenther@suse.de>
5259 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5260 of fixed offset from memset VN.
5262 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5264 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5266 (alloc_cand_and_find_basis): Initialize first_interp field.
5267 (slsr_process_mul): Modify first_interp field.
5268 (slsr_process_add): Likewise.
5269 (slsr_process_cast): Modify first_interp field for each new
5271 (slsr_process_copy): Likewise.
5272 (dump_candidate): Dump first_interp field.
5273 (replace_mult_candidate): Process all interpretations, not just
5275 (replace_rhs_if_not_dup): Likewise.
5276 (replace_one_candidate): Likewise.
5278 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5280 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5282 (aarch64_needs_frame_chain): New function.
5283 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5285 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5288 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5289 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5290 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5291 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5292 as true for strict-align.
5293 (aarch64_can_inline_p): Perform checks even when callee has no
5294 attributes to check for strict alignment.
5295 * doc/extend.texi (AArch64 Function Attributes): Document
5297 * doc/invoke.texi: (AArch64 Options): Likewise.
5299 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5301 PR tree-optimization/85853
5302 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5303 the handling of the root of the node to...
5304 (vect_slp_analyze_node_operations_1): ...this new function,
5305 and run the whole thing with the child nodes' def types
5306 set according to their SLP node's def type.
5308 2018-05-23 Richard Biener <rguenther@suse.de>
5311 * tree-data-ref.c (create_runtime_alias_checks): Defer
5312 and ignore overflow warnings.
5314 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5316 PR tree-optimization/85822
5317 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5320 2018-05-23 Richard Biener <rguenther@suse.de>
5322 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5323 memset constants via native_interpret_expr.
5325 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5328 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5330 (cgraph_node::create_alias): Likewise.
5331 (cgraph_node::get_availability): Check ifunc_resolver instead
5332 of looking up ifunc attribute.
5333 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5334 * varasm.c (do_assemble_alias): Likewise.
5335 (assemble_alias): Likewise.
5336 (default_binds_local_p_3): Likewise.
5337 * cgraph.h (cgraph_node): Add ifunc_resolver.
5338 (cgraph_node::only_called_directly_or_aliased_p): Return false
5340 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5342 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5343 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5344 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5345 instead of looking up ifunc attribute.
5347 2018-05-22 Luis Machado <luis.machado@linaro.org>
5349 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5351 2018-05-22 Martin Sebor <msebor@redhat.com>
5354 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5355 only when expasion succeeds.
5356 (expand_builtin_strcmp): Same.
5357 (expand_builtin_strncmp): Same.
5359 2018-05-22 Martin Sebor <msebor@redhat.com>
5361 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5363 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5364 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5366 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5367 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5369 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5370 Add check for legitimate_address.
5371 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5372 (aarch64_swap_ldrstr_operands): New.
5373 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5376 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5377 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5379 * config/aarch64/aarch64.md: New patterns to generate stp
5381 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5382 single words and double words.
5383 (load_pair_sw, load_pair_dw): Likewise.
5384 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5386 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5388 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5389 for different mode ldpstp and add peephole for merged zero stores.
5391 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5393 (aarch64_gen_store_pair): Rename calls to match new patterns.
5394 (aarch64_gen_load_pair): Rename calls to match new patterns.
5395 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5396 (load_pair<DREG:mode><DREG2:mode>): ... This.
5397 (store_pair<mode>): Rename to...
5398 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5399 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5401 (V_INT_EQUIV): Handle SImode.
5402 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5405 2018-05-22 Martin Sebor <msebor@redhat.com>
5408 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5409 or ajust the presumed bound on an operation to avoid unnecessary
5412 2018-05-22 Martin Sebor <msebor@redhat.com>
5414 PR tree-optimization/85826
5415 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5416 assuming that a DECL necesarily has a constant size.
5418 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5421 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5423 2018-05-22 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/85834
5426 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5427 non-constant and non-zero memset arguments.
5429 2018-05-22 Martin Liska <mliska@suse.cz>
5432 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5434 2018-05-22 Richard Biener <rguenther@suse.de>
5436 PR tree-optimization/85863
5437 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5438 comparisons when vectype is specified.
5439 (vectorizable_condition): Do not specify vectype for
5440 vect_is_simple_cond when SLP vectorizing.
5442 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5445 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5446 define __ibm128 as long double.
5447 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5448 as a distinct type when IEEE 128-bit support is enabled.
5449 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5450 128-bit types to use the correct functions.
5451 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5452 convert between 128-bit floating point types that have different
5453 modes but the same representation, instead of using gen_lowpart to
5455 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5457 (IFKF_reg): New attributes to give the register constraints for
5459 (extend<mode>tf2_internal): New insns to mark an explicit
5460 conversion between 128-bit floating point types that have a
5461 different mode but share the same representation.
5463 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5465 PR tree-optimization/85814
5466 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5467 a null return from get_strinfo when unsharing the next
5468 strinfo in the chain.
5470 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5473 * varasm.c (weak_finish): Clean up weak_decls.
5475 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5477 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5478 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5479 UNSPEC_UADALP values.
5480 * config/aarch64/iterators.md (ABAL): New int iterator.
5483 (sur): Add mappings for the above.
5484 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5486 (aarch64_<sur>abal<mode>_4): Likewise.
5487 (aarch64_<sur>adalp<mode>_3): Likewise.
5488 (<sur>sadv16qi): New define_expand.
5490 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5492 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5493 (*movdf_internal): Ditto.
5494 (*rcpsf2_sse): Ditto.
5495 (*rsqrtsf2_sse): Ditto.
5496 (*sqrt<mode>2_sse): Ditto.
5498 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5500 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5502 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5503 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5504 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5505 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5507 * config/aarch64/arm_neon.h: Likewise.
5508 * config/aarch64/iterators.md (VQ_I): New.
5510 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5512 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5514 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5517 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5519 * config/nds32/constraints.md (S): New constraint.
5520 * config/nds32/nds32.md (call_internal): Use constraint S.
5521 (call_value_internal): Likewise.
5522 (sibcall_internal): Likewise.
5523 (sibcall_value_internal): Likewise.
5525 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5526 Chung-Ju Wu <jasonwucj@gmail.com>
5528 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5531 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5532 Chung-Ju Wu <jasonwucj@gmail.com>
5534 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5535 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5536 (nds32_rtx_costs_impl): Simplify.
5537 (nds32_address_cost_impl): Simplify.
5538 (nds32_init_rtx_costs): New function.
5539 (nds32_rtx_costs_speed_prefer): Likewise.
5540 (nds32_rtx_costs_size_prefer): Likewise.
5541 (nds32_address_cost_speed_prefer): Likewise.
5542 (nds32_address_cost_speed_fwprop): Likewise.
5543 (nds32_address_cost_size_prefer): Likewise.
5544 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5545 * config/nds32/nds32.c (nds32_option_override): Use
5546 nds32_init_rtx_costs function.
5548 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5550 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5551 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5552 (TARGET_PIPELINE_N8): Likewise.
5553 (TARGET_PIPELINE_N10): Likewise.
5554 (TARGET_PIPELINE_N13): Likewise.
5555 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5557 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5559 * config/nds32/nds32-fpu.md: Update copyright year.
5561 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5563 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5565 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5567 * config/nds32/nds32.c
5568 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5569 * config/nds32/nds32.opt (minline-asm-r15): New option.
5571 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5573 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5575 * config/nds32/nds32.md (abssi2): New pattern.
5577 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5579 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5580 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5581 Merge insn pattern from sse_cvtsi2ss<round_name> and
5582 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5583 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5584 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5585 using SWI48 mode iterator.
5586 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5587 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5588 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5589 pattern from sse_cvttss2si<round_saeonly_name>
5590 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5591 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5592 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5593 using SWI48 mode iterator.
5594 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5595 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5596 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5597 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5598 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5599 using SWI48 mode iterator.
5600 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5601 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5602 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5603 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5604 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5605 SWI48 mode iterator.
5606 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5607 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5608 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5609 pattern from sse_cvttsd2si<round_saeonly_name>
5610 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5612 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5614 * config/nds32/nds32-md-auxiliary.c
5615 (nds32_valid_smw_lwm_base_p): Refine.
5616 (nds32_output_smw_single_word): Refine.
5617 (nds32_output_smw_double_word): New.
5618 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5620 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5622 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5623 (nds32_output_stack_pop): Refine.
5624 (nds32_expand_unaligned_load): Refine.
5625 (nds32_expand_unaligned_store): Refine.
5627 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5628 Chung-Ju Wu <jasonwucj@gmail.com>
5630 * config/nds32/constants.md: Add TP_REGNUM constant.
5631 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5632 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5634 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5635 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5636 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5637 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5638 and PIC code generation.
5639 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5641 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5643 * config/nds32/nds32.md: Support TLS and PIC.
5644 * config/nds32/nds32.c: Support TLS and PIC.
5645 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5646 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5649 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5651 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5652 mode with E_ prefix.
5654 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5655 Chung-Ju Wu <jasonwucj@gmail.com>
5657 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5658 * config/nds32/nds32-md-auxiliary.c
5659 (symbolic_reference_mentioned_p): New.
5660 (nds32_legitimize_ict_address): New.
5661 (nds32_expand_ict_move): New.
5662 (nds32_indirect_call_referenced_p): New.
5663 (nds32_symbol_binds_local_p): Delete.
5664 (nds32_long_call_p): Modify.
5665 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5666 * config/nds32/nds32-protos.h
5667 (symbolic_reference_mentioned_p): Declare.
5668 (nds32_legitimize_ict_address): Declare.
5669 (nds32_expand_ict_move): Declare.
5670 (nds32_indirect_call_referenced_p): Declare.
5671 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5672 (nds32_relax_group): Use nds32_ict_const_p as condition.
5673 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5674 (nds32_asm_file_start): Output ict_model directive in asm code.
5675 (nds32_legitimate_address_p): Consider indirect call.
5676 (nds32_print_operand): Consider indirect call.
5677 (nds32_print_operand_address): Consider indirect call.
5678 (nds32_insert_attributes): Handle "indirect_call" attribute.
5679 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5680 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5681 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5682 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5683 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5684 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5685 (TARGET_ICT_MODEL_SMALL): Define.
5686 (TARGET_ICT_MODEL_LARGE): Define.
5687 * config/nds32/nds32.md (movsi): Consider ict model.
5688 (call, call_value): Consider ict model.
5689 (sibcall, sibcall_value): Consider ict model.
5690 * config/nds32/nds32.opt (mict-model): New option.
5691 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5694 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5695 Monk Chiang <sh.chiang04@gmail.com>
5696 Jim Wilson <jimw@sifive.com>
5698 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5699 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5700 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5701 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5702 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5703 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5704 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5705 compute save_libcall_adjustment properly.
5706 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5707 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5708 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5709 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5710 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5711 (ABI_SPEC): Handle mabi=ilp32e.
5712 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5713 (RVE): Add RVE mask.
5714 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5715 <-march>: Add rv32e as an example.
5717 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5720 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5721 (intra_create_variable_infos): Handle C++ constructors.
5723 2018-05-18 Martin Liska <mliska@suse.cz>
5725 * passes.def: Remove a redundant pass.
5727 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5730 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5732 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5734 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5736 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5738 (armv2, armv2a, armv3, armv3m): Delete architectures.
5739 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5740 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5741 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5743 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5744 (*mulsidi3adddi): Likewise.
5745 (mulsidi3): Likewise.
5746 (*mulsidi3_nov6): Likewise.
5747 (umulsidi3): Likewise.
5748 (umulsidi3_nov6): Likewise.
5749 (umaddsidi4): Likewise.
5750 (*umulsidi3adddi): Likewise.
5751 (smulsi3_highpart): Likewise.
5752 (*smulsi3_highpart_nov6): Likewise.
5753 (umulsi3_highpart): Likewise.
5754 (*umulsi3_highpart_nov6): Likewise.
5755 * config/arm/arm.h (arm_arch3m): Delete.
5756 * config/arm/arm.c (arm_arch3m): Delete.
5757 (arm_option_override_internal): Update armv3-related comment.
5758 (arm_configure_build_target): Delete use of isa_bit_mode32.
5759 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5760 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5761 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5763 (mulusa3): Likewise.
5764 * config/arm/arm-protos.h (arm_arch3m): Delete.
5765 * config/arm/arm-tables.opt: Regenerate.
5766 * config/arm/arm-tune.md: Likewise.
5767 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5768 deleted architectures.
5770 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5772 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5773 (armv5t, armv5te): New features.
5774 (ARMv5, ARMv5e): Delete fgroups.
5775 (ARMv5t, ARMv5te): Adjust for above changes.
5777 (armv5, armv5e): Delete arches.
5778 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5780 (*call_reg_arm): Likewise.
5781 (*call_value_reg_armv5): Likewise.
5782 (*call_value_reg_arm): Likewise.
5783 (*call_symbol): Likewise.
5784 (*call_value_symbol): Likewise.
5785 (*sibcall_insn): Likewise.
5786 (*sibcall_value_insn): Likewise.
5788 (prefetch): Likewise.
5789 (define_split and define_peephole2 dependent on arm_arch5):
5791 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5793 (TARGET_ARM_QBIT): Likewise.
5794 (TARGET_DSP_MULTIPLY): Likewise.
5795 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5796 (arm_arch5, arm_arch5e): Delete.
5797 (arm_arch5t, arm_arch5te): Declare.
5798 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5799 (arm_arch5t): Declare.
5800 (arm_option_reconfigure_globals): Update for the above.
5801 (arm_options_perform_arch_sanity_checks): Update comment, replace
5802 use of arm_arch5 with arm_arch5t.
5803 (use_return_insn): Likewise.
5804 (arm_emit_call_insn): Likewise.
5805 (output_return_instruction): Likewise.
5806 (arm_final_prescan_insn): Likewise.
5807 (arm_coproc_builtin_available): Likewise.
5808 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5809 arm_arch5e with arm_arch5t and arm_arch5te.
5810 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5811 (arm_arch5t, arm_arch5te): Declare.
5812 * config/arm/arm-tables.opt: Regenerate.
5813 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5814 * config/arm/t-multilib: Likewise.
5815 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5816 instead of arm_arch5.
5817 (*call_reg_thumb1): Likewise.
5818 (*call_value_reg_thumb1_v5): Likewise.
5819 (*call_value_reg_thumb1): Likewise.
5820 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5822 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5824 2018-05-18 Martin Liska <mliska@suse.cz>
5826 PR gcov-profile/84846
5827 * doc/gcov.texi: Document -t option of gcov tool.
5829 2018-05-18 Martin Liska <mliska@suse.cz>
5831 PR gcov-profile/84846
5832 * gcov.c (print_usage): Add new -t option.
5833 (process_args): Handle the option.
5834 (generate_results): Use stdout as output when requested by
5837 2018-05-18 Martin Liska <mliska@suse.cz>
5839 PR gcov-profile/84846
5840 * coverage.c (coverage_init): Write PWD to .gcno file.
5841 * doc/gcov.texi: Document how working directory is printed.
5842 * gcov-dump.c (dump_gcov_file): Print PWD.
5843 * gcov.c (output_intermediate_file): Likewise.
5844 (read_graph_file): Read PWD string.
5845 (output_lines): Print PWD.
5847 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5850 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5851 for retval and return false if all args to phi are zero.
5853 2018-05-18 Richard Biener <rguenther@suse.de>
5855 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5857 (evrp_dom_walker::before_dom_children): Call it.
5859 2018-05-18 Richard Biener <rguenther@suse.de>
5861 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5862 results when processing array refs with variable index.
5864 2018-05-18 Toon Moene <toon@moene.org>
5866 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5867 directly after that of -floop-interchange. Indicate that both
5868 options are enabled by default when specifying -O3.
5870 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5872 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5873 iterator. Delete separate integer-mode vec_set<mode> expander.
5874 (aarch64_simd_vec_setv2di): Delete.
5875 (vec_setv2di): Delete.
5876 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5877 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5878 the "w, r" alternative.
5880 2018-05-18 Martin Liska <mliska@suse.cz>
5882 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5883 * tree-pass.h (make_pass_lower_switch_O0): New function.
5884 * tree-switch-conversion.c (node_has_low_bound): Remove.
5885 (node_has_high_bound): Likewise.
5886 (node_is_bounded): Likewise.
5887 (class pass_lower_switch): Make it a template type and create
5889 (pass_lower_switch::execute): Add template argument.
5890 (make_pass_lower_switch): New function.
5891 (make_pass_lower_switch_O0): New function.
5892 (do_jump_if_equal): Remove.
5893 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5894 all the hard work to tree optimization passes.
5896 2018-05-18 Martin Liska <mliska@suse.cz>
5898 * dbgcnt.c (limit_low): Renamed from limit.
5899 (limit_high): New variable.
5900 (dbg_cnt_is_enabled): Check for upper limit.
5901 (dbg_cnt): Adjust dumping.
5902 (dbg_cnt_set_limit_by_index): Add new argument for high
5904 (dbg_cnt_set_limit_by_name): Likewise.
5905 (dbg_cnt_process_single_pair): Parse new format.
5906 (dbg_cnt_process_opt): Use strtok.
5907 (dbg_cnt_list_all_counters): Remove 'value' and add
5909 * doc/invoke.texi: Document changes.
5911 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5913 * doc/sourcebuild.texi (scalar_all_fma): Document.
5914 * tree.def (FMA_EXPR): Delete.
5915 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5916 * internal-fn.c (ternary_direct): New macro.
5917 (expand_ternary_optab_fn): Likewise.
5918 (direct_ternary_optab_supported_p): Likewise.
5919 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5920 * builtins.c (fold_builtin_fma): Delete.
5921 (fold_builtin_3): Don't call it.
5922 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5923 * expr.c (expand_expr_real_2): Likewise.
5924 * fold-const.c (operand_equal_p): Likewise.
5925 (fold_ternary_loc): Likewise.
5926 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5927 * gimple.c (DEFTREECODE): Likewise.
5928 * gimplify.c (gimplify_expr): Likewise.
5929 * optabs-tree.c (optab_for_tree_code): Likewise.
5930 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5931 * tree-eh.c (operation_could_trap_p): Likewise.
5932 (stmt_could_throw_1_p): Likewise.
5933 * tree-inline.c (estimate_operator_cost): Likewise.
5934 * tree-pretty-print.c (dump_generic_node): Likewise.
5935 (op_code_prio): Likewise.
5936 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5937 * tree-ssa-operands.c (get_expr_operands): Likewise.
5938 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5939 * fold-const-call.h (fold_fma): Delete.
5940 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5941 CFN_FNMA and CFN_FNMS.
5943 * genmatch.c (combined_fn): New enum.
5944 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5945 (commutative_op): New function.
5946 (commutate): Use it. Handle more than 2 operands.
5947 (dt_operand::gen_gimple_expr): Use commutative_op.
5948 (parser::parse_expr): Allow :c to be used with non-binary
5949 operators if the commutative operand is known.
5950 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5951 CFN_FMS, CFN_FNMA and CFN_FNMS.
5952 (backprop::process_assign_use): Remove FMA_EXPR handling.
5953 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5954 (gen_hsa_fma): New function.
5955 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5956 IFN_FNMA and IFN_FNMS.
5957 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5958 * gimple-fold.h (follow_all_ssa_edges): Declare.
5959 * gimple-fold.c (follow_all_ssa_edges): New function.
5960 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5961 gimple_build interface and use follow_all_ssa_edges to fold the result.
5962 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5963 instead of checking for optabs directly.
5964 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5965 rather than FMA_EXPRs.
5966 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5967 call to IFN_FMA instead of an FMA_EXPR.
5969 2018-05-17 Jim Wilson <jimw@sifive.com>
5971 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5972 sign extended promoted subreg, and the range does not have the sign bit
5973 set, then do a sign extend.
5975 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5976 test, check for sign extended subreg and/or constant operands, and
5977 do a sign extend in that case.
5979 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5981 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5982 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5984 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5985 Change logics_shift_reg to logics_shift_imm.
5986 (thunderx2t99_fp_loadpair_basic): Delete.
5987 (thunderx2t99_fp_storepair_basic): Delete.
5988 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5989 (thunderx2t99_asimd_polynomial): Delete.
5990 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5991 and neon_fp_mul_d_scalar_q.
5992 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5993 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5994 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5995 (thunderx2t99_asimd_lut): Add missing tbl types.
5996 (thunderx2t99_asimd_ext): Delete.
5997 (thunderx2t99_asimd_load1_1_mult): Delete.
5998 (thunderx2t99_asimd_load1_2_mult): Delete.
5999 (thunderx2t99_asimd_load1_ldp): New.
6000 (thunderx2t99_asimd_load1): New.
6001 (thunderx2t99_asimd_load2): Add missing *load2* types.
6002 (thunderx2t99_asimd_load3): New.
6003 (thunderx2t99_asimd_load4): New.
6004 (thunderx2t99_asimd_store1_1_mult): Delete.
6005 (thunderx2t99_asimd_store1_2_mult): Delete.
6006 (thunderx2t99_asimd_store2_mult): Delete.
6007 (thunderx2t99_asimd_store2_onelane): Delete.
6008 (thunderx2t99_asimd_store_stp): New.
6009 (thunderx2t99_asimd_store1): New.
6010 (thunderx2t99_asimd_store2): New.
6011 (thunderx2t99_asimd_store3): New.
6012 (thunderx2t99_asimd_store4): New.
6014 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6016 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6017 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6019 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6020 Segher Boessenkool <segher@kernel.crashing.org>
6023 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6026 2018-05-17 Richard Biener <rguenther@suse.de>
6028 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6029 for pruning loop and prune defs feeding only already visited PHIs.
6031 2018-05-17 Richard Biener <rguenther@suse.de>
6033 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6035 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6036 Richard Biener <rguenther@suse.de>
6038 PR tree-optimization/85793
6039 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6040 for VMAT_ELEMENTWISE.
6042 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6044 * internal-fn.h (lookup_internal_fn): Declare
6045 * internal-fn.c (lookup_internal_fn): New function.
6046 * gimple.c (gimple_build_call_from_tree): Handle calls to
6048 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6049 internal function names.
6050 * tree-pretty-print.c (dump_generic_node): Likewise.
6051 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6053 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6055 * gimple-fold.h (gimple_build): Make the function forms take
6056 combined_fn rather than built_in_function.
6057 (gimple_simplify): Likewise.
6058 * gimple-match-head.c (gimple_simplify): Likewise.
6059 * gimple-fold.c (gimple_build): Likewise.
6060 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6061 rather than gimple_build_call_internal.
6062 (get_initial_defs_for_reduction): Likewise.
6063 (vect_create_epilog_for_reduction): Likewise.
6064 (vectorizable_live_operation): Likewise.
6066 2018-05-17 Martin Liska <mliska@suse.cz>
6068 * gimple-ssa-sprintf.c (format_directive): Do not use
6069 space in between 'G_' and '('.
6071 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6074 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6075 even if the mask is not all ones.
6078 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6080 (ix86_gimple_fold_builtin): Likewise.
6083 * config/i386/i386.c: Include tree-vector-builder.h.
6084 (ix86_vector_shift_count): New function.
6085 (ix86_fold_builtin): Fold shift builtins by scalar count.
6086 (ix86_gimple_fold_builtin): Likewise.
6088 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6089 _mm512_setzero): New intrinsics.
6091 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6092 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6094 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6095 code generation for cases where splatting a value is not useful.
6096 * simplify-rtx.c (simplify_ternary_operation): Simplify
6097 vec_merge across a vec_duplicate and a paradoxical subreg forming
6098 a vector mode to a vec_concat.
6100 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6102 * config.gcc: Support "goldmont-plus".
6103 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6105 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6106 PROCESSOR_GOLDMONT_PLUS.
6107 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6108 (processor_target_table): Add "goldmont-plus".
6109 (PTA_GOLDMONT_PLUS): Define.
6110 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6111 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6112 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6113 (fold_builtin_cpu): Add "goldmont-plus".
6114 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6115 (ix86_option_override_internal): Add "goldmont-plus".
6116 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6117 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6118 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6119 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6121 2018-05-17 Richard Biener <rguenther@suse.de>
6123 PR tree-optimization/85757
6124 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6125 remove defs that only feed that PHI from further processing.
6127 2018-05-16 Jim Wilson <jimw@sifive.com>
6129 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6131 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6133 2018-05-16 Mark Wielaard <mark@klomp.org>
6135 * dwarf2out.c (count_index_strings): New function.
6136 (output_indirect_strings): Call count_index_strings and generate
6137 header for dwarf_version >= 5.
6139 2018-05-16 Mark Wielaard <mark@klomp.org>
6141 * dwarf2out.c (dwarf_FORM): New function.
6142 (set_indirect_string): Use dwarf_FORM.
6143 (reset_indirect_string): Likewise.
6144 (size_of_die): Likewise.
6145 (value_format): Likewise.
6146 (output_die): Likewise.
6147 (add_skeleton_AT_string): Likewise.
6148 (output_macinfo_op): Likewise.
6149 (index_string): Likewise.
6150 (output_index_string_offset): Likewise.
6151 (output_index_string): Likewise.
6152 (count_index_strings): Likewise.
6154 2018-05-16 Carl Love <cel@us.ibm.com>
6156 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6157 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6159 2018-05-16 Martin Jambor <mjambor@suse.cz>
6161 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6162 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6164 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6166 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6167 (fnma<mode>4): Likewise.
6168 (fms<mode>4): Likewise.
6169 (fnms<mode>4): Likewise.
6170 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6171 (aarch64_fnma<mode>4): Likewise.
6172 (aarch64_fms<mode>4): Likewise.
6173 (aarch64_fnms<mode>4): Likewise.
6174 (aarch64_fnmadd<mode>4): Likewise.
6176 2018-05-16 Jason Merrill <jason@redhat.com>
6178 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6180 2018-05-16 Richard Biener <rguenther@suse.de>
6182 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6183 (dump_stmt_cost): Declare.
6184 (add_stmt_cost): Dump cost we add.
6185 (add_stmt_costs): New function.
6186 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6188 (vect_analyze_stmt): Adjust prototype.
6189 (vectorizable_condition): Likewise.
6190 (vectorizable_live_operation): Likewise.
6191 (vectorizable_reduction): Likewise.
6192 (vectorizable_induction): Likewise.
6193 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6194 cost vector to pass to vectorizable_ and record afterwards.
6195 (vect_model_reduction_cost): Take cost vector argument and adjust.
6196 (vect_model_induction_cost): Likewise.
6197 (vectorizable_reduction): Likewise.
6198 (vectorizable_induction): Likewise.
6199 (vectorizable_live_operation): Likewise.
6200 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6201 SLP_TREE_NUMBER_OF_VEC_STMTS.
6202 (vect_analyze_slp_cost_1): Remove.
6203 (vect_analyze_slp_cost): Likewise.
6204 (vect_slp_analyze_node_operations): Take visited args and
6205 a target cost vector. Avoid processing already visited stmt sets.
6206 (vect_slp_analyze_operations): Use a local cost vector to gather
6207 costs and register those of non-discarded instances.
6208 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6209 (vect_schedule_slp_instance): Remove copying of
6210 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6212 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6213 adding cost. Record cost entry location.
6214 (vect_prologue_cost_for_slp_op): Function to compute cost of
6215 a constant or invariant generated for SLP vect in the prologue,
6216 split out from vect_analyze_slp_cost_1.
6217 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6218 (vect_model_promotion_demotion_cost): Likewise.
6219 (vect_model_store_cost): Likewise, make static.
6220 (vect_model_load_cost): Likewise.
6221 (vectorizable_bswap): Add cost vector arg and adjust.
6222 (vectorizable_call): Likewise.
6223 (vectorizable_simd_clone_call): Likewise.
6224 (vectorizable_conversion): Likewise.
6225 (vectorizable_assignment): Likewise.
6226 (vectorizable_shift): Likewise.
6227 (vectorizable_operation): Likewise.
6228 (vectorizable_store): Likewise.
6229 (vectorizable_load): Likewise.
6230 (vectorizable_condition): Likewise.
6231 (vectorizable_comparison): Likewise.
6232 (can_vectorize_live_stmts): Likewise.
6233 (vect_analyze_stmt): Likewise.
6234 (vect_transform_stmt): Adjust calls to vectorizable_*.
6235 * tree-vectorizer.c: Include gimple-pretty-print.h.
6236 (dump_stmt_cost): New function.
6238 2018-05-16 Richard Biener <rguenther@suse.de>
6240 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6241 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6242 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6243 (check_name): New callback.
6244 (dse_classify_store): Track cycles via a visited bitmap of PHI
6245 defs and simplify handling of in-loop and across loop dead stores
6246 and properly fail for loop-variant refs. Handle byte-tracking with
6247 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6250 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6252 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6253 (vect_get_mask_type_for_stmt): Likewise.
6254 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6256 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6257 to determine the statement's vector type and the vector type that
6258 should be used for calculating nunits. Deal with cases in which
6259 the type has to be deferred.
6260 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6261 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6262 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6263 (vect_determine_vf_for_stmt): New functions, split out from...
6264 (vect_determine_vectorization_factor): ...here.
6265 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6266 (vect_get_mask_type_for_stmt): New functions, split out from
6267 vect_determine_vectorization_factor.
6269 2018-05-16 Richard Biener <rguenther@suse.de>
6271 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6272 verify the [VEC_]COND_EXPR embedded comparison.
6274 2018-05-15 Martin Sebor <msebor@redhat.com>
6276 PR tree-optimization/85753
6277 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6278 RECORD_TYPE in addition to ARRAY_TYPE.
6280 2018-05-15 Martin Sebor <msebor@redhat.com>
6283 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6285 2018-05-15 Richard Biener <rguenther@suse.de>
6287 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6288 add by_clobber_p one. Change algorithm to collect all defs
6289 representing uses we need to walk and try reducing them to
6290 a single one before failing.
6291 (dse_dom_walker::dse_optimize_stmt): Adjust.
6293 2018-05-13 Mark Wielaard <mark@klomp.org>
6295 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6296 (size_of_loc_descr): Likewise.
6297 (output_loc_operands): Likewise.
6298 (output_loc_operands_raw): Likewise.
6299 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6300 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6301 (hash_loc_operands): Likewise.
6302 (compare_loc_operands): Likewise.
6304 2018-05-14 Mark Wielaard <mark@klomp.org>
6306 * dwarf2out.c (count_index_addrs): New function.
6307 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6309 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6311 PR tree-optimization/83648
6312 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6313 return value as malloc candidate.
6315 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6318 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6319 param as true in call to suggest_attribute.
6321 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6323 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6324 -mreadonly-in-sdata.
6326 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6328 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6330 (aarch64_crypto_aesd_fused): Likewise.
6332 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6334 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6335 (movsi_aarch64): Likewise.
6336 (load_pairsi): Likewise.
6337 (load_pairdi): Likewise.
6338 (store_pairsi): Likewise.
6339 (store_pairdi): Likewise.
6340 (load_pairsf): Likewise.
6341 (load_pairdf): Likewise.
6342 (store_pairsf): Likewise.
6343 (store_pairdf): Likewise.
6344 (zero_extend): Likewise.
6345 (trunc): Swap alternatives.
6346 (fcvt_target): Add '?' to prefer w over r.
6348 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6351 * config/i386/i386.md: Disallow non-commutative arithmetics in
6352 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6353 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6354 in the peephole2 before it.
6356 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6358 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6359 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6360 (ix86_handle_option): Handle -mcldemote.
6361 * config.gcc: New header.
6362 * config/i386/cldemoteintrin.h: New file.
6363 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6364 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6366 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6367 OPTION_MASK_ISA_CLDEMOTE.
6368 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6369 (ix86_valid_target_attribute_inner_p): Ditto.
6370 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6371 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6372 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6373 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6374 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6376 * config/i386/i386.opt: Add -mcldemote.
6377 * config/i386/x86intrin.h: New header.
6378 * doc/invoke.texi: Add -mcldemote.
6380 2018-05-14 Richard Biener <rguenther@suse.de>
6382 * doc/match-and-simplify.texi: Adjust :s documentation.
6384 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6386 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6387 intended memcpy size.
6388 (REORDER_45): Likewise.
6390 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6392 * sort.cc: New file.
6393 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6394 * vec.c (qsort_chk): Use gcc_qsort.
6395 * Makefile.in (OBJS-libcommon): Add sort.o.
6396 (build/sort.o): New target. Use it...
6397 (BUILD_RTL): ... here, and...
6398 (build/gencfn-macros): ... here, and...
6399 (build/genmatch): ... here.
6401 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6402 Chung-Ju Wu <jasonwucj@gmail.com>
6404 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6405 * config/nds32/nds32-graywolf.md: New file.
6406 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6407 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6409 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6410 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6411 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6412 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6413 * config/nds32/pipelines.md: Include n15 settings.
6415 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6416 Chung-Ju Wu <jasonwucj@gmail.com>
6418 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6419 * config/nds32/nds32-n13.md: New file.
6420 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6421 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6423 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6424 * config/nds32/nds32.md (pipeline_model): Add n13.
6425 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6426 * config/nds32/pipelines.md: Include n13 settings.
6428 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6429 Chung-Ju Wu <jasonwucj@gmail.com>
6431 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6432 * config/nds32/nds32-n10.md: New file.
6433 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6434 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6436 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6437 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6438 * config/nds32/nds32.md (pipeline_model): Add n10.
6439 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6440 * config/nds32/pipelines.md: Include n10 settings.
6442 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6443 Kito Cheng <kito.cheng@gmail.com>
6444 Chung-Ju Wu <jasonwucj@gmail.com>
6446 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6447 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6448 Add enum values for DSP extension instructions.
6449 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6451 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6452 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6454 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6455 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6456 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6457 * config/nds32/nds32-intrinsic.md: Likewise.
6458 * config/nds32/nds32_intrinsic.h: Likewise.
6459 * config/nds32/nds32-md-auxiliary.c: Likewise.
6460 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6461 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6462 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6463 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6464 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6465 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6466 TYPE_DMAC in switch statement.
6467 * config/nds32/nds32.c: New checking and implementation for DSP
6468 extension instructions.
6469 * config/nds32/nds32.h: Likewise.
6470 * config/nds32/nds32.md: Likewise.
6471 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6472 * config/nds32/predicates.md: Implement new predicates for DSP
6475 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6477 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6478 Reformat alternatives and attributes so it is easier to identify
6479 which constraints/attributes go with which instruction.
6480 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6481 (mov<mode>_softfloat32, FMOVE64): Likewise.
6482 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6483 (mov<mode>_softfloat64, FMOVE64): Likewise.
6485 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6487 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6489 (Basic PowerPC Built-in Functions): The new name of the
6490 subsection previously known as "PowerPC Built-in Functions".
6491 (Basic PowerPC Built-in Functions Available on all Configurations):
6493 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6494 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6495 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6496 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6498 2018-05-11 Martin Jambor <mjambor@suse.cz>
6501 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6504 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6507 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6509 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6511 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6512 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6513 (ix86_handle_option): Handle -mwaitpkg.
6514 * config.gcc: New header.
6515 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6516 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6517 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6519 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6520 OPTION_MASK_ISA_WAITPKG.
6521 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6522 (ix86_option_override_internal): Add PTA_WAITPKG.
6523 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6524 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6525 IX86_BUILTIN_TPAUSE.
6526 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6527 __builtin_ia32_umwait and __builtin_ia32_tpause.
6528 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6529 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6530 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6531 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6532 UNSPECV_TPAUSE): New.
6533 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6534 * config/i386/i386.opt: Add -mwaitpkg.
6535 * config/i386/waitpkgintrin.h: New file.
6536 * config/i386/x86intrin.h: New header.
6537 * doc/invoke.texi: Add -mwaitpkg.
6539 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6542 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6544 (cortex-m0): Use armv6s-m isa.
6545 (cortex-m0plus): Likewise.
6546 (cortex-m1): Likewise.
6547 (cortex-m0.small-multiply): Likewise.
6548 (cortex-m0plus.small-multiply): Likewise.
6549 (cortex-m1.small-multiply): Likewise.
6551 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6552 Jakub Jelinek <jakub@redhat.com>
6554 PR tree-optimization/85692
6555 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6556 source permute as well.
6558 2018-05-11 Martin Liska <mliska@suse.cz>
6561 * doc/extend.texi: Document LLVM style format for no_sanitize
6564 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6566 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6567 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6568 (mode_supports_vsx_dform_quad): Likewise.
6569 (mode_supports_vmx_dform): Move these functions to be next to the
6570 other mode_supports functions.
6571 (mode_supports_dq_form): Likewise.
6572 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6573 mode_supports_dq_form.
6574 (reg_offset_addressing_ok_p): Likewise.
6575 (offsettable_ok_by_alignment): Likewise.
6576 (rs6000_legitimate_offset_address_p): Likewise.
6577 (legitimate_lo_sum_address_p): Likewise.
6578 (rs6000_legitimize_address): Likewise.
6579 (rs6000_legitimize_reload_address): Likewise.
6580 (rs6000_secondary_reload_inner): Likewise.
6581 (rs6000_preferred_reload_class): Likewise.
6582 (rs6000_output_move_128bit): Likewise.
6584 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6586 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6587 Generate SImode target register for null target.
6588 <case IX86_BUILTIN_XGETBV>: Ditto.
6589 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6590 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6592 2018-05-10 Carl Love <cel@us.ibm.com>
6594 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6595 dcbtt and dcbtstt if operands[2] is 0.
6597 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6600 * config/i386/sse.md (usadv64qi): New expander.
6602 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6604 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6605 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6606 -maltivec=be support.
6607 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6608 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6609 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6610 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6611 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6612 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6613 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6614 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6615 altivec_vsumsws): Adjust.
6616 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6617 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6618 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6620 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6621 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6622 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6623 (altivec_lve<VI_char>x): Delete expand.
6624 (*altivec_lve<VI_char>x_internal): Rename to...
6625 (altivec_lve<VI_char>x): ... this.
6626 (altivec_lvxl_<mode>): Delete expand.
6627 (*altivec_lvxl_<mode>_internal): Rename to ...
6628 (altivec_lvxl_<mode>): ... this.
6629 (altivec_stvxl_<mode>): Delete expand.
6630 (*altivec_stvxl_<mode>_internal): Rename to ...
6631 (altivec_stvxl_<mode>): ... this.
6632 (altivec_stve<VI_char>x): Delete expand.
6633 (*altivec_stve<VI_char>x_internal): Rename to ...
6634 (altivec_stve<VI_char>x): ... this.
6635 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6636 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6637 reduc_plus_scal_<mode>): Adjust.
6638 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6640 (rs6000_cpu_cpp_builtins): Adjust.
6641 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6642 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6643 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6644 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6645 -maltivec=be support.
6646 (rs6000_split_vec_extract_var): Adjust.
6647 (rs6000_split_v4si_init): Adjust.
6648 (swap_selector_for_mode): Delete.
6649 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6650 altivec_expand_stvex_be): Delete.
6651 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6652 -maltivec=be support.
6653 (rs6000_gimple_fold_builtin): Ditto.
6654 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6656 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6657 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6658 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6659 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6660 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6661 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6662 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6663 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6664 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6665 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6666 anonymous split): Adjust.
6667 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6668 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6670 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6672 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6673 when --with-gxx-include-dir is also specified.
6674 * configure: Regenerate.
6676 2018-05-09 Jim Wilson <jimw@sifive.com>
6679 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6680 * config/riscv/t-withmultilib: New.
6681 * config/riscv/withmultilib.h: New.
6682 * doc/install.texi: Document RISC-V --with-multilib-list support.
6684 2018-05-09 Richard Biener <rguenther@suse.de>
6686 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6688 (vect_bb_vectorization_profitable_p): Adjust. Compute
6689 actual scalar cost using the cost vector and the add_stmt_cost
6692 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6694 PR rtl-optimization/85645
6695 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6696 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6698 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6700 PR rtl-optimization/85645
6701 * shrink-wrap.c (spread_components): Return a boolean saying if
6702 anything was changed.
6703 (try_shrink_wrapping_separate): Iterate spread_components until
6704 nothing changes anymore.
6706 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6708 PR rtl-optimization/85645
6709 * regrename.c (build_def_use): Also kill the chains that include the
6710 destination of a REG_CFA_REGISTER note.
6712 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6714 PR rtl-optimization/85645
6715 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6716 insn that has a REG_CFA_REGISTER note.
6718 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6720 * cfgexpand.c (expand_clobber): New function.
6721 (expand_gimple_stmt_1): Use it.
6722 * tree-vect-stmts.c (vect_clobber_variable): New function,
6724 (vectorizable_simd_clone_call): ...here.
6725 (vectorizable_store): Emit a clobber either side of an
6726 IFN_STORE_LANES sequence.
6727 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6729 2018-05-09 Tom de Vries <tom@codesourcery.com>
6732 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6733 (define_insn "trap_if_false"): Add exit after trap.
6735 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6737 PR rtl-optimization/85638
6738 * bb-reorder.c: Include common/common-target.h.
6739 (create_forwarder_block): New function extracted from...
6740 (fix_up_crossing_landing_pad): ...here. Rename into...
6741 (dw2_fix_up_crossing_landing_pad): ...this.
6742 (sjlj_fix_up_crossing_landing_pad): New function.
6743 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6744 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6745 from both partitions and exit the loop after one iteration.
6747 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6750 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6752 (Basic PowerPC Built-in Functions): The new name of the
6753 subsection previously known as "PowerPC Built-in Functions".
6754 (Basic PowerPC Built-in Functions Available on all Configurations):
6756 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6758 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6760 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6762 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6765 2018-05-08 Jim Wilson <jimw@sifive.com>
6767 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6768 (LD_EMUL_SUFFIX): New.
6769 (LINK_SPEC): Use it.
6771 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6773 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6775 (Basic PowerPC Built-in Functions): The new name of the
6776 subsection previously known as "PowerPC Built-in Functions".
6777 (Basic PowerPC Built-in Functions Available on all Configurations):
6779 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6781 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6783 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6785 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6788 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6791 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6792 after cmpelim optimization.
6794 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6796 * config.gcc: Support "goldmont".
6797 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6798 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6800 * config/i386/i386.c (m_GOLDMONT): Define.
6801 (processor_target_table): Add "goldmont".
6802 (PTA_GOLDMONT): Define.
6803 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6804 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6805 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6806 (fold_builtin_cpu): Add "goldmont".
6807 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6808 (ix86_option_override_internal): Add "goldmont".
6809 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6810 (processor_type): Add PROCESSOR_GOLDMONT.
6811 * config/i386/i386.md: Add CPU "glm".
6812 * config/i386/glm.md: New file.
6813 * config/i386/x86-tune.def: Add m_GOLDMONT.
6814 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6816 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6819 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6821 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6822 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6823 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6826 * config/i386/i386.c (ix86_fold_builtin): Handle
6827 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6830 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6831 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6833 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6836 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6837 (check_arch): Likewise.
6838 (check_fpu): Return the result rather than printing it.
6839 (end arch): Fix operator precedence.
6840 (end cpu): Likewise.
6841 (END): Print the result from check_fpu.
6843 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6844 Alan Hayward <alan.hayward@arm.com>
6845 David Sherwood <david.sherwood@arm.com>
6847 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6848 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6849 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6850 (*fcmuo<mode>_and): New patterns.
6852 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6854 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6855 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6856 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6857 (cmp_op, sve_imm_con): New code attributes.
6858 (SVE_COND_INT_CMP, imm_con): Delete.
6859 (cmp_op): Remove above unspecs from int attribute.
6860 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6862 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6863 comparison-specific unspecs.
6864 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6865 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6866 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6867 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6868 (*vec_fcm<cmp_op><mode>): Rename to...
6869 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6870 (*vec_fcmuo<mode>): Rename to...
6871 (*fcmuo<mode>): ...this and adjust likewise.
6872 (*pred_fcm<cmp_op><mode>): New pattern.
6873 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6874 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6876 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6878 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6879 (aarch64_emit_sve_predicated_cond): New function.
6880 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6881 (aarch64_emit_unspec_cond_or): Replace with...
6882 (aarch64_emit_sve_or_conds): ...this new function. Use
6883 aarch64_emit_sve_ptrue_op for the individual comparisons and
6884 aarch64_emit_binop to OR them together.
6885 (aarch64_emit_inverted_unspec_cond): Replace with...
6886 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6887 aarch64_emit_sve_ptrue_op for the comparison and
6888 aarch64_emit_unop to invert the result.
6889 (aarch64_expand_sve_vec_cmp_float): Update after the above
6890 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6892 2018-05-07 Nathan Sidwell <nathan@acm.org>
6894 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6895 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6896 (Backwards Compatibility): Likewise.
6898 2018-05-07 Luis Machado <luis.machado@linaro.org>
6902 2018-05-07 Luis Machado <luis.machado@linaro.org>
6904 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6905 <prefetch_dynamic_strides>: New const bool field.
6906 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6907 prefetch_dynamic_strides.
6908 (exynosm1_prefetch_tune): Likewise.
6909 (thunderxt88_prefetch_tune): Likewise.
6910 (thunderx_prefetch_tune): Likewise.
6911 (thunderx2t99_prefetch_tune): Likewise.
6912 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6914 (aarch64_override_options_internal): Update to set
6915 PARAM_PREFETCH_DYNAMIC_STRIDES.
6916 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6917 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6918 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6919 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6920 prefetch-dynamic-strides setting.
6922 2018-05-07 Luis Machado <luis.machado@linaro.org>
6924 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6925 <minimum_stride>: New const int field.
6926 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6927 minimum_stride field.
6928 (exynosm1_prefetch_tune): Likewise.
6929 (thunderxt88_prefetch_tune): Likewise.
6930 (thunderx_prefetch_tune): Likewise.
6931 (thunderx2t99_prefetch_tune): Likewise.
6932 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6933 (aarch64_override_options_internal): Update to set
6934 PARAM_PREFETCH_MINIMUM_STRIDE.
6935 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6936 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6937 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6938 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6939 stride is constant and is below the minimum stride threshold.
6941 2018-05-07 Luis Machado <luis.machado@linaro.org>
6943 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6946 2018-05-07 Luis Machado <luis.machado@linaro.org>
6948 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6949 <prefetch_dynamic_strides>: New const bool field.
6950 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6951 prefetch_dynamic_strides.
6952 (exynosm1_prefetch_tune): Likewise.
6953 (thunderxt88_prefetch_tune): Likewise.
6954 (thunderx_prefetch_tune): Likewise.
6955 (thunderx2t99_prefetch_tune): Likewise.
6956 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6958 (aarch64_override_options_internal): Update to set
6959 PARAM_PREFETCH_DYNAMIC_STRIDES.
6960 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6961 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6962 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6963 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6964 prefetch-dynamic-strides setting.
6966 2018-05-07 Luis Machado <luis.machado@linaro.org>
6968 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6969 <minimum_stride>: New const int field.
6970 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6971 minimum_stride field.
6972 (exynosm1_prefetch_tune): Likewise.
6973 (thunderxt88_prefetch_tune): Likewise.
6974 (thunderx_prefetch_tune): Likewise.
6975 (thunderx2t99_prefetch_tune): Likewise.
6976 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6977 (aarch64_override_options_internal): Update to set
6978 PARAM_PREFETCH_MINIMUM_STRIDE.
6979 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6980 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6981 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6982 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6983 stride is constant and is below the minimum stride threshold.
6985 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6988 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6989 the type is addressable. Don't force op into register if it has
6992 2018-05-05 Roland McGrath <mcgrathr@google.com>
6995 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6996 any section for which we don't know a specific type it should have,
6997 regardless of name. Previously this was done only for the exact
6998 names ".init_array", ".fini_array", and ".preinit_array".
6999 (default_elf_asm_named_section): Add comment about
7000 relationship with default_section_type_flags and SECTION_NOTYPE.
7001 (get_section): Don't consider it a type conflict if one side has
7002 SECTION_NOTYPE and the other doesn't, as long as neither has the
7003 SECTION_BSS et al used in the default_section_type_flags logic.
7005 2018-05-05 Tom de Vries <tom@codesourcery.com>
7008 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7009 (workaround_barsyncs): New function.
7010 (nvptx_reorg): Use workaround_barsyncs.
7011 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7012 (define_expand "nvptx_membar_cta"): New define_expand.
7013 (define_insn "*nvptx_membar_cta"): New insn.
7015 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7017 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7018 To improve optimization opportunities.
7019 * builtin-types.def: The new needed builtin types for the above.
7021 2018-05-04 Richard Biener <rguenther@suse.de>
7023 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7024 * gimple-ssa-store-merging.c
7025 (imm_store_chain_info::output_merged_store): Remove redundant create,
7026 release split_store vector contents on failure.
7027 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7028 scalar stmt vector on cache hit.
7030 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7032 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7034 * config.gcc (powerpc-xilinx-eabi*): Remove.
7035 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7037 (fusion_addis_mem_combo_load): Ditto.
7038 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7040 (rs6000_cpu_cpp_builtins): Ditto.
7041 * config/rs6000/rs6000-linux.c
7042 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7043 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7044 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7046 (rs6000_setup_reg_addr_masks): Ditto.
7047 (rs6000_init_hard_regno_mode_ok): Ditto.
7048 (rs6000_option_override_internal): Ditto.
7049 (legitimate_lo_sum_address_p): Ditto.
7050 (rs6000_legitimize_address): Ditto.
7051 (rs6000_legitimize_reload_address): Ditto.
7052 (rs6000_legitimate_address_p): Ditto.
7053 (abi_v4_pass_in_fpr): Ditto.
7054 (setup_incoming_varargs): Ditto.
7055 (rs6000_gimplify_va_arg): Ditto.
7056 (rs6000_split_multireg_move): Ditto.
7057 (rs6000_savres_strategy): Ditto.
7058 (rs6000_emit_prologue_components): Ditto.
7059 (rs6000_emit_epilogue_components): Ditto.
7060 (rs6000_emit_prologue): Ditto.
7061 (rs6000_emit_epilogue): Ditto.
7062 (rs6000_elf_file_end): Ditto.
7063 (rs6000_function_value): Ditto.
7064 (rs6000_libcall_value): Ditto.
7065 * config/rs6000/rs6000.h: Ditto.
7066 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7067 (TARGET_MINMAX): ... this. New.
7068 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7069 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7070 (*movsi_internal1_single): Delete.
7071 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7072 mfpu=, mxilinx-fpu): Delete.
7073 * config/rs6000/singlefp.h: Delete.
7074 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7075 * config/rs6000/t-rs6000: Ditto.
7076 * config/rs6000/t-xilinx: Delete.
7077 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7078 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7079 (VStype_simple): Delete.
7080 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7081 * config/rs6000/xfpu.h: Delete.
7082 * config/rs6000/xfpu.md: Delete.
7083 * config/rs6000/xilinx.h: Delete.
7084 * config/rs6000/xilinx.opt: Delete.
7085 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7086 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7088 2018-05-04 Tom de Vries <tom@codesourcery.com>
7091 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7094 2018-05-04 Richard Biener <rguenther@suse.de>
7097 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7098 (expand_complex_div_wide): Likewise.
7099 (expand_complex_operations_1): Likewise.
7100 (expand_complex_libcall): Preserve EH info of the original stmt.
7101 (tree_lower_complex): Handle removed blocks.
7102 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7103 on complex multiplication and division libcall builtins.
7105 2018-05-04 Richard Biener <rguenther@suse.de>
7108 * fold-const.c (negate_expr_p): Restrict negation of operand
7109 zero of a division to when we know that can happen without
7111 (fold_negate_expr_1): Likewise.
7113 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7116 * real.h (real_nextafter): Declare.
7117 * real.c (real_nextafter): New function.
7118 * fold-const-call.c (fold_const_nextafter): New function.
7119 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7120 CASE_CFN_NEXTTOWARD.
7121 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7122 even when arg1_mode is different from arg0_mode.
7124 2018-05-03 Nathan Sidwell <nathan@acm.org>
7126 * doc/extend.texi (Deprecated Features): Remove
7128 (Backwards Compatibility): Likewise.
7129 * doc/invoke.texi (C++ Language Options): Likewise.
7130 (C++ Dialect Options): Likewise.
7132 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7135 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7136 _mm512_mask_mullox_epi64): New intrinsics.
7138 2018-05-03 Tom de Vries <tom@codesourcery.com>
7141 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7142 dump files): Add offload-tree.
7144 2018-05-03 Richard Biener <rguenther@suse.de>
7146 PR tree-optimization/85615
7147 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7148 to loops not nested in BBs loop father to avoid creating multi-entry
7151 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7153 PR tree-optimization/70291
7154 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7155 arguments. Change return type to tree. Emit libcall as a new
7156 statement rather than replacing existing one when inplace_p is true.
7157 (expand_complex_multiplication_components): New function.
7158 (expand_complex_multiplication): Expand floating-point complex
7159 multiplication using the above.
7160 (expand_complex_division): Rename inner_type parameter to type.
7161 Update expand_complex_libcall call-site.
7162 (expand_complex_operations_1): Update expand_complex_multiplication
7163 and expand_complex_division call-sites.
7165 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7168 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7169 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7170 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7171 the highest significant bit of the shift count mask is clear. In
7172 check whether and[sq]i3 is needed verify that all significant bits
7173 of the shift count other than the highest are set.
7175 2018-05-02 Tom de Vries <tom@codesourcery.com>
7178 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7179 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7180 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7181 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7182 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7183 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7184 __builtin_goacc_parlevel_size.
7186 2018-05-02 Richard Biener <rguenther@suse.de>
7188 PR tree-optimization/85597
7189 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7190 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7193 2018-05-02 Tom de Vries <tom@codesourcery.com>
7196 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7197 dump files): Add ltrans-tree.
7199 2018-05-02 Tom de Vries <tom@codesourcery.com>
7202 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7203 dump files): Add wpa-ipa.
7205 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7207 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7208 powerpc*-*-linux*paired* target.
7209 * config/rs6000/750cl.h: Delete.
7210 * config/rs6000/paired.h: Delete.
7211 * config/rs6000/paired.md: Delete.
7212 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7214 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7215 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7216 comment. Remove paired float support.
7217 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7218 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7220 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7221 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7223 * config/rs6000/rs6000.c: Remove paired float support.
7224 (paired_expand_vector_init, paired_expand_vector_move,
7225 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7226 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7227 paired_expand_builtin, paired_expand_predicate_builtin,
7228 paired_init_builtins): Delete.
7229 * config/rs6000/rs6000.h: Remove paired float support.
7230 * config/rs6000/rs6000.md: Remove paired float support.
7231 (move_from_CR_ov_bit): Delete.
7232 * config/rs6000/rs6000.opt (mpaired): Delete.
7233 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7234 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7236 2018-05-02 Richard Biener <rguenther@suse.de>
7239 * gimplify.c (gimplify_save_expr): When in SSA form allow
7240 SAVE_EXPRs to compute to SSA vars.
7242 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7245 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7246 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7247 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7248 clobber operands[2], instead use a new pseudo. Formatting fixes.
7250 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7252 PR tree-optimization/85586
7253 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7254 exit early for statements in the same group if the accesses are
7257 2018-05-02 Tom de Vries <tom@codesourcery.com>
7260 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7263 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7265 PR tree-optimization/85143
7266 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7268 2018-05-01 Tom de Vries <tom@codesourcery.com>
7271 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7272 not found" error message.
7274 2018-05-01 Tom de Vries <tom@codesourcery.com>
7277 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7278 * vec.c (test_ordered_remove_if): New function.
7279 (vec_c_tests): Call test_ordered_remove_if.
7280 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7281 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7282 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7283 VEC_ORDERED_REMOVE_IF.
7285 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7287 PR tree-optimization/82665
7288 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7289 pointer subtraction where arguments come from a memchr call.
7291 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7293 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7294 --push-state --as-needed and --pop-state instead of --as-needed and
7295 --no-as-needed if ld supports it.
7296 * configure: Regenerated.
7299 * doc/install.texi2html: Replace _002d with - and _002a with * in
7300 generated html files using sed.
7302 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7305 * gcc-rich-location.c (blank_line_before_p): New function.
7306 (use_new_line): New function.
7307 (gcc_rich_location::add_fixit_insert_formatted): New function.
7308 * gcc-rich-location.h
7309 (gcc_rich_location::add_fixit_insert_formatted): New function.
7311 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7313 * selftest.c (assert_streq): Rename "expected" and "actual" to
7314 "val1" and "val2". Extend NULL-handling to cover both inputs
7315 symmetrically, while still requiring both to be non-NULL for a pass.
7316 * selftest.h (assert_streq): Rename "expected" and "actual" to
7318 (ASSERT_EQ): Likewise.
7319 (ASSERT_EQ_AT): Likewise.
7320 (ASSERT_KNOWN_EQ): Likewise.
7321 (ASSERT_KNOWN_EQ_AT): Likewise.
7322 (ASSERT_NE): Likewise.
7323 (ASSERT_MAYBE_NE): Likewise.
7324 (ASSERT_MAYBE_NE_AT): Likewise.
7325 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7326 the assertion to pass.
7327 (ASSERT_STREQ_AT): Likewise.
7329 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7331 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7332 interaction with -pie.
7334 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7336 * selftest.h: Fix alphabetization of per-source-file selftest
7339 2018-04-30 Jason Merrill <jason@redhat.com>
7341 PR c++/61982 - dead stores to destroyed objects.
7342 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7345 2018-04-30 Jason Merrill <jason@redhat.com>
7347 * tree.c (build_clobber): New.
7348 * tree.h: Declare it.
7349 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7351 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7353 * diagnostic-show-locus.c (layout::layout): Update for
7354 location_get_source_line returning a char_span.
7355 (struct char_span): Move to input.h.
7356 (struct correction): Update for fields in char_span becoming
7358 (struct source_line): Update for location_get_source_line
7359 returning a char_span.
7360 (layout::print_line): Likewise.
7361 * edit-context.c (edited_file::print_content): Likewise.
7362 (edited_file::print_diff_hunk): Likewise.
7363 (edited_file::print_run_of_changed_lines): Likewise.
7364 (edited_file::get_num_lines): Likewise.
7365 (edited_line::edited_line): Likewise.
7366 * final.c (asm_show_source): Likewise.
7367 * input.c (location_get_source_line): Convert return type
7368 from const char * to char_span, losing the final "line_len"
7370 (dump_location_info): Update for the above.
7371 (get_substring_ranges_for_loc): Likewise. Use a char_span
7372 when handling the literal within the line.
7373 (test_reading_source_line): Update for location_get_source_line
7374 returning a char_span.
7375 * input.h (class char_span): Move here from
7376 diagnostic-show-locus.c, converting from a struct to a class.
7377 Make data members private.
7378 (char_span::operator bool): New.
7379 (char_span::length): New.
7380 (char_span::get_buffer): New.
7381 (char_span::operator[]): New.
7382 (char_span::subspan): Make const.
7383 (char_span::xstrdup): New.
7384 (location_get_source_line): Convert return type from const char *
7385 to char_span, losing the final "line_size" param.
7387 2018-04-30 Jan Hubicka <jh@suse.cz>
7389 * lto-wrapper.c (ltrans_priorities): New static var.
7390 (cmp_priority): New.
7391 (run_gcc): Read priorities and if doing parallel build order
7392 the Makefile by them.
7394 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7396 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7398 2018-04-30 Richard Biener <rguenther@suse.de>
7400 * tree-cfg.c (verify_address): Remove base argument, add
7401 flag whether to check TREE_ADDRESSABLE and do that.
7402 (verify_expr): Remove.
7403 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7404 (verify_gimple_assign_single): Likewise.
7405 (verify_gimple_switch): Likewise.
7406 (verify_expr_location_1): Dereference tp once. Add (disabled)
7407 piece from verify_expr.
7408 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7410 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7412 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7414 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7416 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7417 (small_data_pattern): Likewise.
7418 (arc_rewrite_small_data): Likewise.
7419 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7420 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7421 (get_symbol_alignment): New function.
7422 (legitimate_small_data_address_p): Likewise.
7423 (legitimate_scaled_address): Update, call
7424 legitimate_small_data_address_p.
7425 (output_sdata): New static variable.
7426 (arc_print_operand): Update how we handle small data operands.
7427 (arc_print_operand_address): Likewise.
7428 (arc_legitimate_address_p): Update, use
7429 legitimate_small_data_address_p.
7430 (arc_rewrite_small_data_p): Remove.
7431 (arc_rewrite_small_data_1): Likewise.
7432 (arc_rewrite_small_data): Likewise.
7433 (small_data_pattern): Likewise.
7434 (compact_sda_memory_operand): Update to use
7435 legitimate_small_data_address_p and get_symbol_alignment.
7436 (prepare_move_operands): Don't rewite sdata pattern.
7437 (prepare_extend_operands): Remove.
7438 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7440 (zero_extendqisi2): Likewise.
7441 (zero_extendhisi2): Likewise.
7442 (extendqihi2): Likewise.
7443 (extendqisi2): Likewise.
7444 (extendhisi2): Likewise.
7448 * config/arc/constraints.md (Usd): Change it to memory constraint.
7450 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7452 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7453 as source of std instructions.
7454 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7455 allow 6-bit constants as source for store instructions.
7456 (movdi_insn): Update instruction pattern to allow 6-bit constants
7457 as source for store instructions.
7459 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7461 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7463 2018-04-30 Nathan Sidwell <nathan@acm.org>
7464 Sandra Loosemore <sandra@codesourcery.com>
7466 * dumpfile.c (dump_open): Allow '-' for stdout.
7467 * doc/invoke.texi (Developer Options): Document dump filename
7468 determination early. Document stdin/stdout selection.
7470 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7472 Microblaze Target: PIC data text relative
7474 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7475 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7477 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7478 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7479 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7480 New addressing mode for data-text relative position indepenedent code.
7481 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7482 'ADDRESS_SYMBOLIC_TXT_REL'.
7483 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7484 (microblaze_legitimate_pic_operand): Exclude function calls from
7485 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7486 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7488 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7489 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7490 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7491 for 'address + offset'.
7492 (microblaze_expand_prologue): Add new function prologue call for
7494 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7495 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7496 table in case of TARGET_PIC_DATA_TEXT_REL.
7497 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7498 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7499 Add new macros 'UNSPEC_TEXT',
7500 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7501 + exclude function calls from 'UNSPEC_PLT' in case of data text
7503 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7504 new target hook for generating address diff vector tables in case of
7506 * doc/tm.texi : Regenerate.
7507 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7508 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7509 of addr diff vector generation.
7510 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7511 target hook definition.
7512 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7513 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7514 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7515 option for data text relative.
7517 2018-04-30 Richard Biener <rguenther@suse.de>
7519 * tree-chrec.h (evolution_function_is_constant_p): Remove
7521 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7523 2018-04-30 Richard Biener <rguenther@suse.de>
7526 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7528 2018-04-30 Richard Biener <rguenther@suse.de>
7530 PR tree-optimization/28364
7531 PR tree-optimization/85275
7532 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7533 copying first exit test.
7535 2018-04-28 Mark Wielaard <mark@klomp.org>
7537 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7539 (dwarf_AT): Handle DW_AT_addr_base.
7540 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7542 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7545 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7546 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7547 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7548 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7550 2018-04-28 Richard Biener <rguenther@suse.de>
7552 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7553 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7554 to reflect use. Only add interesting stmts.
7556 2018-04-27 Martin Jambor <mjambor@suse.cz>
7559 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7560 the jump function allows for passing through aggregate values.
7562 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7564 * input.h (in_system_header_at): Convert from macro to inline
7566 (from_macro_expansion_at): Likewise.
7567 (from_macro_definition_at): Likewise.
7569 2018-04-27 Jeff Law <law@redhat.com>
7571 * config.gcc: Mark tile* targets as deprecated/obsolete.
7573 2018-04-27 Richard Biener <rguenther@suse.de>
7575 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7578 2018-04-27 Richard Biener <rguenther@suse.de>
7580 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7582 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7584 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7585 with Yd constraint. Set "preferred_for_speed" attribute from
7586 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7588 (*movdi_internal): Ditto.
7589 (movti_interunit splitters): Remove
7590 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7591 (movdi_interunit splitters): Ditto.
7592 * config/i386/constraints.md (Ye): Remove.
7593 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7595 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7598 * config/aarch64/constraints.md (Usg): Limit to 31.
7601 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7603 PR tree-optimization/85529
7604 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7605 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7606 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7607 zero extension or masking of the MSB bit.
7608 (optimize_range_tests): Add FIRST_BB argument, pass it through
7609 to optimize_range_tests_var_bound.
7610 (maybe_optimize_range_tests, reassociate_bb): Adjust
7611 optimize_range_tests callers.
7613 2018-04-26 Richard Biener <rguenther@suse.de>
7614 Jakub Jelinek <jakub@redhat.com>
7616 * cgraph.h (symbol_table): Just declare debug method here.
7617 * symtab.c (symbol_table::debug): Define.
7619 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7621 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7623 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7625 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7626 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7627 (*movdi_internal): Substitute Yi and Yj constraint with x
7628 and Ym and Yn constraint with y constraint. Update "isa"
7629 attribute and set "preferred_for_speed" attribute from
7630 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7631 (*movsi_internal): Ditto.
7632 (*movdf_internal): Ditto.
7633 (*movsf_internal): Ditto.
7634 (*zero_extendsidi2): Ditto.
7635 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7636 (sse2_loadld): Ditto.
7637 (*vec_extract<ssevecmodelower>_0): Ditto.
7638 (*vec_extractv4si_0_zext_sse4): Ditto.
7639 (vec_concatv2di): Ditto.
7640 (*vec_dup<mode>): Ditto.
7641 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7642 * config/i386/constraints.md (Yi): Remove.
7647 2018-04-26 Nathan Sidwell <nathan@acm.org>
7649 * dumpfile.c (dump_open): New.
7650 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7651 (dump_finish): Detect stdio/stderr by value not name.
7653 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7655 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7657 2018-04-26 Tom de Vries <tom@codesourcery.com>
7660 * config/nvptx/nvptx.c (verify_neutering_jumps)
7661 (verify_neutering_labels): New function
7662 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7664 2018-04-26 Tom de Vries <tom@codesourcery.com>
7667 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7668 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7671 2018-04-26 Richard Biener <rguenther@suse.de>
7672 Tom de Vries <tom@codesourcery.com>
7675 * lto-streamer-out.c (output_function): Fixup loops if required to match
7676 discovery done in the reader.
7678 2018-04-26 Richard Biener <rguenther@suse.de>
7680 PR tree-optimization/85116
7681 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7682 have a loop exit from the single latch predecessor. Remove
7683 case of header with just condition.
7684 (ch_base::copy_headers): Exclude infinite loops from any
7686 (pass_ch::execute): Record exits.
7688 2018-04-26 Richard Biener <rguenther@suse.de>
7690 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7691 prologue cost vector and pass it to vect_get_load_cost.
7692 (vect_get_peeling_costs_all_drs): Likewise.
7693 (vect_peeling_hash_get_lowest_cost): Likewise.
7694 (vect_enhance_data_refs_alignment): Likewise.
7696 2018-04-26 Richard Biener <rguenther@suse.de>
7699 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7700 checking of integer<->pointer conversions.
7701 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7702 sign-/zero-extending pointer types.
7703 (expand_omp_for_static_chunk): Likewise.
7705 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7706 Jean Lee <xiaoyur347@gmail.com>
7708 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7709 (TARGET_ASAN_SHADOW_OFFSET): Define.
7710 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7711 true for -fsanitize=address.
7713 2018-04-25 Mark Wielaard <mark@klomp.org>
7715 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7718 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7720 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7721 than "alu", remove explicit "memory" and "imm_disp" attributes.
7722 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7725 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7726 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7727 gen_lowpart_no_emit.
7729 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7732 * config/i386/i386.c (ix86_expand_builtin): Change memory
7733 operand to XI, extend p0 to Pmode.
7734 * config/i386/i386.md: Change unspec volatile and operand
7735 1 mode to XI, change operand 0 mode to P.
7737 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7739 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7740 GET_MODE_MASK before any checking.
7741 (nds32_can_use_bset_p): Likewise.
7742 (nds32_can_use_btgl_p): Likewise.
7744 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7746 * config/nds32/nds32-doubleword.md: New define_split pattern for
7747 illegal register number.
7749 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7751 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7753 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7755 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7757 2018-04-25 Richard Biener <rguenther@suse.de>
7759 * lto-streamer.h (LTO_major_version): Bump to 8.
7761 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7763 * BASE-VER: Set to 9.0.0.
7765 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7767 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7768 in __abskf2 and __powikf2.
7770 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7773 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7774 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7775 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7776 Use the above on operand 2. Reindent.
7777 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7779 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7782 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7784 * config/i386/i386.opt (mcet): Removed.
7785 * doc/install.texi: Remove -mcet documentation.
7786 * doc/invoke.texi: Likewise.
7788 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7791 * doc/install.texi: Remove -mcet from bootstrap-cet.
7793 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7796 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7797 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7801 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7802 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7803 containing a CONST_VECTOR.
7805 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7807 * doc/install.texi: Update newlib dependency for nvptx.
7809 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7812 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7813 instead of INTVAL when shifting x left.
7815 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7817 PR tree-optimization/85478
7818 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7819 vect_grouped_store_supported for single element vectors.
7821 2018-04-24 Richard Biener <rguenther@suse.de>
7824 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7825 load cost increase to the case of non-constant step.
7827 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7830 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7831 destination if any_malformed_asm.
7833 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7836 * expr.c (store_field): In the bitfield case, if the value comes from
7837 a function call and is returned in registers by means of a PARALLEL,
7838 do not change the mode of the temporary unless BLKmode and VOIDmode.
7840 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7842 PR rtl-optimization/85423
7843 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7844 dependencies to debug insns when the previous insn is non-debug.
7846 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7848 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7849 enums into a single definition.
7850 (fls): Fix predicates and printing.
7853 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7855 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7856 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7857 and short u6 immediate.
7858 (check_if_valid_sleep_operand): Remove.
7859 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7861 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7863 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7864 flag_always_save_lp condition.
7865 * config/nds32/nds32.opt (malways-save-lp): New option.
7867 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7869 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7870 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7871 * config/nds32/nds32.h
7872 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7873 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7875 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7877 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7878 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7880 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7881 Chung-Ju Wu <jasonwucj@gmail.com>
7883 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7885 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7886 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7888 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7890 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7892 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7894 * config/nds32/nds32-protos.h (nds32_data_alignment,
7895 nds32_local_alignment): Declare.
7896 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7897 nds32_local_alignment): New functions.
7898 (TARGET_CONSTANT_ALIGNMENT): Define.
7899 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7901 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7903 * config/nds32/nds32.c
7904 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7905 (TARGET_MODES_TIEABLE_P): Likewise.
7907 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7909 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7912 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7913 Chung-Ju Wu <jasonwucj@gmail.com>
7915 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7916 for unaligned access.
7917 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7919 * config/nds32/nds32-intrinsic.md: Likewise.
7920 * config/nds32/nds32_intrinsic.h: Likewise.
7921 * config/nds32/nds32.h (nds32_builtins): Likewise.
7922 * config/nds32/nds32.opt (munaligned-access): New option.
7923 * config/nds32/nds32.c (nds32_asm_file_start): Display
7924 flag_unaligned_access status.
7926 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7928 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7929 -mno-relax is present.
7930 * config/riscv/linux.h (LINK_SPEC): Ditto.
7932 2018-04-20 Martin Sebor <msebor@redhat.com>
7935 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7937 (gimple_fold_builtin_stxcpy_chk): Same.
7938 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7940 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7943 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7944 __powikf2 when long double is IEEE 128-bit.
7946 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7948 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7949 step to make sure stack always aligned.
7951 2018-04-20 Carl Love <cel@us.ibm.com>
7954 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7955 size check for arg0.
7957 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7958 Tom de Vries <tom@codesourcery.com>
7961 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7962 Emit insns for calls too.
7963 (nvptx_find_par): Always look for worker-level predecessor insn.
7964 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7966 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7967 (nvptx_process_pars): Propagate frames for calls.
7969 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7972 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7974 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7975 (ix86_handle_option): Don't handle OPT_mibt.
7976 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7978 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7980 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7981 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7982 (ix86_target_macros): Define __CET__ with flag_cf_protection
7983 for -fcf-protection.
7984 * config/i386/i386.c (isa2_opts): Remove -mibt.
7985 * config/i386/i386.h (TARGET_IBT): Removed.
7986 (TARGET_IBT_P): Likewise.
7987 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7988 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7989 * config/i386/i386.opt (mcet): Update help message.
7992 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7993 -mcet as an alias for -mshstk.
7995 2018-04-20 Richard Biener <rguenther@suse.de>
7998 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7999 complexity by forcing a single use of the multiply operand.
8001 2018-04-20 Martin Jambor <mjambor@suse.cz>
8004 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8005 recursion dependency to only apply to non-clones.
8007 2018-04-20 Martin Jambor <mjambor@suse.cz>
8010 * ipa-cp.c (create_specialized_node): Check that clones of
8011 self-recursive edges exist during IPA-CP.
8013 2018-04-19 Toon Moene <toon@moene.org>
8015 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8018 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8020 PR tree-optimization/85467
8021 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8022 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8023 VECTOR_CST element to type.
8025 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8028 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8029 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8030 (builtin_longjmp): Likewise.
8031 (save_stack_nonlocal): New pattern.
8032 (restore_stack_nonlocal): Likewise.
8034 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8037 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8038 Replace ASM_OUTPUT_LABEL with fprintf.
8040 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8043 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8044 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8045 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8046 define __IBT__ and __SHSTK__ for -fcf-protection.
8047 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8049 (ix86_trampoline_init): Likewise.
8050 (x86_output_mi_thunk): Likewise.
8051 (ix86_notrack_prefixed_insn_p): Likewise.
8052 (ix86_option_override_internal): Don't disallow -fcf-protection.
8053 * config/i386/i386.md (rdssp<mode>): Also enable for
8055 (incssp<mode>): Likewise.
8056 (nop_endbr): Likewise.
8057 * config/i386/i386.opt (mcet): Change help message to built-in
8061 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8062 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8063 enable CET built-in functions.
8065 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8067 * common/config/i386/i386-common.c
8068 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8069 OPTION_MASK_ISA_MOVDIRI_UNSET,
8070 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8071 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8072 * config.gcc (movdirintrin.h): New header.
8073 * config/i386/cpuid.h (bit_MOVDIRI,
8074 bit_MOVDIR64B): New bits.
8075 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8077 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8078 (VOID, PVOID, PCVOID)): New function types.
8079 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8080 __builtin_ia32_directstoreu_u64,
8081 __builtin_ia32_movdir64b): New builtins.
8082 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8083 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8085 (ix86_valid_target_attribute_inner_p): Ditto.
8086 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8087 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8088 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8089 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8090 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8091 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8092 (movdiri<mode>, movdir64b_<mode>): New.
8093 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8094 * config/i386/immintrin.h: Include movdirintrin.h.
8095 * config/i386/movdirintrin.h: New file.
8096 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8098 2018-04-19 Richard Biener <rguenther@suse.de>
8101 * cfg.c (clear_bb_flags): When loop state says we have
8102 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8104 2018-04-19 Richard Biener <rguenther@suse.de>
8106 PR tree-optimization/84737
8107 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8108 copying restrict info.
8109 (vect_setup_realignment): Use it.
8110 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8111 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8112 the first DR to all generated stores.
8113 (vectorizable_load): Likewise for loads.
8115 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8117 PR tree-optimization/85446
8118 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8119 the integral and pointer types to have the same precision.
8121 * doc/install.texi: Document --disable-cet being the default and
8124 2018-04-18 Martin Liska <mliska@suse.cz>
8126 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8129 2018-04-18 Martin Liska <mliska@suse.cz>
8132 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8135 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8136 arguments if they are comparable.
8138 2018-04-18 Martin Liska <mliska@suse.cz>
8141 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8144 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8147 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8150 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8151 ENDBR after calling __morestack.
8153 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8156 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8157 by using gcc_base_ver to generate a gcc_driver_version, and use
8158 it when generating GCC_DRIVER_NAME.
8159 * configure: Regenerate.
8161 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8164 * config.gcc: Obsolete powerpc*-*-*spe*.
8166 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8169 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8170 (stabstr_D): Change type of unum from unsigned int to
8171 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8174 2018-04-17 Jim Wilson <jimw@sifive.com>
8177 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8178 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8179 Set arg_pointer_offset after using pretend_args_size.
8181 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8183 PR rtl-optimization/85431
8184 * dse.c (record_store): Ignore zero width stores.
8187 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8188 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8189 __builtin_stack_restore rather than after it.
8190 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8191 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8192 argument instead of virtual_dynamic_stack_rtx.
8194 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8196 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8198 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8199 Add note to error message to explain internal mapping of overloaded
8200 built-in function name to non-overloaded built-in function name.
8201 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8204 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8207 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8208 where the inputs overlap with the output.
8210 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8212 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8213 (=v, v) alternative and explicit "memory" attribute.
8214 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8215 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8217 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8218 "sselog1" type instead of "sselog".
8219 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8220 "sselog". Remove explicit "memory" attribute.
8221 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8222 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8224 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8225 "sselog1" type instead of "sselog". Remove explicit "memory"
8227 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8228 use "sselog1" type instead of "sselog". Remove explicit "memory"
8230 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8231 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8233 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8234 "sselog1" type instead of "sselog". Remove explicit "memory"
8236 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8237 use "sselog1" type instead of "sselog". Remove explicit "memory"
8241 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8244 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8247 2018-04-17 Martin Jambor <mjambor@suse.cz>
8250 * ipa-cp.c (create_specialized_node): Call
8251 expand_all_artificial_thunks if necessary.
8253 2018-04-17 Martin Liska <mliska@suse.cz>
8256 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8257 in message, remote space in between '_G' and '('.
8259 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8262 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8263 avx512f_vmcmp<mode>3<round_saeonly_name>,
8264 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8265 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8266 avx512f_rndscale<mode><round_saeonly_name>,
8267 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8268 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8269 Use %<iptr>2 instead of %2 for -masm=intel.
8270 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8271 avx512f_vcvttss2usi<round_saeonly_name>,
8272 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8274 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8275 avx512f_vcvttsd2usi<round_saeonly_name>,
8276 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8277 Use %q1 instead of %1 for -masm=intel.
8278 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8279 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8280 of %3 for -masm=intel.
8281 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8283 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8285 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8287 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8288 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8290 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8292 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8293 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8294 %g1 and one with %0 and %1.
8295 (avx512er_vmrcp28<mode><round_saeonly_name>,
8296 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8298 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8299 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8300 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8301 of %0 and %{%4%} for -masm=intel.
8302 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8303 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8304 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8305 order of %0 and %{%5%}%{z%} for -masm=intel.
8307 2018-04-17 Jan Hubicka <jh@suse.cz>
8310 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8312 2018-04-17 Martin Liska <mliska@suse.cz>
8315 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8316 for target_clone error message. Make default implementation
8317 clone to be a local declaration.
8318 (separate_attrs): Add new argument and check for an empty
8320 (expand_target_clones): Handle it.
8321 (ipa_target_clone): Make redirection just for target_clones
8324 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8325 Tom de Vries <tom@codesourcery.com>
8328 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8329 tiled basic blocks without omp continue statements.
8331 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8334 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8335 vec_extract expression as having side effects to make sure it gets
8338 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8341 * config/i386/i386.c (get_builtin_code_for_version): Check
8344 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8347 * gcc/config.gcc: Support "skylake".
8348 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8350 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8351 (processor_target_table): Add "skylake".
8352 (ix86_option_override_internal): Add "skylake".
8353 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8354 PROCESSOR_CANNONLAKE.
8355 (get_builtin_code_for_version): Fix priority for
8356 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8357 PROCESSOR_SKYLAKE-AVX512.
8358 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8359 (processor_type): Add PROCESSOR_SKYLAKE.
8361 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8362 Jason Merrill <jason@redhat.com>
8365 * convert.c (convert_to_integer_1): Use direct recursion for
8366 enumeral types and types with a precision less than the number
8367 of bits in their mode.
8369 2018-04-16 Julia Koval <julia.koval@intel.com>
8372 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8373 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8375 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8378 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8379 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8380 and -mno-direct-move.
8382 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8385 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8386 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8387 Ensure negative shifts result in {0}.
8389 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8391 PR rtl-optimization/79916
8392 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8393 regs (if any) to define how to gnerate SD moves when LRA is in
8396 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8398 PR rtl-optimization/85393
8399 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8400 * except.c (expand_dw2_landing_pad_for_region): Make static.
8401 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8402 a label and unconditional jump to old_bb, rather than
8403 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8406 PR rtl-optimization/85376
8407 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8408 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8409 instead of a specific value.
8411 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8412 Bin Cheng <bin.cheng@arm.com>
8414 PR tree-optimization/82965
8415 PR tree-optimization/83991
8416 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8417 by_profile_only parameter.
8418 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8419 information if the loop was predicted to iterate too many times.
8420 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8422 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8425 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8428 2018-04-13 Martin Liska <mliska@suse.cz>
8429 Jakub Jelinek <jakub@redhat.com>
8432 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8433 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8434 * builtins.c (expand_builtin_memory_copy_args): Use
8435 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8436 handle dest_addr == pc_rtx.
8438 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8441 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8442 asked to not generate direct moves.
8443 (fix_trunc<mode>si2_stfiwx): Similar.
8444 (fix_trunc<mode>si2_internal): Similar.
8446 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8449 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8450 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8451 lookup if dest in some wider mode is known to be const0_rtx and
8452 if so, record permanent equivalence for it to be ZERO_EXTEND of
8453 the narrower mode destination.
8455 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8457 * lto-streamer-out.c (output_function): Revert 259346.
8458 * omp-expand.c (expand_oacc_for): Likewise.
8460 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8462 PR rtl-optimization/85354
8463 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8464 * sel-sched.c (sel_global_init): ... here.
8466 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8469 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8470 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8471 mode for PE-COFF targets.
8472 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8473 (i386_pe_asm_lto_end): Likewise.
8474 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8475 (TARGET_ASM_LTO_END): Likewise.
8476 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8477 (i386_pe_asm_lto_start): New function.
8478 (i386_pe_asm_lto_end): Likewise.
8480 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8481 Richard Biener <rguenther@suse.de>
8484 * lto-streamer-out.c (output_function): Fix CFG loop state before
8486 * omp-expand.c (expand_oacc_for): Handle calls to internal
8487 functions like regular functions.
8489 2018-04-12 Richard Biener <rguenther@suse.de>
8492 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8493 for the early LTO debug to properly generate references to it
8494 during DIE emission. Do not re-use that for the skeleton for
8496 (dwarf2out_early_finish): Likewise.
8498 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8501 * config/i386/sse.md
8502 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8503 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8504 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8505 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8506 and output is a reg, avoid creating invalid lowpart subreg, but
8507 instead split into a 512-bit move. Don't split if not AVX512VL,
8508 input is xmm16+ reg and output is a mem.
8509 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8510 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8511 xmm16+ reg and output is a mem.
8513 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8515 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8516 also for flag_dwarf2_cfi_asm.
8518 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8520 PR rtl-optimization/85342
8521 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8522 a bool scalar var inside of the loop instead. Don't try to update
8523 recog_data.operand after failed apply_change_group.
8525 2018-04-12 Tom de Vries <tom@codesourcery.com>
8528 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8529 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8530 array with flexible array member as array without given dimension.
8531 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8532 argument for undefined param to true.
8534 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8537 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8538 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8539 from PowerPC section.
8540 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8541 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8542 help text that is too long.
8543 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8544 help text that is too long.
8545 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8546 help text that is too long.
8548 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8550 * config/alpha/alpha.md (stack_probe_internal): Rename
8551 from "probe_stack". Update all callers.
8553 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8555 PR rtl-optimization/84566
8556 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8557 sched_macro_fuse_insns.
8559 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8562 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8563 (compute_block_dependences): ... from here.
8565 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8567 PR tree-optimization/85331
8568 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8569 from int to HOST_WIDE_INT.
8571 2018-04-11 Martin Jambor <mjambor@suse.cz>
8574 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8575 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8576 not the same as the source val.
8577 (cgraph_edge_brings_value_p): New parameter.
8578 (gather_edges_for_value): Pass destination value to
8579 cgraph_edge_brings_value_p.
8580 (perhaps_add_new_callers): Likewise.
8581 (get_info_about_necessary_edges): Likewise and exclude values brought
8582 only by self-recursive edges.
8583 (create_specialized_node): Redirect only clones of self-calling edges.
8584 (+self_recursive_pass_through_p): New function.
8585 (find_more_scalar_values_for_callers_subset): Use it.
8586 (find_aggregate_values_for_callers_subset): Likewise.
8587 (known_aggs_to_agg_replacement_list): Removed.
8588 (decide_whether_version_node): Re-calculate known constants for all
8589 remaining context clones.
8591 2018-04-11 Richard Biener <rguenther@suse.de>
8594 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8595 from early DWARF output.
8596 (dwarf2out_early_finish): Output line info unconditionally into
8597 early DWARF and add reference to it.
8599 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8602 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8603 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8604 other than V2DFmode using iptr mode attribute.
8605 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8607 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8609 PR rtl-optimization/84659
8610 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8612 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8615 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8617 (output_loc_list): Pass address of a dummy size variable even in the
8618 locview handling loop.
8619 (index_location_lists): Add comment on why skip_loc_list_entry can't
8622 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8625 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8628 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8631 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8632 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8633 and -mstring-compare-inline-limit.
8635 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8638 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8639 for stack clash protection in a register whenever we need it to be in
8642 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8644 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8645 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8647 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8650 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8652 (mlong-double-): Ditto.
8653 * config/rs6000/sysv4.opt (msdata=): Ditto.
8654 (mtls-size=): Ditto.
8656 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8658 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8659 erroneous entries for
8660 "vector int vec_ldl (int, long int *)", and
8661 "vector unsigned int vec_ldl (int, unsigned long int *)".
8662 Add comments and entries for
8663 "vector bool char vec_ldl (int, bool char *)",
8664 "vector bool short vec_ldl (int, bool short *)",
8665 "vector bool int vec_ldl (int, bool int *)",
8666 "vector bool long long vec_ldl (int, bool long long *)",
8667 "vector pixel vec_ldl (int, pixel *)",
8668 "vector long long vec_ldl (int, long long *)",
8669 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8670 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8671 type tree bool_long_long_type_node and correct definition of
8672 bool_V2DI_type_node to make reference to this new type tree.
8673 (rs6000_mangle_type): Replace erroneous reference to
8674 bool_long_type_node with bool_long_long_type_node.
8675 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8676 comments to emphasize sign distinctions for char and int types and
8677 replace RS6000_BTI_bool_long constant with
8678 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8679 use of RS6000_BTI_pixel.
8680 (bool_long_type_node): Remove this macro definition.
8681 (bool_long_long_type_node): New macro definition
8683 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8685 PR rtl-optimization/85300
8686 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8687 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8688 simplify_unary_operation fails.
8690 2018-04-10 Martin Liska <mliska@suse.cz>
8692 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8693 cgraph_edge and ipa_ref.
8695 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8699 * config/i386/sse.md
8700 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8701 computation of the VEC_MERGE selector from mask.
8702 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8703 Fix decoding of the VEC_MERGE selector into mask.
8705 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8707 PR tree-optimization/85286
8708 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8710 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8712 * final.c (final_1): Set insn_last_address as well as
8713 insn_current_address.
8715 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8718 * explow.c (emit_stack_probe): Call validize_mem on memory location
8719 before passing it to gen_probe_stack. Create address operand and
8720 legitimize it for the probe_stack_address case.
8722 2018-04-09 Jan Hubicka <jh@suse.cz>
8725 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8726 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8727 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8729 (free_lang_data): Rebuild type inheritance graph.
8731 2018-04-09 Martin Sebor <msebor@redhat.com>
8733 * invoke.texi (-finline-small-functions): Mention other optimization
8735 (-findirect-inlining, -fpartial-inlining): Same.
8736 (-finline-functions-called-once): Same.
8737 (-freorder-blocks-and-partition): Same.
8739 2018-04-09 Jan Hubicka <jh@suse.cz>
8742 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8743 jumps; choose last target that matches the criteria (i.e.
8744 no partition changes for non-crossing jumps).
8745 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8746 support for redirecting crossing jumps to non-crossing.
8748 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8750 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8751 also for naked functions.
8753 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8755 * config/arc/arc.md (add_shift): New pattern.
8756 (add_shift2): Likewise.
8757 (sub_shift): Likewise.
8758 (sub_shift_cmp0_noout): Likewise.
8759 (compare_si_ashiftsi): Likewise.
8760 (xbfu_cmp0_noout): New combine pattern.
8761 (xbfu_cmp0"): Likewise.
8762 (movsi_set_cc_insn): Place the predicable variant first.
8763 (commutative_binary_cmp0_noout): Remove clobber.
8764 (commutative_binary_cmp0): New pattern.
8765 (noncommutative_binary_cmp0): Likewise.
8766 (noncommutative_binary_cmp0_noout): Likewise.
8767 (noncommutative_binary_comparison_result_used): Removed.
8768 (rsub_cmp0): New pattern.
8769 (rsub_cmp0_noout): Likewise.
8770 (extzvsi): Changed, keep only meaningful variants.
8771 (SQH, SEZ): New iterators.
8772 (SQH_postfix): New mode attribute.
8773 (SEZ_prefix): New code attribute.
8774 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8775 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8776 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8778 (noncommutative_operator): Check the availability of barrel
8781 2018-04-09 Richard Biener <rguenther@suse.de>
8783 PR tree-optimization/85284
8784 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8785 Only use the niter constraining form of simple_iv when the exit
8788 2018-04-09 Tom de Vries <tom@codesourcery.com>
8791 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8792 (define_expand "*memory_barrier"): New define_expand.
8793 (define_insn "memory_barrier"): New insn.
8795 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8797 PR rtl-optimization/80463
8798 PR rtl-optimization/83972
8799 PR rtl-optimization/83480
8801 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8802 correct producer for the insn.
8803 (tidy_control_flow): Fixup seqnos in case of debug insns.
8805 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8807 PR rtl-optimization/83913
8809 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8810 different sched-times when merging exprs.
8812 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8814 PR rtl-optimization/83962
8816 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8817 tidy_fallthru_edge and tidy_control_flow.
8819 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8821 PR rtl-optimization/83530
8823 * sel-sched.c (force_next_insn): New global variable.
8824 (remove_insn_for_debug): When force_next_insn is true, also leave only
8825 next insn in the ready list.
8826 (sel_sched_region): When the region wasn't scheduled, make another pass
8827 over it with force_next_insn set to 1.
8829 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8831 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8833 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8834 for interrupt control.
8835 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8836 functions for interrupt control.
8837 * config/nds32/nds32-intrinsic.md: Likewise.
8838 * config/nds32/nds32_intrinsic.h: Likewise.
8839 * config/nds32/nds32.h (nds32_builtins): Likewise.
8841 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8843 * config/nds32/nds32.c (nds32_init_machine_status,
8844 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8845 strict_aligned_p field.
8846 (nds32_expand_to_rtl_hook): New function.
8847 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8848 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8850 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8851 Chung-Ju Wu <jasonwucj@gmail.com>
8853 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8854 * config/nds32/nds32-n7.md: New file.
8855 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8856 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8858 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8859 * config/nds32/nds32.md (pipeline_model): Add n7.
8860 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8861 * config/nds32/pipelines.md: Include n7 settings.
8863 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8864 Chung-Ju Wu <jasonwucj@gmail.com>
8866 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8867 * config/nds32/nds32-e8.md: New file.
8868 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8869 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8871 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8872 * config/nds32/nds32.md (pipeline_model): Add e8.
8873 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8874 * config/nds32/pipelines.md: Include e8 settings.
8876 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8877 Chung-Ju Wu <jasonwucj@gmail.com>
8879 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8880 * config/nds32/nds32-n8.md: New file.
8881 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8882 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8884 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8885 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8886 * config/nds32/nds32.md (pipeline_model): Add n8.
8887 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8888 * config/nds32/pipelines.md: Include n8 settings.
8890 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8891 Chung-Ju Wu <jasonwucj@gmail.com>
8893 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8894 * config/nds32/nds32-n9-2r1w.md: New file.
8895 * config/nds32/nds32-n9-3r2w.md: New file.
8896 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8897 nds32_register_ports): New or modify for cpu n9.
8898 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8900 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8901 * config/nds32/nds32-utils.c: New file.
8902 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8903 TARGET_MUL_SLOW): Define.
8904 * config/nds32/nds32.md (pipeline_model): New attribute.
8905 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8906 New options that support cpu n9.
8907 * config/nds32/pipelines.md: Include n9 settings.
8908 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8910 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8912 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8913 information if necessary.
8914 (output_cond_branch_compare_zero): Likewise.
8915 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8916 (nds32_target_alignment): Refine for alignment.
8917 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8918 (FUNCTION_BOUNDARY): Modify.
8919 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8921 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8923 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8925 * config/nds32/constants.md (unspec_volatile_element): Add values for
8926 TLB operation and data prefetch.
8927 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8928 functions for TLB operation and data prefetch.
8929 * config/nds32/nds32-intrinsic.md: Likewise.
8930 * config/nds32/nds32_intrinsic.h: Likewise.
8931 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8932 (nds32_print_operand): Likewise.
8933 * config/nds32/nds32.h (nds32_builtins): Likewise.
8935 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8936 Andrew Pinski <pinsika@gcc.gnu.org>
8939 * match.pd: Use constant_boolean_node of correct type instead of
8940 boolean_true_node or boolean_false_node for simplifying
8941 pointer comparisons to zero.
8943 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8945 PR tree-optimization/80021
8946 * tree.c (verify_type_variant): Make error call in verify_variant_match
8947 translatable and remove final full stop.
8949 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8951 * config/nds32/constants.md (unspec_volatile_element): Add
8952 UNSPEC_VOLATILE_EH_RETURN.
8953 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8954 nds32_output_stack_pop): Support dwarf exception handling process.
8955 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8956 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8957 exception handling process.
8958 (nds32_compute_stack_frame): Likewise.
8959 (nds32_return_addr_rtx): Likewise.
8960 (nds32_initial_elimination_offset): Likewise.
8961 (nds32_expand_prologue): Likewise.
8962 (nds32_expand_epilogue): Likewise.
8963 (nds32_dynamic_chain_address): New function.
8964 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8966 (DYNAMIC_CHAIN_ADDRESS): Define.
8967 (EH_RETURN_DATA_REGNO): Define.
8968 (EH_RETURN_STACKADJ_RTX): Define.
8969 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8970 patterns for dwarf exception handling.
8972 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8974 * config/nds32/nds32.h: Clean up obsolete macros.
8976 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8978 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8979 Add enum values for particular instructions.
8980 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8981 particular intrinsic functions.
8982 * config/nds32/nds32-intrinsic.md: Likewise.
8983 * config/nds32/nds32_intrinsic.h: Likewise.
8984 * config/nds32/nds32.h (nds32_builtins): Likewise.
8985 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8986 (btst, ave): New patterns for particular instructions.
8988 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8990 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8991 Add enum values for atomic load/store and memory sync.
8992 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8994 * config/nds32/nds32-intrinsic.md: Likewise.
8995 * config/nds32/nds32_intrinsic.h: Likewise.
8996 * config/nds32/nds32.h (nds32_builtins): Likewise.
8998 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9000 PR tree-optimization/85257
9001 * fold-const.c (native_encode_vector): If not all elts could fit
9002 and off is -1, return 0 rather than offset.
9003 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9004 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9005 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9006 adjust buffer in native_interpret_expr call.
9008 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9010 * config/nds32/constants.md (unspec_volatile_element): Add cache
9011 control enum values.
9012 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9013 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9014 * config/nds32/nds32.c (nds32_cctl_names): New.
9015 (nds32_print_operand): Handle cache control register names.
9016 * config/nds32/nds32.h (nds32_builtins): New enum values.
9017 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9019 * config/nds32/nds32.md (type): Add mmu.
9020 * config/nds32/pipelines.md (simple_insn): Add mmu.
9022 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9024 * config/nds32/nds32.md (type): Remove call.
9025 * config/nds32/pipelines.md (simple_insn): Likewise.
9027 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9029 * config/nds32/constants.md (unspec_volatile_element): Add
9030 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9031 UNSPEC_VOLATILE_FMFCFG.
9032 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9033 description for fmfcfg and fmfcsr.
9034 (bdesc_1arg): Add fmtcsr.
9035 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9036 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9037 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9038 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9039 unspec_fmfcfg): New patterns.
9040 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9041 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9042 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9043 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9044 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9045 __nds32__fmfcfg): Define.
9047 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9049 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9050 intrinsic register names.
9051 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9052 intrinsic register enum values and macros.
9054 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9056 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9057 for load/store addressing form.
9058 (nds32_print_operand_address): Likewise.
9060 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9063 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9064 based on LABEL_REF. Remove useless assertion.
9065 (pic_address_needs_scratch): Fix formatting.
9066 (sparc_legitimize_pic_address): Minor tweaks.
9067 (sparc_delegitimize_address): Adjust assertion accordingly.
9068 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9069 into symbolic_operand.
9070 (movsi_high_pic_label_ref): Likewise.
9071 (movsi_lo_sum_pic_label_ref): Likewise.
9072 (movdi_pic_label_ref): Likewise.
9073 (movdi_high_pic_label_ref): Likewise.
9074 (movdi_lo_sum_pic_label_ref): Likewise.
9076 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9078 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9079 custom LIB_SPEC setup.
9081 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9082 Kito Cheng <kito.cheng@gmail.com>
9084 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9085 * config/riscv/freebsd.h: New.
9087 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9089 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9090 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9093 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9094 Kito Cheng <kito.cheng@gmail.com>
9096 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9097 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9098 * config/nds32/nds32-protos.h (nds32_output_call,
9099 nds32_output_return): Declare.
9100 * config/nds32/nds32.md: Refine all the call and return patterns.
9102 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9105 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9106 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9108 PR rtl-optimization/84872
9109 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9110 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9113 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9115 * expr.c (copy_blkmode_to_reg): Revert 254862.
9116 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9118 2018-04-06 Richard Biener <rguenther@suse.de>
9121 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9122 after seeing a component reference with an adjacent field. Treat
9123 refs to arrays at struct end of external decls similar to
9124 refs to unconstrained commons.
9126 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9129 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9130 look through SAVE_EXPRs with non-side-effects argument. Adjust
9132 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9135 2018-04-06 Richard Biener <rguenther@suse.de>
9138 * alias.c (find_base_term): New wrapper around find_base_term
9139 unwinding CSELIB_VAL_PTR changes.
9140 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9143 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9145 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9147 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9148 constant definitions.
9149 ("nop"): lr 0,0 -> nopr r0
9150 ("nop_lr0", "nop_lr1"): New insn definitions.
9152 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9154 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9155 NDS32_V3PUSH_AVAILABLE_P macro.
9157 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9158 Chung-Ju Wu <jasonwucj@gmail.com>
9160 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9161 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9162 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9163 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9164 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9165 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9166 * config/nds32/constraints.md: New constraints and checking for hard
9167 float configuration.
9168 * config/nds32/iterators.md: New mode iterator and attribute for hard
9169 float configuration.
9170 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9172 * config/nds32/nds32-fpu.md: New file.
9173 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9174 deal with hard float code generation.
9175 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9177 (abi_type, float_reg_number): New enum type.
9178 * config/nds32/nds32-predicates.c: New predicates for hard float.
9179 * config/nds32/nds32-protos.h: Declare functions for hard float.
9180 * config/nds32/nds32.c: Implementation for hard float configuration.
9181 * config/nds32/nds32.h: Definitions for hard float configuration.
9182 * config/nds32/nds32.md: Include hard float machine description and
9183 modify patterns for hard float configuration.
9184 * config/nds32/nds32.opt: New options for hard float configuration.
9185 * config/nds32/predicates.md: New predicates for hard float
9188 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9190 * common/config/nds32/nds32-common.c
9191 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9193 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9196 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9197 CONSTRUCTOR_ELT (ctor, ...)->value.
9199 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9202 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9204 2018-04-05 Tom de Vries <tom@codesourcery.com>
9207 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9210 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9211 Kito Cheng <kito.cheng@gmail.com>
9213 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9214 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9215 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9216 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9218 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9219 Kito Cheng <kito.cheng@gmail.com>
9221 * config/nds32/constraints.md (Ufe): New memory constraint.
9222 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9223 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9224 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9226 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9227 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9229 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9231 * config/nds32/nds32.md: Use optimize_size in the condition for
9232 alu-shift instructions.
9234 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9236 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9238 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9240 * config/nds32/nds32.md (negsi2): Refine pattern.
9242 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9243 Chung-Ju Wu <jasonwucj@gmail.com>
9245 * config/nds32/iterators.md (shift_rotate): New code iterator.
9246 (shift): New code attribute.
9247 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9248 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9249 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9250 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9251 bit-wise operations.
9252 (andsi3, *andsi3): Ditto.
9253 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9254 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9255 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9256 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9257 nds32_ior_operand, nds32_xor_operand): New predicates.
9259 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9261 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9262 (addsi3, subsi3): ... this.
9264 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9266 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9268 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9270 * config/nds32/nds32.md: Adjust indention.
9272 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9274 * config/nds32/nds32.md (feature): New attribute.
9276 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9278 * config/nds32/nds32.md (subtype): New attribute.
9280 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9283 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9284 expansion to perform a bitwise AND of the argument followed by a
9285 boolean negation of the result.
9287 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9289 PR rtl-optimization/84878
9290 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9291 the basic block. Assert the use reference is not artificial and that
9292 it has an associated insn.
9294 2018-04-04 Michael Matz <matz@suse.de>
9296 * builtins.c (compute_objsize): Pass correct operand
9297 to array_at_struct_end_p.
9299 2018-04-04 Richard Biener <rguenther@suse.de>
9302 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9303 from contexts for DINFO_LEVEL_TERSE and below.
9305 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9307 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9308 resiter_operand condition.
9309 * config/nds32/nds32.md (*move<mode>): Ditto.
9311 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9312 Monk Chiang <sh.chiang04@gmail.com>
9314 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9316 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9318 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9320 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9321 Kito Cheng <kito.cheng@gmail.com>
9323 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9324 nds32_cond_code_str, output_cond_branch,
9325 output_cond_branch_compare_zero, nds32_expand_cbranch,
9326 nds32_expand_cstore, nds32_expand_movcc,
9327 nds32_output_cbranchsi4_equality_zero,
9328 nds32_output_cbranchsi4_equality_reg,
9329 nds32_output_cbranchsi4_equality_reg_or_const_int,
9330 nds32_output_cbranchsi4_greater_less_zero: New functions.
9331 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9332 nds32_expand_cstore, nds32_expand_movcc,
9333 nds32_output_cbranchsi4_equality_zero,
9334 nds32_output_cbranchsi4_equality_reg,
9335 nds32_output_cbranchsi4_equality_reg_or_const_int,
9336 nds32_output_cbranchsi4_greater_less_zero): Declare.
9337 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9338 nds32_rimm11s_operand): New predicates.
9339 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9340 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9343 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9345 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9346 * config/nds32/nds32.md: Ditto.
9347 * config/nds32/pipelines.md: Ditto.
9349 2018-04-04 Richard Biener <rguenther@suse.de>
9351 PR tree-optimization/85168
9352 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9353 propagating abnormals.
9355 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9357 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9359 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9360 Kito Cheng <kito.cheng@gmail.com>
9362 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9363 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9364 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9365 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9366 * config/nds32/nds32.md (sibcall_internal): New.
9367 (sibcall_register): Remove.
9368 (sibcall_immediate): Remove.
9369 (sibcall_value_internal): New.
9370 (sibcall_value_register): Remove.
9371 (sibcall_value_immediate): Remove.
9372 * config/nds32/predicates.md (nds32_general_register_operand): New.
9373 (nds32_call_address_operand): New.
9375 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9377 PR rtl-optimization/85167
9378 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9379 bb_defs if *split_p, instead preinitialize it to NULL.
9381 PR tree-optimization/85156
9382 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9383 evaluating the argument multiple times.
9385 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9387 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9389 (_mm_cvtpd_ps): Likewise.
9390 (_mm_cvttpd_epi32): Likewise.
9391 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9392 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9393 vector, pixel, and bool following altivec.h include.
9395 2018-04-03 Martin Sebor <msebor@redhat.com>
9397 * doc/extend.texi (Common Function Attributes): Clarify.
9398 (const attribute): Likewise.
9399 (pure attribute): Likewise.
9401 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9404 * config/i386/i386.c (ix86_expand_vector_set): Use
9405 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9407 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9409 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9410 instructions when changing rounding bits to preserve precision bits
9411 in the x87 control word.
9413 2018-04-03 Martin Liska <mliska@suse.cz>
9415 PR tree-optimization/82491
9416 * rtl.h (strip_offset_and_add): Replace += suboffset with
9417 poly_uint64 () + suboffset.
9419 2018-03-29 Martin Liska <mliska@suse.cz>
9420 Martin Jambor <mjambor@suse.cz>
9423 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9424 param_type is not an integral or pointer type.
9426 2018-04-03 Richard Biener <rguenther@suse.de>
9428 * sese.h (recompute_all_dominators): Remove.
9430 2018-04-02 Martin Sebor <msebor@redhat.com>
9432 * doc/invoke.texi (-Wrestrict): Fix typos.
9434 2018-04-02 Jim Wilson <jimw@sifive.com>
9436 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9437 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9438 (<optab>di3, <optab>si3_extend): Likewise.
9439 (<optab>si3_mask, <optab>si3_mask_1): New.
9440 (<optab>di3_mask, <optab>di3_mask_1): New.
9441 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9442 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9443 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9445 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9447 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9450 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9452 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9453 (nds32_canonicalize_comparison): New function.
9455 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9456 Kito Cheng <kito.cheng@gmail.com>
9457 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9459 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9460 * config/nds32/constants.md (unspec_volatile_element): Add
9461 UNSPEC_VOLATILE_RELAX_GROUP.
9462 * config/nds32/nds32-relax-opt.c: New file.
9463 * config/nds32/nds32-predicates.c
9464 (nds32_symbol_load_store_p): New function.
9465 * config/nds32/nds32-protos.h
9466 (nds32_symbol_load_store_p): Declare function.
9467 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9468 * config/nds32/nds32.c
9469 (nds32_register_pass): New function to register pass.
9470 (nds32_register_passes): New function to register passes.
9471 * config/nds32/nds32.md (relax_group): New pattern.
9472 * config/nds32/nds32.opt (mrelax-hint): New option.
9473 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9475 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9477 * config/nds32/t-nds32: Modify files dependency.
9479 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9481 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9482 (PROFILE_HOOK): Define its implementation.
9484 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9486 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9487 type and 32-bit size.
9489 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9492 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9493 (V_128_256): New mode iterator.
9494 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9495 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9496 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9498 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9499 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9501 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9504 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9505 NaN inputs correctly.
9507 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9510 * config/rs6000/vsx.md (??r): New mode attribute.
9511 (*vsx_mov<mode>_64bit): Use it.
9512 (*vsx_mov<mode>_32bit): Likewise.
9514 2018-03-30 Martin Sebor <msebor@redhat.com>
9516 PR tree-optimization/84818
9517 * builtins.c (check_access): Use warning_n.
9519 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9522 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9524 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9527 2018-03-30 Julia Koval <julia.koval@intel.com>
9530 * x86-tune.def (movx, partial_reg_dependency): Enable for
9533 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9536 * lra-constraints.c (process_alt_operands): Move setting
9537 this_alternative_matches below.
9539 2018-03-29 Martin Liska <mliska@suse.cz>
9542 * doc/invoke.texi: Document how LTO works with debug info.
9543 Describe auto-load support of binutils. Mention 'x86-64'
9544 as valid option value of -march option.
9546 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9548 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9551 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9552 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9553 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9556 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9559 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9560 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9561 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9562 for RS6000_BTM_POWERPC64.
9563 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9564 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9565 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9570 2018-03-28 Carl Love <cel@us.ibm.com>
9573 2017-09-27 Carl Love <cel@us.ibm.com>
9575 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9576 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9577 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9580 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9582 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9583 instead of __vector bool.
9584 (_mm_max_pu8): Likewise.
9585 (_mm_min_pi16): Likewise.
9587 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9590 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9591 (DIVWEUO): Likewise.
9593 (DIVDEUO): Likewise.
9594 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9595 DIVWEUO and DIVDEUO.
9596 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9597 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9598 (div_extend): Likewise.
9599 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9601 (__builtin_divweuo): Likewise.
9602 (__builtin_divdeo): Likewise.
9603 (__builtin_divdeuo): Likewise.
9605 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9608 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9609 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9611 PR tree-optimization/82004
9612 * gimple-match-head.c (optimize_pow_to_exp): New function.
9613 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9614 Don't fold to exp if optimize_pow_to_exp is false.
9616 2018-03-28 Martin Liska <mliska@suse.cz>
9619 * calls.c (initialize_argument_information): Fix trailing space.
9620 * common.opt: Fix typo and provide better explanation for
9621 -fsanitize-coverage option.
9622 * config/i386/i386.opt: Fix typo.
9624 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9625 Martin Liska <mliska@suse.cz>
9628 * gimplify.c (asan_poison_variable): Don't do the check for
9629 gimplify_omp_ctxp here.
9630 (gimplify_decl_expr): Do it here.
9631 (gimplify_target_expr): Likewise.
9633 2018-03-28 Martin Liska <mliska@suse.cz>
9636 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9637 chkp_type_bounds_count if MPX is not enabled.
9639 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9641 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9643 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9646 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9647 function to create the function decl for complex long double
9648 multiply and divide for -mabi=ieeelongdouble.
9649 (init_float128_ieee): Call it.
9651 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9654 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9655 -fcf-protection=branch -mibt.
9656 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9658 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9661 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9662 arm_word_relocations.
9664 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9667 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9668 extern array declarations.
9670 2018-03-27 Richard Biener <rguenther@suse.de>
9673 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9674 explicit single_use checks.
9676 2018-03-27 Richard Biener <rguenther@suse.de>
9678 PR tree-optimization/85082
9679 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9682 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9684 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9685 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9686 Turn on fasynchronous-unwind-tables and funwind-tables.
9688 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9691 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9692 (*bmi_blsr_<mode>_ccz): Ditto.
9694 2018-03-26 Tom de Vries <tom@codesourcery.com>
9696 PR tree-optimization/85063
9697 * omp-general.c (offloading_function_p): New function. Factor out
9699 * omp-offload.c (pass_omp_target_link::gate): ... here.
9700 * omp-general.h (offloading_function_p): Declare.
9701 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9702 with attribute omp declare target for offloading functions.
9704 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9706 PR tree-optimization/84005
9707 * tree-data-ref.h (get_base_for_alignment): Declare.
9708 * tree-data-ref.c (get_base_for_alignment_1): New function.
9709 (get_base_for_alignment): Likewise.
9710 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9711 get_base_for_alignment to find a suitable base object, instead
9712 of always using drb->base_address.
9714 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9717 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9718 known size by default.
9720 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9723 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9724 and non BLKmode operands.
9726 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9729 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9730 Clean up attributes.
9732 2018-03-23 Richard Biener <rguenther@suse.de>
9735 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9736 we are going to emit early debug for LTO.
9738 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9741 * function.c (match_asm_constraints_1): Don't optimize if input
9742 doesn't satisfy general_operand predicate for output's mode.
9745 * alias.c (write_dependence_p): Don't require for x_canonicalized
9746 non-VOIDmode if x has VOIDmode.
9749 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9750 just don't try to optimize it rather than assert it never happens.
9752 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9754 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9755 macro expansions for definition of ST_INTERNAL_<mode> and
9756 LD_INTERNAL_<mode> builtins.
9757 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9759 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9761 (altivec_expand_st_builtin): Likewise.
9762 (altivec_expand_builtin): Remove calls to deleted functions.
9763 (rs6000_address_for_altivec): Delete this function.
9764 * config/rs6000/vector.md: Remove expands for
9765 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9767 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9770 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9771 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9772 re-computing once computed.
9773 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9774 (arm_init_machine_status): Initialize
9775 machine->static_chain_stack_bytes.
9777 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9780 * doc/extend.texi: Add four new prototypes for vec_ld.
9781 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9782 definitions for more logical presentation.
9783 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9784 entries for V1TI variants of __builtin_altivec_ld builtin.
9785 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9786 handling of V1TI variant of LVX icode pattern.
9787 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9788 (rs6000_gimple_fold_builtin): Likewise.
9789 (altivec_init_builtins): Add code to define
9790 __builtin_altivec_lvx_v1ti function.
9792 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9795 * function.c (match_asm_constraints_1): Don't do the optimization
9796 if input isn't a REG, SUBREG, MEM or constant.
9798 2018-03-22 Tom de Vries <tom@codesourcery.com>
9800 PR tree-optimization/84956
9801 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9802 bb_has_abnormal_pred.
9804 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9807 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9808 DECL_INITIAL (decl) to decl at the end.
9809 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9812 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9814 * doc/extend.texi (__builtin_tgmath): Document when complex
9815 integer types are treated as _Complex _Float64.
9817 2018-03-21 Tom de Vries <tom@codesourcery.com>
9819 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9821 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9823 PR tree-optimization/84960
9824 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9825 if it is ENTRY block, move them into single succ of ENTRY in that case.
9827 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9829 PR tree-optimization/84811
9830 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9831 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9832 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9833 (known_subrange_p): Update accordingly. Cast each value involved
9834 in the size comparison, rather than casting the result of the
9837 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9839 PR tree-optimization/84982
9840 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9841 by flipping the least significant bit rather than all bits from
9842 bitpos to bitpos + bitsize - 1.
9844 2018-03-21 Nathan Sidwell <nathan@acm.org>
9846 * doc/extend.texi (Deprecated Features): Remove mention of
9847 long-deleted deprecations.
9849 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9852 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9853 * configure: Regenerate.
9855 2018-03-21 Tom de Vries <tom@codesourcery.com>
9857 PR tree-optimization/83126
9858 * tree-parloops.c (num_phis): New function.
9859 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9861 2018-03-21 Nathan Sidwell <nathan@acm.org>
9863 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9864 mention anon-struct/union members and trailing attributes.
9866 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9868 PR tree-optimization/84969
9869 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9870 builtin memset partitions if they set different rhs values.
9872 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9874 PR rtl-optimization/84989
9875 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9876 VEC_DUPLICATE with scalar result mode.
9878 2018-03-21 Martin Liska <mliska@suse.cz>
9881 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9882 not intended return statement.
9884 2018-03-21 Martin Liska <mliska@suse.cz>
9887 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9888 (chkp_find_bound_slots_1): Limit number of iterations.
9890 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9893 * Minor grammar fixes for x86 options.
9895 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9898 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9899 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9901 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9904 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9905 (altivec_lvx_<mode>_1op): Likewise.
9906 (altivec_stvx_<mode>_2op): Likewise.
9907 (altivec_stvx_<mode>_1op): Likewise.
9908 (altivec_lvx_<VM2:mode>): New define_expand.
9909 (altivec_stvx_<VM2:mode>): Likewise.
9910 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9911 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9912 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9913 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9914 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9915 (rs6000_gen_lvx): Likewise.
9916 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9917 (altivec_expand_stv_builtin): Likewise.
9918 (altivec_expand_builtin): Likewise.
9919 * config/rs6000/vector.md: Likewise.
9921 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9924 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9927 2018-03-20 Richard Biener <rguenther@suse.de>
9930 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9931 sign-conversions as zero, fall back to standard scalar_stmt
9934 2018-03-20 Martin Liska <mliska@suse.cz>
9937 * predict.c (rebuild_frequencies): Handle case when we have
9938 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9940 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9943 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9944 flag_section_anchors.
9945 * varasm.c (use_blocks_for_decl_p): Remove hack for
9946 dw2_force_const_mem.
9949 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9951 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9952 be created, use lowpart_subreg of operands[0] rather than operands[0]
9954 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9955 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9956 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9957 and n constraint instead of aarch64_shift_imm_di and Usd.
9958 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9959 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9961 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9964 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9965 to favor GPR over NEON registers.
9966 (<shift>di3_neon): Likewise.
9968 2018-03-20 Tom de Vries <tom@codesourcery.com>
9971 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9972 (nvptx_process_pars): Emit bar.sync asap and alap.
9974 2018-03-20 Tom de Vries <tom@codesourcery.com>
9977 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9978 seen_label if seen_label is already set.
9980 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9983 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9984 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9985 Use 1U instead of 1. Formatting fixes.
9988 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9989 instead of TREE_TYPE (s1) for the return value.
9991 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9993 PR tree-optimization/84946
9994 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9995 bitsize + bitsize in poly_uint64 rather than poly_int64.
9998 * dwarf2asm.c: Include fold-const.c.
9999 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10000 of decl rather than decl itself.
10002 PR rtl-optimization/84643
10003 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10005 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10008 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10009 calling assemble_variable.
10011 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10014 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10015 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10017 2018-03-19 Jim Wilson <jimw@sifive.com>
10020 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10021 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10022 (riscv_first_stack_step): Likewise.
10023 (riscv_option_override): Use STACK_BOUNDARY instead of
10024 MIN_STACK_BOUNDARY.
10025 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10026 MIN_STACK_BOUNDARY.
10027 (BIGGEST_ALIGNMENT): Set to 128.
10028 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10029 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10032 2018-03-19 Richard Biener <rguenther@suse.de>
10034 PR tree-optimization/84933
10035 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10036 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10038 2018-03-19 Richard Biener <rguenther@suse.de>
10040 PR tree-optimization/84859
10041 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10042 (cond_if_else_store_replacement): Perform sinking operation on
10043 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10044 Generalize what a BB with a single eligible store is.
10046 2018-03-19 Richard Biener <rguenther@suse.de>
10048 PR tree-optimization/84929
10049 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10050 chrec_is_positive against non-chrec arg.
10052 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10055 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10057 2018-03-18 Martin Liska <mliska@suse.cz>
10059 PR rtl-optimization/84635
10060 * regrename.c (build_def_use): Use matches_mode only when
10063 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10065 PR tree-optimization/84913
10066 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10067 vectorize chains of COND_EXPRs.
10069 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10071 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10073 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10075 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10077 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10079 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10081 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10082 Kito Cheng <kito.cheng@gmail.com>
10084 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10085 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10086 (nds32_adjust_reg_alloc_order): New function.
10087 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10089 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10091 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10092 nds32_print_operand, nds32_print_operand_address): Use
10093 HOST_WIDE_INT_PRINT_DEC instead.
10095 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10097 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10099 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10102 * config/i386/i386.c (initial_ix86_tune_features,
10103 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10104 unsigned long long.
10105 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10106 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10107 rather than 1u << ix86_tune. Formatting fix.
10108 (ix86_option_override_internal): Change ix86_arch_mask from
10109 unsigned int to unsigned HOST_WIDE_INT, initialize to
10110 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10111 (ix86_function_specific_restore): Likewise.
10113 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10116 * postreload.c (reload_combine_recognize_pattern): Perform
10117 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10118 truncate_int_for_mode the result for the destination's mode.
10121 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10122 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10124 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10127 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10130 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10131 regno_allocno_class_array and sorted_pseudos.
10132 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10133 insns where regno is used.
10135 2018-03-16 Martin Liska <mliska@suse.cz>
10138 * multiple_target.c (create_dispatcher_calls): Redirect
10139 reference in the symbol table.
10141 2018-03-16 Martin Liska <mliska@suse.cz>
10144 * multiple_target.c (create_dispatcher_calls): Redirect also
10147 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10151 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10152 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10153 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10155 2018-03-16 Julia Koval <julia.koval@intel.com>
10157 * doc/invoke.texi (Skylake Server): Add CLWB.
10158 Cannonlake): Remove CLWB.
10160 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10162 PR tree-optimization/84841
10163 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10165 (FLOAT_ONE_CONST_TYPE): Define.
10166 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10167 (sort_by_operand_rank): Put entries with higher constant_type last
10168 rather than first to match comments.
10170 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10172 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10175 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10178 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10179 check and use address of target always.
10181 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10184 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10185 (indirect_thunk_bnd_needed): Likewise.
10186 (indirect_thunks_used): Likewise.
10187 (indirect_thunks_bnd_used): Likewise.
10188 (indirect_return_needed): New.
10189 (indirect_return_bnd_needed): Likewise.
10190 (output_indirect_thunk_function): Add a bool argument for
10192 (output_indirect_thunk_function): Don't generate alias for
10193 function return thunk.
10194 (ix86_code_end): Call output_indirect_thunk_function to generate
10195 function return thunks.
10196 (ix86_output_function_return): Set indirect_return_bnd_needed
10197 and indirect_return_needed instead of indirect_thunk_bnd_needed
10198 and indirect_thunk_needed.
10200 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10202 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10203 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10204 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10206 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10207 Paul Hua <paul.hua.gm@gmail.com>
10210 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10212 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10214 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10215 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10216 resp. SFmode cases.
10218 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10221 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10222 instead of GET_MODE_SIZE when comparing Units.
10224 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10227 * varasm.c (hash_section): Return an unchangeble hash value
10228 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10229 Return !aarch64_can_use_per_function_literal_pools_p ().
10231 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10234 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10235 rather than address of cmode as last argument to prepare_cmp_insn.
10237 2018-03-15 Julia Koval <julia.koval@intel.com>
10239 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10240 F_AVX512VNNI, F_AVX512BITALG): New.
10242 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10245 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10246 insn for floating-point loads and stores.
10248 2018-03-14 Carl Love <cel@us.ibm.com>
10250 * config/rs6000/rs6000-c.c: Add macro definitions for
10251 ALTIVEC_BUILTIN_VEC_PERMXOR.
10252 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10253 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10254 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10255 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10257 * config/doc/extend.texi: Add prototypes for vec_permxor.
10259 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10262 * diagnostic-show-locus.c (class layout_point): Convert m_line
10263 from int to linenum_type.
10264 (line_span::comparator): Use linenum "compare" function when
10265 comparing line numbers.
10266 (test_line_span): New function.
10267 (layout_range::contains_point): Convert param "row" from int to
10269 (layout_range::intersects_line_p): Likewise.
10270 (layout::will_show_line_p): Likewise.
10271 (layout::print_source_line): Likewise.
10272 (layout::should_print_annotation_line_p): Likewise.
10273 (layout::print_annotation_line): Likewise.
10274 (layout::print_leading_fixits): Likewise.
10275 (layout::annotation_line_showed_range_p): Likewise.
10276 (struct line_corrections): Likewise for field m_row.
10277 (line_corrections::line_corrections): Likewise for param "row".
10278 (layout::print_trailing_fixits): Likewise.
10279 (layout::get_state_at_point): Likewise.
10280 (layout::get_x_bound_for_row): Likewise.
10281 (layout::print_line): Likewise.
10282 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10283 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10284 * input.c (selftest::test_linenum_comparisons): New function.
10285 (selftest::input_c_tests): Call it.
10286 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10287 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10288 * selftest.h (ASSERT_GT): New macro.
10289 (ASSERT_GT_AT): New macro.
10290 (ASSERT_LT): New macro.
10291 (ASSERT_LT_AT): New macro.
10293 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10295 PR rtl-optimization/84780
10296 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10298 2018-03-14 Martin Liska <mliska@suse.cz>
10300 * tree.c (record_node_allocation_statistics): Use
10301 get_stats_node_kind.
10302 (get_stats_node_kind): New function extracted from
10303 record_node_allocation_statistics.
10304 (free_node): Use get_stats_node_kind.
10306 2018-03-14 Richard Biener <rguenther@suse.de>
10308 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10309 that the value-set of ANTIC_IN doesn't grow.
10312 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10314 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10315 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10317 2018-03-14 Julia Koval <julia.koval@intel.com>
10319 * config.gcc (icelake-client, icelake-server): New.
10321 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10322 (initial_ix86_arch_features): Ditto.
10323 (PTA_SKYLAKE): Add SGX.
10324 (PTA_ICELAKE): Remove.
10325 (PTA_ICELAKE_CLIENT): New.
10326 (PTA_ICELAKE_SERVER): New.
10327 (ix86_option_override_internal): Split up icelake on icelake client and
10329 (get_builtin_code_for_version): Ditto.
10330 (fold_builtin_cpu): Ditto.
10331 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10332 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10333 * config/i386/i386.h (processor_type): Ditto.
10334 * doc/invoke.texi: Ditto.
10336 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10339 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10340 INTEGER_CST offset, add it together with bitpos / 8 and
10341 sign extend based on POINTER_SIZE.
10345 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10348 * config/i386/constraints.md (Yc): New register constraint.
10349 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10350 Use Yc constraint for alternative 2 of operand 0. Remove
10351 preferred_for_speed attribute.
10353 2018-03-14 Richard Biener <rguenther@suse.de>
10355 PR tree-optimization/84830
10356 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10357 with the old one to avoid oscillations.
10359 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10362 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10364 (assign_by_spills): Return a flag of reload assignment failure.
10365 Do not process the reload assignment failures. Do not spill other
10366 reload pseudos if they has the same reg class. Update n if
10368 (lra_assign): Add a return arg. Set up from the result of
10369 assign_by_spills call.
10370 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10371 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10372 usage_insns if it is not NULL.
10373 (spill_hard_reg_in_range): New function.
10374 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10375 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10376 function prototypes.
10377 (lra_assign): Change prototype.
10378 * lra.c (lra): Add code to deal with fails by splitting hard reg
10381 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10383 * config/riscv/riscv.opt (mrelax): New option.
10384 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10385 "norelax" when riscv_mrelax is disabled.
10386 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10388 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10391 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10392 reassociation for int modes.
10394 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10396 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10397 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10399 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10400 * config/aarch64/aarch64-sve.md
10401 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10402 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10403 (*extend<mode><Vwide>2): Rename to...
10404 (aarch64_sve_extend<mode><Vwide>2): ...this.
10405 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10406 renaming the old pattern to...
10407 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10409 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10410 define_expand, renaming the old pattern to...
10411 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10412 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10413 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10414 account when deciding which SVE instruction the optab should use.
10415 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10417 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10419 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10420 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10421 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10422 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10423 (tlsdesc_small_<mode>): Turn a define_expand and use
10424 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10425 (tlsdesc_small_advsimd_<mode>): ...this.
10426 (tlsdesc_small_sve_<mode>): New pattern.
10428 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10430 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10431 (UNSPEC_UMUL_HIGHPART): New constants.
10432 (MUL_HIGHPART): New int iteraor.
10433 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10434 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10436 (*<su>mul<mode>3_highpart): New define_insn.
10438 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10441 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10444 2018-03-13 Martin Liska <mliska@suse.cz>
10447 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10449 (sem_item_optimizer::~sem_item_optimizer): Release it.
10450 (sem_item_optimizer::merge_classes): Register variable aliases.
10451 (sem_item_optimizer::fixup_pt_set): New function.
10452 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10453 * ipa-icf.h: Declare new variables and functions.
10455 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10457 PR middle-end/84834
10458 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10459 integer_pow2p@2 and test integer_pow2p in condition.
10460 (A < 0 ? C : 0): Similarly for @1.
10462 PR middle-end/84831
10463 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10464 characters starting at p contain '\0' character, don't look beyond
10468 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10469 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10472 * reg-stack.c (change_stack): Change update_end var from int to
10473 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10474 also call set_block_for_insn on the newly added insns and rescan.
10477 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10478 on the last operand.
10481 * tree.c (stabilize_reference_1): Return save_expr (e) for
10482 STATEMENT_LIST even if it doesn't have side-effects.
10484 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10486 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10488 2018-03-12 Renlin Li <renlin.li@arm.com>
10490 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10491 aarch64_output_scalar_simd_mov_immediate.
10493 2018-03-12 Martin Sebor <msebor@redhat.com>
10495 PR tree-optimization/83456
10496 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10497 for perfectly overlapping calls to memcpy.
10498 (gimple_fold_builtin_memory_chk): Same.
10499 (gimple_fold_builtin_strcpy): Handle no-warning.
10500 (gimple_fold_builtin_stxcpy_chk): Same.
10501 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10503 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10505 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10506 parameter. Use it for SFmode.
10507 (rs6000_function_arg_advance_1): Adjust.
10508 (rs6000_function_arg): Adjust.
10509 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10511 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10513 PR rtl-optimization/84169
10514 PR rtl-optimization/84780
10515 * combine.c (can_combine_p): Check for a 2-insn combination whether
10516 the destination register is used between the two insns, too.
10518 2018-03-12 Richard Biener <rguenther@suse.de>
10520 PR tree-optimization/84803
10521 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10522 for refs DR analysis didn't process.
10524 2018-03-12 Richard Biener <rguenther@suse.de>
10526 PR tree-optimization/84777
10527 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10528 force-vectorize loops ignore whether we are optimizing for size.
10530 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10532 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10533 (TARGET_MD_ASM_ADJUST): Define.
10535 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10536 Kito Cheng <kito.cheng@gmail.com>
10537 Chung-Ju Wu <jasonwucj@gmail.com>
10539 * config/nds32/nds32.c (nds32_compute_stack_frame,
10540 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10541 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10542 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10543 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10544 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10545 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10546 * config/nds32/nds32.md (prologue, epilogue): Use macro
10547 NDS32_V3PUSH_AVAILABLE_P to do checking.
10549 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10552 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10553 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10554 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10555 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10556 addition of most attributes on !orig_type_die or the attribute not
10557 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10559 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10560 Chung-Ju Wu <jasonwucj@gmail.com>
10562 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10563 __NDS32_VH__ macro.
10564 * config/nds32/nds32.opt (mvh): New option.
10566 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10567 Chung-Ju Wu <jasonwucj@gmail.com>
10569 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10571 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10572 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10575 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10576 Chung-Ju Wu <jasonwucj@gmail.com>
10578 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10580 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10581 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10583 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10584 Kito Cheng <kito.cheng@gmail.com>
10585 Chung-Ju Wu <jasonwucj@gmail.com>
10587 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10588 UNSPEC_FFMISM and UNSPEC_FLMISM.
10589 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10590 for ffb, ffmism and flmism.
10591 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10592 (unspec_ffmism): Ditto.
10593 (unspec_flmism): Ditto.
10594 (nds32_expand_builtin_impl): Check if string extension is available.
10595 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10596 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10598 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10601 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10604 * lra-assigns.c (assign_by_spills): Return a flag of reload
10605 assignment failure. Do not process the reload assignment
10606 failures. Do not spill other reload pseudos if they has the same
10608 (lra_assign): Add a return arg. Set up from the result of
10609 assign_by_spills call.
10610 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10611 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10612 usage_insns if it is not NULL.
10613 (spill_hard_reg_in_range): New function.
10614 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10615 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10616 function prototypes.
10617 (lra_assign): Change prototype.
10618 * lra.c (lra): Add code to deal with fails by splitting hard reg
10621 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10624 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10626 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10629 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10630 (dw_val_node): Add val_symbolic_view.
10631 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10632 (symview_upper_bound): New.
10633 (new_line_info_table): Initialize symviews_since_reset.
10634 (dwarf2out_source_line): Count symviews_since_reset and set
10635 symview_upper_bound.
10636 (dw_val_equal_p): Handle symview.
10637 (add_AT_symview): New.
10638 (print_dw_val): Handle symview.
10639 (attr_checksum, attr_checksum_ordered): Likewise.
10640 (same_dw_val_p, size_of_die): Likewise.
10641 (value_format, output_die): Likewise.
10642 (add_high_low_attributes): Use add_AT_symview for entry_view.
10643 (dwarf2out_finish): Reset symview_upper_bound, clear
10646 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10649 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10650 Add strict argument and use it.
10651 (rs6000_split_multireg_move): Update for new strict argument.
10652 (mem_operand_gpr): Disallow all non-offsettable addresses.
10653 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10655 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10658 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10659 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10660 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10663 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10664 decl, use remap_type if we want to use the type.
10666 2018-03-09 Martin Sebor <msebor@redhat.com>
10668 PR tree-optimization/84526
10669 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10671 (builtin_access::generic_overlap): Be prepared to handle non-array
10674 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10676 PR rtl-optimization/84682
10677 * lra-constraints.c (process_address_1): Check is_address flag
10678 for address constraints.
10679 (process_alt_operands): Likewise.
10680 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10681 preprocess_constraints.
10682 * recog.h (preprocess_constraints): Add oploc parameter.
10684 * recog.c (preprocess_constraints): Test address_operand for
10685 CT_ADDRESS constraints.
10687 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10690 * lra-assigns.c (assign_by_spills): Return a flag of reload
10691 assignment failure. Do not process the reload assignment
10692 failures. Do not spill other reload pseudos if they has the same
10694 (lra_assign): Add a return arg. Set up from the result of
10695 assign_by_spills call.
10696 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10697 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10698 usage_insns if it is not NULL.
10699 (spill_hard_reg_in_range): New function.
10700 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10701 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10702 function prototypes.
10703 (lra_assign): Change prototype.
10704 * lra.c (lra): Add code to deal with fails by splitting hard reg
10707 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10710 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10711 Accept complain bool parameter. Only emit errors if it is true.
10712 (arm_parse_cpu_option_name): Likewise.
10713 (arm_target_thumb_only): Adjust callers of the above.
10714 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10715 prototype to take a default true bool parameter.
10716 (arm_parse_arch_option_name): Likewise.
10718 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10719 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10723 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10724 * configure: Regenerate.
10725 * configure.ac ("linker --version-script option"): New.
10726 ("linker soname option"): New.
10728 2018-03-09 Richard Biener <rguenther@suse.de>
10730 PR tree-optimization/84775
10731 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10732 immediate uses of predicate stmts and mark them modified.
10735 PR tree-optimization/84178
10736 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10738 (version_loop_for_if_conversion): Delay update_ssa call.
10739 (tree_if_conversion): Delay update_ssa until after predicate
10742 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10745 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10746 when the function accesses prior frames.
10748 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10751 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10752 gen_llsym, otherwise call maybe_gen_llsym.
10754 PR inline-asm/84742
10755 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10756 has ',' character inside of it.
10758 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10761 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10762 as clobbering CC_REGNUM.
10764 2018-03-08 Richard Biener <rguenther@suse.de>
10766 PR middle-end/84552
10767 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10768 (follow_copies_to_constant): Do not follow SSA names registered
10771 2018-03-08 Richard Biener <rguenther@suse.de>
10773 PR tree-optimization/84178
10774 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10776 (version_loop_for_if_conversion): Delay update_ssa call.
10777 (tree_if_conversion): Delay update_ssa until after predicate
10780 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10782 PR tree-optimization/84178
10783 * tree-if-conv.c (release_bb_predicate): Remove the
10784 the assertion that the stmts have NULL use_ops.
10785 Discard the statements, asserting that they haven't
10786 yet been added to a BB.
10788 2018-03-08 Richard Biener <rguenther@suse.de>
10790 PR tree-optimization/84746
10791 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10792 (phi_translate): Pass in destination ANTIC_OUT set.
10793 (phi_translate_1): Likewise. For a simplified result lookup
10794 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10795 (phi_translate_set): Adjust.
10796 (do_pre_regular_insertion): Likewise.
10797 (do_pre_partial_partial_insertion): Likewise.
10799 2018-03-08 Martin Liska <mliska@suse.cz>
10801 PR gcov-profile/84735
10802 * doc/gcov.texi: Document usage of profile files.
10803 * gcov-io.h: Document changes in the format.
10805 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10809 * dwarf2out.c (struct dw_line_info_table): Update comments for
10811 (FORCE_RESET_NEXT_VIEW): New.
10812 (FORCE_RESETTING_VIEW_P): New.
10813 (RESETTING_VIEW_P): Check for -1 too.
10814 (ZERO_VIEW_P): Likewise.
10815 (new_line_info_table): Force-reset next view.
10816 (dwarf2out_begin_function): Likewise.
10817 (dwarf2out_source_line): Simplify zero_view_p initialization.
10818 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10819 view directly. Omit view when omitting .loc at line 0.
10821 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10823 PR tree-optimization/84740
10824 * tree-switch-conversion.c (process_switch): Call build_constructors
10825 only if info.phi_count is non-zero.
10827 PR tree-optimization/84739
10828 * tree-tailcall.c (find_tail_calls): Check call arguments against
10829 DECL_ARGUMENTS (current_function_decl) rather than
10830 DECL_ARGUMENTS (func) when checking for tail recursion.
10832 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10834 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10835 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10836 Volker Reichelt's entry and add entries for people that perform
10837 GCC fuzzy testing and report numerous bugs.
10839 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10842 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10843 readonly data in sdata, if that is disabled.
10844 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10845 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10846 -mreadonly-in-sdata option.
10848 2018-03-07 Martin Sebor <msebor@redhat.com>
10850 PR tree-optimization/84468
10851 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10852 basic block when looking for nul assignment.
10854 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10857 * except.h (output_function_exception_table): Adjust prototype.
10858 * except.c (output_function_exception_table): Remove FNNAME parameter
10859 and add SECTION parameter. Ouput one part of the table at a time.
10860 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10861 the first part of the exception table and emit unwind directives.
10862 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10863 (i386_pe_seh_cold_init): Likewise.
10864 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10865 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10866 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10867 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10868 * config/i386/winnt.c: Include except.h.
10869 (struct seh_frame_state): Add reg_offset, after_prologue and
10870 in_cold_section fields.
10871 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10872 (i386_pe_seh_cold_init): New function.
10873 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10874 to seh->in_cold_section.
10875 (seh_emit_push): Record the offset of the push.
10876 (seh_emit_save): Record the offet of the save.
10877 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10878 Test seh->after_prologue to disregard the epilogue.
10879 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10880 (i386_pe_end_cold_function): New function.
10882 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10885 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10886 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10889 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10891 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10892 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10894 PR middle-end/84723
10895 * multiple_target.c: Include tree-inline.h and intl.h.
10896 (expand_target_clones): Diagnose and fail if node->definition and
10897 !tree_versionable_function_p (node->decl).
10899 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10901 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10903 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10904 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10905 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10907 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10910 * combine.c (try_combine): Use reg_or_subregno instead of handling
10911 just paradoxical SUBREGs and REGs.
10913 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10915 * config/arc/arc.c (arc_finalize_pic): Remove function.
10916 (arc_must_save_register): We use single base PIC register, remove
10917 checks to save/restore the PIC register.
10918 (arc_expand_prologue): Likewise.
10919 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10921 (arc_verify_short): Likewise.
10922 (arc_attr_type): Likewise.
10923 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10924 (walk_stores): Likewise.
10925 (arc_address_cost): Make it static.
10926 (arc_verify_short): Likewise.
10927 (branch_dest): Likewise.
10928 (arc_attr_type): Likewise.
10929 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10930 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10931 (arc_final_prescan_insn): Remove inserting the nops due to
10932 hardware hazards. It is done in reorg step.
10933 (insn_length_variant_t): Remove.
10934 (insn_length_parameters_t): Likewise.
10935 (arc_insn_length_parameters): Likewise.
10936 (arc_get_insn_variants): Likewise.
10937 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10939 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10941 PR inline-asm/84683
10942 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10945 PR tree-optimization/84687
10946 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10948 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10950 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10952 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10953 Rename to ppc_speculation_barrier.
10954 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10955 __builtin_ppc_speculation_barrier.
10957 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10960 * combine.c (combine_simplify_rtx): Don't try to simplify if
10961 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10964 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10966 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10967 to 32 bytes when compiling for POWER9.
10969 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10972 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10973 regparm >= 3 with no arg reg available also for calls with
10974 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10977 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10979 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10981 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10984 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10986 2018-03-05 Richard Biener <rguenther@suse.de>
10988 PR tree-optimization/84486
10989 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10990 When inserting a __builtin_assume_aligned call set the LHS
10991 SSA name alignment info accordingly.
10993 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10995 PR tree-optimization/84114
10996 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10997 Avoid reassociation of FLOAT_MODE addition.
10999 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11001 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11002 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11003 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11004 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11005 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11006 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11007 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11009 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11010 __builtin_ia32_wbinvd): New builtins.
11011 (SPECIAL_ARGS2): New.
11012 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11013 (SPECIAL_ARGS2): New.
11014 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11015 (ix86_valid_target_attribute_inner_p): Ditto.
11016 (ix86_init_mmx_sse_builtins): Add special_args2.
11017 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11018 TARGET_WBNOINVD_P): New.
11019 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11020 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11021 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11022 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11023 * config/i386/pconfigintrin.h: New file.
11024 * config/i386/wbnoinvdintrin.h: Ditto.
11025 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11027 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11029 2018-03-05 Richard Biener <rguenther@suse.de>
11031 PR tree-optimization/84670
11032 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11034 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11035 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11036 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11037 doesn't grow if all successors (recursively) were visited at least
11040 2018-03-05 Richard Biener <rguenther@suse.de>
11042 PR tree-optimization/84650
11043 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11044 if executed in the loop pipeline.
11046 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11048 * doc/configfiles.texi (Configuration Files): Move info about
11049 conditionalizing $target-protos.h to...
11050 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11051 differs from $target-protos.h.
11053 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11054 Chung-Ju Wu <jasonwucj@gmail.com>
11056 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11057 * config/nds32/nds32-multiple.md (setmemsi): Define.
11058 * config/nds32/nds32-memory-manipulation.c
11059 (nds32_gen_dup_4_byte_to_word_value): New.
11060 (emit_setmem_word_loop): New.
11061 (emit_setmem_byte_loop): New.
11062 (nds32_expand_setmem_loop): New.
11063 (nds32_expand_setmem_loop_v3m): New.
11064 (nds32_expand_setmem_unroll): New.
11065 (nds32_expand_setmem): New.
11067 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11068 Chung-Ju Wu <jasonwucj@gmail.com>
11070 * config/nds32/nds32-memory-manipulation.c
11071 (nds32_emit_load_store): New.
11072 (nds32_emit_post_inc_load_store): New.
11073 (nds32_emit_mem_move): New.
11074 (nds32_emit_mem_move_block): New.
11075 (nds32_expand_movmemsi_loop_unknown_size): New.
11076 (nds32_expand_movmemsi_loop_known_size): New.
11077 (nds32_expand_movmemsi_loop): New.
11078 (nds32_expand_movmemsi_unroll): New.
11079 (nds32_expand_movmemqi): Rename ...
11080 (nds32_expand_movmemsi): ... to this.
11081 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11082 (movmemsi): ... to this.
11083 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11084 (nds32_expand_movmemsi): ... to this.
11086 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11087 Monk Chiang <sh.chiang04@gmail.com>
11088 Chung-Ju Wu <jasonwucj@gmail.com>
11090 * config/nds32/nds32-protos.h
11091 (nds32_expand_load_multiple): New arguments.
11092 (nds32_expand_store_multiple): Ditto.
11093 (nds32_valid_multiple_load_store): Rename ...
11094 (nds32_valid_multiple_load_store_p): ... to this.
11095 * config/nds32/nds32-memory-manipulation.c
11096 (nds32_expand_load_multiple): Refine implementation.
11097 (nds32_expand_store_multiple): Ditto.
11098 * config/nds32/nds32-multiple.md
11099 (load_multiple): Update nds32_expand_load_multiple interface.
11100 (store_multiple): Update nds32_expand_store_multiple interface.
11101 * config/nds32/nds32-predicates.c
11102 (nds32_valid_multiple_load_store): Rename ...
11103 (nds32_valid_multiple_load_store_p): ... to this and refine
11105 * config/nds32/predicates.md
11106 (nds32_load_multiple_and_update_address_operation): New predicate.
11107 (nds32_store_multiple_and_update_address_operation): New predicate.
11109 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11110 Chung-Ju Wu <jasonwucj@gmail.com>
11112 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11113 (combo): New attribute.
11114 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11116 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11118 * config/nds32/nds32.opt: Change -mcmodel= default value.
11120 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11121 Monk Chiang <sh.chiang04@gmail.com>
11122 Chung-Ju Wu <jasonwucj@gmail.com>
11124 * config/nds32/constants.md (unspec_element): New enum.
11125 * config/nds32/constraints.md (Umw): New constraint.
11126 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11127 * config/nds32/nds32-intrinsic.md: Likewise.
11128 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11129 (nds32_valid_smw_lwm_base_p): New.
11130 (nds32_output_smw_single_word): New.
11131 (nds32_output_lmw_single_word): New.
11132 (nds32_expand_unaligned_load): New.
11133 (nds32_expand_unaligned_store): New.
11134 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11135 (nds32_output_smw_single_word): Declare.
11136 (nds32_output_lmw_single_word): Declare.
11137 (nds32_expand_unaligned_load): Declare.
11138 (nds32_expand_unaligned_store): Declare.
11139 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11140 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11141 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11142 NDS32_BUILTIN_UASTORE_DW.
11143 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11146 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11147 Kito Cheng <kito.cheng@gmail.com>
11148 Chung-Ju Wu <jasonwucj@gmail.com>
11150 * config/nds32/nds32-intrinsic.c
11151 (nds32_expand_builtin_null_ftype_reg): Delete.
11152 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11153 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11154 (nds32_read_argument): New.
11155 (nds32_legitimize_target): Ditto.
11156 (nds32_legitimize_argument): Ditto.
11157 (nds32_check_constant_argument): Ditto.
11158 (nds32_expand_unop_builtin): Ditto.
11159 (nds32_expand_unopimm_builtin): Ditto.
11160 (nds32_expand_binop_builtin): Ditto.
11161 (nds32_builtin_decl_impl): Ditto.
11162 (builtin_description): Ditto.
11163 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11164 (nds32_init_builtins_impl): Ditto.
11165 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11166 (nds32_builtin_decl): New.
11167 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11168 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11170 2018-03-02 Jeff Law <law@redhat.com>
11172 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11173 (redundant_insn, fill_simple_delay_slots): Likewise.
11174 (fill_slots_from_thread): Likewise.
11175 * resource.c (mark_referenced_resources): Likewise.
11176 (mark_set_resources, find_dead_or_set_registers): Likewise.
11178 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11180 * substring-locations.h (format_warning_va): Formatting fix for
11181 ATTRIBUTE_GCC_DIAG.
11182 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11184 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11185 * substring-locations.c: Include intl.h.
11186 (format_warning_va): Turned into small wrapper around
11187 format_warning_n_va, renamed to ...
11188 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11189 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11191 (format_warning_at_substring_n): New function.
11192 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11193 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11194 format_warning_at_substring with just a shorter name instead of
11195 const function pointer.
11196 (fmtwarn_n): New function.
11197 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11198 appropriate, get rid of all the fmtstr temporaries, move conditionals
11199 with G_() wrapped string literals directly into fmtwarn arguments,
11200 cast dir.len to (int), formatting fixes.
11202 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11204 * doc/invoke.texi: Remove "Cilk Plus" references.
11206 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11207 Richard Biener <rguenther@suse.de>
11210 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11211 for error or warning attributes if CALL_FROM_THUNK_P is set.
11214 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11217 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11218 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11221 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11222 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11224 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11225 instead of -1U in last predictors element's probability member.
11227 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11230 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11231 arguments if they are comparable.
11233 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11235 PR tree-optimization/84634
11236 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11237 masks and masked_loop_p with a single loop_masks, making sure it's
11238 null for bb vectorization.
11240 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11242 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11243 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11244 loop->force_vectorize to check whether there is no alias.
11246 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11249 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11251 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11253 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11254 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11255 instead of a loop around prev_real_insn.
11256 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11259 PR inline-asm/84625
11260 * config/i386/i386.c (ix86_print_operand): Use conditional
11261 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11264 2018-03-02 Richard Biener <rguenther@suse.de>
11266 PR tree-optimization/84427
11267 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11268 (bitmap_set_subtract_values): Rewrite to handle multiple
11271 (prune_clobbered_mems): Likewise.
11272 (phi_translate): Take edge instead of pred/phiblock.
11273 (phi_translate_1): Likewise.
11274 (phi_translate_set): Likewise. Insert all translated
11275 exprs for a value into the set, keeping possibly multiple
11276 expressions per value.
11277 (compute_antic_aux): Adjust for phi_translate changes.
11278 When intersecting union the expressions and prune those
11279 not in the final value set, keeping possibly multiple
11280 expressions per value. Do not use value-insertion
11281 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11282 all expressions. Add verification that the value-sets
11283 only shrink during iteration.
11284 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11285 (do_pre_regular_insertion): Likewise.
11286 (do_pre_partial_partial_insertion): Likewise.
11288 2018-03-02 Richard Biener <rguenther@suse.de>
11291 * config/darwin.c (saved_debug_info_level): New static global.
11292 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11293 (darwin_asm_lto_end): Restore debug info generation settings.
11295 2018-03-01 Martin Liska <mliska@suse.cz>
11298 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11299 volatile arguments.
11301 2018-03-01 Richard Biener <rguenther@suse.de>
11304 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11305 type annotation with LTO when debug was disabled at compile-time.
11307 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11309 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11311 (mips_final_postscan_insn): Likewise.
11313 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11315 PR rtl-optimization/84528
11316 * alias.c (init_alias_target): Add commentary.
11317 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11318 a unique base value if the frame pointer is not eliminated
11319 to the stack pointer.
11321 2018-03-01 Tom de Vries <tom@codesourcery.com>
11323 PR rtl-optimization/83327
11324 * lra-int.h (hard_regs_spilled_into): Declare.
11325 * lra.c (hard_regs_spilled_into): Define.
11326 (init_reg_info): Init hard_regs_spilled_into.
11327 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11328 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11329 (process_bb_lives): Handle hard_regs_spilled_into.
11330 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11331 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11333 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11335 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11336 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11337 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11338 * config/rs6000/aix72.h: New file.
11340 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11342 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11343 instead of warning_at with conditional singular and plural messages
11347 * stor-layout.c (update_alignment_for_field): For
11348 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11349 && !DECL_PACKED (field), do the alignment update, just use
11350 only desired_align instead of MAX (type_align, desired_align)
11352 (place_field): Don't do known_align < desired_align handling
11353 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11354 is non-NULL, instead do it after rli->prev_field handling and
11355 only if not within a bitfield word. For DECL_PACKED (field)
11356 use type_align of BITS_PER_UNIT.
11358 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11360 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11361 superfluous parentheses and trailing spaces.
11363 2018-02-28 Richard Biener <rguenther@suse.de>
11365 PR tree-optimization/84584
11366 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11367 SCoPs with fake exit edge.
11369 2018-02-28 Martin Liska <mliska@suse.cz>
11372 * timevar.c (timer::print): Fix format to properly print 100%
11375 2018-02-28 Richard Biener <rguenther@suse.de>
11377 PR middle-end/84607
11378 * genmatch.c (capture_info::walk_match): Do not mark
11379 captured expressions without operands as expr_p given
11380 they act more like predicates and should be subject to
11381 "lost tail" side-effect preserving.
11383 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11385 PR rtl-optimization/81611
11386 * auto-inc-dec.c (attempt_change): Move dead note from
11387 mem_insn if it's the next use of regno
11388 (find_address): Take address use of reg holding
11389 non-incremented value. Add parm to limit search to the named
11391 (merge_in_block): Attempt to use a mem insn that is the next
11392 use of the original regno.
11394 2018-02-27 Martin Sebor <msebor@redhat.com>
11397 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11398 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11400 2018-02-27 Martin Sebor <msebor@redhat.com>
11402 PR translation/84207
11403 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11404 n argument to unsigned HOST_WIDE_INT.
11405 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11406 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11407 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11408 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11410 2018-02-27 Richard Biener <rguenther@suse.de>
11412 PR tree-optimization/84512
11413 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11414 Do not use the estimate returned from record_stmt_cost for
11415 the scalar iteration cost but sum properly using add_stmt_cost.
11417 2018-02-27 Richard Biener <rguenther@suse.de>
11419 PR tree-optimization/84466
11420 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11421 Adjust last change to less strictly validate use operands.
11423 2018-02-27 Martin Liska <mliska@suse.cz>
11425 PR gcov-profile/84548
11426 * gcov.c (process_file): Allow partial overlap and consider it
11427 also as group functions.
11428 (output_lines): Properly calculate range of lines for a group.
11430 2018-02-27 Martin Liska <mliska@suse.cz>
11432 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11433 'ggc' suffixes. Change first column width.
11434 (timer::print): Fix formatting of the column.
11436 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11438 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11439 preserve inline entry blocks for the sake of debug inline
11440 entry point markers alone.
11441 (remove_unused_locals): Suggest in comments a better place to
11442 force the preservation of inline entry blocks that are
11443 otherwise unused, but do not preserve them.
11445 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11447 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11449 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11452 * config/i386/constraints.md (Bs): Replace
11453 ix86_indirect_branch_register with
11454 TARGET_INDIRECT_BRANCH_REGISTER.
11456 * config/i386/i386.md (indirect_jump): Likewise.
11457 (tablejump): Likewise.
11458 (*sibcall_memory): Likewise.
11459 (*sibcall_value_memory): Likewise.
11460 Peepholes of indirect call and jump via memory: Likewise.
11461 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11462 (*sibcall_value_GOT_32): Likewise.
11463 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11464 (GOT_memory_operand): Likewise.
11465 (call_insn_operand): Likewise.
11466 (sibcall_insn_operand): Likewise.
11467 (GOT32_symbol_operand): Likewise.
11468 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11470 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11472 PR rtl-optimization/83496
11473 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11474 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11475 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11476 redundant insn, if any.
11477 (relax_delay_slots): Likewise.
11478 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11480 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11482 PR tree-optimization/83965
11483 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11484 that grouped statements are part of a reduction chain. Return
11485 true if the statement is not marked as a reduction itself but
11486 is part of a group.
11487 (vect_recog_dot_prod_pattern): Don't check whether the statement
11488 is part of a group here.
11489 (vect_recog_sad_pattern): Likewise.
11490 (vect_recog_widen_sum_pattern): Likewise.
11492 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11495 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11497 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11500 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11502 (ix86_output_indirect_function_return): New prototype.
11503 (ix86_split_simple_return_pop_internal): Likewise.
11504 * config/i386/i386.c (indirect_return_via_cx): New.
11505 (indirect_return_via_cx_bnd): Likewise.
11506 (indirect_thunk_name): Handle return va CX_REG.
11507 (output_indirect_thunk_function): Create alias for
11508 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11509 (ix86_output_indirect_jmp): Remove the bool argument.
11510 (ix86_output_indirect_function_return): New function.
11511 (ix86_split_simple_return_pop_internal): Likewise.
11512 * config/i386/i386.md (*indirect_jump): Don't pass false
11513 to ix86_output_indirect_jmp.
11514 (*tablejump_1): Likewise.
11515 (simple_return_pop_internal): Change it to define_insn_and_split.
11516 Call ix86_split_simple_return_pop_internal to split it for
11517 -mfunction-return=.
11518 (simple_return_indirect_internal): Call
11519 ix86_output_indirect_function_return instead of
11520 ix86_output_indirect_jmp.
11522 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11525 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11526 memset and value initialization afterwards.
11528 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11530 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11532 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11535 * common/config/aarch64/aarch64-common.c
11536 (aarch_option_optimization_table[]): Switch
11537 off fomit-frame-pointer
11539 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11540 Chung-Ju Wu <jasonwucj@gmail.com>
11542 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11544 (store_multiple): Ditto.
11546 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11548 * config.gcc: Add --with-cpu support for nds32 target.
11549 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11550 * config/nds32/nds32.opt: Add -mcpu= option.
11552 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11554 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11555 isel=yes): Warn for these deprecated options.
11557 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11559 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11560 ISA_2_5_MASKS_EMBEDDED.
11562 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11564 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11565 p->max as pointers rather than using iterative_hash_expr.
11567 2018-02-23 Carl Love <cel@us.ibm.com>
11569 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11570 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11572 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11573 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11574 P8V_BUILTIN_VEC_VUNSIGNED2.
11576 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11579 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11580 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11582 * lra-constraints.c (curr_insn_transform): Set up
11583 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11585 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11586 * lra-eliminations.c (spill_pseudos): Ditto.
11587 (process_insn_for_elimination): Ditto.
11588 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11589 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11590 LRA_NON_CLOBBERED_ALT.
11592 2018-02-22 Martin Sebor <msebor@redhat.com>
11594 PR tree-optimization/84480
11595 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11596 to maybe_diag_stxncpy_trunc. Call it.
11597 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11598 from gimple_fold_builtin_strcpy. Print inlining stack.
11599 (handle_builtin_stxncpy): Print inlining stack.
11600 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11602 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11605 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11606 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11607 and -fcheck-pointer-bounds are used together.
11608 (indirect_thunk_prefix): New enum.
11609 (indirect_thunk_need_prefix): New function.
11610 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11611 "_nt" instead of "_bnd" for NOTRACK prefix.
11612 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11613 (output_indirect_thunk_function): Likewise.
11615 (ix86_code_end): Update output_indirect_thunk_function calls.
11616 (ix86_output_indirect_branch_via_reg): Replace
11617 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11618 (ix86_output_indirect_branch_via_push): Likewise.
11619 (ix86_output_function_return): Likewise.
11620 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11621 incompatible with -fcf-protection=branch and
11622 -fcheck-pointer-bounds.
11624 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11627 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11628 Change gcc_assert call to output_operand_lossage.
11630 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11632 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11634 2018-02-22 DJ Delorie <dj@redhat.com>
11635 Sebastian Perta <sebastian.perta@renesas.com>
11636 Oleg Endo <olegendo@gcc.gnu.org>
11638 * config/rx/rx.c (rx_rtx_costs): New function.
11639 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11641 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11643 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11645 2018-02-22 Martin Liska <mliska@suse.cz>
11648 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11649 Add "native" as a possible value.
11651 2018-02-22 Martin Liska <mliska@suse.cz>
11654 * config/i386/i386.c (ix86_option_override_internal):
11655 Add "native" as a possible value for -march and -mtune.
11657 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11660 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11661 to all type variants.
11663 PR tree-optimization/84503
11664 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11665 width as info->bitpos + info->bitsize - start.
11666 (merged_store_group::merge_overlapping): Simplify width computation.
11667 (check_no_overlap): New function.
11668 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11669 start + width and last_order of the group, fail if check_no_overlap
11671 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11672 to group if check_no_overlap fails.
11674 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11676 * config/rs6000/altivec.md: Delete contraint arguments to
11677 define_expand, define_split, and define_peephole2, and in
11678 define_insn_and_split if always unused.
11679 * config/rs6000/darwin.md: Ditto.
11680 * config/rs6000/dfp.md: Ditto.
11681 * config/rs6000/rs6000.md: Ditto.
11682 * config/rs6000/sync.md: Ditto.
11683 * config/rs6000/vector.md: Ditto.
11684 * config/rs6000/vsx.md: Ditto.
11686 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11688 * config/rs6000/altivec.md: Write output control strings as braced
11689 blocks instead of double-quoted strings.
11690 * config/rs6000/darwin.md: Ditto.
11691 * config/rs6000/rs6000.md: Ditto.
11692 * config/rs6000/vector.md: Ditto.
11693 * config/rs6000/vsx.md: Ditto.
11695 2018-02-21 Jason Merrill <jason@redhat.com>
11697 PR c++/84314 - ICE with templates and fastcall attribute.
11698 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11700 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11702 * ipa-cp.c (determine_versionability): Fix comment typos.
11704 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11707 * ipa-cp.c (determine_versionability): Do not version functions caling
11710 2018-02-21 Martin Liska <mliska@suse.cz>
11713 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11714 Add "native" as a possible value.
11715 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11716 the macro when native cpu detection is available.
11718 2018-02-21 Martin Liska <mliska@suse.cz>
11721 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11722 Add "native" as a possible value.
11723 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11724 when native cpu detection is available.
11726 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11727 Martin Sebor <msebor@redhat.com>
11729 PR tree-optimization/84478
11730 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11732 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11733 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11734 support which is conservatively correct, for 2 only stay conservative
11735 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11736 argument to the 2 argument get_range_strlen, adjust 6 arg
11737 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11739 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11740 (gimple_fold_builtin_strlen): Pass true as last argument to
11743 2018-02-20 Martin Sebor <msebor@redhat.com>
11745 PR middle-end/84095
11746 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11747 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11748 (builtin_memref::builtin_memref): Factor out parts into
11749 set_base_and_offset and call it.
11751 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11753 PR middle-end/84406
11754 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11755 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11756 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11757 search at the associated MODE_INT.
11759 2018-02-20 Jeff Law <law@redhat.com>
11761 PR middle-end/82123
11762 PR tree-optimization/81592
11763 PR middle-end/79257
11764 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11765 for range data rather than using global data.
11766 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11767 range data rather than using global data.
11768 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11769 pass it to children as needed.
11770 (struct directive::fmtresult): Similarly.
11771 (struct directive::set_width): Similarly.
11772 (struct directive::set_precision): Similarly.
11773 (format_integer, format_directive, parse_directive): Similarly.
11774 (format_none): Accept unnamed vr_values parameter.
11775 (format_percent, format_floating, format_character): Similarly.
11776 (format_string, format_plain): Similarly.
11777 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11778 the EVRP range analyzer for range data rather than using global data.
11779 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11780 gimple-ssa-evrp-analyze.h
11781 (class sprintf_dom_walker): Add after_dom_children member function.
11782 Add evrp_range_analyzer member.
11783 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11784 range analyzer as needed.
11785 (sprintf_dom_walker::after_dom_children): New member function.
11786 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11788 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11789 (evrp_range_analyzer::pop_to_marker): Likewise.
11791 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11793 PR tree-optimization/84419
11794 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11795 with the required type if its current type is compatible but
11798 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11800 PR middle-end/82004
11801 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11802 after vectorization.
11804 2018-02-20 Martin Liska <mliska@suse.cz>
11807 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11808 possible values if we don't have a hint.
11810 2018-02-20 Martin Liska <mliska@suse.cz>
11814 * final.c (shorten_branches): Build align_tab array with one
11816 * opts.c (finish_options): Add alignment option limit check.
11817 (MAX_CODE_ALIGN): Likewise.
11818 (MAX_CODE_ALIGN_VALUE): Likewise.
11819 * doc/invoke.texi: Document maximum allowed option value for
11820 all -falign-* options.
11822 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11825 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11826 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11827 * var-tracking.c (emit_note_insn_var_location): Remove all references
11828 to NOTE_INSN_CALL_ARG_LOCATION.
11829 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11830 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11831 Use copy_rtx_if_shared.
11832 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11833 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11834 (dwarf2out_var_location): Remove handling of
11835 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11837 * final.c (final_scan_insn): Remove all references to
11838 NOTE_INSN_CALL_ARG_LOCATION.
11839 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11840 before dumping final insns.
11841 * except.c (emit_note_eh_region_end): Remove all references to
11842 NOTE_INSN_CALL_ARG_LOCATION.
11843 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11844 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11845 * config/arc/arc.c (hwloop_optimize): Likewise.
11846 * config/arm/arm.c (create_fix_barrier): Likewise.
11847 * config/s390/s390.c (s390_chunkify_start): Likewise.
11848 * config/sh/sh.c (find_barrier): Likewise.
11849 * config/i386/i386.c (rest_of_insert_endbranch,
11850 ix86_seh_fixup_eh_fallthru): Likewise.
11851 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11852 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11853 * config/frv/frv.c (frv_function_prologue): Likewise.
11854 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11856 (note_outside_basic_block_p): Remove all references to
11857 NOTE_INSN_CALL_ARG_LOCATION.
11858 * gengtype.c (adjust_field_rtx_def): Likewise.
11859 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11861 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11862 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11865 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11868 PR tree-optimization/84452
11869 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11870 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11873 2018-02-19 Martin Liska <mliska@suse.cz>
11876 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11878 2018-02-19 Martin Liska <mliska@suse.cz>
11879 Richard Sandiford <richard.sandiford@linaro.org>
11881 PR tree-optimization/82491
11882 * gimple-fold.c (get_base_constructor): Make earlier bail out
11885 2018-02-19 Carl Love <cel@us.ibm.com>
11887 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11888 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11890 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11891 P8V_BUILTIN_VEC_NEG.
11893 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11895 * config/rl78/rl78.md (movdf): New define expand.
11897 2018-02-19 Martin Liska <mliska@suse.cz>
11900 * doc/invoke.texi: Fix typo.
11901 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11903 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11905 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11906 handle rs6000_single_float and rs6000_double_float specially for
11909 2018-02-16 Jeff Law <law@redhat.com>
11911 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11912 (pop_regs): Use it.
11914 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11917 * ipa-inline.c (inline_small_functions): Fix a typo.
11919 2018-02-16 Nathan Sidwell <nathan@acm.org>
11921 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11923 2018-02-16 Carl Love <cel@us.ibm.com>
11925 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11926 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11927 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11928 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11929 expansion to P8V_BUILTIN_VEC_FLOAT2.
11931 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11933 PR rtl-optimization/70023
11934 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11935 src_regno into account.
11937 2018-02-16 Carl Love <cel@us.ibm.com>
11939 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11940 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11941 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11942 * config/rs6000/rs6000.c: Remove case statements for
11943 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11944 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11945 and P9V_BUILTIN_VEC_VINSERT4B.
11946 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11947 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11948 * config/rs6000/vsx.md:
11949 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11952 2018-02-16 Carl Love <cel@us.ibm.com>
11954 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11956 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11958 * config/rs6000/rs6000-c.c: Add the definitions for
11959 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11960 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11961 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11962 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11963 definition for insert4b and define insn *insert3b_internal.
11964 * doc/extend.texi: Add documentation for vec_extract4b.
11966 2018-02-16 Nathan Sidwell <nathan@acm.org>
11968 * doc/extend.texi (Backwards Compatibility): Mention friend
11969 injection. Note for-scope is deprecated.
11970 * doc/invoke.texi (-ffriend-injection): Deprecate.
11972 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11974 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11975 that moved to I2, also allow destinations that are a paradoxical
11976 subreg (instead of a normal reg).
11978 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11981 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11984 2018-02-16 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/84037
11987 PR tree-optimization/84016
11989 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11990 Adjust vec_construct for the fact we need additional higher latency
11991 128bit inserts for AVX256 and AVX512 vector builds.
11992 (ix86_add_stmt_cost): Scale vector construction cost for
11995 2018-02-16 Richard Biener <rguenther@suse.de>
11997 PR tree-optimization/84417
11998 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11999 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12000 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12002 2018-02-16 Martin Liska <mliska@suse.cz>
12005 * internal-fn.def (ASAN_CHECK): Set proper flags.
12006 (ASAN_MARK): Likewise.
12008 2018-02-16 Julia Koval <julia.koval@intel.com>
12010 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12011 from PTA_CANNONLAKE.
12013 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12016 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12017 Use ++iter rather than iter++ for std::list iterators.
12018 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12019 defer deleting them until all nodes in the forest are processed. Do
12020 free even leaf nodes. Change to_process into auto_vec.
12023 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12024 * vec.h (vec_default_construct): Use memset instead of placement new
12025 if BROKEN_VALUE_INITIALIZATION is defined.
12026 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12027 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12030 PR rtl-optimization/83723
12031 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12032 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12033 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12035 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12037 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12039 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12041 PR rtl-optimization/81443
12042 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12043 from inner REGs to paradoxical SUBREGs.
12045 2018-02-16 Richard Biener <rguenther@suse.de>
12047 PR tree-optimization/84399
12048 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12049 For operands we can analyze at their definition make sure we can
12050 analyze them at each use as well.
12052 2018-02-16 Richard Biener <rguenther@suse.de>
12054 PR tree-optimization/84190
12055 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12056 volatile accesses if the decl isn't volatile.
12058 2018-02-15 Jason Merrill <jason@redhat.com>
12060 PR c++/84314 - ICE with templates and fastcall attribute.
12061 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12062 TYPE_CANONICAL on an existing type.
12064 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12066 PR tree-optimization/84383
12067 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12068 dstoff nor call operand_equal_p if dstbase is NULL.
12070 PR tree-optimization/84334
12071 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12072 also a CONSTANT_CLASS_P, punt.
12074 2018-02-14 Jim Wilson <jimw@sifive.com>
12076 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12077 first SMALL_OPERAND check. New local min_second_step. Move assert
12078 to where locals are set. Add TARGET_RVC support.
12079 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12081 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12083 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12085 2018-02-14 Martin Sebor <msebor@redhat.com>
12087 PR tree-optimization/83698
12088 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12089 arrays constrain the offset range to their bounds.
12090 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12091 (builtin_access::overlap): Avoid setting the size of overlap if it's
12093 (maybe_diag_overlap): Also consider arrays when deciding what values
12094 of offsets to include in diagnostics.
12096 2018-02-14 Martin Sebor <msebor@redhat.com>
12099 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12100 that correspond to the kind of a declaration.
12102 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12105 * config/pa/pa.md: Load address of PIC label using the linkage table
12106 if the label is nonlocal.
12108 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12110 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12111 warning message if user requests -maltivec=be.
12112 * doc/invoke.texi: Document deprecation of -maltivec=be.
12114 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12117 * config/rs6000/rs6000-c.c: Update definitions for
12118 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12119 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12121 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12124 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12125 add _get_ssp intrinsics. Remove argument from
12126 __builtin_ia32_rdssp[d|q].
12127 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12128 * config/i386/i386-builtin.def: Remove argument from
12129 __builtin_ia32_rdssp[d|q].
12130 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12131 ix86_expand_special_args_builtin for _rdssp[d|q].
12132 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12133 Clear register before usage.
12134 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12135 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12137 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12139 PR tree-optimization/84357
12140 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12141 operand 1 of an ARRAY_REF too.
12143 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12146 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12147 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12149 (set_of_reg): New struct.
12150 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12151 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12152 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12154 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12155 Split into bitclr, bitset, bitinvert patterns if appropriate.
12156 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12157 use rx_fuse_in_memory_bitop.
12158 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12159 to named insn, correct maximum insn length.
12161 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12164 * machmode.def: Define a complex mode for PARTIAL_INT.
12165 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12167 * doc/rtl.texi: Document CSPImode.
12168 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12170 (msp430_hard_regno_nregs_with_padding): Likewise.
12172 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12175 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12177 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12179 PR rtl-optimization/84169
12180 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12181 we generated a parallel as new i3 and we split that to new i2 and i3
12182 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12183 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12184 those to i2, not i1. Partially rewrite this scan code.
12186 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12189 * stor-layout.c (place_field): For variable length fields, adjust
12190 offset_align afterwards not just based on the field's alignment,
12191 but also on the size.
12193 PR middle-end/84309
12194 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12195 of exps and logs in the use_exp2 case.
12197 2018-02-13 Jeff Law <law@redhat.com>
12199 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12200 entry for "vector".
12202 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12205 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12209 * common.opt (gas-loc-support, gas-locview-support): New.
12210 (ginline-points, ginternal-reset-location-views): New.
12211 * doc/invoke.texi: Document them. Use @itemx where intended.
12212 (gvariable-location-views): Adjust.
12213 * target.def (reset_location_view): New.
12214 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12215 (TARGET_RESET_LOCATION_VIEW): New.
12216 * doc/tm.texi: Rebuilt.
12217 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12218 (dwarf2out_default_as_locview_support): New.
12219 (output_asm_line_debug_info): Use option variables.
12220 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12221 (output_loc_list): Likewise.
12222 (add_high_low_attributes): Check option variables.
12223 Don't output entry view attribute in strict mode.
12224 (gen_inlined_subroutine_die): Check option variables.
12225 (dwarf2out_inline_entry): Likewise.
12226 (init_sections_and_labels): Likewise.
12227 (dwarf2out_early_finish): Likewise.
12228 (maybe_reset_location_view): New, from...
12229 (dwarf2out_var_location): ... here. Call it.
12230 * debug.h (dwarf2out_default_as_loc_support): Declare.
12231 (dwarf2out_default_as_locview_support): Declare.
12232 * hooks.c (hook_int_rtx_insn_0): New.
12233 * hooks.h (hook_int_rtx_insn_0): Declare.
12234 * toplev.c (process_options): Take -gas-loc-support and
12235 -gas-locview-support from dwarf2out. Enable
12236 -gvariable-location-views by default only with locview
12237 assembler support. Enable -ginternal-reset-location-views by
12238 default only if the target defines the corresponding hook.
12239 Enable -ginline-points by default if location views are
12240 enabled; force it disabled if statement frontiers are
12242 * tree-inline.c (expand_call_inline): Check option variables.
12243 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12245 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12247 PR tree-optimization/84321
12248 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12249 handling. Also check whether the anti-range contains any values
12250 that satisfy the mask; switch to a VR_RANGE if not.
12252 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12255 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12257 2018-02-13 Martin Jambor <mjambor@suse.cz>
12260 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12261 of call statements, also set location of a load to a temporary.
12263 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12265 * config/rl78/rl78.c (add_vector_labels): New function.
12266 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12267 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12268 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12269 which checks that no arguments are passed.
12270 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12271 * doc/extend.texi: Documentation for the new attribute.
12273 2018-02-13 Andreas Schwab <schwab@suse.de>
12275 * config/riscv/linux.h (CPP_SPEC): Define.
12277 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12280 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12281 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12282 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12283 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12284 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12285 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12286 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12287 temporarily for AES and PCLMUL builtins.
12289 PR tree-optimization/84339
12290 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12291 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12294 PR middle-end/84309
12295 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12296 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12297 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12299 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12301 * omp-simd-clone.h: New file.
12302 * omp-simd-clone.c: Include omp-simd-clone.h.
12303 (expand_simd_clones): No longer static.
12304 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12305 cgraph.h and omp-simd-clone.h.
12306 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12307 (vect_recog_widen_shift_pattern): Formatting fix.
12308 (vect_pattern_recog_1): Don't check optab for calls.
12311 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12312 operands[2] into a REG before using gen_lowpart on it.
12314 2018-02-12 Jeff Law <law@redhat.com>
12317 * config/sh/sh.c (find_barrier): Consider a sibling call
12320 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12321 successfully back substituting a reg.
12323 2018-02-12 Richard Biener <rguenther@suse.de>
12325 PR tree-optimization/84037
12326 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12327 parameter, move visited init to caller.
12328 (vect_slp_analyze_operations): Separate cost from validity
12329 check, initialize visited once for all instances.
12330 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12332 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12334 (vect_model_promotion_demotion_cost): Likewise.
12335 (vectorizable_bswap): Guard cost modeling with !slp_node
12336 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12338 (vectorizable_call): Likewise.
12339 (vectorizable_conversion): Likewise.
12340 (vectorizable_assignment): Likewise.
12341 (vectorizable_shift): Likewise.
12342 (vectorizable_operation): Likewise.
12343 (vectorizable_store): Likewise.
12344 (vectorizable_load): Likewise.
12345 (vectorizable_condition): Likewise.
12346 (vectorizable_comparison): Likewise.
12348 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12351 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12352 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12353 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12355 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12357 PR middle-end/83665
12358 * params.def (inline-min-speedup): Increase from 8 to 15.
12359 (max-inline-insns-auto): Decrease from 40 to 30.
12360 * ipa-split.c (consider_split): Add some buffer for function to
12361 be considered inlining candidate.
12362 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12365 2018-02-12 Richard Biener <rguenther@suse.de>
12367 PR tree-optimization/84037
12368 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12369 matched stmts if we cannot swap the non-matched ones.
12371 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12373 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12374 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12375 _mm_maskz_scalef_round_ss): New intrinsics.
12376 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12377 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12378 __builtin_ia32_scalefss_round): Remove.
12379 (__builtin_ia32_scalefsd_mask_round,
12380 __builtin_ia32_scalefss_mask_round): New intrinsics.
12381 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12382 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12383 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12384 "<round_constraint>")): Changed to ...
12385 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12386 "<round_scalar_constraint>")): ... this.
12387 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12388 %0, %1, %2<round_op3>}"): Changed to ...
12389 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12390 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12391 %2<round_scalar_mask_op3>}"): ... this.
12392 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12394 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12396 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12397 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12398 (_mm_maskz_sqrt_round_ss): New intrinsics.
12399 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12400 (__builtin_ia32_sqrtsd_mask_round)
12401 (__builtin_ia32_sqrtss_mask_round): New builtins.
12402 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12403 (__builtin_ia32_sqrtss_round): Remove.
12404 (__builtin_ia32_sqrtsd_mask_round)
12405 (__builtin_ia32_sqrtss_mask_round): New builtins.
12406 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12407 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12408 ((match_operand:VF_128 1 "vector_operand"
12409 "xBm,<round_constraint>")): Changed to ...
12410 ((match_operand:VF_128 1 "vector_operand"
12411 "xBm,<round_scalar_constraint>")): ... this.
12412 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12413 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12414 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12415 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12416 %<iptr>1<round_scalar_mask_op3>}): ... this.
12417 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12418 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12420 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12423 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12424 Cast vec_cmpeq result to correct type.
12425 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12426 Cast vec_cmpgt result to correct type.
12428 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12430 * final.c (final_scan_insn_1): Renamed from...
12431 (final_scan_insn): ... this. New wrapper, to recover
12432 seen from the outermost call in recursive ones.
12433 * config/sparc/sparc.c (output_return): Drop seen from call.
12434 (output_sibcall): Likewise.
12435 * config/visium/visium.c (output_branch): Likewise.
12437 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12439 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12442 2018-02-10 Alan Modra <amodra@gmail.com>
12445 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12446 Specify LR as an input.
12448 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12451 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12452 remove_member_access_dummy_vars): New functions.
12453 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12454 lower_omp_1, execute_lower_omp): Use them.
12456 PR rtl-optimization/84308
12457 * shrink-wrap.c (spread_components): Release todo vector.
12459 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12461 PR rtl-optimization/57193
12462 * ira-color.c (struct allocno_color_data): Add member
12463 conflict_allocno_hard_prefs.
12464 (update_conflict_allocno_hard_prefs): New.
12465 (bucket_allocno_compare_func): Add a preference based on
12466 conflict_allocno_hard_prefs.
12467 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12468 (color_allocnos): Remove a dead code. Initiate
12469 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12471 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12474 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12475 constraint from =wa to wa. Avoid a subreg on the output operand,
12476 instead use a pseudo and subreg it in a move.
12477 (p9_xxbrd_<mode>): Changed to ...
12478 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12479 (p9_xxbrd_v2df): New expander.
12480 (p9_xxbrw_<mode>): Changed to ...
12481 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12482 (p9_xxbrw_v4sf): New expander.
12484 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12486 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12488 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12491 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12492 multiply in 32-bit mode.
12493 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12494 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12497 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12499 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12500 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12501 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12502 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12504 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12507 * dwarf2out.c (is_trivial_indirect_ref): New function.
12508 (dwarf2out_late_global_decl): Do not generate a location
12509 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12510 and that are not defined in the current unit.
12512 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12514 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12515 instead of a libcall for UNORDERED.
12517 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12520 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12521 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12523 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12526 * config/s390/s390.c (s390_set_current_function): Invoke
12527 s390_indirect_branch_settings also if fndecl didn't change.
12529 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12531 * config/rs6000/rs6000.md (blockage): Set length to zero.
12533 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12535 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12537 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12540 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12541 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12545 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12546 PARALLEL incoming that failed vt_get_decl_and_offset check.
12548 PR middle-end/84237
12549 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12550 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12552 (get_variable_section): For decls in named .bss* sections pass true as
12553 second argument to bss_initializer_p.
12555 2018-02-09 Marek Polacek <polacek@redhat.com>
12556 Jakub Jelinek <jakub@redhat.com>
12559 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12560 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12561 Sync some changes from cxx_fold_indirect_ref.
12563 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12565 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12567 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12568 (BLOCK_INLINE_ENTRY_LABEL): New.
12569 (dwarf2out_var_location): Disregard inline entry markers.
12570 (inline_entry_data): New struct.
12571 (inline_entry_data_hasher): New hashtable type.
12572 (inline_entry_data_hasher::hash): New.
12573 (inline_entry_data_hasher::equal): New.
12574 (inline_entry_data_table): New variable.
12575 (add_high_low_attributes): Add DW_AT_entry_pc and
12576 DW_AT_GNU_entry_view attributes if a pending entry is found
12577 in inline_entry_data_table. Add old entry_pc attribute only
12578 if debug nonbinding markers are disabled.
12579 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12580 markers are enabled.
12581 (block_within_block_p, dwarf2out_inline_entry): New.
12582 (dwarf2out_finish): Check that no entries remained in
12583 inline_entry_data_table.
12584 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12585 (final_scan_insn, notice_source_line): Likewise.
12586 (rest_of_clean_state): Skip inline entry markers.
12587 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12589 * gimple.c (gimple_build_debug_inline_entry): New.
12590 * gimple.h (enum gimple_debug_subcode): Add
12591 GIMPLE_DEBUG_INLINE_ENTRY.
12592 (gimple_build_debug_inline_entry): Declare.
12593 (gimple_debug_inline_entry_p): New.
12594 (gimple_debug_nonbind_marker_p): Adjust.
12595 * insn-notes.def (INLINE_ENTRY): New.
12596 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12597 inline entry marker notes.
12598 (print_insn): Likewise.
12599 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12600 (INSN_DEBUG_MARKER_KIND): Likewise.
12601 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12602 * tree-inline.c (expand_call_inline): Build and insert
12603 debug_inline_entry stmt.
12604 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12605 inline entry blocks early, if nonbind markers are enabled.
12606 (dump_scope_block): Dump fragment info.
12607 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12608 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12609 (gimple_build_debug_inline_entry): New.
12610 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12611 Enable/disable inline entry points too.
12612 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12613 (DEBUG_INSN): Describe inline entry markers.
12615 * common.opt (gvariable-location-views): New.
12616 (gvariable-location-views=incompat5): New.
12617 * config.in: Rebuilt.
12618 * configure: Rebuilt.
12619 * configure.ac: Test assembler for view support.
12620 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12621 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12622 * dwarf2out.c (var_loc_view): New typedef.
12623 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12624 (dwarf2out_locviews_in_attribute): New.
12625 (dwarf2out_locviews_in_loclist): New.
12626 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12627 (enum dw_line_info_opcode): Add LI_adv_address.
12628 (struct dw_line_info_table): Add view.
12629 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12630 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12631 (zero_view_p): New variable.
12632 (ZERO_VIEW_P): New macro.
12633 (output_asm_line_debug_info): New.
12634 (struct var_loc_node): Add view.
12635 (add_AT_view_list, AT_loc_list): New.
12636 (add_var_loc_to_decl): Add view param. Test it against last.
12637 (new_loc_list): Add view params. Record them.
12638 (AT_loc_list_ptr): Handle loc and view lists.
12639 (view_list_to_loc_list_val_node): New.
12640 (print_dw_val): Handle dw_val_class_view_list.
12641 (size_of_die): Likewise.
12642 (value_format): Likewise.
12643 (loc_list_has_views): New.
12644 (gen_llsym): Set vl_symbol too.
12645 (maybe_gen_llsym, skip_loc_list_entry): New.
12646 (dwarf2out_maybe_output_loclist_view_pair): New.
12647 (output_loc_list): Output view list or entries too.
12648 (output_view_list_offset): New.
12649 (output_die): Handle dw_val_class_view_list.
12650 (output_dwarf_version): New.
12651 (output_compilation_unit_header): Use it.
12652 (output_skeleton_debug_sections): Likewise.
12653 (output_rnglists, output_line_info): Likewise.
12654 (output_pubnames, output_aranges): Update version comments.
12655 (output_one_line_info_table): Output view numbers in asm comments.
12656 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12657 Call maybe_gen_llsym.
12658 (loc_list_from_tree_1): Adjust.
12659 (add_AT_location_description): Create view list attribute if
12660 needed, check it's absent otherwise.
12661 (convert_cfa_to_fb_loc_list): Adjust.
12662 (maybe_emit_file): Call output_asm_line_debug_info for test.
12663 (dwarf2out_var_location): Reset views as needed. Precompute
12664 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12665 attribute. Set view.
12666 (new_line_info_table): Reset next view.
12667 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12668 (dwarf2out_source_line): Likewise. Output view resets and labels to
12669 the assembler, or select appropriate line info opcodes.
12670 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12671 (optimize_string_length): Catch it. Adjust.
12672 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12673 dw_val_class_view_list, and remove it if no longer needed.
12674 (hash_loc_list): Hash view numbers.
12675 (loc_list_hasher::equal): Compare them.
12676 (optimize_location_lists): Check whether a view list symbol is
12677 needed, and whether the locview attribute is present, and
12678 whether they match. Remove the locview attribute if no longer
12680 (index_location_lists): Call skip_loc_list_entry for test.
12681 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12682 Use output_dwarf_version.
12683 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12684 (struct dw_val_node): Add val_view_list.
12685 * final.c (SEEN_NEXT_VIEW): New.
12686 (set_next_view_needed): New.
12687 (clear_next_view_needed): New.
12688 (maybe_output_next_view): New.
12689 (final_start_function): Rename to...
12690 (final_start_function_1): ... this. Take pointer to FIRST,
12691 add SEEN parameter. Emit param bindings in the initial view.
12692 (final_start_function): Reintroduce SEEN-less interface.
12693 (final): Rename to...
12694 (final_1): ... this. Take SEEN parameter. Output final pending
12695 next view at the end.
12696 (final): Reintroduce seen-less interface.
12697 (final_scan_insn): Output pending next view before switching
12698 sections or ending a block. Mark the next view as needed when
12699 outputting variable locations. Notify debug backend of section
12700 changes, and of location view changes.
12701 (rest_of_handle_final): Adjust.
12702 * toplev.c (process_options): Autodetect value for debug variable
12703 location views option. Warn on incompat5 without -gdwarf-5.
12704 * doc/invoke.texi (gvariable-location-views): New.
12705 (gvariable-location-views=incompat5): New.
12706 (gno-variable-location-views): New.
12708 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12710 PR tree-optimization/84136
12711 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12712 that the result of find_edge is non-NULL.
12714 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12717 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12718 storing integer register in SImode. Fix cost of 256 and 512
12719 byte aligned SSE register store.
12721 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12723 * config/i386/i386.c (ix86_multiplication_cost): Fix
12724 multiplication cost for TARGET_AVX512DQ.
12726 2018-02-08 Marek Polacek <polacek@redhat.com>
12728 PR tree-optimization/84238
12729 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12732 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12734 PR tree-optimization/84265
12735 * tree-vect-stmts.c (vectorizable_store): Don't treat
12736 VMAT_CONTIGUOUS accesses as grouped.
12737 (vectorizable_load): Likewise.
12739 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12741 PR tree-optimization/81635
12742 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12743 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12744 (test_round_for_mask): New functions.
12745 (wide_int_cc_tests): Call test_round_for_mask.
12746 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12747 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12748 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12749 range returned by get_range_info.
12751 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12754 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12755 * symtab.c: Include builtins.h
12756 (symtab_node::output_to_lto_symbol_table_p): Move here
12757 from lto-streamer-out.c:output_symbol_p.
12758 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12759 (output_symbol_p): Move all logic to symtab.c
12760 (produce_symtab): Update.
12762 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12764 * config/s390/s390-opts.h (enum indirect_branch): Define.
12765 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12766 (s390_indirect_branch_via_thunk)
12767 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12768 (enum s390_indirect_branch_type): Define.
12769 * config/s390/s390.c (struct s390_frame_layout, struct
12770 machine_function): Remove.
12771 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12772 (indirect_branch_table_label_no, indirect_branch_table_name):
12774 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12775 (enum s390_indirect_branch_option): Define.
12776 (s390_return_addr_from_memory): New function.
12777 (s390_handle_string_attribute): New function.
12778 (s390_attribute_table): Add new attribute handler.
12779 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12780 (s390_indirect_branch_via_thunk): New function.
12781 (s390_indirect_branch_via_inline_thunk): New function.
12782 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12783 sibling call optimization for non z10 compiles.
12784 (s390_emit_call): Force indirect branch target to be a single
12785 register. Add r1 clobber for non-z10 compiles.
12786 (s390_emit_epilogue): Emit return jump via return_use expander.
12787 (s390_reorg): Handle JUMP_INSNs as execute targets.
12788 (s390_option_override_internal): Perform validity checks for the
12789 new command line options.
12790 (s390_indirect_branch_attrvalue): New function.
12791 (s390_indirect_branch_settings): New function.
12792 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12793 (s390_output_indirect_thunk_function): New function.
12794 (s390_code_end): Implement target hook.
12795 (s390_case_values_threshold): Implement target hook.
12796 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12798 * config/s390/s390.h (struct s390_frame_layout)
12799 (struct machine_function): Move here from s390.c.
12800 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12801 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12802 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12803 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12804 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12805 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12806 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12807 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12808 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12809 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12810 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12811 (mnemonic attribute): Add values which aren't recognized
12813 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12814 pattern for branch conversion. Fix mnemonic attribute.
12815 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12816 indirect branch via thunk if requested.
12817 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12818 ("*indirect_jump"): Disable for branch conversion using out of
12820 ("indirect_jump_via_thunk<mode>_z10")
12821 ("indirect_jump_via_thunk<mode>")
12822 ("indirect_jump_via_inlinethunk<mode>_z10")
12823 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12824 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12825 ("casesi_jump_via_inlinethunk<mode>_z10")
12826 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12827 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12828 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12829 ("*indirect2_jump"): Disable for branch conversion.
12830 ("casesi_jump"): Turn into expander and expand patterns for branch
12832 ("return_use"): New expander.
12833 ("*return"): Emit return via thunk and rename it to ...
12834 ("*return<mode>"): ... this one.
12835 * config/s390/s390.opt: Add new options and and enum for the
12838 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12840 * lra-constraints.c (match_reload): Unconditionally use
12841 gen_lowpart_SUBREG, rather than selecting between that
12842 and equivalent gen_rtx_SUBREG code.
12844 2018-02-08 Richard Biener <rguenther@suse.de>
12846 PR tree-optimization/84233
12847 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12848 changed flag instead of boguously re-using phi_inserted.
12850 2018-02-08 Martin Jambor <mjambor@suse.cz>
12852 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12853 static local variables.
12855 2018-02-08 Richard Biener <rguenther@suse.de>
12857 PR tree-optimization/84278
12858 * tree-vect-stmts.c (vectorizable_store): When looking for
12859 smaller vector types to perform grouped strided loads/stores
12860 make sure the mode is supported by the target.
12861 (vectorizable_load): Likewise.
12863 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12865 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12866 Increase LDP/STP opportunities by adding adjacent callee-saves.
12868 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12870 PR rtl-optimization/84068
12871 PR rtl-optimization/83459
12872 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12874 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12876 PR tree-optimization/84224
12877 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12878 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12879 non-zero arguments.
12881 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12884 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12885 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12886 count, remove one USE.
12888 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12890 * doc/install.texi (Configuration): Document the
12891 --with-long-double-format={ibm,ieee} PowerPC configuration
12895 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12896 Convert from define_expand to be define_insn_and_split. Rework
12897 float/double/_Float128 conversions to QI/HI/SImode to work with
12898 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12899 conversions to QI/HImode types did a store and then a load to
12900 truncate the value. For conversions to VSX registers, don't split
12901 the insn, instead emit the code directly. Use the code iterator
12902 any_fix to combine signed and unsigned conversions.
12903 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12904 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12905 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12906 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12907 (fix_<mode>di2_hw): Likewise.
12908 (fixuns_<mode>di2_hw): Likewise.
12909 (fix_<mode>si2_hw): Likewise.
12910 (fixuns_<mode>si2_hw): Likewise.
12911 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12912 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12913 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12914 fix<uns>_trunc<SFDF:mode>si2_p8.
12915 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12916 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12917 (fix<uns>_<mode>_mem): Likewise.
12918 (fctiw<u>z_<mode>_mem): Likewise.
12919 (fix<uns>_<mode>_mem): Likewise.
12920 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12921 the register allocator from doing a direct move to the GPRs to do
12922 a store, and instead use the ISA 3.0 store byte/half-word from
12923 vector register instruction. For IEEE 128-bit floating point,
12924 also optimize stores of 32-bit ints.
12925 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12927 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12929 * genextract.c (push_pathstr_operand): New function to support
12931 (walk_rtx): Call push_pathstr_operand.
12932 (print_path): Support [a-zA-Z].
12934 2018-02-07 Richard Biener <rguenther@suse.de>
12936 PR tree-optimization/84037
12937 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12938 (cse_and_gimplify_to_preheader): Declare.
12939 (vect_get_place_in_interleaving_chain): Likewise.
12940 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12942 (_loop_vec_info::~_loop_vec_info): Delete it.
12943 (cse_and_gimplify_to_preheader): New function.
12944 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12945 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12946 (vectorizable_load): Likewise. For grouped stores always base
12947 the IV on the first element.
12948 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12949 condition before gimplifying.
12951 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12953 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12954 *DIV_EXPR and *MOD_EXPR.
12956 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12959 * config/i386/i386.c (ix86_option_override_internal): Mask out
12960 the CF_SET bit when checking -fcf-protection.
12962 2018-02-07 Tom de Vries <tom@codesourcery.com>
12965 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12968 2018-02-07 Richard Biener <rguenther@suse.de>
12970 PR tree-optimization/84204
12971 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12974 PR tree-optimization/84205
12975 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12976 special-case isl_ast_op_zdiv_r.
12978 PR tree-optimization/84223
12979 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12980 Only add conditions from within the region.
12981 (gather_bbs::after_dom_children): Adjust.
12983 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12986 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12987 * config/avr/avr.md: Only post-reload split REG-REG moves if
12988 either register is GENERAL_REG_P.
12990 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12992 PR tree-optimization/84235
12993 * tree-ssa-scopedtables.c
12994 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12995 if the subtraction is performed in floating point type where NaNs are
12996 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12997 build 1. Formatting fix.
12999 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13002 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13003 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13004 and skip it regardless of bb boundaries. Use CALL_P macro,
13005 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13006 unnecessarily, formatting fix.
13008 2018-02-06 Michael Collison <michael.collison@arm.com>
13010 * config/arm/thumb2.md:
13011 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13012 (*thumb_mov_notscc): Ditto.
13014 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13017 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13018 unsigned_fix, not "s".
13020 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13022 * configure.ac (gcc_fn_eh_frame_ro): New function.
13023 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13024 correct .eh_frame permissions.
13025 * configure: Regenerate.
13027 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13029 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13030 irrelevant options.
13032 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13034 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13035 Display warning message for -mno-speculate-indirect-jumps.
13037 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13039 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13041 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13043 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13045 PR tree-optimization/84225
13046 * tree-eh.c (find_trapping_overflow): Only call
13047 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13049 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13052 * config/i386/i386.c: Reimplement the check of possible options
13053 -mibt/-mshstk conbination. Change error messages.
13054 * doc/invoke.texi: Fix a typo: remove extra '='.
13056 2018-02-06 Marek Polacek <polacek@redhat.com>
13058 PR tree-optimization/84228
13059 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13061 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13064 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13065 emitted arch directives.
13066 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13067 __ARM_FEATURE_COPROC before changing architectures.
13069 2018-02-06 Richard Biener <rguenther@suse.de>
13071 * config/i386/i386.c (print_reg): Fix typo.
13072 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13074 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13076 * configure: Regenerate.
13078 2018-02-05 Martin Sebor <msebor@redhat.com>
13080 PR tree-optimization/83369
13081 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13084 2018-02-05 Martin Liska <mliska@suse.cz>
13086 * doc/invoke.texi: Cherry-pick upstream r323995.
13088 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13090 * ira.c (ira_init_register_move_cost): Adjust comment.
13092 2018-02-05 Martin Liska <mliska@suse.cz>
13094 PR gcov-profile/84137
13095 * doc/gcov.texi: Fix typo in documentation.
13097 2018-02-05 Martin Liska <mliska@suse.cz>
13099 PR gcov-profile/83879
13100 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13101 using dlopen functionality.
13103 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13105 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13106 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13107 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13108 _mm_maskz_range_round_ss): New intrinsics.
13109 (__builtin_ia32_rangesd128_round)
13110 (__builtin_ia32_rangess128_round): Remove.
13111 (__builtin_ia32_rangesd128_mask_round,
13112 __builtin_ia32_rangess128_mask_round): New builtins.
13113 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13114 __builtin_ia32_rangess128_round): Remove.
13115 (__builtin_ia32_rangesd128_mask_round,
13116 __builtin_ia32_rangess128_mask_round): New builtins.
13117 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13118 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13119 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13120 "<round_saeonly_constraint>")): Changed to ...
13121 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13122 "<round_saeonly_scalar_constraint>")): ... this.
13123 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13124 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13125 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13126 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13127 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13129 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13131 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13133 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13134 Remove all values except native, 8540 and 8548.
13136 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13138 * config/i386/i386.c (ix86_output_function_return): Pass
13139 INVALID_REGNUM, instead of -1, as invalid register number to
13140 indirect_thunk_name and output_indirect_thunk.
13142 2018-02-02 Julia Koval <julia.koval@intel.com>
13144 * config.gcc: Add -march=icelake.
13145 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13146 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13147 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13148 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13149 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13150 (processor_target_table): Add icelake.
13151 (ix86_option_override_internal): Handle new PTAs.
13152 (get_builtin_code_for_version): Handle icelake.
13153 (M_INTEL_COREI7_ICELAKE): New.
13154 (fold_builtin_cpu): Handle icelake.
13155 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13156 * doc/invoke.texi: Add -march=icelake.
13158 2018-02-02 Julia Koval <julia.koval@intel.com>
13160 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13161 to wide_int_bitmask.
13162 * wide-int-bitmask.h: New.
13164 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13167 * config/i386/i386.md: Replace Pmode with word_mode in
13168 builtin_setjmp_setup and builtin_longjmp to support x32.
13170 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13174 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13176 (rs6000_supported_cpu_names): New static variable.
13177 (linux_cpu_translation_table): Likewise.
13178 (elf_platform) <cpu>: Define new static variable and use it.
13179 Translate kernel AT_PLATFORM name to canonical name if needed.
13180 Error if platform name is unknown.
13182 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13185 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13187 2018-02-01 Jeff Law <law@redhat.com>
13190 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13191 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13192 the scratch if RELEASE_VIA_POP is false.
13193 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13194 If we have to save a temporary register, decrement SIZE appropriately.
13195 Pass new arguments to release_scratch_register_on_entry.
13196 (ix86_adjust_stack_and_probe): Likewise.
13197 (ix86_emit_probe_stack_range): Pass new arguments to
13198 release_scratch_register_on_entry.
13200 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13202 PR rtl-optimization/84157
13203 * combine.c (change_zero_ext): Use REG_P predicate in
13204 front of HARD_REGISTER_P predicate.
13206 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13208 * config/avr/avr.c (avr_option_override): Move disabling of
13209 -fdelete-null-pointer-checks to...
13210 * common/config/avr/avr-common.c (avr_option_optimization_table):
13213 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13215 PR tree-optimization/81635
13216 * tree-data-ref.c (split_constant_offset_1): For types that
13217 wrap on overflow, try to use range info to prove that wrapping
13220 2018-02-01 Renlin Li <renlin.li@arm.com>
13223 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13224 TAILCALL_ADDR_REGS.
13225 (aarch64_register_move_cost): Likewise.
13226 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13227 TAILCALL_ADDR_REGS.
13228 (REG_CLASS_NAMES): Likewise.
13229 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13230 TAILCALL_ADDR_REGS. Remove IP registers.
13231 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13233 2018-02-01 Richard Biener <rguenther@suse.de>
13235 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13236 for specifying RPO order and allow NULL for that.
13237 * domwalk.c (dom_walker::dom_walker): Likewise.
13238 (dom_walker::walk): Handle NULL RPO order.
13239 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13241 (rewrite_update_dom_walker): Likewise.
13242 (mark_def_dom_walker): Likewise.
13244 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13246 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13247 (aarch64_maybe_expand_sve_subreg_move): Declare.
13248 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13249 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13251 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13252 that are semantically a reverse operation.
13253 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13254 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13255 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13257 (aarch64_can_change_mode_class): For big-endian, forbid changes
13258 between two SVE modes if they have different element sizes.
13260 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13262 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13263 the TImode handling for big-endian targets.
13265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13267 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13268 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13270 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13271 Remove BSWAP handing for big-endian targets and use the form of
13272 LD1RQ appropariate for the mode.
13274 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13276 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13277 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13278 duplicated element.
13280 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13283 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13284 check for operands that need to go through aarch64_sve_reload_be.
13286 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13288 PR tree-optimization/81661
13289 PR tree-optimization/84117
13290 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13291 * tree-eh.c: Include gimplify.h.
13292 (find_trapping_overflow, replace_trapping_overflow,
13293 rewrite_to_non_trapping_overflow): New functions.
13294 * tree-vect-loop.c: Include tree-eh.h.
13295 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13296 * tree-data-ref.c: Include tree-eh.h.
13297 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13299 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13301 PR rtl-optimization/84123
13302 * combine.c (change_zero_ext): Check if hard register satisfies
13303 can_change_dest_mode before calling gen_lowpart_SUBREG.
13305 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13308 * ira.c (ira_init_register_move_cost): Remove assert.
13310 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13312 PR rtl-optimization/84071
13313 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13314 * doc/tm.texi: Regenerate.
13316 2018-01-31 Richard Biener <rguenther@suse.de>
13318 PR tree-optimization/84132
13319 * tree-data-ref.c (analyze_miv_subscript): Properly
13320 check whether evolution_function_is_affine_multivariate_p
13321 before calling gcd_of_steps_may_divide_p.
13323 2018-01-31 Julia Koval <julia.koval@intel.com>
13326 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13327 * config/i386/i386.md (rdpid_rex64) New.
13328 (rdpid): Make 32bit only.
13330 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13333 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13334 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13336 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13339 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13341 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13343 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13345 PR rtl-optimization/84071
13346 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13347 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13349 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13351 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13352 (arc_attribute_table): Add 'aux' attribute.
13353 (arc_in_small_data_p): Consider aux like variables.
13354 (arc_is_aux_reg_p): New function.
13355 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13356 (arc_get_aux_arg): New function.
13357 (prepare_move_operands): Handle aux-register access.
13358 (arc_handle_aux_attribute): New function.
13359 * doc/extend.texi (ARC Variable attributes): Add subsection.
13361 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13363 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13364 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13365 (arc_attribute_table): Add 'uncached' attribute.
13366 (arc_print_operand): Print '.di' flag for uncached memory
13368 (arc_in_small_data_p): Do not consider for small data the uncached
13370 (arc_is_uncached_mem_p): New function.
13371 * config/arc/predicates.md (compact_store_memory_operand): Check
13372 for uncached memory accesses.
13373 (nonvol_nonimm_operand): Likewise.
13374 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13376 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13379 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13380 falign-loops=): Add Optimization flag.
13382 2018-01-30 Jeff Law <law@redhat.com>
13385 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13386 INT_REGISTERS_SAVED. Check it prior to calling
13387 get_scratch_register_on_entry.
13388 (ix86_adjust_stack_and_probe): Similarly.
13389 (ix86_emit_probe_stack_range): Similarly.
13390 (ix86_expand_prologue): Corresponding changes.
13392 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13395 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13396 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13398 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13401 * lra-constraints.c (curr_insn_transform): Process AND in the
13404 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13406 PR rtl-optimization/83986
13407 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13408 dependence against last_pending_memory_flush in addition to
13409 pending_jump_insns.
13411 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13413 PR tree-optimization/81611
13414 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13417 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13420 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13423 2018-01-30 Richard Biener <rguenther@suse.de>
13424 Jakub Jelinek <jakub@redhat.com>
13426 PR tree-optimization/84111
13427 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13428 inner loops added during recursion, as they don't have up-to-date
13431 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13434 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13435 (can_inline_edge_by_limits_p): ... here.
13436 (can_early_inline_edge_p, check_callers,
13437 update_caller_keys, update_callee_keys, recursive_inlining,
13438 add_new_edges_to_heap, speculation_useful_p,
13439 inline_small_functions,
13440 inline_small_functions, flatten_function,
13441 inline_to_all_callers_1): Update.
13443 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13445 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13448 2018-01-30 Richard Biener <rguenther@suse.de>
13450 PR tree-optimization/83008
13451 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13452 invariant and constant vector uses in stmts when they need
13453 more than one stmt.
13455 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13458 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13459 * configure: Regenerate.
13461 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13463 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13465 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13466 Use gen_rtx_REG rather than gen_lowpart.
13468 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13470 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13471 rather than 0 when creating partial subregs.
13473 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13475 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13478 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13481 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13482 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13483 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13484 flags. This restores the settings used before the 2017-07-24.
13485 Turning off pre increment/decrement/modify allows IVOPTS to
13486 optimize DF/SF loops where the index is an int.
13488 2018-01-29 Richard Biener <rguenther@suse.de>
13489 Kelvin Nilsen <kelvin@gcc.gnu.org>
13492 * tree-vect-stmts.c (vectorizable_call): Don't call
13493 targetm.vectorize_builtin_md_vectorized_function if callee is
13496 2018-01-22 Carl Love <cel@us.ibm.com>
13498 * doc/extend.tex: Fix typo in second arg in
13499 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13501 2018-01-29 Richard Biener <rguenther@suse.de>
13503 PR tree-optimization/84086
13504 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13505 (flush_ssaname_freelist): When SSA names were released reset
13506 the SCEV hash table.
13508 2018-01-29 Richard Biener <rguenther@suse.de>
13510 PR tree-optimization/84057
13511 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13512 removed paths when removing edges.
13514 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13516 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13517 -mfunction-return=@var{choice}.
13519 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13521 PR diagnostic/84034
13522 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13523 Handle CR like TAB.
13524 (layout::print_source_line): Likewise.
13525 (test_get_line_width_without_trailing_whitespace): Add test cases.
13527 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13529 PR middle-end/84040
13530 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13533 2018-01-26 Jim Wilson <jimw@sifive.com>
13535 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13537 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13540 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13542 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13543 and CMP + SUB-immediate -> SUBS.
13545 2018-01-26 Martin Sebor <msebor@redhat.com>
13547 PR tree-optimization/83896
13548 * tree-ssa-strlen.c (get_string_len): Rename...
13549 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13550 Avoid assuming length is constant.
13551 (handle_char_store): Use HOST_WIDE_INT for string length.
13553 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13556 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13557 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13559 2018-01-26 Richard Biener <rguenther@suse.de>
13561 PR rtl-optimization/84003
13562 * dse.c (record_store): Only record redundant stores when
13563 the earlier store aliases at least all accesses the later one does.
13565 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13567 PR rtl-optimization/83985
13568 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13569 REG_CFA_RESTORE insns.
13570 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13571 REG_CFA_RESTORE insns here.
13574 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13575 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13577 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13579 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13580 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13581 (arc_init): Likewise.
13582 (arc_override_options): Likewise.
13583 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13585 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13587 * config/arc/arc.h (TARGET_DBNZ): Define.
13588 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13589 properly set the tune attribute.
13590 (dbnz): Use TARGET_DBNZ guard.
13591 * config/arc/arc.opt (mtune): Add core3 option.
13593 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13595 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13596 recognize new pic like addresses.
13597 (arc_delegitimize_address): Clean up.
13599 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13601 * config/arc/arc-arches.def: Option mrf16 valid for all
13603 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13604 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13605 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13606 * config/arc/arc-tables.opt: Regenerate.
13607 * config/arc/arc.c (arc_conditional_register_usage): Handle
13608 reduced register file case.
13609 (arc_file_start): Set must have build attributes.
13610 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13611 mrf16 option value.
13612 * config/arc/arc.opt (mrf16): Add new option.
13613 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13614 * config/arc/genmultilib.awk: Handle new mrf16 option.
13615 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13616 * config/arc/t-multilib: Regenerate.
13617 * doc/invoke.texi (ARC Options): Document mrf16 option.
13619 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13621 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13622 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13623 (arc_attribute_table): Add 'secure_call' attribute.
13624 (arc_print_operand): Print secure call operand.
13625 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13627 (arc_is_secure_call_p): New function. * config/arc/arc.md
13628 (call_i): Add support for sjli instruction.
13629 (call_value_i): Likewise.
13630 * config/arc/constraints.md (Csc): New constraint.
13632 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13633 John Eric Martin <John.Martin@emmicro-us.com>
13635 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13636 * config/arc/arc.c (_arc_jli_section): New struct.
13637 (arc_jli_section): New type.
13638 (rc_jli_sections): New static variable.
13639 (arc_handle_jli_attribute): New function.
13640 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13641 (arc_file_end): New function.
13642 (TARGET_ASM_FILE_END): Define.
13643 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13644 (arc_add_jli_section): New function.
13645 (jli_call_scan): Likewise.
13646 (arc_reorg): Call jli_call_scan.
13647 (arc_output_addsi): Remove 'S' from printing asm operand.
13648 (arc_is_jli_call_p): New function.
13649 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13651 (movhi_insn): Likewise.
13652 (movsi_insn): Likewise.
13653 (movsi_set_cc_insn): Likewise.
13654 (loadqi_update): Likewise.
13655 (load_zeroextendqisi_update): Likewise.
13656 (load_signextendqisi_update): Likewise.
13657 (loadhi_update): Likewise.
13658 (load_zeroextendhisi_update): Likewise.
13659 (load_signextendhisi_update): Likewise.
13660 (loadsi_update): Likewise.
13661 (loadsf_update): Likewise.
13662 (movsicc_insn): Likewise.
13663 (bset_insn): Likewise.
13664 (bxor_insn): Likewise.
13665 (bclr_insn): Likewise.
13666 (bmsk_insn): Likewise.
13667 (bicsi3_insn): Likewise.
13668 (cmpsi_cc_c_insn): Likewise.
13669 (movsi_ne): Likewise.
13670 (movsi_cond_exec): Likewise.
13671 (clrsbsi2): Likewise.
13672 (norm_f): Likewise.
13682 (call_i): Remove 'S' asm letter, add jli instruction.
13683 (call_value_i): Likewise.
13684 * config/arc/arc.op (mjli-always): New option.
13685 * config/arc/constraints.md (Cji): New constraint.
13686 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13688 (subsf3_fpx): Likewise.
13689 (mulsf3_fpx): Likewise.
13690 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13692 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13693 function attrbutes.
13694 * doc/invoke.texi (ARC): Document mjli-always option.
13696 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13698 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13699 avoid addition with 0 and use incw and decw where possible.
13701 2018-01-26 Richard Biener <rguenther@suse.de>
13703 PR tree-optimization/81082
13704 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13705 association if it requires casting to unsigned.
13706 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13707 from fold_plusminus_mult_expr to catch important cases late when
13708 range info is available.
13710 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13712 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13713 * configure.ac (hidden_linkonce): New test.
13714 * configure: Regenerate.
13715 * config.in: Regenerate.
13717 2018-01-26 Julia Koval <julia.koval@intel.com>
13719 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13720 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13721 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13722 _mm_mask_bitshuffle_epi64_mask): Fix type.
13723 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13724 USI_FTYPE_V4DI_V4DI_USI): Remove.
13725 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13726 __builtin_ia32_vpshufbitqmb256_mask,
13727 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13728 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13729 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13731 2018-01-26 Alan Modra <amodra@gmail.com>
13734 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13735 UNSPEC_VBPERMQ. Sort other unspecs.
13737 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13739 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13741 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13743 PR middle-end/83055
13744 * predict.c (drop_profile): Do not push/pop cfun; update also
13746 (handle_missing_profiles): Fix logic looking for zero profiles.
13748 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13750 PR middle-end/83977
13751 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13752 on functions with #pragma omp declare simd or functions with simd
13754 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13755 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13756 Remove trailing \n from warning_at calls.
13758 2018-01-25 Tom de Vries <tom@codesourcery.com>
13761 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13762 for neutered workers.
13764 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13767 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13768 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13770 2018-01-24 Jeff Law <law@redhat.com>
13773 * i386.c (get_probe_interval): Move to earlier point.
13774 (ix86_compute_frame_layout): If -fstack-clash-protection and
13775 the frame is larger than the probe interval, then use pushes
13776 to save registers rather than reg->mem moves.
13777 (ix86_expand_prologue): Remove conditional for int_registers_saved
13780 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13783 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13784 min/max for never referenced object.
13786 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13788 PR middle-end/83977
13789 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13791 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13792 attributes from DECL_ATTRIBUTES (decl) without affecting
13793 DECL_ATTRIBUTES (current_function_decl).
13794 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13795 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13797 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13799 PR tree-optimization/83979
13800 * fold-const.c (fold_comparison): Use constant_boolean_node
13801 instead of boolean_{true,false}_node.
13803 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13805 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13808 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13810 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13811 Simplify the clause that sets the length attribute.
13812 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13813 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13814 clause that sets the length attribute.
13815 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13817 2018-01-24 Tom de Vries <tom@codesourcery.com>
13820 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13821 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13822 Add strict parameter.
13823 (prevent_branch_around_nothing): Insert dummy insn between branch to
13824 label and label with no ptx insn inbetween.
13825 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13827 2018-01-24 Tom de Vries <tom@codesourcery.com>
13830 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13831 for neutered threads in warp.
13832 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13834 2018-01-24 Richard Biener <rguenther@suse.de>
13836 PR tree-optimization/83176
13837 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13840 2018-01-24 Richard Biener <rguenther@suse.de>
13842 PR tree-optimization/82819
13843 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13844 code generating pluses that are no-ops in the target precision.
13846 2018-01-24 Richard Biener <rguenther@suse.de>
13848 PR middle-end/84000
13849 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13851 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13853 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13854 to merge probabilities.
13855 * predict.c (probably_never_executed): Also mark as cold functions
13856 with global 0 profile and guessed local profile.
13857 * profile-count.c (profile_probability::combine_with_count): New
13859 * profile-count.h (profile_probability::operator*,
13860 profile_probability::operator*=, profile_probability::operator/,
13861 profile_probability::operator/=): Reduce precision to adjusted
13862 and set value to guessed on contradictory divisions.
13863 (profile_probability::combine_with_freq): Remove.
13864 (profile_probability::combine_wiht_count): Declare.
13865 (profile_count::force_nonzero):: Set to adjusted.
13866 (profile_count::probability_in):: Set quality to adjusted.
13867 * tree-ssa-tail-merge.c (replace_block_by): Use
13868 combine_with_count.
13870 2018-01-23 Andrew Waterman <andrew@sifive.com>
13871 Jim Wilson <jimw@sifive.com>
13873 * config/riscv/riscv.c (riscv_stack_boundary): New.
13874 (riscv_option_override): Set riscv_stack_boundary. Handle
13875 riscv_preferred_stack_boundary_arg.
13876 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13877 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13878 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13879 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13880 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13881 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13883 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13886 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13887 of struct ix86_frame.
13888 (ix86_expand_epilogue): Likewise. Add a local variable for
13889 the reg_save_offset field in struct ix86_frame.
13891 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13893 PR tree-optimization/82604
13894 * tree-loop-distribution.c (enum partition_kind): New enum item
13895 PKIND_PARTIAL_MEMSET.
13896 (partition_builtin_p): Support above new enum item.
13897 (generate_code_for_partition): Ditto.
13898 (compute_access_range): Differentiate cases that equality can be
13899 proven at all loops, the innermost loops or no loops.
13900 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13901 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13902 (finalize_partitions, distribute_loop): Don't fuse partition of
13903 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13904 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13905 parloop is enabled.
13907 2018-01-23 Martin Liska <mliska@suse.cz>
13909 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13910 order to ignore the predictor.
13911 (PRED_POLYMORPHIC_CALL): Likewise.
13912 (PRED_RECURSIVE_CALL): Likewise.
13914 2018-01-23 Martin Liska <mliska@suse.cz>
13916 * tree-profile.c (tree_profiling): Print function header to
13917 aware reader which function we are working on.
13918 * value-prof.c (gimple_find_values_to_profile): Do not print
13919 not interesting value histograms.
13921 2018-01-23 Martin Liska <mliska@suse.cz>
13923 * profile-count.h (enum profile_quality): Add
13924 profile_uninitialized as the first value. Do not number values
13925 as they are zero based.
13926 (profile_count::verify): Update sanity check.
13927 (profile_probability::verify): Likewise.
13929 2018-01-23 Nathan Sidwell <nathan@acm.org>
13931 * doc/invoke.texi (ffor-scope): Deprecate.
13933 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13935 PR tree-optimization/83510
13936 * domwalk.c (set_all_edges_as_executable): New function.
13937 (dom_walker::dom_walker): Convert bool param
13938 "skip_unreachable_blocks" to enum reachability. Move setup of
13939 edge flags to set_all_edges_as_executable and only do it when
13940 reachability is REACHABLE_BLOCKS.
13941 * domwalk.h (enum dom_walker::reachability): New enum.
13942 (dom_walker::dom_walker): Convert bool param
13943 "skip_unreachable_blocks" to enum reachability.
13944 (set_all_edges_as_executable): New decl.
13945 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13946 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13948 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13949 but converting true to REACHABLE_BLOCKS.
13950 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13952 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13953 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13954 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13956 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13957 if check_all_array_refs will be called.
13959 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13961 * tree.c (selftest::test_location_wrappers): Add more test
13964 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13966 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13967 (selftest::test_bit_in_range): Likewise.
13969 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13972 * doc/sourcebuild.texi (vect_float): Say that the selector
13973 only describes the situation when -funsafe-math-optimizations is on.
13974 (vect_float_strict): Document.
13976 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13978 PR tree-optimization/83965
13979 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13980 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13981 instead of checking only for a reduction.
13982 (vect_recog_widen_sum_pattern): Likewise.
13984 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13986 * predict.c (probably_never_executed): Only use precise profile info.
13987 (compute_function_frequency): Skip after inlining hack since we now
13988 have quality checking.
13990 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13992 * profile-count.h (profile_probability::very_unlikely,
13993 profile_probability::unlikely, profile_probability::even): Set
13994 precision to guessed.
13996 2018-01-23 Richard Biener <rguenther@suse.de>
13998 PR tree-optimization/83963
13999 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14000 Properly terminate dominator walk when crossing the exit edge not
14001 when visiting its source block.
14003 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14006 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14007 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14009 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14011 PR tree-optimization/83957
14012 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14013 semicolon after for body surrounded by braces.
14015 PR tree-optimization/83081
14016 * profile-count.h (profile_probability::split): New method.
14017 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14018 Use profile_probability::split.
14019 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14020 when splitting a single conditional jump into 2.
14022 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14024 PR tree-optimization/69452
14025 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14028 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14030 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14031 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14032 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14034 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14036 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14038 * config/rl78/rl78.md (movdi): New define_expand.
14039 * config/rl78/rl78.c (rl78_split_movdi): New function.
14041 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14044 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14046 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14047 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14048 128-bit to produce an UNSPEC move to get the double word with the
14049 signbit and then a shift directly to do signbit.
14050 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14051 implementation with a new version that just does either a direct
14052 move or a regular move. Move memory interface to separate insns.
14053 Move insns so they are next to the expander.
14054 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14055 with signbit move. Split big and little endian case.
14056 (signbit<mode>2_dm_mem_le): Likewise.
14057 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14058 (signbit<mode>2_dm2): Likewise.
14060 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14062 * config/rl78/rl78.md (anddi3): New define_expand.
14064 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14066 * config/rl78/rl78.md (umindi3): New define_expand.
14068 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14070 * config/rl78/rl78.md (smindi3): New define_expand.
14072 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14074 * config/rl78/rl78.md (smaxdi3): New define_expand.
14076 2018-01-22 Carl Love <cel@us.ibm.com>
14078 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14079 LVX_V1TI): Add macro expansion.
14080 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14081 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14082 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14083 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14084 Change check to determine if the instruction is a byte reversing
14085 entry. Fix typo in comment.
14086 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14087 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14088 Add def_builtin calls for new builtins.
14089 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14090 Add define_insn expansion.
14092 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14094 * config/rl78/rl78.md (umaxdi3): New define_expand.
14096 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14098 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14099 for non-QImode registers.
14101 2018-01-22 Richard Biener <rguenther@suse.de>
14103 PR tree-optimization/83963
14104 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14105 including the loop exit block.
14106 (scop_detection::merge_sese): Likewise.
14107 (scop_detection::add_scop): Do it here instead.
14109 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14111 * doc/sourcebuild.texi (arm_softfloat): Document.
14113 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14116 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14117 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14118 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14120 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14121 David Edelsohn <dje.gcc@gmail.com>
14124 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14125 Change "crset eq" to "crset 2".
14126 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14127 (*call_indirect_aix<mode>_nospec): Likewise.
14128 (*call_value_indirect_aix<mode>_nospec): Likewise.
14129 (*call_indirect_elfv2<mode>_nospec): Likewise.
14130 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14131 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14132 change assembly output from . to $.
14133 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14134 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14135 (*tablejump<mode>_internal1_nospec): Likewise.
14137 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14140 * config/sh/sh_optimize_sett_clrt.cc:
14141 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14143 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14145 PR tree-optimization/83940
14146 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14147 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14148 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14149 use it to pass back the definition type.
14150 (vect_check_store_rhs): Likewise rhs_dt_out.
14151 (vect_build_gather_load_calls): Add a mask_dt argument and use
14152 it instead of a call to vect_is_simple_use.
14153 (vectorizable_store): Update calls to vect_check_load_store_mask
14154 and vect_check_store_rhs. Use the dt returned by the latter instead
14155 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14156 instead of calls to vect_is_simple_use. Pass the scalar rather
14157 than the vector operand to vect_is_simple_use when handling
14158 second and subsequent copies of an rhs value.
14159 (vectorizable_load): Update calls to vect_check_load_store_mask
14160 and vect_build_gather_load_calls. Use the cached mask_dt and
14161 gs_info.offset_dt instead of calls to vect_is_simple_use.
14163 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14165 PR middle-end/83945
14166 * tree-emutls.c: Include gimplify.h.
14167 (lower_emutls_2): New function.
14168 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14169 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14170 it before further processing.
14173 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14174 UINTVAL (trueop1) instead of INTVAL (op1).
14176 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14180 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14181 INCOMING_FRAME_SP_OFFSET if not defined.
14182 (scan_trace): Add ENTRY argument. If true and
14183 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14184 emit a note to adjust the CFA offset.
14185 (create_cfi_notes): Adjust scan_trace callers.
14186 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14187 INCOMING_FRAME_SP_OFFSET in the CIE.
14188 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14189 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14191 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14192 * doc/tm.texi: Regenerated.
14194 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14196 PR rtl-optimization/83147
14197 * lra-constraints.c (remove_inheritance_pseudos): Use
14198 lra_substitute_pseudo_within_insn.
14200 2018-01-19 Tom de Vries <tom@codesourcery.com>
14201 Cesar Philippidis <cesar@codesourcery.com>
14204 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14206 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14209 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14210 spaces for function labels.
14212 2018-01-19 Martin Liska <mliska@suse.cz>
14214 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14215 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14216 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14217 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14218 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14219 (PRED_CONST_RETURN): Change from 69 to 65.
14220 (PRED_NULL_RETURN): Change from 91 to 71.
14221 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14222 (PRED_LOOP_GUARD): Change from 66 to 73.
14224 2018-01-19 Martin Liska <mliska@suse.cz>
14226 * predict.c (predict_insn_def): Add new assert.
14227 (struct branch_predictor): Change type to signed integer.
14228 (test_prediction_value_range): Amend test to cover
14229 PROB_UNINITIALIZED.
14230 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14231 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14232 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14233 (PRED_LOOP_IV_COMPARE): Likewise.
14234 * predict.h (PROB_UNINITIALIZED): Define new constant.
14236 2018-01-19 Martin Liska <mliska@suse.cz>
14238 * predict.c (dump_prediction): Add new format for
14239 analyze_brprob.py script which is enabled with -details
14241 * profile-count.h (precise_p): New function.
14243 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14245 PR tree-optimization/83922
14246 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14247 there are no statements that need masking.
14248 (vect_active_double_reduction_p): New function.
14249 (vect_analyze_loop_operations): Use it when handling phis that
14250 are not in the loop header.
14252 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14254 PR tree-optimization/83914
14255 * tree-vect-loop.c (vectorizable_induction): Don't convert
14256 init_expr or apply the peeling adjustment for inductions
14257 that are nested within the vectorized loop.
14259 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14261 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14264 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14268 * function.h (gimplify_parameters): Add gimple_seq * argument.
14269 * function.c: Include gimple.h and options.h.
14270 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14271 for the added local temporaries if needed.
14272 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14273 if there are any parameter cleanups, wrap whole body into a
14274 try/finally with the cleanups.
14276 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14279 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14280 Use GET_MODE_CLASS for scalar floating point.
14282 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14286 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14287 Fix call of call_cgraph_insertion_hooks.
14289 2018-01-18 Martin Sebor <msebor@redhat.com>
14291 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14293 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14296 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14299 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14302 * common.opt: (-ffile-prefix-map): New option.
14303 * opts.c (common_handle_option): Defer it.
14304 * opts-global.c (handle_common_deferred_options): Handle it.
14305 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14306 * file-prefix-map.h: New file.
14307 (remap_debug_filename, add_debug_prefix_map): ...here.
14308 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14309 * final.c (debug_prefix_map, add_debug_prefix_map
14310 remap_debug_filename): Move to...
14311 * file-prefix-map.c: New file.
14312 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14313 generalize, get rid of alloca(), use strrchr() instead of strchr().
14314 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14315 Implement in terms of add_prefix_map().
14316 (remap_macro_filename, remap_debug_filename): Implement in term of
14318 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14319 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14320 * dbxout.c: Include file-prefix-map.h.
14321 * varasm.c: Likewise.
14322 * vmsdbgout.c: Likewise.
14323 * xcoffout.c: Likewise.
14324 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14325 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14326 * doc/invoke.texi (-ffile-prefix-map): Document.
14327 (-fdebug-prefix-map): Update description.
14329 2018-01-18 Martin Liska <mliska@suse.cz>
14331 * config/i386/i386.c (indirect_thunk_name): Document that also
14333 (output_indirect_thunk): Document why both instructions
14334 (pause and lfence) are generated.
14336 2018-01-18 Richard Biener <rguenther@suse.de>
14338 PR tree-optimization/83887
14339 * graphite-scop-detection.c
14340 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14341 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14342 (scop_detection::merge_sese): Re-implement with a flood-fill
14343 algorithm that properly finds a SESE region if it exists.
14345 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14348 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14349 pointer_diff optimizations use view_convert instead of convert.
14351 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14353 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14354 Generate different code for -mno-speculate-indirect-jumps.
14355 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14356 (*call_indirect_aix<mode>): Disable for
14357 -mno-speculate-indirect-jumps.
14358 (*call_indirect_aix<mode>_nospec): New define_insn.
14359 (*call_value_indirect_aix<mode>): Disable for
14360 -mno-speculate-indirect-jumps.
14361 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14362 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14363 -mno-speculate-indirect-jumps.
14364 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14366 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14368 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14369 long double type, set the flags for noting the default long double
14370 type, even if we don't pass or return a long double type.
14372 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14375 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14378 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14380 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14381 support for merge[hl].
14382 (fold_mergehl_helper): New helper function.
14383 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14384 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14385 (altivec_vmrglw_direct): Add xxmrglw insn.
14387 2018-01-17 Andrew Waterman <andrew@sifive.com>
14389 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14390 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14392 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14395 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14396 call the lto_location_cache before reading the
14397 DECL_SOURCE_LOCATION of the types.
14399 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14400 Richard Sandiford <richard.sandiford@linaro.org>
14402 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14403 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14404 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14405 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14406 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14408 * config/aarch64/constraints.md (aarch64_movti_operand):
14410 * config/aarch64/predicates.md (Uti): Add new constraint.
14412 2018-01-17 Carl Love <cel@us.ibm.com>
14414 * config/rs6000/vsx.md (define_expand xl_len_r,
14415 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14416 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14418 (define_expand, define_insn): Move the shift left from the
14419 define_insn to the define_expand for lxvl and stxvl instructions.
14420 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14421 and XL_LEN_R definitions to PURE.
14423 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14425 * config/i386/i386.c (indirect_thunk_name): Declare regno
14426 as unsigned int. Compare regno with INVALID_REGNUM.
14427 (output_indirect_thunk): Ditto.
14428 (output_indirect_thunk_function): Ditto.
14429 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14430 in the call to output_indirect_thunk_function.
14432 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14434 PR middle-end/83884
14435 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14436 rather than the size of inner_type to determine the stack slot size
14437 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14439 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14442 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14445 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14447 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14448 endian Linux systems to optionally enable multilibs for selecting
14449 the long double type if the user configured an explicit type.
14450 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14451 have no long double multilibs if not defined.
14452 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14453 warn if the user used -mabi={ieee,ibm}longdouble and we built
14454 multilibs for long double.
14455 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14456 appropriate multilib option.
14457 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14459 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14460 for building long double multilibs.
14461 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14463 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14465 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14468 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14470 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14473 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14476 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14479 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14481 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14482 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14484 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14486 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14487 different rtl trees depending on TARGET_64BIT.
14488 (rs6000_gen_lvx): Likewise.
14490 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14492 * config/visium/visium.md (nop): Tweak comment.
14493 (hazard_nop): Likewise.
14495 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14497 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14498 -mspeculate-indirect-jumps.
14499 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14500 for -mno-speculate-indirect-jumps.
14501 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14502 (*call_value_indirect_elfv2<mode>): Disable for
14503 -mno-speculate-indirect-jumps.
14504 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14505 (indirect_jump): Emit different RTL for
14506 -mno-speculate-indirect-jumps.
14507 (*indirect_jump<mode>): Disable for
14508 -mno-speculate-indirect-jumps.
14509 (*indirect_jump<mode>_nospec): New define_insn.
14510 (tablejump): Emit different RTL for
14511 -mno-speculate-indirect-jumps.
14512 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14513 (tablejumpsi_nospec): New define_expand.
14514 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14515 (tablejumpdi_nospec): New define_expand.
14516 (*tablejump<mode>_internal1): Disable for
14517 -mno-speculate-indirect-jumps.
14518 (*tablejump<mode>_internal1_nospec): New define_insn.
14519 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14522 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14524 * caller-save.c (insert_save): Drop unnecessary parameter. All
14527 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14528 Richard Biener <rguenth@suse.de>
14531 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14532 return early, inline manually is_gimple_sizepos. Make sure if we
14533 call gimplify_expr we don't end up with a gimple constant.
14534 * tree.c (variably_modified_type_p): Don't return true for
14535 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14536 * gimplify.h (is_gimple_sizepos): Remove.
14538 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14540 PR tree-optimization/83857
14541 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14542 vectorizable_live_operation for pure SLP statements.
14543 (vectorizable_live_operation): Handle PHIs.
14545 2018-01-16 Richard Biener <rguenther@suse.de>
14547 PR tree-optimization/83867
14548 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14549 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14551 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14554 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14555 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14556 If off is not INTEGER_CST, issue a may not be aligned warning
14557 rather than isn't aligned. Use isn%'t rather than isn't.
14558 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14560 <case MULT_EXPR>: Improve the case when bottom and one of the
14561 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14562 operand, in that case check if the other operand is multiple of
14563 bottom divided by the INTEGER_CST operand.
14565 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14568 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14569 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14570 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14571 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14572 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14573 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14574 (pa_function_arg_size): New function.
14576 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14578 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14579 in a separate statement.
14581 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14583 PR tree-optimization/83847
14584 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14585 group gathers and scatters.
14587 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14589 PR rtl-optimization/86620
14590 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14592 PR rtl-optimization/83213
14593 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14594 to last if both are JUMP_INSNs.
14596 PR tree-optimization/83843
14597 * gimple-ssa-store-merging.c
14598 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14599 store_immediate_info for bswap/nop orig_stores.
14601 2018-01-15 Andrew Waterman <andrew@sifive.com>
14603 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14605 <UDIV>: Increase cost if !TARGET_DIV.
14607 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14609 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14610 (define_attr "cr_logical_3op"): New.
14611 (cceq_ior_compare): Adjust.
14612 (cceq_ior_compare_complement): Adjust.
14613 (*cceq_rev_compare): Adjust.
14614 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14615 (is_cracked_insn): Adjust.
14616 (insn_must_be_first_in_group): Adjust.
14617 * config/rs6000/40x.md: Adjust.
14618 * config/rs6000/440.md: Adjust.
14619 * config/rs6000/476.md: Adjust.
14620 * config/rs6000/601.md: Adjust.
14621 * config/rs6000/603.md: Adjust.
14622 * config/rs6000/6xx.md: Adjust.
14623 * config/rs6000/7450.md: Adjust.
14624 * config/rs6000/7xx.md: Adjust.
14625 * config/rs6000/8540.md: Adjust.
14626 * config/rs6000/cell.md: Adjust.
14627 * config/rs6000/e300c2c3.md: Adjust.
14628 * config/rs6000/e500mc.md: Adjust.
14629 * config/rs6000/e500mc64.md: Adjust.
14630 * config/rs6000/e5500.md: Adjust.
14631 * config/rs6000/e6500.md: Adjust.
14632 * config/rs6000/mpc.md: Adjust.
14633 * config/rs6000/power4.md: Adjust.
14634 * config/rs6000/power5.md: Adjust.
14635 * config/rs6000/power6.md: Adjust.
14636 * config/rs6000/power7.md: Adjust.
14637 * config/rs6000/power8.md: Adjust.
14638 * config/rs6000/power9.md: Adjust.
14639 * config/rs6000/rs64.md: Adjust.
14640 * config/rs6000/titan.md: Adjust.
14642 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14644 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14645 ix86_indirect_branch_register logic.
14647 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14649 * config/i386/constraints.md (Bs): Update
14650 ix86_indirect_branch_register check. Don't check
14651 ix86_indirect_branch_register with GOT_memory_operand.
14653 * config/i386/predicates.md (GOT_memory_operand): Don't check
14654 ix86_indirect_branch_register here.
14655 (GOT32_symbol_operand): Likewise.
14657 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14659 * config/i386/predicates.md (constant_call_address_operand):
14660 Rewrite ix86_indirect_branch_register logic.
14661 (sibcall_insn_operand): Likewise.
14663 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14665 * config/i386/constraints.md (Bs): Replace
14666 ix86_indirect_branch_thunk_register with
14667 ix86_indirect_branch_register.
14669 * config/i386/i386.md (indirect_jump): Likewise.
14670 (tablejump): Likewise.
14671 (*sibcall_memory): Likewise.
14672 (*sibcall_value_memory): Likewise.
14673 Peepholes of indirect call and jump via memory: Likewise.
14674 * config/i386/i386.opt: Likewise.
14675 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14676 (GOT_memory_operand): Likewise.
14677 (call_insn_operand): Likewise.
14678 (sibcall_insn_operand): Likewise.
14679 (GOT32_symbol_operand): Likewise.
14681 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14683 PR middle-end/83837
14684 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14685 type rather than type addr's type points to.
14686 (expand_omp_atomic_mutex): Likewise.
14687 (expand_omp_atomic): Likewise.
14689 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14692 * config/i386/i386.c (output_indirect_thunk_function): Use
14693 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14694 for __x86_return_thunk.
14696 2018-01-15 Richard Biener <rguenther@suse.de>
14698 PR middle-end/83850
14699 * expmed.c (extract_bit_field_1): Fix typo.
14701 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14704 * config/arm/iterators.md (VF): New mode iterator.
14705 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14706 Remove integer-related logic from pattern.
14707 (neon_vabd<mode>_3): Likewise.
14709 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14711 PR middle-end/82694
14712 * common.opt (fstrict-overflow): No longer an alias.
14713 (fwrapv-pointer): New option.
14714 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14715 also for pointer types based on flag_wrapv_pointer.
14716 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14717 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14718 opts->x_flag_wrapv got set.
14719 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14720 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14721 POINTER_TYPE_OVERFLOW_UNDEFINED.
14722 * match.pd: Likewise in address comparison pattern.
14723 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14725 2018-01-15 Richard Biener <rguenther@suse.de>
14728 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14729 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14730 Reset type names to their identifier if their TYPE_DECL doesn't
14731 have linkage (and thus is used for ODR and devirt).
14732 (save_debug_info_for_decl): Remove.
14733 (save_debug_info_for_type): Likewise.
14734 (add_tree_to_fld_list): Adjust.
14735 * tree-pretty-print.c (dump_generic_node): Make dumping of
14736 type names more robust.
14738 2018-01-15 Richard Biener <rguenther@suse.de>
14740 * BASE-VER: Bump to 8.0.1.
14742 2018-01-14 Martin Sebor <msebor@redhat.com>
14745 * builtins.c (check_access): Avoid warning when the no-warning bit
14748 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14750 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14751 * ira-color (allocno_hard_regs_compare): Likewise.
14753 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14756 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14757 Use .pushsection/.popsection.
14759 2018-01-14 Martin Sebor <msebor@redhat.com>
14762 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14764 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14766 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14767 entry from extra_headers.
14768 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14769 extra_headers, make the list bitwise identical to the i?86-*-* one.
14771 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14773 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14774 -mcmodel=large with -mindirect-branch=thunk,
14775 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14776 -mfunction-return=thunk-extern.
14777 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14778 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14779 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14781 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14783 * config/i386/i386.c (print_reg): Print the name of the full
14784 integer register without '%'.
14785 (ix86_print_operand): Handle 'V'.
14786 * doc/extend.texi: Document 'V' modifier.
14788 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14790 * config/i386/constraints.md (Bs): Disallow memory operand for
14791 -mindirect-branch-register.
14793 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14794 (GOT_memory_operand): Likewise.
14795 (call_insn_operand): Likewise.
14796 (sibcall_insn_operand): Likewise.
14797 (GOT32_symbol_operand): Likewise.
14798 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14799 for -mindirect-branch-register.
14800 (tablejump): Likewise.
14801 (*sibcall_memory): Likewise.
14802 (*sibcall_value_memory): Likewise.
14803 Disallow peepholes of indirect call and jump via memory for
14804 -mindirect-branch-register.
14805 (*call_pop): Replace m with Bw.
14806 (*call_value_pop): Likewise.
14807 (*sibcall_pop_memory): Replace m with Bs.
14808 * config/i386/i386.opt (mindirect-branch-register): New option.
14809 * doc/invoke.texi: Document -mindirect-branch-register option.
14811 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14813 * config/i386/i386-protos.h (ix86_output_function_return): New.
14814 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14815 set function_return_type.
14816 (indirect_thunk_name): Add ret_p to indicate thunk for function
14818 (output_indirect_thunk_function): Pass false to
14819 indirect_thunk_name.
14820 (ix86_output_indirect_branch_via_reg): Likewise.
14821 (ix86_output_indirect_branch_via_push): Likewise.
14822 (output_indirect_thunk_function): Create alias for function
14823 return thunk if regno < 0.
14824 (ix86_output_function_return): New function.
14825 (ix86_handle_fndecl_attribute): Handle function_return.
14826 (ix86_attribute_table): Add function_return.
14827 * config/i386/i386.h (machine_function): Add
14828 function_return_type.
14829 * config/i386/i386.md (simple_return_internal): Use
14830 ix86_output_function_return.
14831 (simple_return_internal_long): Likewise.
14832 * config/i386/i386.opt (mfunction-return=): New option.
14833 (indirect_branch): Mention -mfunction-return=.
14834 * doc/extend.texi: Document function_return function attribute.
14835 * doc/invoke.texi: Document -mfunction-return= option.
14837 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14839 * config/i386/i386-opts.h (indirect_branch): New.
14840 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14841 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14842 with local indirect jump when converting indirect call and jump.
14843 (ix86_set_indirect_branch_type): New.
14844 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14845 (indirectlabelno): New.
14846 (indirect_thunk_needed): Likewise.
14847 (indirect_thunk_bnd_needed): Likewise.
14848 (indirect_thunks_used): Likewise.
14849 (indirect_thunks_bnd_used): Likewise.
14850 (INDIRECT_LABEL): Likewise.
14851 (indirect_thunk_name): Likewise.
14852 (output_indirect_thunk): Likewise.
14853 (output_indirect_thunk_function): Likewise.
14854 (ix86_output_indirect_branch_via_reg): Likewise.
14855 (ix86_output_indirect_branch_via_push): Likewise.
14856 (ix86_output_indirect_branch): Likewise.
14857 (ix86_output_indirect_jmp): Likewise.
14858 (ix86_code_end): Call output_indirect_thunk_function if needed.
14859 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14861 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14862 (ix86_attribute_table): Add indirect_branch.
14863 * config/i386/i386.h (machine_function): Add indirect_branch_type
14864 and has_local_indirect_jump.
14865 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14867 (tablejump): Likewise.
14868 (*indirect_jump): Use ix86_output_indirect_jmp.
14869 (*tablejump_1): Likewise.
14870 (simple_return_indirect_internal): Likewise.
14871 * config/i386/i386.opt (mindirect-branch=): New option.
14872 (indirect_branch): New.
14875 (thunk-inline): Likewise.
14876 (thunk-extern): Likewise.
14877 * doc/extend.texi: Document indirect_branch function attribute.
14878 * doc/invoke.texi: Document -mindirect-branch= option.
14880 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14883 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14885 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14887 * ipa-inline.c (want_inline_small_function_p): Return false if
14888 inlining has already failed with CIF_FINAL_ERROR.
14889 (update_caller_keys): Call want_inline_small_function_p before
14891 (update_callee_keys): Likewise.
14893 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14895 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14897 (rs6000_quadword_masked_address_p): Likewise.
14898 (quad_aligned_load_p): Likewise.
14899 (quad_aligned_store_p): Likewise.
14900 (const_load_sequence_p): Add comment to describe the outer-most loop.
14901 (mimic_memory_attributes_and_flags): New function.
14902 (rs6000_gen_stvx): Likewise.
14903 (replace_swapped_aligned_store): Likewise.
14904 (rs6000_gen_lvx): Likewise.
14905 (replace_swapped_aligned_load): Likewise.
14906 (replace_swapped_load_constant): Capitalize argument name in
14907 comment describing this function.
14908 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14909 and stores that access quad-word aligned addresses and replace
14910 with stvx or lvx instructions when appropriate.
14911 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14912 New function prototype.
14913 (rs6000_quadword_masked_address_p): Likewise.
14914 (rs6000_gen_lvx): Likewise.
14915 (rs6000_gen_stvx): Likewise.
14916 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14917 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14918 when memory address is aligned.
14919 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14920 this split to select lvx instruction when memory address is aligned.
14921 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14922 instruction when memory address is aligned.
14923 (*vsx_le_perm_load_v16qi): Likewise.
14924 (four unnamed splitters): Modify to select the stvx instruction
14925 when memory is aligned.
14927 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14929 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14930 which appears in the queue multiple times.
14932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14933 Alan Hayward <alan.hayward@arm.com>
14934 David Sherwood <david.sherwood@arm.com>
14936 * tree-vectorizer.h (vec_lower_bound): New structure.
14937 (_loop_vec_info): Add check_nonzero and lower_bounds.
14938 (LOOP_VINFO_CHECK_NONZERO): New macro.
14939 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14940 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14941 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14942 fields. Make seg_len the distance travelled, not including the
14944 (dr_direction_indicator): Declare.
14945 (dr_zero_step_indicator): Likewise.
14946 (dr_known_forward_stride_p): Likewise.
14947 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14949 (runtime_alias_check_p): Allow runtime alias checks with
14951 (operator ==): Compare access_size and align.
14952 (prune_runtime_alias_test_list): Rework for new distinction between
14953 the access_size and seg_len.
14954 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14956 (get_segment_min_max): New function.
14957 (create_intersect_range_checks): Use it.
14958 (dr_step_indicator): New function.
14959 (dr_direction_indicator): Likewise.
14960 (dr_zero_step_indicator): Likewise.
14961 (dr_known_forward_stride_p): Likewise.
14962 * tree-loop-distribution.c (data_ref_segment_size): Return
14963 DR_STEP * (niters - 1).
14964 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14966 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14967 (vect_preserves_scalar_order_p): New function, split out from...
14968 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14969 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14970 (vect_vfa_access_size): New function.
14971 (vect_vfa_align): Likewise.
14972 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14973 (dump_lower_bound): New function.
14974 (vect_check_lower_bound): Likewise.
14975 (vect_small_gap_p): Likewise.
14976 (vectorizable_with_step_bound_p): Likewise.
14977 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14978 depencies if the vectorization factor is 1. Convert the checks
14979 for nonzero steps into checks on the bounds of DR_STEP. Try using
14980 a bunds check for variable steps if the minimum required step is
14981 relatively small. Update calls to the dr_with_seg_len
14982 constructor and to vect_compile_time_alias.
14983 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14985 (vect_loop_versioning): Call it.
14986 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14988 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14991 Alan Hayward <alan.hayward@arm.com>
14992 David Sherwood <david.sherwood@arm.com>
14994 * doc/sourcebuild.texi (vect_scatter_store): Document.
14995 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14997 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14999 * genopinit.c (main): Add supports_vec_scatter_store and
15000 supports_vec_scatter_store_cached to target_optabs.
15001 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15002 IFN_MASK_SCATTER_STORE.
15003 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15005 * internal-fn.h (internal_store_fn_p): Declare.
15006 (internal_fn_stored_value_index): Likewise.
15007 * internal-fn.c (scatter_store_direct): New macro.
15008 (expand_scatter_store_optab_fn): New function.
15009 (direct_scatter_store_optab_supported_p): New macro.
15010 (internal_store_fn_p): New function.
15011 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15012 IFN_MASK_SCATTER_STORE.
15013 (internal_fn_mask_index): Likewise.
15014 (internal_fn_stored_value_index): New function.
15015 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15016 for scatter stores.
15017 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15018 * optabs-query.c (supports_vec_scatter_store_p): New function.
15019 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15020 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15021 true for scatter stores.
15022 (vect_gather_scatter_fn_p): Handle scatter stores too.
15023 (vect_check_gather_scatter): Consider using scatter stores if
15024 supports_vec_scatter_store_p.
15025 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15026 scatter stores too.
15027 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15028 internal_fn_stored_value_index.
15029 (check_load_store_masking): Handle scatter stores too.
15030 (vect_get_store_rhs): Make public.
15031 (vectorizable_call): Use internal_store_fn_p.
15032 (vectorizable_store): Handle scatter store internal functions.
15033 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15034 when deciding whether the end of the group has been reached.
15035 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15036 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15037 (mask_scatter_store<mode>): New insns.
15039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15040 Alan Hayward <alan.hayward@arm.com>
15041 David Sherwood <david.sherwood@arm.com>
15043 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15044 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15045 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15047 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15048 Use vect_truncate_gather_scatter_offset if we can't treat the
15049 operation as a normal gather load or scatter store.
15050 (get_group_load_store_type): Take the gather_scatter_info
15051 as argument. Try using a gather load or scatter store for
15052 single-element groups.
15053 (get_load_store_type): Update calls to get_group_load_store_type
15054 and vect_use_strided_gather_scatters_p.
15056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15057 Alan Hayward <alan.hayward@arm.com>
15058 David Sherwood <david.sherwood@arm.com>
15060 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15061 optional tree argument.
15062 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15064 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15065 but continue to use the current value as a fallback.
15066 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15067 to compare the updates.
15068 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15069 (get_load_store_type): Use it when handling a strided access.
15070 (vect_get_strided_load_store_ops): New function.
15071 (vect_get_data_ptr_increment): Likewise.
15072 (vectorizable_load): Handle strided gather loads. Always pass
15073 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15076 Alan Hayward <alan.hayward@arm.com>
15077 David Sherwood <david.sherwood@arm.com>
15079 * doc/md.texi (gather_load@var{m}): Document.
15080 (mask_gather_load@var{m}): Likewise.
15081 * genopinit.c (main): Add supports_vec_gather_load and
15082 supports_vec_gather_load_cached to target_optabs.
15083 * optabs-tree.c (init_tree_optimization_optabs): Use
15084 ggc_cleared_alloc to allocate target_optabs.
15085 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15086 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15088 * internal-fn.h (internal_load_fn_p): Declare.
15089 (internal_gather_scatter_fn_p): Likewise.
15090 (internal_fn_mask_index): Likewise.
15091 (internal_gather_scatter_fn_supported_p): Likewise.
15092 * internal-fn.c (gather_load_direct): New macro.
15093 (expand_gather_load_optab_fn): New function.
15094 (direct_gather_load_optab_supported_p): New macro.
15095 (direct_internal_fn_optab): New function.
15096 (internal_load_fn_p): Likewise.
15097 (internal_gather_scatter_fn_p): Likewise.
15098 (internal_fn_mask_index): Likewise.
15099 (internal_gather_scatter_fn_supported_p): Likewise.
15100 * optabs-query.c (supports_at_least_one_mode_p): New function.
15101 (supports_vec_gather_load_p): Likewise.
15102 * optabs-query.h (supports_vec_gather_load_p): Declare.
15103 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15104 and memory_type field.
15105 (NUM_PATTERNS): Bump to 15.
15106 * tree-vect-data-refs.c: Include internal-fn.h.
15107 (vect_gather_scatter_fn_p): New function.
15108 (vect_describe_gather_scatter_call): Likewise.
15109 (vect_check_gather_scatter): Try using internal functions for
15110 gather loads. Recognize existing calls to a gather load function.
15111 (vect_analyze_data_refs): Consider using gather loads if
15112 supports_vec_gather_load_p.
15113 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15114 (vect_get_gather_scatter_offset_type): Likewise.
15115 (vect_convert_mask_for_vectype): Likewise.
15116 (vect_add_conversion_to_patterm): Likewise.
15117 (vect_try_gather_scatter_pattern): Likewise.
15118 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15119 (vect_vect_recog_func_ptrs): Add it.
15120 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15121 internal_fn_mask_index and internal_gather_scatter_fn_p.
15122 (check_load_store_masking): Take the gather_scatter_info as an
15123 argument and handle gather loads.
15124 (vect_get_gather_scatter_ops): New function.
15125 (vectorizable_call): Check internal_load_fn_p.
15126 (vectorizable_load): Likewise. Handle gather load internal
15128 (vectorizable_store): Update call to check_load_store_masking.
15129 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15130 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15131 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15132 (aarch64_gather_scale_operand_d): New predicates.
15133 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15134 (mask_gather_load<mode>): New insns.
15136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15137 Alan Hayward <alan.hayward@arm.com>
15138 David Sherwood <david.sherwood@arm.com>
15140 * optabs.def (fold_left_plus_optab): New optab.
15141 * doc/md.texi (fold_left_plus_@var{m}): Document.
15142 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15143 * internal-fn.c (fold_left_direct): Define.
15144 (expand_fold_left_optab_fn): Likewise.
15145 (direct_fold_left_optab_supported_p): Likewise.
15146 * fold-const-call.c (fold_const_fold_left): New function.
15147 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15148 * tree-parloops.c (valid_reduction_p): New function.
15149 (gather_scalar_reductions): Use it.
15150 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15151 (vect_finish_replace_stmt): Declare.
15152 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15153 (needs_fold_left_reduction_p): New function, split out from...
15154 (vect_is_simple_reduction): ...here. Accept reductions that
15155 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15156 (vect_force_simple_reduction): Also store the reduction type in
15157 the assignment's STMT_VINFO_REDUC_TYPE.
15158 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15159 (merge_with_identity): New function.
15160 (vect_expand_fold_left): Likewise.
15161 (vectorize_fold_left_reduction): Likewise.
15162 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15163 scalar phi in place for it. Check for target support and reject
15164 cases that would reassociate the operation. Defer the transform
15165 phase to vectorize_fold_left_reduction.
15166 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15167 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15168 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15172 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15173 call to ifc_temp_var.
15175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15176 Alan Hayward <alan.hayward@arm.com>
15177 David Sherwood <david.sherwood@arm.com>
15179 * target.def (legitimize_address_displacement): Take the original
15180 offset as a poly_int.
15181 * targhooks.h (default_legitimize_address_displacement): Update
15183 * targhooks.c (default_legitimize_address_displacement): Likewise.
15184 * doc/tm.texi: Regenerate.
15185 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15186 as an argument, moving assert of ad->disp == ad->disp_term to...
15187 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15188 Try calling targetm.legitimize_address_displacement before expanding
15189 the address rather than afterwards, and adjust for the new interface.
15190 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15191 Match the new hook interface. Handle SVE addresses.
15192 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15193 new hook interface.
15195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15197 * Makefile.in (OBJS): Add early-remat.o.
15198 * target.def (select_early_remat_modes): New hook.
15199 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15200 * doc/tm.texi: Regenerate.
15201 * targhooks.h (default_select_early_remat_modes): Declare.
15202 * targhooks.c (default_select_early_remat_modes): New function.
15203 * timevar.def (TV_EARLY_REMAT): New timevar.
15204 * passes.def (pass_early_remat): New pass.
15205 * tree-pass.h (make_pass_early_remat): Declare.
15206 * early-remat.c: New file.
15207 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15209 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15212 Alan Hayward <alan.hayward@arm.com>
15213 David Sherwood <david.sherwood@arm.com>
15215 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15216 vfm1 with a bound_epilog parameter.
15217 (vect_do_peeling): Update calls accordingly, and move the prologue
15218 call earlier in the function. Treat the base bound_epilog as 0 for
15219 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15220 this base when peeling for gaps.
15221 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15222 with fully-masked loops.
15223 (vect_estimate_min_profitable_iters): Handle the single peeled
15224 iteration in that case.
15226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15227 Alan Hayward <alan.hayward@arm.com>
15228 David Sherwood <david.sherwood@arm.com>
15230 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15231 single-element interleaving even if the size is not a power of 2.
15232 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15233 accesses for single-element interleaving if the group size is
15236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15237 Alan Hayward <alan.hayward@arm.com>
15238 David Sherwood <david.sherwood@arm.com>
15240 * doc/md.texi (fold_extract_last_@var{m}): Document.
15241 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15242 * optabs.def (fold_extract_last_optab): New optab.
15243 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15244 * internal-fn.c (fold_extract_direct): New macro.
15245 (expand_fold_extract_optab_fn): Likewise.
15246 (direct_fold_extract_optab_supported_p): Likewise.
15247 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15248 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15249 EXTRACT_LAST_REDUCTION.
15250 (get_initial_def_for_reduction): Do not create an initial vector
15251 for EXTRACT_LAST_REDUCTION reductions.
15252 (vectorizable_reduction): Leave the scalar phi in place for
15253 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15254 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15255 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15256 transform phase to vectorizable_condition.
15257 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15259 (vect_finish_stmt_generation): ...here.
15260 (vect_finish_replace_stmt): New function.
15261 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15262 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15264 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15267 Alan Hayward <alan.hayward@arm.com>
15268 David Sherwood <david.sherwood@arm.com>
15270 * doc/md.texi (extract_last_@var{m}): Document.
15271 * optabs.def (extract_last_optab): New optab.
15272 * internal-fn.def (EXTRACT_LAST): New internal function.
15273 * internal-fn.c (cond_unary_direct): New macro.
15274 (expand_cond_unary_optab_fn): Likewise.
15275 (direct_cond_unary_optab_supported_p): Likewise.
15276 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15277 loops using EXTRACT_LAST.
15278 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15279 (extract_last_<mode>): ...this optab.
15280 (vec_extract<mode><Vel>): Update accordingly.
15282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15283 Alan Hayward <alan.hayward@arm.com>
15284 David Sherwood <david.sherwood@arm.com>
15286 * target.def (empty_mask_is_expensive): New hook.
15287 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15288 * doc/tm.texi: Regenerate.
15289 * targhooks.h (default_empty_mask_is_expensive): Declare.
15290 * targhooks.c (default_empty_mask_is_expensive): New function.
15291 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15292 if the target says that empty masks are expensive.
15293 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15295 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15298 Alan Hayward <alan.hayward@arm.com>
15299 David Sherwood <david.sherwood@arm.com>
15301 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15302 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15303 (vect_use_loop_mask_for_alignment_p): New function.
15304 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15305 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15306 niters_skip argument. Make sure that the first niters_skip elements
15307 of the first iteration are inactive.
15308 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15309 Update call to vect_set_loop_masks_directly.
15310 (get_misalign_in_elems): New function, split out from...
15311 (vect_gen_prolog_loop_niters): ...here.
15312 (vect_update_init_of_dr): Take a code argument that specifies whether
15313 the adjustment should be added or subtracted.
15314 (vect_update_init_of_drs): Likewise.
15315 (vect_prepare_for_masked_peels): New function.
15316 (vect_do_peeling): Skip prologue peeling if we're using a mask
15317 instead. Update call to vect_update_inits_of_drs.
15318 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15320 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15321 alignment. Do not include the number of peeled iterations in
15322 the minimum threshold in that case.
15323 (vectorizable_induction): Adjust the start value down by
15324 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15325 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15326 Take the number of skipped iterations into account when calculating
15328 * tree-vect-stmts.c (vect_gen_while_not): New function.
15330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15331 Alan Hayward <alan.hayward@arm.com>
15332 David Sherwood <david.sherwood@arm.com>
15334 * doc/sourcebuild.texi (vect_fully_masked): Document.
15335 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15336 default value to 0.
15337 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15339 (vect_analyze_loop_2): ...here. Don't check the vectorization
15340 factor against the number of loop iterations if the loop is
15343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15344 Alan Hayward <alan.hayward@arm.com>
15345 David Sherwood <david.sherwood@arm.com>
15347 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15348 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15349 (dump_groups): Update accordingly.
15350 (iv_use::mem_type): New member variable.
15351 (address_p): New function.
15352 (record_use): Add a mem_type argument and initialize the new
15354 (record_group_use): Add a mem_type argument. Use address_p.
15355 Remove obsolete null checks of base_object. Update call to record_use.
15356 (find_interesting_uses_op): Update call to record_group_use.
15357 (find_interesting_uses_cond): Likewise.
15358 (find_interesting_uses_address): Likewise.
15359 (get_mem_type_for_internal_fn): New function.
15360 (find_address_like_use): Likewise.
15361 (find_interesting_uses_stmt): Try find_address_like_use before
15362 calling find_interesting_uses_op.
15363 (addr_offset_valid_p): Use the iv mem_type field as the type
15364 of the addressed memory.
15365 (add_autoinc_candidates): Likewise.
15366 (get_address_cost): Likewise.
15367 (split_small_address_groups_p): Use address_p.
15368 (split_address_groups): Likewise.
15369 (add_iv_candidate_for_use): Likewise.
15370 (autoinc_possible_for_pair): Likewise.
15371 (rewrite_groups): Likewise.
15372 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15373 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15374 (get_alias_ptr_type_for_ptr_address): New function.
15375 (rewrite_use_address): Rewrite address uses in calls that were
15376 identified by find_address_like_use.
15378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15379 Alan Hayward <alan.hayward@arm.com>
15380 David Sherwood <david.sherwood@arm.com>
15382 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15384 * gimple-expr.h (is_gimple_addressable: Likewise.
15385 * gimple-expr.c (is_gimple_address): Likewise.
15386 * internal-fn.c (expand_call_mem_ref): New function.
15387 (expand_mask_load_optab_fn): Use it.
15388 (expand_mask_store_optab_fn): Likewise.
15390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15391 Alan Hayward <alan.hayward@arm.com>
15392 David Sherwood <david.sherwood@arm.com>
15394 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15395 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15396 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15397 (cond_umax@var{mode}): Document.
15398 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15399 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15400 (cond_umin_optab, cond_umax_optab): New optabs.
15401 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15402 (COND_IOR, COND_XOR): New internal functions.
15403 * internal-fn.h (get_conditional_internal_fn): Declare.
15404 * internal-fn.c (cond_binary_direct): New macro.
15405 (expand_cond_binary_optab_fn): Likewise.
15406 (direct_cond_binary_optab_supported_p): Likewise.
15407 (get_conditional_internal_fn): New function.
15408 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15409 Cope with reduction statements that are vectorized as calls rather
15411 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15412 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15413 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15414 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15415 (UNSPEC_COND_EOR): New unspecs.
15416 (optab): Add mappings for them.
15417 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15418 (sve_int_op, sve_fp_op): New int attributes.
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 (while_ult_optab): New optab.
15425 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15426 * internal-fn.def (WHILE_ULT): New internal function.
15427 * internal-fn.h (direct_internal_fn_supported_p): New override
15428 that takes two types as argument.
15429 * internal-fn.c (while_direct): New macro.
15430 (expand_while_optab_fn): New function.
15431 (convert_optab_supported_p): Likewise.
15432 (direct_while_optab_supported_p): New macro.
15433 * wide-int.h (wi::udiv_ceil): New function.
15434 * tree-vectorizer.h (rgroup_masks): New structure.
15435 (vec_loop_masks): New typedef.
15436 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15437 and fully_masked_p.
15438 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15439 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15440 (vect_max_vf): New function.
15441 (slpeel_make_loop_iterate_ntimes): Delete.
15442 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15443 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15444 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15445 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15446 internal-fn.h, stor-layout.h and optabs-query.h.
15447 (vect_set_loop_mask): New function.
15448 (add_preheader_seq): Likewise.
15449 (add_header_seq): Likewise.
15450 (interleave_supported_p): Likewise.
15451 (vect_maybe_permute_loop_masks): Likewise.
15452 (vect_set_loop_masks_directly): Likewise.
15453 (vect_set_loop_condition_masked): Likewise.
15454 (vect_set_loop_condition_unmasked): New function, split out from
15455 slpeel_make_loop_iterate_ntimes.
15456 (slpeel_make_loop_iterate_ntimes): Rename to..
15457 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15458 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15459 (vect_do_peeling): Update call accordingly.
15460 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15462 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15463 mask_compare_type, can_fully_mask_p and fully_masked_p.
15464 (release_vec_loop_masks): New function.
15465 (_loop_vec_info): Use it to free the loop masks.
15466 (can_produce_all_loop_masks_p): New function.
15467 (vect_get_max_nscalars_per_iter): Likewise.
15468 (vect_verify_full_masking): Likewise.
15469 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15470 retries, and free the mask rgroups before retrying. Check loop-wide
15471 reasons for disallowing fully-masked loops. Make the final decision
15472 about whether use a fully-masked loop or not.
15473 (vect_estimate_min_profitable_iters): Do not assume that peeling
15474 for the number of iterations will be needed for fully-masked loops.
15475 (vectorizable_reduction): Disable fully-masked loops.
15476 (vectorizable_live_operation): Likewise.
15477 (vect_halve_mask_nunits): New function.
15478 (vect_double_mask_nunits): Likewise.
15479 (vect_record_loop_mask): Likewise.
15480 (vect_get_loop_mask): Likewise.
15481 (vect_transform_loop): Handle the case in which the final loop
15482 iteration might handle a partial vector. Call vect_set_loop_condition
15483 instead of slpeel_make_loop_iterate_ntimes.
15484 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15485 (check_load_store_masking): New function.
15486 (prepare_load_store_mask): Likewise.
15487 (vectorizable_store): Handle fully-masked loops.
15488 (vectorizable_load): Likewise.
15489 (supportable_widening_operation): Use vect_halve_mask_nunits for
15491 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15492 (vect_gen_while): New function.
15493 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15494 (aarch64_uqdec<mode>): New insn.
15496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15497 Alan Hayward <alan.hayward@arm.com>
15498 David Sherwood <david.sherwood@arm.com>
15500 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15501 (reduc_xor_scal_optab): New optabs.
15502 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15503 (reduc_xor_scal_@var{m}): Document.
15504 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15505 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15506 internal functions.
15507 * fold-const-call.c (fold_const_call): Handle them.
15508 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15509 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15510 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15511 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15512 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15513 (UNSPEC_XORV): New unspecs.
15514 (optab): Add entries for them.
15515 (BITWISEV): New int iterator.
15516 (bit_reduc_op): New int attributes.
15518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15519 Alan Hayward <alan.hayward@arm.com>
15520 David Sherwood <david.sherwood@arm.com>
15522 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15523 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15524 * optabs.def (vec_shl_insert_optab): New optab.
15525 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15526 (duplicate_and_interleave): Likewise.
15527 * tree-vect-loop.c: Include internal-fn.h.
15528 (neutral_op_for_slp_reduction): New function, split out from
15529 get_initial_defs_for_reduction.
15530 (get_initial_def_for_reduction): Handle option 2 for variable-length
15531 vectors by loading the neutral value into a vector and then shifting
15532 the initial value into element 0.
15533 (get_initial_defs_for_reduction): Replace the code argument with
15534 the neutral value calculated by neutral_op_for_slp_reduction.
15535 Use gimple_build_vector for constant-length vectors.
15536 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15537 but the first group_size elements have a neutral value.
15538 Use duplicate_and_interleave otherwise.
15539 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15540 Update call to get_initial_defs_for_reduction. Handle SLP
15541 reductions for variable-length vectors by creating one vector
15542 result for each scalar result, with the elements associated
15543 with other scalar results stubbed out with the neutral value.
15544 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15545 Require IFN_VEC_SHL_INSERT for double reductions on
15546 variable-length vectors, or SLP reductions that have
15547 a neutral value. Require can_duplicate_and_interleave_p
15548 support for variable-length unchained SLP reductions if there
15549 is no neutral value, such as for MIN/MAX reductions. Also require
15550 the number of vector elements to be a multiple of the number of
15551 SLP statements when doing variable-length unchained SLP reductions.
15552 Update call to vect_create_epilog_for_reduction.
15553 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15554 and remove initial values.
15555 (duplicate_and_interleave): Make public.
15556 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15557 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15560 Alan Hayward <alan.hayward@arm.com>
15561 David Sherwood <david.sherwood@arm.com>
15563 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15564 (can_duplicate_and_interleave_p): New function.
15565 (vect_get_and_check_slp_defs): Take the vector of statements
15566 rather than just the current one. Remove excess parentheses.
15567 Restriction rejectinon of vect_constant_def and vect_external_def
15568 for variable-length vectors to boolean types, or types for which
15569 can_duplicate_and_interleave_p is false.
15570 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15571 (duplicate_and_interleave): New function.
15572 (vect_get_constant_vectors): Use gimple_build_vector for
15573 constant-length vectors and suitable variable-length constant
15574 vectors. Use duplicate_and_interleave for other variable-length
15575 vectors. Don't defer the update when inserting new statements.
15577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15578 Alan Hayward <alan.hayward@arm.com>
15579 David Sherwood <david.sherwood@arm.com>
15581 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15582 min_profitable_iters doesn't go negative.
15584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15585 Alan Hayward <alan.hayward@arm.com>
15586 David Sherwood <david.sherwood@arm.com>
15588 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15589 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15590 * optabs.def (vec_mask_load_lanes_optab): New optab.
15591 (vec_mask_store_lanes_optab): Likewise.
15592 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15593 (MASK_STORE_LANES): Likewise.
15594 * internal-fn.c (mask_load_lanes_direct): New macro.
15595 (mask_store_lanes_direct): Likewise.
15596 (expand_mask_load_optab_fn): Handle masked operations.
15597 (expand_mask_load_lanes_optab_fn): New macro.
15598 (expand_mask_store_optab_fn): Handle masked operations.
15599 (expand_mask_store_lanes_optab_fn): New macro.
15600 (direct_mask_load_lanes_optab_supported_p): Likewise.
15601 (direct_mask_store_lanes_optab_supported_p): Likewise.
15602 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15604 (vect_load_lanes_supported): Likewise.
15605 * tree-vect-data-refs.c (strip_conversion): New function.
15606 (can_group_stmts_p): Likewise.
15607 (vect_analyze_data_ref_accesses): Use it instead of checking
15608 for a pair of assignments.
15609 (vect_store_lanes_supported): Take a masked_p parameter.
15610 (vect_load_lanes_supported): Likewise.
15611 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15612 vect_store_lanes_supported and vect_load_lanes_supported.
15613 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15614 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15615 parameter. Don't allow gaps for masked accesses.
15616 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15617 and vect_load_lanes_supported.
15618 (get_load_store_type): Take a masked_p parameter and update
15619 call to get_group_load_store_type.
15620 (vectorizable_store): Update call to get_load_store_type.
15621 Handle IFN_MASK_STORE_LANES.
15622 (vectorizable_load): Update call to get_load_store_type.
15623 Handle IFN_MASK_LOAD_LANES.
15625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15626 Alan Hayward <alan.hayward@arm.com>
15627 David Sherwood <david.sherwood@arm.com>
15629 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15631 * config/aarch64/aarch64-protos.h
15632 (aarch64_sve_struct_memory_operand_p): Declare.
15633 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15634 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15635 (VPRED, vpred): Handle SVE structure modes.
15636 * config/aarch64/constraints.md (Utx): New constraint.
15637 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15638 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15639 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15640 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15641 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15642 structure modes. Split into pieces after RA.
15643 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15644 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15646 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15647 SVE structure modes.
15648 (aarch64_classify_address): Likewise.
15649 (sizetochar): Move earlier in file.
15650 (aarch64_print_operand): Handle SVE register lists.
15651 (aarch64_array_mode): New function.
15652 (aarch64_sve_struct_memory_operand_p): Likewise.
15653 (TARGET_ARRAY_MODE): Redefine.
15655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15656 Alan Hayward <alan.hayward@arm.com>
15657 David Sherwood <david.sherwood@arm.com>
15659 * target.def (array_mode): New target hook.
15660 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15661 * doc/tm.texi: Regenerate.
15662 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15663 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15664 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15665 targetm.array_mode.
15666 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15670 Alan Hayward <alan.hayward@arm.com>
15671 David Sherwood <david.sherwood@arm.com>
15673 * fold-const.c (fold_binary_loc): Check the argument types
15674 rather than the result type when testing for a vector operation.
15676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15678 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15679 * doc/tm.texi: Regenerate.
15681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15682 Alan Hayward <alan.hayward@arm.com>
15683 David Sherwood <david.sherwood@arm.com>
15685 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15686 (sve): Document new AArch64 extension.
15687 * doc/md.texi (w): Extend the description of the AArch64
15688 constraint to include SVE vectors.
15689 (Upl, Upa): Document new AArch64 predicate constraints.
15690 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15692 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15693 (msve-vector-bits=): New option.
15694 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15695 SVE when these are disabled.
15696 (sve): New extension.
15697 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15698 modes. Adjust their number of units based on aarch64_sve_vg.
15699 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15700 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15701 aarch64_addr_query_type.
15702 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15703 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15704 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15705 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15706 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15707 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15708 (aarch64_simd_imm_zero_p): Delete.
15709 (aarch64_check_zero_based_sve_index_immediate): Declare.
15710 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15711 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15712 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15713 (aarch64_sve_float_mul_immediate_p): Likewise.
15714 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15715 rather than an rtx.
15716 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15717 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15718 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15719 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15720 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15721 (aarch64_regmode_natural_size): Likewise.
15722 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15723 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15725 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15726 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15727 for VG and the SVE predicate registers.
15728 (V_ALIASES): Add a "z"-prefixed alias.
15729 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15730 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15731 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15732 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15733 (REG_CLASS_NAMES): Add entries for them.
15734 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15735 and the predicate registers.
15736 (aarch64_sve_vg): Declare.
15737 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15738 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15739 (REGMODE_NATURAL_SIZE): Define.
15740 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15742 * config/aarch64/aarch64.c: Include cfgrtl.h.
15743 (simd_immediate_info): Add a constructor for series vectors,
15744 and an associated step field.
15745 (aarch64_sve_vg): New variable.
15746 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15747 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15748 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15749 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15750 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15751 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15752 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15753 (aarch64_get_mask_mode): New functions.
15754 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15755 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15756 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15757 predicate modes and predicate registers. Explicitly restrict
15758 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15759 to store a vector mode if it is recognized by
15760 aarch64_classify_vector_mode.
15761 (aarch64_regmode_natural_size): New function.
15762 (aarch64_hard_regno_caller_save_mode): Return the original mode
15764 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15765 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15766 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15767 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15769 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15770 does not overlap dest if the function is frame-related. Handle
15772 (aarch64_split_add_offset): New function.
15773 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15774 them aarch64_add_offset.
15775 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15776 and update call to aarch64_sub_sp.
15777 (aarch64_add_cfa_expression): New function.
15778 (aarch64_expand_prologue): Pass extra temporary registers to the
15779 functions above. Handle the case in which we need to emit new
15780 DW_CFA_expressions for registers that were originally saved
15781 relative to the stack pointer, but now have to be expressed
15782 relative to the frame pointer.
15783 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15785 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15786 IP0 and IP1 values for SVE frames.
15787 (aarch64_expand_vec_series): New function.
15788 (aarch64_expand_sve_widened_duplicate): Likewise.
15789 (aarch64_expand_sve_const_vector): Likewise.
15790 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15791 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15792 into the register, rather than emitting a SET directly.
15793 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15794 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15795 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15796 (offset_9bit_signed_scaled_p): New functions.
15797 (aarch64_replicate_bitmask_imm): New function.
15798 (aarch64_bitmask_imm): Use it.
15799 (aarch64_cannot_force_const_mem): Reject expressions involving
15800 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15801 (aarch64_classify_index): Handle SVE indices, by requiring
15802 a plain register index with a scale that matches the element size.
15803 (aarch64_classify_address): Handle SVE addresses. Assert that
15804 the mode of the address is VOIDmode or an integer mode.
15805 Update call to aarch64_classify_symbol.
15806 (aarch64_classify_symbolic_expression): Update call to
15807 aarch64_classify_symbol.
15808 (aarch64_const_vec_all_in_range_p): New function.
15809 (aarch64_print_vector_float_operand): Likewise.
15810 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15811 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15812 and the FP immediates 1.0 and 0.5.
15813 (aarch64_print_address_internal): Handle SVE addresses.
15814 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15815 (aarch64_regno_regclass): Handle predicate registers.
15816 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15818 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15819 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15820 (aarch64_convert_sve_vector_bits): New function.
15821 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15822 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15823 rather than an rtx.
15824 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15825 Handle SVE vector and predicate modes. Accept VL-based constants
15826 that need only one temporary register, and VL offsets that require
15827 no temporary registers.
15828 (aarch64_conditional_register_usage): Mark the predicate registers
15829 as fixed if SVE isn't available.
15830 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15831 Return true for SVE vector and predicate modes.
15832 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15833 rather than an unsigned int. Handle SVE modes.
15834 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15836 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15838 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15839 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15840 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15841 (aarch64_sve_float_mul_immediate_p): New functions.
15842 (aarch64_sve_valid_immediate): New function.
15843 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15844 Explicitly reject structure modes. Check for INDEX constants.
15845 Handle PTRUE and PFALSE constants.
15846 (aarch64_check_zero_based_sve_index_immediate): New function.
15847 (aarch64_simd_imm_zero_p): Delete.
15848 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15849 vector modes. Accept constants in the range of CNT[BHWD].
15850 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15851 ask for an Advanced SIMD mode.
15852 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15853 (aarch64_simd_vector_alignment): Handle SVE predicates.
15854 (aarch64_vectorize_preferred_vector_alignment): New function.
15855 (aarch64_simd_vector_alignment_reachable): Use it instead of
15857 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15858 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15860 (MAX_VECT_LEN): Delete.
15861 (expand_vec_perm_d): Add a vec_flags field.
15862 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15863 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15864 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15866 (aarch64_evpc_rev): Rename to...
15867 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15868 (aarch64_evpc_rev_global): New function.
15869 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15870 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15872 (aarch64_evpc_sve_tbl): New function.
15873 (aarch64_expand_vec_perm_const_1): Update after rename of
15874 aarch64_evpc_rev. Handle SVE permutes too, trying
15875 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15876 than aarch64_evpc_tbl.
15877 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15878 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15879 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15880 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15881 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15882 (aarch64_expand_sve_vcond): New functions.
15883 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15884 of aarch64_vector_mode_p.
15885 (aarch64_dwarf_poly_indeterminate_value): New function.
15886 (aarch64_compute_pressure_classes): Likewise.
15887 (aarch64_can_change_mode_class): Likewise.
15888 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15889 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15890 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15891 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15892 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15893 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15894 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15895 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15897 (Dn, Dl, Dr): Accept const as well as const_vector.
15898 (Dz): Likewise. Compare against CONST0_RTX.
15899 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15900 of "vector" where appropriate.
15901 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15902 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15903 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15904 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15905 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15906 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15907 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15908 (v_int_equiv): Extend to SVE modes.
15909 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15911 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15912 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15913 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15914 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15915 (SVE_COND_FP_CMP): New int iterators.
15916 (perm_hilo): Handle the new unpack unspecs.
15917 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15919 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15920 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15921 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15922 (aarch64_equality_operator, aarch64_constant_vector_operand)
15923 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15924 (aarch64_sve_nonimmediate_operand): Likewise.
15925 (aarch64_sve_general_operand): Likewise.
15926 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15927 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15928 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15929 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15930 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15931 (aarch64_sve_float_arith_immediate): Likewise.
15932 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15933 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15934 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15935 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15936 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15937 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15938 (aarch64_sve_float_arith_operand): Likewise.
15939 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15940 (aarch64_sve_float_mul_operand): Likewise.
15941 (aarch64_sve_vec_perm_operand): Likewise.
15942 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15943 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15944 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15945 as well as const_vector.
15946 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15947 in file. Use CONST0_RTX and CONSTM1_RTX.
15948 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15949 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15950 Use aarch64_simd_imm_zero.
15951 * config/aarch64/aarch64-sve.md: New file.
15952 * config/aarch64/aarch64.md: Include it.
15953 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15954 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15955 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15956 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15957 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15958 (sve): New attribute.
15959 (enabled): Disable instructions with the sve attribute unless
15961 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15962 aarch64_expand_mov_immediate.
15963 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15964 CNT[BHSD] immediates.
15965 (movti): Split CONST_POLY_INT moves into two halves.
15966 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15967 Split additions that need a temporary here if the destination
15968 is the stack pointer.
15969 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15970 (*add<mode>3_poly_1): New instruction.
15971 (set_clobber_cc): New expander.
15973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15975 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15976 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15977 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15978 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15979 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15980 Change innermode from fixed_mode_size to machine_mode.
15981 (simplify_subreg): Update call accordingly. Handle a constant-sized
15982 subreg of a variable-length CONST_VECTOR.
15984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15985 Alan Hayward <alan.hayward@arm.com>
15986 David Sherwood <david.sherwood@arm.com>
15988 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15989 (add_offset_to_base): New function, split out from...
15990 (create_mem_ref): ...here. When handling a scale other than 1,
15991 check first whether the address is valid without the offset.
15992 Add it into the base if so, leaving the index and scale as-is.
15994 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15997 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15998 fold_for_warn before checking if arg2 is INTEGER_CST.
16000 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16002 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16003 (store_multiple_operation): Delete.
16004 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16005 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16006 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16007 guarded by TARGET_STRING.
16008 (rs6000_output_load_multiple): Delete.
16009 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16010 OPTION_MASK_STRING / TARGET_STRING handling.
16011 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16012 (const rs6000_opt_masks) <"string">: Change mask to 0.
16013 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16014 (MASK_STRING): Delete.
16015 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16017 (load_multiple): Delete.
16024 (store_multiple): Delete.
16031 (movmemsi_8reg): Delete.
16032 (corresponding unnamed define_insn): Delete.
16033 (movmemsi_6reg): Delete.
16034 (corresponding unnamed define_insn): Delete.
16035 (movmemsi_4reg): Delete.
16036 (corresponding unnamed define_insn): Delete.
16037 (movmemsi_2reg): Delete.
16038 (corresponding unnamed define_insn): Delete.
16039 (movmemsi_1reg): Delete.
16040 (corresponding unnamed define_insn): Delete.
16041 * config/rs6000/rs6000.opt (mno-string): New.
16042 (mstring): Replace by deprecation warning stub.
16043 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16045 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16047 * regrename.c (regrename_do_replace): If replacing the same
16048 reg multiple times, try to reuse last created gen_raw_REG.
16051 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16052 main to workaround a bug in GDB.
16054 2018-01-12 Tom de Vries <tom@codesourcery.com>
16057 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16059 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16061 PR rtl-optimization/80481
16062 * ira-color.c (get_cap_member): New function.
16063 (allocnos_conflict_by_live_ranges_p): Use it.
16064 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16065 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16067 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16070 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16071 (*saddl_se_1): Ditto.
16072 (*ssubsi_1): Ditto.
16073 (*ssubl_se_1): Ditto.
16075 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16077 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16078 rather than wi::to_widest for DR_INITs.
16079 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16080 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16081 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16083 (vect_analyze_group_access_1): Note that here.
16085 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16087 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16088 polynomial type sizes.
16090 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16092 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16093 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16094 (gimple_add_tmp_var): Likewise.
16096 2018-01-12 Martin Liska <mliska@suse.cz>
16098 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16099 (gimple_alloc_sizes): Likewise.
16100 (dump_gimple_statistics): Use PRIu64 in printf format.
16101 * gimple.h: Change uint64_t to int.
16103 2018-01-12 Martin Liska <mliska@suse.cz>
16105 * tree-core.h: Use uint64_t instead of int.
16106 * tree.c (tree_node_counts): Likewise.
16107 (tree_node_sizes): Likewise.
16108 (dump_tree_statistics): Use PRIu64 in printf format.
16110 2018-01-12 Martin Liska <mliska@suse.cz>
16112 * Makefile.in: As qsort_chk is implemented in vec.c, add
16113 vec.o to linkage of gencfn-macros.
16114 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16115 passing the info to record_node_allocation_statistics.
16116 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16118 * ggc-common.c (struct ggc_usage): Add operator== and use
16119 it in operator< and compare function.
16120 * mem-stats.h (struct mem_usage): Likewise.
16121 * vec.c (struct vec_usage): Remove operator< and compare
16122 function. Can be simply inherited.
16124 2018-01-12 Martin Jambor <mjambor@suse.cz>
16127 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16128 * tree-ssa-math-opts.c: Include domwalk.h.
16129 (convert_mult_to_fma_1): New function.
16130 (fma_transformation_info): New type.
16131 (fma_deferring_state): Likewise.
16132 (cancel_fma_deferring): New function.
16133 (result_of_phi): Likewise.
16134 (last_fma_candidate_feeds_initial_phi): Likewise.
16135 (convert_mult_to_fma): Added deferring logic, split actual
16136 transformation to convert_mult_to_fma_1.
16137 (math_opts_dom_walker): New type.
16138 (math_opts_dom_walker::after_dom_children): New method, body moved
16139 here from pass_optimize_widening_mul::execute, added deferring logic
16141 (pass_optimize_widening_mul::execute): Moved most of code to
16142 math_opts_dom_walker::after_dom_children.
16143 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16144 * config/i386/i386.c (ix86_option_override_internal): Added
16145 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16147 2018-01-12 Richard Biener <rguenther@suse.de>
16150 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16151 inline instance vars.
16153 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16156 * config/rx/rx.c (rx_is_restricted_memory_address):
16157 Handle SUBREG case.
16159 2018-01-12 Richard Biener <rguenther@suse.de>
16161 PR tree-optimization/80846
16162 * target.def (split_reduction): New target hook.
16163 * targhooks.c (default_split_reduction): New function.
16164 * targhooks.h (default_split_reduction): Declare.
16165 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16166 target requests first reduce vectors by combining low and high
16168 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16169 (get_vectype_for_scalar_type_and_size): Export.
16170 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16171 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16172 * doc/tm.texi: Regenerate.
16173 * config/i386/i386.c (ix86_split_reduction): Implement
16174 TARGET_VECTORIZE_SPLIT_REDUCTION.
16176 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16179 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16180 in PIC mode except for TARGET_VXWORKS_RTP.
16181 * config/sparc/sparc.c: Include cfgrtl.h.
16182 (TARGET_INIT_PIC_REG): Define.
16183 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16184 (sparc_pic_register_p): New predicate.
16185 (sparc_legitimate_address_p): Use it.
16186 (sparc_legitimize_pic_address): Likewise.
16187 (sparc_delegitimize_address): Likewise.
16188 (sparc_mode_dependent_address_p): Likewise.
16189 (gen_load_pcrel_sym): Remove 4th parameter.
16190 (load_got_register): Adjust call to above. Remove obsolete stuff.
16191 (sparc_expand_prologue): Do not call load_got_register here.
16192 (sparc_flat_expand_prologue): Likewise.
16193 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16194 (sparc_use_pseudo_pic_reg): New function.
16195 (sparc_init_pic_reg): Likewise.
16196 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16197 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16199 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16201 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16202 Add item for branch_cost.
16204 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16206 PR rtl-optimization/83565
16207 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16208 not extend the result to a larger mode for rotate operations.
16209 (num_sign_bit_copies1): Likewise.
16211 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16214 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16216 Use values-Xc.o for -pedantic.
16217 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16219 2018-01-12 Martin Liska <mliska@suse.cz>
16222 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16224 (possible_polymorphic_call_targets): Use it.
16225 (ipa_devirt): Likewise.
16227 2018-01-12 Martin Liska <mliska@suse.cz>
16229 * profile-count.h (enum profile_quality): Use 0 as invalid
16230 enum value of profile_quality.
16232 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16234 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16235 -mext-string options.
16237 2018-01-12 Richard Biener <rguenther@suse.de>
16239 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16240 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16241 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16243 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16245 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16247 * configure.ac (--with-long-double-format): Add support for the
16248 configuration option to change the default long double format on
16250 * config.gcc (powerpc*-linux*-*): Likewise.
16251 * configure: Regenerate.
16252 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16253 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16254 used without modification.
16256 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16258 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16259 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16260 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16261 MISC_BUILTIN_SPEC_BARRIER.
16262 (rs6000_init_builtins): Likewise.
16263 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16265 (speculation_barrier): New define_insn.
16266 * doc/extend.texi: Document __builtin_speculation_barrier.
16268 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16271 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16272 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16273 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16275 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16276 integral modes instead of "ss" and "sd".
16277 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16278 vectors with 32-bit and 64-bit elements.
16279 (vecdupssescalarmodesuffix): New mode attribute.
16280 (vec_dup<mode>): Use it.
16282 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16285 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16286 frame if argument is passed on stack.
16288 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16291 * ree.c (combine_reaching_defs): Optimize also
16292 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16293 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16295 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16297 PR middle-end/83189
16298 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16300 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16302 PR middle-end/83718
16303 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16304 after they are computed.
16306 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16308 PR tree-optimization/83695
16309 * gimple-loop-linterchange.cc
16310 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16311 reset cached scev information after interchange.
16312 (pass_linterchange::execute): Remove call to scev_reset_htab.
16314 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16316 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16317 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16318 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16319 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16320 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16321 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16322 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16323 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16324 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16325 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16326 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16327 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16328 (V_lane_reg): Likewise.
16329 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16331 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16332 (vfmal_lane_low<mode>_intrinsic,
16333 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16334 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16335 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16336 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16337 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16338 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16340 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16342 * config/arm/arm-cpus.in (fp16fml): New feature.
16343 (ALL_SIMD): Add fp16fml.
16344 (armv8.2-a): Add fp16fml as an option.
16345 (armv8.3-a): Likewise.
16346 (armv8.4-a): Add fp16fml as part of fp16.
16347 * config/arm/arm.h (TARGET_FP16FML): Define.
16348 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16350 * config/arm/arm-modes.def (V2HF): Define.
16351 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16352 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16353 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16354 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16355 vfmsl_low, vfmsl_high): New set of builtins.
16356 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16357 (vfml_op): New code attribute.
16358 (VFMLHALVES): New int iterator.
16359 (VFML, VFMLSEL): New mode attributes.
16360 (V_reg): Define mapping for V2HF.
16361 (V_hi, V_lo): New mode attributes.
16362 (VF_constraint): Likewise.
16363 (vfml_half, vfml_half_selector): New int attributes.
16364 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16366 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16367 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16369 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16370 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16371 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16372 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16374 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16375 Document new effective target and option set.
16377 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16379 * config/arm/arm-cpus.in (armv8_4): New feature.
16380 (ARMv8_4a): New fgroup.
16381 (armv8.4-a): New arch.
16382 * config/arm/arm-tables.opt: Regenerate.
16383 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16384 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16385 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16386 Add matching rules for -march=armv8.4-a and extensions.
16387 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16389 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16392 * config/rx/rx.md (BW): New mode attribute.
16393 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16395 2018-01-11 Richard Biener <rguenther@suse.de>
16397 PR tree-optimization/83435
16398 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16399 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16400 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16402 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16403 Alan Hayward <alan.hayward@arm.com>
16404 David Sherwood <david.sherwood@arm.com>
16406 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16408 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16409 (aarch64_print_address_internal): Use it to check for a zero offset.
16411 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16412 Alan Hayward <alan.hayward@arm.com>
16413 David Sherwood <david.sherwood@arm.com>
16415 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16416 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16417 Return a poly_int64 rather than a HOST_WIDE_INT.
16418 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16419 rather than a HOST_WIDE_INT.
16420 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16421 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16422 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16423 final_offset from HOST_WIDE_INT to poly_int64.
16424 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16425 to_constant when getting the number of units in an Advanced SIMD
16427 (aarch64_builtin_vectorized_function): Check for a constant number
16429 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16431 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16432 attribute instead of GET_MODE_NUNITS.
16433 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16434 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16435 GET_MODE_SIZE for fixed-size registers.
16436 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16437 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16438 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16439 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16440 (aarch64_print_operand, aarch64_print_address_internal)
16441 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16442 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16443 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16444 Handle polynomial GET_MODE_SIZE.
16445 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16446 wider than SImode without modification.
16447 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16448 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16449 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16450 passing and returning SVE modes.
16451 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16452 rather than GEN_INT.
16453 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16454 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16455 (aarch64_allocate_and_probe_stack_space): Likewise.
16456 (aarch64_layout_frame): Cope with polynomial offsets.
16457 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16458 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16459 polynomial offsets.
16460 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16461 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16462 poly_int64 rather than a HOST_WIDE_INT.
16463 (aarch64_get_separate_components, aarch64_process_components)
16464 (aarch64_expand_prologue, aarch64_expand_epilogue)
16465 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16466 (aarch64_anchor_offset): New function, split out from...
16467 (aarch64_legitimize_address): ...here.
16468 (aarch64_builtin_vectorization_cost): Handle polynomial
16469 TYPE_VECTOR_SUBPARTS.
16470 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16472 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16473 number of elements from the PARALLEL rather than the mode.
16474 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16475 rather than GET_MODE_BITSIZE.
16476 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16477 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16478 (aarch64_expand_vec_perm_const_1): Handle polynomial
16479 d->perm.length () and d->perm elements.
16480 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16481 Apply to_constant to d->perm elements.
16482 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16483 polynomial CONST_VECTOR_NUNITS.
16484 (aarch64_move_pointer): Take amount as a poly_int64 rather
16486 (aarch64_progress_pointer): Avoid temporary variable.
16487 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16488 the mode attribute instead of GET_MODE.
16490 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16491 Alan Hayward <alan.hayward@arm.com>
16492 David Sherwood <david.sherwood@arm.com>
16494 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16495 x exists before using it.
16496 (aarch64_add_constant_internal): Rename to...
16497 (aarch64_add_offset_1): ...this. Replace regnum with separate
16498 src and dest rtxes. Handle the case in which they're different,
16499 including when the offset is zero. Replace scratchreg with an rtx.
16500 Use 2 additions if there is no spare register into which we can
16501 move a 16-bit constant.
16502 (aarch64_add_constant): Delete.
16503 (aarch64_add_offset): Replace reg with separate src and dest
16504 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16505 Use aarch64_add_offset_1.
16506 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16507 an rtx rather than an int. Take the delta as a poly_int64
16508 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16509 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16510 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16511 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16512 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16513 and aarch64_add_sp.
16514 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16515 aarch64_add_constant.
16517 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16519 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16520 Use scalar_float_mode.
16522 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16524 * config/aarch64/aarch64-simd.md
16525 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16526 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16527 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16528 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16529 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16530 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16531 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16532 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16533 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16534 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16536 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16539 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16540 targ_options->x_arm_arch_string is non NULL.
16542 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16544 * config/aarch64/aarch64.h
16545 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16547 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16550 * expmed.c (emit_store_flag_force): Swap if const op0
16551 and change VOIDmode to mode of op0.
16553 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16555 PR rtl-optimization/83761
16556 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16557 than bytes to mode_for_size.
16559 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16561 PR middle-end/83189
16562 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16563 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16566 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16568 PR middle-end/83575
16569 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16570 when in layout mode.
16571 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16572 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16575 2018-01-10 Michael Collison <michael.collison@arm.com>
16577 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16578 * config/aarch64/aarch64-option-extension.def: Add
16579 AARCH64_OPT_EXTENSION of 'fp16fml'.
16580 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16581 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16582 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16583 * config/aarch64/constraints.md (Ui7): New constraint.
16584 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16585 (VFMLA_SEL_W): Ditto.
16588 (VFMLA16_LOW): New int iterator.
16589 (VFMLA16_HIGH): Ditto.
16590 (UNSPEC_FMLAL): New unspec.
16591 (UNSPEC_FMLSL): Ditto.
16592 (UNSPEC_FMLAL2): Ditto.
16593 (UNSPEC_FMLSL2): Ditto.
16594 (f16mac): New code attribute.
16595 * config/aarch64/aarch64-simd-builtins.def
16596 (aarch64_fmlal_lowv2sf): Ditto.
16597 (aarch64_fmlsl_lowv2sf): Ditto.
16598 (aarch64_fmlalq_lowv4sf): Ditto.
16599 (aarch64_fmlslq_lowv4sf): Ditto.
16600 (aarch64_fmlal_highv2sf): Ditto.
16601 (aarch64_fmlsl_highv2sf): Ditto.
16602 (aarch64_fmlalq_highv4sf): Ditto.
16603 (aarch64_fmlslq_highv4sf): Ditto.
16604 (aarch64_fmlal_lane_lowv2sf): Ditto.
16605 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16606 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16607 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16608 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16609 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16610 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16611 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16612 (aarch64_fmlal_lane_highv2sf): Ditto.
16613 (aarch64_fmlsl_lane_highv2sf): Ditto.
16614 (aarch64_fmlal_laneq_highv2sf): Ditto.
16615 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16616 (aarch64_fmlalq_lane_highv4sf): Ditto.
16617 (aarch64_fmlsl_lane_highv4sf): Ditto.
16618 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16619 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16620 * config/aarch64/aarch64-simd.md:
16621 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16622 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16623 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16624 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16625 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16626 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16627 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16628 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16629 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16630 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16631 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16632 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16633 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16634 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16635 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16636 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16637 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16638 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16639 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16640 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16641 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16642 (vfmlsl_low_u32): Ditto.
16643 (vfmlalq_low_u32): Ditto.
16644 (vfmlslq_low_u32): Ditto.
16645 (vfmlal_high_u32): Ditto.
16646 (vfmlsl_high_u32): Ditto.
16647 (vfmlalq_high_u32): Ditto.
16648 (vfmlslq_high_u32): Ditto.
16649 (vfmlal_lane_low_u32): Ditto.
16650 (vfmlsl_lane_low_u32): Ditto.
16651 (vfmlal_laneq_low_u32): Ditto.
16652 (vfmlsl_laneq_low_u32): Ditto.
16653 (vfmlalq_lane_low_u32): Ditto.
16654 (vfmlslq_lane_low_u32): Ditto.
16655 (vfmlalq_laneq_low_u32): Ditto.
16656 (vfmlslq_laneq_low_u32): Ditto.
16657 (vfmlal_lane_high_u32): Ditto.
16658 (vfmlsl_lane_high_u32): Ditto.
16659 (vfmlal_laneq_high_u32): Ditto.
16660 (vfmlsl_laneq_high_u32): Ditto.
16661 (vfmlalq_lane_high_u32): Ditto.
16662 (vfmlslq_lane_high_u32): Ditto.
16663 (vfmlalq_laneq_high_u32): Ditto.
16664 (vfmlslq_laneq_high_u32): Ditto.
16665 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16666 (AARCH64_FL_FOR_ARCH8_4): New.
16667 (AARCH64_ISA_F16FML): New ISA flag.
16668 (TARGET_F16FML): New feature flag for fp16fml.
16669 (doc/invoke.texi): Document new fp16fml option.
16671 2018-01-10 Michael Collison <michael.collison@arm.com>
16673 * config/aarch64/aarch64-builtins.c:
16674 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16675 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16676 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16677 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16678 (AARCH64_ISA_SHA3): New ISA flag.
16679 (TARGET_SHA3): New feature flag for sha3.
16680 * config/aarch64/iterators.md (sha512_op): New int attribute.
16681 (CRYPTO_SHA512): New int iterator.
16682 (UNSPEC_SHA512H): New unspec.
16683 (UNSPEC_SHA512H2): Ditto.
16684 (UNSPEC_SHA512SU0): Ditto.
16685 (UNSPEC_SHA512SU1): Ditto.
16686 * config/aarch64/aarch64-simd-builtins.def
16687 (aarch64_crypto_sha512hqv2di): New builtin.
16688 (aarch64_crypto_sha512h2qv2di): Ditto.
16689 (aarch64_crypto_sha512su0qv2di): Ditto.
16690 (aarch64_crypto_sha512su1qv2di): Ditto.
16691 (aarch64_eor3qv8hi): Ditto.
16692 (aarch64_rax1qv2di): Ditto.
16693 (aarch64_xarqv2di): Ditto.
16694 (aarch64_bcaxqv8hi): Ditto.
16695 * config/aarch64/aarch64-simd.md:
16696 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16697 (aarch64_crypto_sha512su0qv2di): Ditto.
16698 (aarch64_crypto_sha512su1qv2di): Ditto.
16699 (aarch64_eor3qv8hi): Ditto.
16700 (aarch64_rax1qv2di): Ditto.
16701 (aarch64_xarqv2di): Ditto.
16702 (aarch64_bcaxqv8hi): Ditto.
16703 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16704 (vsha512h2q_u64): Ditto.
16705 (vsha512su0q_u64): Ditto.
16706 (vsha512su1q_u64): Ditto.
16707 (veor3q_u16): Ditto.
16708 (vrax1q_u64): Ditto.
16709 (vxarq_u64): Ditto.
16710 (vbcaxq_u16): Ditto.
16711 * config/arm/types.md (crypto_sha512): New type attribute.
16712 (crypto_sha3): Ditto.
16713 (doc/invoke.texi): Document new sha3 option.
16715 2018-01-10 Michael Collison <michael.collison@arm.com>
16717 * config/aarch64/aarch64-builtins.c:
16718 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16719 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16720 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16721 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16722 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16723 (AARCH64_ISA_SM4): New ISA flag.
16724 (TARGET_SM4): New feature flag for sm4.
16725 * config/aarch64/aarch64-simd-builtins.def
16726 (aarch64_sm3ss1qv4si): Ditto.
16727 (aarch64_sm3tt1aq4si): Ditto.
16728 (aarch64_sm3tt1bq4si): Ditto.
16729 (aarch64_sm3tt2aq4si): Ditto.
16730 (aarch64_sm3tt2bq4si): Ditto.
16731 (aarch64_sm3partw1qv4si): Ditto.
16732 (aarch64_sm3partw2qv4si): Ditto.
16733 (aarch64_sm4eqv4si): Ditto.
16734 (aarch64_sm4ekeyqv4si): Ditto.
16735 * config/aarch64/aarch64-simd.md:
16736 (aarch64_sm3ss1qv4si): Ditto.
16737 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16738 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16739 (aarch64_sm4eqv4si): Ditto.
16740 (aarch64_sm4ekeyqv4si): Ditto.
16741 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16742 (sm3part_op): Ditto.
16743 (CRYPTO_SM3TT): Ditto.
16744 (CRYPTO_SM3PART): Ditto.
16745 (UNSPEC_SM3SS1): New unspec.
16746 (UNSPEC_SM3TT1A): Ditto.
16747 (UNSPEC_SM3TT1B): Ditto.
16748 (UNSPEC_SM3TT2A): Ditto.
16749 (UNSPEC_SM3TT2B): Ditto.
16750 (UNSPEC_SM3PARTW1): Ditto.
16751 (UNSPEC_SM3PARTW2): Ditto.
16752 (UNSPEC_SM4E): Ditto.
16753 (UNSPEC_SM4EKEY): Ditto.
16754 * config/aarch64/constraints.md (Ui2): New constraint.
16755 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16756 * config/arm/types.md (crypto_sm3): New type attribute.
16757 (crypto_sm4): Ditto.
16758 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16759 (vsm3tt1aq_u32): Ditto.
16760 (vsm3tt1bq_u32): Ditto.
16761 (vsm3tt2aq_u32): Ditto.
16762 (vsm3tt2bq_u32): Ditto.
16763 (vsm3partw1q_u32): Ditto.
16764 (vsm3partw2q_u32): Ditto.
16765 (vsm4eq_u32): Ditto.
16766 (vsm4ekeyq_u32): Ditto.
16767 (doc/invoke.texi): Document new sm4 option.
16769 2018-01-10 Michael Collison <michael.collison@arm.com>
16771 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16772 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16773 (AARCH64_FL_FOR_ARCH8_4): New.
16774 (AARCH64_FL_V8_4): New flag.
16775 (doc/invoke.texi): Document new armv8.4-a option.
16777 2018-01-10 Michael Collison <michael.collison@arm.com>
16779 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16780 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16781 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16782 * config/aarch64/aarch64-option-extension.def: Add
16783 AARCH64_OPT_EXTENSION of 'sha2'.
16784 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16785 (crypto): Disable sha2 and aes if crypto disabled.
16786 (crypto): Enable aes and sha2 if enabled.
16787 (simd): Disable sha2 and aes if simd disabled.
16788 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16790 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16791 (TARGET_SHA2): New feature flag for sha2.
16792 (TARGET_AES): New feature flag for aes.
16793 * config/aarch64/aarch64-simd.md:
16794 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16795 conditional on TARGET_AES.
16796 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16797 (aarch64_crypto_sha1hsi): Make pattern conditional
16799 (aarch64_crypto_sha1hv4si): Ditto.
16800 (aarch64_be_crypto_sha1hv4si): Ditto.
16801 (aarch64_crypto_sha1su1v4si): Ditto.
16802 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16803 (aarch64_crypto_sha1su0v4si): Ditto.
16804 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16805 (aarch64_crypto_sha256su0v4si): Ditto.
16806 (aarch64_crypto_sha256su1v4si): Ditto.
16807 (doc/invoke.texi): Document new aes and sha2 options.
16809 2018-01-10 Martin Sebor <msebor@redhat.com>
16811 PR tree-optimization/83781
16812 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16815 2018-01-11 Martin Sebor <msebor@gmail.com>
16816 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16818 PR tree-optimization/83501
16819 PR tree-optimization/81703
16821 * tree-ssa-strlen.c (get_string_cst): Rename...
16822 (get_string_len): ...to this. Handle global constants.
16823 (handle_char_store): Adjust.
16825 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16826 Jim Wilson <jimw@sifive.com>
16828 * config/riscv/riscv-protos.h (riscv_output_return): New.
16829 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16830 (riscv_attribute_table, riscv_output_return),
16831 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16832 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16833 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16835 (riscv_expand_prologue): Add early return for naked function.
16836 (riscv_expand_epilogue): Likewise.
16837 (riscv_function_ok_for_sibcall): Return false for naked function.
16838 (riscv_set_current_function): New.
16839 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16840 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16841 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16842 * doc/extend.texi (RISC-V Function Attributes): New.
16844 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16846 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16847 check for 128-bit long double before checking TCmode.
16848 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16849 128-bit long doubles before checking TFmode or TCmode.
16850 (FLOAT128_IBM_P): Likewise.
16852 2018-01-10 Martin Sebor <msebor@redhat.com>
16854 PR tree-optimization/83671
16855 * builtins.c (c_strlen): Unconditionally return zero for the empty
16857 Use -Warray-bounds for warnings.
16858 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16859 for non-constant array indices with COMPONENT_REF, arrays of
16860 arrays, and pointers to arrays.
16861 (gimple_fold_builtin_strlen): Determine and set length range for
16862 non-constant character arrays.
16864 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16866 PR middle-end/81897
16867 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16870 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16872 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16874 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16877 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16878 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16879 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16880 indexed_or_indirect_operand predicate.
16881 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16882 (*vsx_le_perm_load_v8hi): Likewise.
16883 (*vsx_le_perm_load_v16qi): Likewise.
16884 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16885 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16886 (*vsx_le_perm_store_v8hi): Likewise.
16887 (*vsx_le_perm_store_v16qi): Likewise.
16888 (eight unnamed splitters): Likewise.
16890 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16892 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16893 * config/rs6000/emmintrin.h: Likewise.
16894 * config/rs6000/mmintrin.h: Likewise.
16895 * config/rs6000/xmmintrin.h: Likewise.
16897 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16900 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16902 * tree.c (tree_nop_conversion): Return true for location wrapper
16904 (maybe_wrap_with_location): New function.
16905 (selftest::check_strip_nops): New function.
16906 (selftest::test_location_wrappers): New function.
16907 (selftest::tree_c_tests): Call it.
16908 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16909 (maybe_wrap_with_location): New decl.
16910 (EXPR_LOCATION_WRAPPER_P): New macro.
16911 (location_wrapper_p): New inline function.
16912 (tree_strip_any_location_wrapper): New inline function.
16914 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16917 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16918 stack_realign_offset for the largest alignment of stack slot
16920 (ix86_find_max_used_stack_alignment): New function.
16921 (ix86_finalize_stack_frame_flags): Use it. Set
16922 max_used_stack_alignment if we don't realign stack.
16923 * config/i386/i386.h (machine_function): Add
16924 max_used_stack_alignment.
16926 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16928 * config/arm/arm.opt (-mbranch-cost): New option.
16929 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16932 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16935 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16936 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16938 2018-01-10 Richard Biener <rguenther@suse.de>
16941 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16942 early out so it also covers the case where we have a non-NULL
16945 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16947 PR tree-optimization/83753
16948 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16949 for non-strided grouped accesses if the number of elements is 1.
16951 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16954 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16955 * i386.h (TARGET_USE_GATHER): Define.
16956 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16958 2018-01-10 Martin Liska <mliska@suse.cz>
16961 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16962 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16964 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16965 CLEANUP_NO_PARTITIONING is not set.
16967 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16969 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16970 for vectors, as a partial revert of r254296.
16971 * rtl.h (const_vec_p): Delete.
16972 (const_vec_duplicate_p): Don't test for vector CONSTs.
16973 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16974 * expmed.c (make_tree): Likewise.
16977 * common.md (E, F): Use CONSTANT_P instead of checking for
16979 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16980 checking for CONST_VECTOR.
16982 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16984 PR middle-end/83575
16985 * predict.c (force_edge_cold): Handle in more sane way edges
16986 with no prediction.
16988 2018-01-09 Carl Love <cel@us.ibm.com>
16990 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16992 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16993 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16994 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16995 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16996 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16997 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16998 * config/rs6000/rs6000-protos.h: Add extern defition for
16999 rs6000_generate_float2_double_code.
17000 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17002 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17003 (float2_v2df): Add define_expand.
17005 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17008 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17009 op_mode in the force_to_mode call.
17011 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17013 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17014 instead of checking each element individually.
17015 (aarch64_evpc_uzp): Likewise.
17016 (aarch64_evpc_zip): Likewise.
17017 (aarch64_evpc_ext): Likewise.
17018 (aarch64_evpc_rev): Likewise.
17019 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17020 instead of checking each element individually. Return true without
17022 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17023 whether all selected elements come from the same input, instead of
17024 checking each element individually. Remove calls to gen_rtx_REG,
17025 start_sequence and end_sequence and instead assert that no rtl is
17028 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17030 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17031 order of HIGH and CONST checks.
17033 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17035 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17036 if the destination isn't an SSA_NAME.
17038 2018-01-09 Richard Biener <rguenther@suse.de>
17040 PR tree-optimization/83668
17041 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17043 (canonicalize_loop_form): ... here, renamed from ...
17044 (canonicalize_loop_closed_ssa_form): ... this and amended to
17045 swap successor edges for loop exit blocks to make us use
17046 the RPO order we need for initial schedule generation.
17048 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17050 PR tree-optimization/64811
17051 * match.pd: When optimizing comparisons with Inf, avoid
17052 introducing or losing exceptions from comparisons with NaN.
17054 2018-01-09 Martin Liska <mliska@suse.cz>
17057 * asan.c (shadow_mem_size): Add gcc_assert.
17059 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17061 Don't save registers in main().
17064 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17065 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17066 * config/avr/avr.c (avr_set_current_function): Don't error if
17067 naked, OS_task or OS_main are specified at the same time.
17068 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17070 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17072 * common/config/avr/avr-common.c (avr_option_optimization_table):
17073 Switch on -mmain-is-OS_task for optimizing compilations.
17075 2018-01-09 Richard Biener <rguenther@suse.de>
17077 PR tree-optimization/83572
17078 * graphite.c: Include cfganal.h.
17079 (graphite_transform_loops): Connect infinite loops to exit
17080 and remove fake edges at the end.
17082 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17084 * ipa-inline.c (edge_badness): Revert accidental checkin.
17086 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17089 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17090 symbols; not inline clones.
17092 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17095 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17096 hard registers. Formatting fixes.
17098 PR preprocessor/83722
17099 * gcc.c (try_generate_repro): Pass
17100 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17101 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17104 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17105 Kito Cheng <kito.cheng@gmail.com>
17107 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17108 (riscv_leaf_function_p): Delete.
17109 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17111 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17113 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17115 (do_ifelse): New function.
17116 (do_isel): New function.
17117 (do_sub3): New function.
17118 (do_add3): New function.
17119 (do_load_mask_compare): New function.
17120 (do_overlap_load_compare): New function.
17121 (expand_compare_loop): New function.
17122 (expand_block_compare): Call expand_compare_loop() when appropriate.
17123 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17124 option description.
17125 (-mblock-compare-inline-loop-limit): New option.
17127 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17130 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17131 Reverse order of second and third operands in first alternative.
17132 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17133 of first and second elements in UNSPEC_VPERMR vector.
17134 (altivec_expand_vec_perm_le): Likewise.
17136 2018-01-08 Jeff Law <law@redhat.com>
17138 PR rtl-optimizatin/81308
17139 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17140 (process_switch): If group_case_labels makes a change, then set
17142 (pass_convert_switch::execute): If a switch is converted, then
17143 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17145 PR rtl-optimization/81308
17146 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17149 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17151 PR target/83663 - Revert r255946
17152 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17153 generation for cases where splatting a value is not useful.
17154 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17155 across a vec_duplicate and a paradoxical subreg forming a vector
17156 mode to a vec_concat.
17158 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17160 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17161 -march=armv8.3-a variants.
17162 * config/arm/t-multilib: Likewise.
17163 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17165 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17167 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17169 (cceq_ior_compare_complement): Give it a name so I can use it, and
17170 change boolean_or_operator predicate to boolean_operator so it can
17171 be used to generate a crand.
17172 (eqne): New code iterator.
17173 (bd/bd_neg): New code_attrs.
17174 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17175 a single define_insn.
17176 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17177 decrement (bdnzt/bdnzf/bdzt/bdzf).
17178 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17179 with the new names of the branch decrement patterns, and added the
17180 names of the branch decrement conditional patterns.
17182 2018-01-08 Richard Biener <rguenther@suse.de>
17184 PR tree-optimization/83563
17185 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17188 2018-01-08 Richard Biener <rguenther@suse.de>
17190 PR middle-end/83713
17191 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17193 2018-01-08 Richard Biener <rguenther@suse.de>
17195 PR tree-optimization/83685
17196 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17197 references to abnormals.
17199 2018-01-08 Richard Biener <rguenther@suse.de>
17202 * dwarf2out.c (output_indirect_strings): Handle empty
17203 skeleton_debug_str_hash.
17204 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17206 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17208 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17209 (emit_store_direct): Likewise.
17210 (arc_trampoline_adjust_address): Likewise.
17211 (arc_asm_trampoline_template): New function.
17212 (arc_initialize_trampoline): Use asm_trampoline_template.
17213 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17214 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17215 * config/arc/arc.md (flush_icache): Delete pattern.
17217 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17219 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17220 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17223 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17226 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17227 by not USED_FOR_TARGET.
17228 (make_pass_resolve_sw_modes): Likewise.
17230 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17232 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17235 2018-01-08 Richard Biener <rguenther@suse.de>
17237 PR middle-end/83580
17238 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17240 2018-01-08 Richard Biener <rguenther@suse.de>
17242 PR middle-end/83517
17243 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17245 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17247 PR middle-end/81897
17248 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17249 basic blocks with a small number of successors.
17250 (convert_control_dep_chain_into_preds): Improve handling of
17252 (dump_predicates): Split apart into...
17253 (dump_pred_chain): ...here...
17254 (dump_pred_info): ...and here.
17255 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17256 (can_chain_union_be_invalidated_p): Improve check for invalidation
17258 (uninit_uses_cannot_happen): Avoid unnecessary if
17259 convert_control_dep_chain_into_preds yielded nothing.
17261 2018-01-06 Martin Sebor <msebor@redhat.com>
17263 PR tree-optimization/83640
17264 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17265 subtracting negative offset from size.
17266 (builtin_access::overlap): Adjust offset bounds of the access to fall
17267 within the size of the object if possible.
17269 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17271 PR rtl-optimization/83699
17272 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17273 extract_bit_field_as_subreg to cases in which the extracted
17274 value is also a vector.
17276 * lra-constraints.c (process_alt_operands): Test for the equivalence
17277 substitutions when detecting a possible reload cycle.
17279 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17282 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17283 by default if flag_selective_schedling{,2}. Formatting fixes.
17285 PR rtl-optimization/83682
17286 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17287 if it has non-VECTOR_MODE element mode.
17288 (vec_duplicate_p): Likewise.
17290 PR middle-end/83694
17291 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17292 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17294 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17297 * config/i386/i386-builtin.def
17298 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17299 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17300 Require also OPTION_MASK_ISA_AVX512F in addition to
17301 OPTION_MASK_ISA_GFNI.
17302 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17303 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17304 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17305 to OPTION_MASK_ISA_GFNI.
17306 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17307 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17308 OPTION_MASK_ISA_AVX512BW.
17309 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17310 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17311 addition to OPTION_MASK_ISA_GFNI.
17312 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17313 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17314 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17315 to OPTION_MASK_ISA_GFNI.
17316 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17317 a requirement for all ISAs rather than any of them with a few
17319 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17321 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17322 bitmasks to be enabled with 3 exceptions, instead of requiring any
17323 enabled ISA with lots of exceptions.
17324 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17325 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17326 Change avx512bw in isa attribute to avx512f.
17327 * config/i386/sgxintrin.h: Add license boilerplate.
17328 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17329 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17330 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17331 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17333 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17334 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17335 temporarily sse2 rather than sse if not enabled already.
17338 * config/i386/sse.md (VI248_VLBW): Rename to ...
17339 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17340 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17341 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17342 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17343 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17344 mode iterator instead of VI248_VLBW.
17346 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17348 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17349 (record_modified): Skip clobbers; add debug output.
17350 (param_change_prob): Use sreal frequencies.
17352 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17354 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17355 punt for user-aligned variables.
17357 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17359 * tree-chrec.c (chrec_contains_symbols): Return true for
17362 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17365 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17366 of (x|y) == x for BICS pattern.
17368 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17370 PR tree-optimization/83605
17371 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17372 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17375 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17377 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17378 * config/epiphany/rtems.h: New file.
17380 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17381 Uros Bizjak <ubizjak@gmail.com>
17384 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17385 QIreg_operand instead of register_operand predicate.
17386 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17387 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17388 comments instead of -fmitigate[-_]rop.
17390 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17393 * cgraphunit.c (symbol_table::compile): Switch to text_section
17394 before calling assembly_start debug hook.
17395 * run-rtl-passes.c (run_rtl_passes): Likewise.
17398 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17400 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17401 range_int_cst_p rather than !symbolic_range_p before calling
17402 extract_range_from_multiplicative_op_1.
17404 2018-01-04 Jeff Law <law@redhat.com>
17406 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17407 redundant test in assertion.
17409 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17411 * doc/rtl.texi: Document machine_mode wrapper classes.
17413 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17415 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17416 using tree_to_uhwi.
17418 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17420 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17421 the VEC_PERM_EXPR fold to fail.
17423 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17426 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17427 to switched_sections.
17429 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17432 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17433 test for d.testing.
17435 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17438 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17439 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17441 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17444 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17445 is BLKmode and bitpos not zero or mode change is needed.
17447 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17450 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17453 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17456 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17457 instead of MULT rtx. Update all corresponding splitters.
17458 (*saddl_se): Ditto.
17459 (*ssub<modesuffix>): Ditto.
17460 (*ssubl_se): Ditto.
17461 (*cmp_sadd_di): Update split patterns.
17462 (*cmp_sadd_si): Ditto.
17463 (*cmp_sadd_sidi): Ditto.
17464 (*cmp_ssub_di): Ditto.
17465 (*cmp_ssub_si): Ditto.
17466 (*cmp_ssub_sidi): Ditto.
17467 * config/alpha/predicates.md (const23_operand): New predicate.
17468 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17469 Look for ASHIFT, not MULT inner operand.
17470 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17472 2018-01-04 Martin Liska <mliska@suse.cz>
17474 PR gcov-profile/83669
17475 * gcov.c (output_intermediate_file): Add version to intermediate
17477 * doc/gcov.texi: Document new field 'version' in intermediate
17478 file format. Fix location of '-k' option of gcov command.
17480 2018-01-04 Martin Liska <mliska@suse.cz>
17483 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17485 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17487 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17489 2018-01-03 Martin Sebor <msebor@redhat.com>
17491 PR tree-optimization/83655
17492 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17493 checking calls with invalid arguments.
17495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17497 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17498 (vectorizable_mask_load_store): Delete.
17499 (vectorizable_call): Return false for masked loads and stores.
17500 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17501 instead of gimple_assign_rhs1.
17502 (vectorizable_load): Handle IFN_MASK_LOAD.
17503 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17507 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17509 (vectorizable_mask_load_store): ...here.
17510 (vectorizable_load): ...and here.
17512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17514 * tree-vect-stmts.c (vect_build_all_ones_mask)
17515 (vect_build_zero_merge_argument): New functions, split out from...
17516 (vectorizable_load): ...here.
17518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17520 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17522 (vectorizable_mask_load_store): ...here.
17523 (vectorizable_store): ...and here.
17525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17527 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17529 (vectorizable_mask_load_store): ...here.
17531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17533 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17534 (vect_model_store_cost): Take a vec_load_store_type instead of a
17536 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17537 (vect_model_store_cost): Take a vec_load_store_type instead of a
17539 (vectorizable_mask_load_store): Update accordingly.
17540 (vectorizable_store): Likewise.
17541 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17545 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17546 IFN_MASK_LOAD calls here rather than...
17547 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17550 Alan Hayward <alan.hayward@arm.com>
17551 David Sherwood <david.sherwood@arm.com>
17553 * expmed.c (extract_bit_field_1): For vector extracts,
17554 fall back to extract_bit_field_as_subreg if vec_extract
17557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17558 Alan Hayward <alan.hayward@arm.com>
17559 David Sherwood <david.sherwood@arm.com>
17561 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17562 they are variable or constant sized.
17563 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17564 slots for constant-sized data.
17566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17567 Alan Hayward <alan.hayward@arm.com>
17568 David Sherwood <david.sherwood@arm.com>
17570 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17571 handling COND_EXPRs with boolean comparisons, try to find a better
17572 basis for the mask type than the boolean itself.
17574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17576 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17577 is calculated and how it can be overridden.
17578 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17579 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17581 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17585 Alan Hayward <alan.hayward@arm.com>
17586 David Sherwood <david.sherwood@arm.com>
17588 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17589 Remove the mode argument.
17590 (aarch64_simd_valid_immediate): Remove the mode and inverse
17592 * config/aarch64/iterators.md (bitsize): New iterator.
17593 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17594 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17595 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17596 aarch64_simd_valid_immediate.
17597 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17598 (aarch64_reg_or_bic_imm): Likewise.
17599 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17600 with an insn_type enum and msl with a modifier_type enum.
17601 Replace element_width with a scalar_mode. Change the shift
17602 to unsigned int. Add constructors for scalar_float_mode and
17603 scalar_int_mode elements.
17604 (aarch64_vect_float_const_representable_p): Delete.
17605 (aarch64_can_const_movi_rtx_p)
17606 (aarch64_simd_scalar_immediate_valid_for_move)
17607 (aarch64_simd_make_constant): Update call to
17608 aarch64_simd_valid_immediate.
17609 (aarch64_advsimd_valid_immediate_hs): New function.
17610 (aarch64_advsimd_valid_immediate): Likewise.
17611 (aarch64_simd_valid_immediate): Remove mode and inverse
17612 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17613 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17614 and aarch64_float_const_representable_p on the result.
17615 (aarch64_output_simd_mov_immediate): Remove mode argument.
17616 Update call to aarch64_simd_valid_immediate and use of
17617 simd_immediate_info.
17618 (aarch64_output_scalar_simd_mov_immediate): Update call
17621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17622 Alan Hayward <alan.hayward@arm.com>
17623 David Sherwood <david.sherwood@arm.com>
17625 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17626 (mode_nunits): Likewise CONST_MODE_NUNITS.
17627 * machmode.def (ADJUST_NUNITS): Document.
17628 * genmodes.c (mode_data::need_nunits_adj): New field.
17629 (blank_mode): Update accordingly.
17630 (adj_nunits): New variable.
17631 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17633 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17634 listed in adj_nunits.
17635 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17636 listed in adj_nunits. Don't emit case statements for such modes.
17637 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17638 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17639 nothing if adj_nunits is nonnull.
17640 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17641 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17642 (emit_mode_fbit): Update use of print_maybe_const_decl.
17643 (emit_move_size): Likewise. Treat the array as non-const
17645 (emit_mode_adjustments): Handle adj_nunits.
17647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17649 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17650 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17651 (VECTOR_MODES): Use it.
17652 (make_vector_modes): Take the prefix as an argument.
17654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17655 Alan Hayward <alan.hayward@arm.com>
17656 David Sherwood <david.sherwood@arm.com>
17658 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17659 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17660 for MODE_VECTOR_BOOL.
17661 * machmode.def (VECTOR_BOOL_MODE): Document.
17662 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17663 (make_vector_bool_mode): New function.
17664 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17666 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17667 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17669 * stor-layout.c (int_mode_for_mode): Likewise.
17670 * tree.c (build_vector_type_for_mode): Likewise.
17671 * varasm.c (output_constant_pool_2): Likewise.
17672 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17673 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17674 for MODE_VECTOR_BOOL.
17675 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17676 of mode class checks.
17677 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17678 instead of a list of mode class checks.
17679 (expand_vector_scalar_condition): Likewise.
17680 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17683 Alan Hayward <alan.hayward@arm.com>
17684 David Sherwood <david.sherwood@arm.com>
17686 * machmode.h (mode_size): Change from unsigned short to
17688 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17689 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17690 or if measurement_type is not polynomial.
17691 (fixed_size_mode::includes_p): Check for constant-sized modes.
17692 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17693 return a poly_uint16 rather than an unsigned short.
17694 (emit_mode_size): Change the type of mode_size from unsigned short
17695 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17696 (emit_mode_adjustments): Cope with polynomial vector sizes.
17697 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17699 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17701 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17702 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17703 * caller-save.c (setup_save_areas): Likewise.
17704 (replace_reg_with_saved_mem): Likewise.
17705 * calls.c (emit_library_call_value_1): Likewise.
17706 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17707 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17708 (gen_lowpart_for_combine): Likewise.
17709 * convert.c (convert_to_integer_1): Likewise.
17710 * cse.c (equiv_constant, cse_insn): Likewise.
17711 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17712 (cselib_subst_to_values): Likewise.
17713 * dce.c (word_dce_process_block): Likewise.
17714 * df-problems.c (df_word_lr_mark_ref): Likewise.
17715 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17716 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17717 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17718 (rtl_for_decl_location): Likewise.
17719 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17720 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17721 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17722 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17723 (expand_expr_real_1): Likewise.
17724 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17725 (pad_below): Likewise.
17726 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17727 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17728 * ira.c (get_subreg_tracking_sizes): Likewise.
17729 * ira-build.c (ira_create_allocno_objects): Likewise.
17730 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17731 (ira_sort_regnos_for_alter_reg): Likewise.
17732 * ira-costs.c (record_operand_costs): Likewise.
17733 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17734 (resolve_simple_move): Likewise.
17735 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17736 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17737 (lra_constraints): Likewise.
17738 (CONST_POOL_OK_P): Reject variable-sized modes.
17739 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17740 (add_pseudo_to_slot, lra_spill): Likewise.
17741 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17742 * optabs-query.c (get_best_extraction_insn): Likewise.
17743 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17744 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17745 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17746 * recog.c (offsettable_address_addr_space_p): Likewise.
17747 * regcprop.c (maybe_mode_change): Likewise.
17748 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17749 * regrename.c (build_def_use): Likewise.
17750 * regstat.c (dump_reg_info): Likewise.
17751 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17752 (find_reloads, find_reloads_subreg_address): Likewise.
17753 * reload1.c (eliminate_regs_1): Likewise.
17754 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17755 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17756 (simplify_binary_operation_1, simplify_subreg): Likewise.
17757 * targhooks.c (default_function_arg_padding): Likewise.
17758 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17759 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17760 (verify_gimple_assign_ternary): Likewise.
17761 * tree-inline.c (estimate_move_cost): Likewise.
17762 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17763 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17764 (get_address_cost_ainc): Likewise.
17765 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17766 (vect_supportable_dr_alignment): Likewise.
17767 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17768 (vectorizable_reduction): Likewise.
17769 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17770 (vectorizable_operation, vectorizable_load): Likewise.
17771 * tree.c (build_same_sized_truth_vector_type): Likewise.
17772 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17773 * var-tracking.c (emit_note_insn_var_location): Likewise.
17774 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17775 (ADDR_VEC_ALIGN): 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 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17783 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17784 or if measurement_type is polynomial.
17785 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17786 * combine.c (make_extraction): Likewise.
17787 * dse.c (find_shift_sequence): Likewise.
17788 * dwarf2out.c (mem_loc_descriptor): Likewise.
17789 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17790 (extract_bit_field, extract_low_bits): Likewise.
17791 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17792 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17793 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17794 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17795 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17796 * reload.c (find_reloads): Likewise.
17797 * reload1.c (alter_reg): Likewise.
17798 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17799 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17800 * tree-if-conv.c (predicate_mem_writes): Likewise.
17801 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17802 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17803 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17804 * valtrack.c (dead_debug_insert_temp): Likewise.
17805 * varasm.c (mergeable_constant_section): Likewise.
17806 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17809 Alan Hayward <alan.hayward@arm.com>
17810 David Sherwood <david.sherwood@arm.com>
17812 * expr.c (expand_assignment): Cope with polynomial mode sizes
17813 when assigning to a CONCAT.
17815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17816 Alan Hayward <alan.hayward@arm.com>
17817 David Sherwood <david.sherwood@arm.com>
17819 * machmode.h (mode_precision): Change from unsigned short to
17821 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17823 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17824 or if measurement_type is not polynomial.
17825 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17826 in which the mode is already known to be a scalar_int_mode.
17827 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17828 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17830 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17831 for GET_MODE_PRECISION.
17832 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17833 for GET_MODE_PRECISION.
17834 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17836 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17837 (expand_field_assignment, make_extraction): Likewise.
17838 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17839 (get_last_value): Likewise.
17840 * convert.c (convert_to_integer_1): Likewise.
17841 * cse.c (cse_insn): Likewise.
17842 * expr.c (expand_expr_real_1): Likewise.
17843 * lra-constraints.c (simplify_operand_subreg): Likewise.
17844 * optabs-query.c (can_atomic_load_p): Likewise.
17845 * optabs.c (expand_atomic_load): Likewise.
17846 (expand_atomic_store): Likewise.
17847 * ree.c (combine_reaching_defs): Likewise.
17848 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17849 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17850 * tree.h (type_has_mode_precision_p): Likewise.
17851 * ubsan.c (instrument_si_overflow): Likewise.
17853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17854 Alan Hayward <alan.hayward@arm.com>
17855 David Sherwood <david.sherwood@arm.com>
17857 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17858 polynomial numbers of units.
17859 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17860 (valid_vector_subparts_p): New function.
17861 (build_vector_type): Remove temporary shim and take the number
17862 of units as a poly_uint64 rather than an int.
17863 (build_opaque_vector_type): Take the number of units as a
17864 poly_uint64 rather than an int.
17865 * tree.c (build_vector_from_ctor): Handle polynomial
17866 TYPE_VECTOR_SUBPARTS.
17867 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17868 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17869 (build_vector_from_val): If the number of units is variable,
17870 use build_vec_duplicate_cst for constant operands and
17871 VEC_DUPLICATE_EXPR otherwise.
17872 (make_vector_type): Remove temporary is_constant ().
17873 (build_vector_type, build_opaque_vector_type): Take the number of
17874 units as a poly_uint64 rather than an int.
17875 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17877 * cfgexpand.c (expand_debug_expr): Likewise.
17878 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17879 (store_constructor, expand_expr_real_1): Likewise.
17880 (const_scalar_mask_from_tree): Likewise.
17881 * fold-const-call.c (fold_const_reduction): Likewise.
17882 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17883 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17884 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17885 (fold_relational_const): Likewise.
17886 (native_interpret_vector): Likewise. Change the size from an
17887 int to an unsigned int.
17888 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17889 TYPE_VECTOR_SUBPARTS.
17890 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17891 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17892 duplicating a non-constant operand into a variable-length vector.
17893 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17894 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17895 * ipa-icf.c (sem_variable::equals): Likewise.
17896 * match.pd: Likewise.
17897 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17898 * print-tree.c (print_node): Likewise.
17899 * stor-layout.c (layout_type): Likewise.
17900 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17901 * tree-cfg.c (verify_gimple_comparison): Likewise.
17902 (verify_gimple_assign_binary): Likewise.
17903 (verify_gimple_assign_ternary): Likewise.
17904 (verify_gimple_assign_single): Likewise.
17905 * tree-pretty-print.c (dump_generic_node): Likewise.
17906 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17907 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17908 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17909 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17910 (vect_shift_permute_load_chain): Likewise.
17911 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17912 (expand_vector_condition, optimize_vector_constructor): Likewise.
17913 (lower_vec_perm, get_compute_type): Likewise.
17914 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17915 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17916 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17917 (vect_recog_mask_conversion_pattern): Likewise.
17918 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17919 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17920 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17921 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17922 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17923 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17924 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17925 (supportable_widening_operation): Likewise.
17926 (supportable_narrowing_operation): Likewise.
17927 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17929 * varasm.c (output_constant): Likewise.
17931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17932 Alan Hayward <alan.hayward@arm.com>
17933 David Sherwood <david.sherwood@arm.com>
17935 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17936 so that both the length == 3 and length != 3 cases set up their
17937 own permute vectors. Add comments explaining why we know the
17938 number of elements is constant.
17939 (vect_permute_load_chain): Likewise.
17941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17942 Alan Hayward <alan.hayward@arm.com>
17943 David Sherwood <david.sherwood@arm.com>
17945 * machmode.h (mode_nunits): Change from unsigned char to
17947 (ONLY_FIXED_SIZE_MODES): New macro.
17948 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17949 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17950 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17952 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17953 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17954 or if measurement_type is not polynomial.
17955 * genmodes.c (ZERO_COEFFS): New macro.
17956 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17958 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17959 Use ZERO_COEFFS when emitting initializers.
17960 * data-streamer.h (bp_pack_poly_value): New function.
17961 (bp_unpack_poly_value): Likewise.
17962 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17963 for GET_MODE_NUNITS.
17964 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17965 for GET_MODE_NUNITS.
17966 * tree.c (make_vector_type): Remove temporary shim and make
17967 the real function take the number of units as a poly_uint64
17968 rather than an int.
17969 (build_vector_type_for_mode): Handle polynomial nunits.
17970 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17971 * emit-rtl.c (const_vec_series_p_1): Likewise.
17972 (gen_rtx_CONST_VECTOR): Likewise.
17973 * fold-const.c (test_vec_duplicate_folding): Likewise.
17974 * genrecog.c (validate_pattern): Likewise.
17975 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17976 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17977 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17978 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17979 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17980 * rtlanal.c (subreg_get_info): Likewise.
17981 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17982 (vect_grouped_load_supported): Likewise.
17983 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17984 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17985 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17986 (simplify_const_unary_operation, simplify_binary_operation_1)
17987 (simplify_const_binary_operation, simplify_ternary_operation)
17988 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17989 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17990 instead of CONST_VECTOR_NUNITS.
17991 * varasm.c (output_constant_pool_2): Likewise.
17992 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17993 explicit-encoded elements in the XVEC for variable-length vectors.
17995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17997 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18000 Alan Hayward <alan.hayward@arm.com>
18001 David Sherwood <david.sherwood@arm.com>
18003 * coretypes.h (fixed_size_mode): Declare.
18004 (fixed_size_mode_pod): New typedef.
18005 * builtins.h (target_builtins::x_apply_args_mode)
18006 (target_builtins::x_apply_result_mode): Change type to
18007 fixed_size_mode_pod.
18008 * builtins.c (apply_args_size, apply_result_size, result_vector)
18009 (expand_builtin_apply_args_1, expand_builtin_apply)
18010 (expand_builtin_return): Update accordingly.
18012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18014 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18015 * cselib.c (cselib_hash_rtx): Likewise.
18016 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18017 CONST_VECTOR encoding.
18019 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18020 Jeff Law <law@redhat.com>
18023 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18024 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18025 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18026 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18029 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18030 explicitly probe *sp in a noreturn function if there were any callee
18031 register saves or frame pointer is needed.
18033 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18036 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18037 BLKmode for ternary, binary or unary expressions.
18040 * var-tracking.c (delete_vta_debug_insn): New inline function.
18041 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18042 insns from get_insns () to NULL instead of each bb separately.
18043 Use delete_vta_debug_insn. No longer static.
18044 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18045 delete_vta_debug_insns callers.
18046 * rtl.h (delete_vta_debug_insns): Declare.
18047 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18048 instead of variable_tracking_main.
18050 2018-01-03 Martin Sebor <msebor@redhat.com>
18052 PR tree-optimization/83603
18053 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18054 arguments past the endof the argument list in functions declared
18055 without a prototype.
18056 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18057 Avoid checking when arguments are null.
18059 2018-01-03 Martin Sebor <msebor@redhat.com>
18062 * doc/extend.texi (attribute const): Fix a typo.
18063 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18064 issuing -Wsuggest-attribute for void functions.
18066 2018-01-03 Martin Sebor <msebor@redhat.com>
18068 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18069 offset_int::from instead of wide_int::to_shwi.
18070 (maybe_diag_overlap): Remove assertion.
18071 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18072 * gimple-ssa-sprintf.c (format_directive): Same.
18073 (parse_directive): Same.
18074 (sprintf_dom_walker::compute_format_length): Same.
18075 (try_substitute_return_value): Same.
18077 2018-01-03 Jeff Law <law@redhat.com>
18079 PR middle-end/83654
18080 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18081 non-constant residual for zero at runtime and avoid probing in
18082 that case. Reorganize code for trailing problem to mirror handling
18085 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18087 PR tree-optimization/83501
18088 * tree-ssa-strlen.c (get_string_cst): New.
18089 (handle_char_store): Call get_string_cst.
18091 2018-01-03 Martin Liska <mliska@suse.cz>
18093 PR tree-optimization/83593
18094 * tree-ssa-strlen.c: Include tree-cfg.h.
18095 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18096 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18097 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18099 (strlen_dom_walker::before_dom_children): Call
18100 gimple_purge_dead_eh_edges. Dump tranformation with details
18102 (strlen_dom_walker::before_dom_children): Update call by adding
18103 new argument cleanup_eh.
18104 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18106 2018-01-03 Martin Liska <mliska@suse.cz>
18109 * cif-code.def (VARIADIC_THUNK): New enum value.
18110 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18113 2018-01-03 Jan Beulich <jbeulich@suse.com>
18115 * sse.md (mov<mode>_internal): Tighten condition for when to use
18116 vmovdqu<ssescalarsize> for TI and OI modes.
18118 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18120 Update copyright years.
18122 2018-01-03 Martin Liska <mliska@suse.cz>
18125 * ipa-visibility.c (function_and_variable_visibility): Skip
18126 functions with noipa attribure.
18128 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18130 * gcc.c (process_command): Update copyright notice dates.
18131 * gcov-dump.c (print_version): Ditto.
18132 * gcov.c (print_version): Ditto.
18133 * gcov-tool.c (print_version): Ditto.
18134 * gengtype.c (create_file): Ditto.
18135 * doc/cpp.texi: Bump @copying's copyright year.
18136 * doc/cppinternals.texi: Ditto.
18137 * doc/gcc.texi: Ditto.
18138 * doc/gccint.texi: Ditto.
18139 * doc/gcov.texi: Ditto.
18140 * doc/install.texi: Ditto.
18141 * doc/invoke.texi: Ditto.
18143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18145 * vector-builder.h (vector_builder::m_full_nelts): Change from
18146 unsigned int to poly_uint64.
18147 (vector_builder::full_nelts): Update prototype accordingly.
18148 (vector_builder::new_vector): Likewise.
18149 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18150 (vector_builder::operator ==): Likewise.
18151 (vector_builder::finalize): Likewise.
18152 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18153 Take the number of elements as a poly_uint64 rather than an
18155 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18156 from unsigned int to poly_uint64.
18157 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18158 (vec_perm_indices::new_vector): Likewise.
18159 (vec_perm_indices::length): Likewise.
18160 (vec_perm_indices::nelts_per_input): Likewise.
18161 (vec_perm_indices::input_nelts): Likewise.
18162 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18163 number of elements per input as a poly_uint64 rather than an
18164 unsigned int. Use the original encoding for variable-length
18165 vectors, rather than clamping each individual element.
18166 For the second and subsequent elements in each pattern,
18167 clamp the step and base before clamping their sum.
18168 (vec_perm_indices::series_p): Handle polynomial element counts.
18169 (vec_perm_indices::all_in_range_p): Likewise.
18170 (vec_perm_indices_to_tree): Likewise.
18171 (vec_perm_indices_to_rtx): Likewise.
18172 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18173 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18174 (tree_vector_builder::new_binary_operation): Handle polynomial
18175 element counts. Return false if we need to know the number
18176 of elements at compile time.
18177 * fold-const.c (fold_vec_perm): Punt if the number of elements
18178 isn't known at compile time.
18180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18182 * vec-perm-indices.h (vec_perm_builder): Change element type
18183 from HOST_WIDE_INT to poly_int64.
18184 (vec_perm_indices::element_type): Update accordingly.
18185 (vec_perm_indices::clamp): Handle polynomial element_types.
18186 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18187 (vec_perm_indices::all_in_range_p): Likewise.
18188 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18190 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18191 polynomial vec_perm_indices element types.
18192 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18193 * fold-const.c (fold_vec_perm): Likewise.
18194 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18195 * tree-vect-generic.c (lower_vec_perm): Likewise.
18196 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18197 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18198 element type to HOST_WIDE_INT.
18200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18201 Alan Hayward <alan.hayward@arm.com>
18202 David Sherwood <david.sherwood@arm.com>
18204 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18205 rather than an int. Use plus_constant.
18206 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18207 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18210 Alan Hayward <alan.hayward@arm.com>
18211 David Sherwood <david.sherwood@arm.com>
18213 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18214 a HOST_WIDE_INT to a poly_int64.
18216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18217 Alan Hayward <alan.hayward@arm.com>
18218 David Sherwood <david.sherwood@arm.com>
18220 * calls.c (load_register_parameters): Cope with polynomial
18221 mode sizes. Require a constant size for BLKmode parameters
18222 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18223 forces a parameter to be padded at the lsb end in order to
18224 fill a complete number of words, require the parameter size
18225 to be ordered wrt UNITS_PER_WORD.
18227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18228 Alan Hayward <alan.hayward@arm.com>
18229 David Sherwood <david.sherwood@arm.com>
18231 * reload1.c (spill_stack_slot_width): Change element type
18232 from unsigned int to poly_uint64_pod.
18233 (alter_reg): Treat mode sizes as polynomial.
18235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18236 Alan Hayward <alan.hayward@arm.com>
18237 David Sherwood <david.sherwood@arm.com>
18239 * reload.c (complex_word_subreg_p): New function.
18240 (reload_inner_reg_of_subreg, push_reload): Use it.
18242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18243 Alan Hayward <alan.hayward@arm.com>
18244 David Sherwood <david.sherwood@arm.com>
18246 * lra-constraints.c (process_alt_operands): Reject matched
18247 operands whose sizes aren't ordered.
18248 (match_reload): Refer to this check here.
18250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18251 Alan Hayward <alan.hayward@arm.com>
18252 David Sherwood <david.sherwood@arm.com>
18254 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18255 that the mode size is in the set {1, 2, 4, 8, 16}.
18257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18258 Alan Hayward <alan.hayward@arm.com>
18259 David Sherwood <david.sherwood@arm.com>
18261 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18262 Use plus_constant instead of gen_rtx_PLUS.
18264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18265 Alan Hayward <alan.hayward@arm.com>
18266 David Sherwood <david.sherwood@arm.com>
18268 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18269 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18270 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18271 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18272 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18273 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18274 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18275 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18276 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18277 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18279 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18280 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18281 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18282 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18283 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18284 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18285 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18286 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18287 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18288 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18290 * expr.c (emit_move_resolve_push): Treat the input and result
18291 of PUSH_ROUNDING as a poly_int64.
18292 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18293 (emit_push_insn): Likewise.
18294 * lra-eliminations.c (mark_not_eliminable): Likewise.
18295 * recog.c (push_operand): Likewise.
18296 * reload1.c (elimination_effects): Likewise.
18297 * rtlanal.c (nonzero_bits1): Likewise.
18298 * calls.c (store_one_arg): Likewise. Require the padding to be
18299 known at compile time.
18301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18302 Alan Hayward <alan.hayward@arm.com>
18303 David Sherwood <david.sherwood@arm.com>
18305 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18306 Use plus_constant instead of gen_rtx_PLUS.
18308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18309 Alan Hayward <alan.hayward@arm.com>
18310 David Sherwood <david.sherwood@arm.com>
18312 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18313 rather than an int.
18315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18316 Alan Hayward <alan.hayward@arm.com>
18317 David Sherwood <david.sherwood@arm.com>
18319 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18320 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18321 via stack temporaries. Treat the mode size as polynomial too.
18323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18324 Alan Hayward <alan.hayward@arm.com>
18325 David Sherwood <david.sherwood@arm.com>
18327 * expr.c (expand_expr_real_2): When handling conversions involving
18328 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18329 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18330 as a poly_uint64 too.
18332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18333 Alan Hayward <alan.hayward@arm.com>
18334 David Sherwood <david.sherwood@arm.com>
18336 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18339 Alan Hayward <alan.hayward@arm.com>
18340 David Sherwood <david.sherwood@arm.com>
18342 * combine.c (can_change_dest_mode): Handle polynomial
18343 REGMODE_NATURAL_SIZE.
18344 * expmed.c (store_bit_field_1): Likewise.
18345 * expr.c (store_constructor): Likewise.
18346 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18347 and polynomial REGMODE_NATURAL_SIZE.
18348 (gen_lowpart_common): Likewise.
18349 * reginfo.c (record_subregs_of_mode): Likewise.
18350 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18357 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
18365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18366 Alan Hayward <alan.hayward@arm.com>
18367 David Sherwood <david.sherwood@arm.com>
18369 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18370 in a POINTER_PLUS_EXPR.
18372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18373 Alan Hayward <alan.hayward@arm.com>
18374 David Sherwood <david.sherwood@arm.com>
18376 * omp-simd-clone.c (simd_clone_subparts): New function.
18377 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18378 (ipa_simd_modify_function_body): Likewise.
18380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18381 Alan Hayward <alan.hayward@arm.com>
18382 David Sherwood <david.sherwood@arm.com>
18384 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18385 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18386 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18387 (expand_vector_condition, vector_element): Likewise.
18388 (subparts_gt): New function.
18389 (get_compute_type): Use subparts_gt.
18390 (count_type_subparts): Delete.
18391 (expand_vector_operations_1): Use subparts_gt instead of
18392 count_type_subparts.
18394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18395 Alan Hayward <alan.hayward@arm.com>
18396 David Sherwood <david.sherwood@arm.com>
18398 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18399 (vect_compile_time_alias): ...this new function. Do the calculation
18400 on poly_ints rather than trees.
18401 (vect_prune_runtime_alias_test_list): Update call accordingly.
18403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18404 Alan Hayward <alan.hayward@arm.com>
18405 David Sherwood <david.sherwood@arm.com>
18407 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18409 (vect_schedule_slp_instance): Likewise.
18411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18412 Alan Hayward <alan.hayward@arm.com>
18413 David Sherwood <david.sherwood@arm.com>
18415 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18416 constant and extern definitions for variable-length vectors.
18417 (vect_get_constant_vectors): Note that the number of units
18418 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18425 of units as polynomial. Choose between WIDE and NARROW based
18428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18429 Alan Hayward <alan.hayward@arm.com>
18430 David Sherwood <david.sherwood@arm.com>
18432 * tree-vect-stmts.c (simd_clone_subparts): New function.
18433 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18436 Alan Hayward <alan.hayward@arm.com>
18437 David Sherwood <david.sherwood@arm.com>
18439 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18440 vectors as polynomial. Use build_index_vector for
18441 IFN_GOMP_SIMD_LANE.
18443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18444 Alan Hayward <alan.hayward@arm.com>
18445 David Sherwood <david.sherwood@arm.com>
18447 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18448 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18449 for variable-length vectors.
18450 (vectorizable_mask_load_store): Treat the number of units as
18451 polynomial, asserting that it is constant if the condition has
18452 already been enforced.
18453 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18460 of units as polynomial. Punt if we can't tell at compile time
18461 which vector contains the final result.
18463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18464 Alan Hayward <alan.hayward@arm.com>
18465 David Sherwood <david.sherwood@arm.com>
18467 * tree-vect-loop.c (vectorizable_induction): Treat the number
18468 of units as polynomial. Punt on SLP inductions. Use an integer
18469 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18470 cast of such a series for variable-length floating-point
18473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18474 Alan Hayward <alan.hayward@arm.com>
18475 David Sherwood <david.sherwood@arm.com>
18477 * tree.h (build_index_vector): Declare.
18478 * tree.c (build_index_vector): New function.
18479 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18480 of units as polynomial, forcibly converting it to a constant if
18481 vectorizable_reduction has already enforced the condition.
18482 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18483 to create a {1,2,3,...} vector.
18484 (vectorizable_reduction): Treat the number of units as polynomial.
18485 Choose vectype_in based on the largest scalar element size rather
18486 than the smallest number of units. Enforce the restrictions
18489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18490 Alan Hayward <alan.hayward@arm.com>
18491 David Sherwood <david.sherwood@arm.com>
18493 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18494 number of units as polynomial.
18496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18497 Alan Hayward <alan.hayward@arm.com>
18498 David Sherwood <david.sherwood@arm.com>
18500 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18501 * target.def (autovectorize_vector_sizes): Return the vector sizes
18502 by pointer, using vector_sizes rather than a bitmask.
18503 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18504 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18505 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18507 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18508 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18509 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18510 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18511 * omp-general.c (omp_max_vf): Likewise.
18512 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18513 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18514 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18515 * tree-vect-slp.c (vect_slp_bb): Likewise.
18516 * doc/tm.texi: Regenerate.
18517 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18519 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18520 the vector size as a poly_uint64 rather than an unsigned int.
18521 (current_vector_size): Change from an unsigned int to a poly_uint64.
18522 (get_vectype_for_scalar_type): Update accordingly.
18523 * tree.h (build_truth_vector_type): Take the size and number of
18524 units as a poly_uint64 rather than an unsigned int.
18525 (build_vector_type): Add a temporary overload that takes
18526 the number of units as a poly_uint64 rather than an unsigned int.
18527 * tree.c (make_vector_type): Likewise.
18528 (build_truth_vector_type): Take the number of units as a poly_uint64
18529 rather than an unsigned int.
18531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18532 Alan Hayward <alan.hayward@arm.com>
18533 David Sherwood <david.sherwood@arm.com>
18535 * target.def (get_mask_mode): Take the number of units and length
18536 as poly_uint64s rather than unsigned ints.
18537 * targhooks.h (default_get_mask_mode): Update accordingly.
18538 * targhooks.c (default_get_mask_mode): Likewise.
18539 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18540 * doc/tm.texi: Regenerate.
18542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18543 Alan Hayward <alan.hayward@arm.com>
18544 David Sherwood <david.sherwood@arm.com>
18546 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18547 * omp-general.c (omp_max_vf): Likewise.
18548 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18549 (expand_omp_simd): Handle polynomial safelen.
18550 * omp-low.c (omplow_simd_context): Add a default constructor.
18551 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18552 (lower_rec_simd_input_clauses): Update accordingly.
18553 (lower_rec_input_clauses): Likewise.
18555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18556 Alan Hayward <alan.hayward@arm.com>
18557 David Sherwood <david.sherwood@arm.com>
18559 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18560 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18561 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18562 (vect_analyze_slp_cost): Likewise.
18563 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18564 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18571 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18572 from an unsigned int * to a poly_uint64_pod *.
18573 (calculate_unrolling_factor): New function.
18574 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18577 Alan Hayward <alan.hayward@arm.com>
18578 David Sherwood <david.sherwood@arm.com>
18580 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18581 from an unsigned int to a poly_uint64.
18582 (_loop_vec_info::slp_unrolling_factor): Likewise.
18583 (_loop_vec_info::vectorization_factor): Change from an int
18585 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18586 (vect_get_num_vectors): New function.
18587 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18588 (vect_get_num_copies): Use vect_get_num_vectors.
18589 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18590 to an unsigned int *.
18591 (vect_analyze_data_refs): Change min_vf from an int * to a
18593 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18594 than an unsigned HOST_WIDE_INT.
18595 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18596 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18597 to an unsigned int *.
18598 (vect_analyze_data_ref_dependences): Likewise.
18599 (vect_compute_data_ref_alignment): Handle polynomial vf.
18600 (vect_enhance_data_refs_alignment): Likewise.
18601 (vect_prune_runtime_alias_test_list): Likewise.
18602 (vect_shift_permute_load_chain): Likewise.
18603 (vect_supportable_dr_alignment): Likewise.
18604 (dependence_distance_ge_vf): Take the vectorization factor as a
18605 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18606 (vect_analyze_data_refs): Change min_vf from an int * to a
18608 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18609 vfm1 as a poly_uint64 rather than an int. Make the same change
18610 for the returned bound_scalar.
18611 (vect_gen_vector_loop_niters): Handle polynomial vf.
18612 (vect_do_peeling): Likewise. Update call to
18613 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18614 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18616 * tree-vect-loop.c (vect_determine_vectorization_factor)
18617 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18618 (vect_get_known_peeling_cost): Likewise.
18619 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18620 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18621 (vect_transform_loop): Likewise. Use the lowest possible VF when
18622 updating the upper bounds of the loop.
18623 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18624 rather than an int.
18625 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18626 polynomial unroll factors.
18627 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18628 (vect_make_slp_decision): Likewise.
18629 (vect_supported_load_permutation_p): Likewise, and polynomial
18631 (vect_analyze_slp_cost): Handle polynomial vf.
18632 (vect_slp_analyze_node_operations): Likewise.
18633 (vect_slp_analyze_bb_1): Likewise.
18634 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18635 than an unsigned HOST_WIDE_INT.
18636 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18637 (vectorizable_load): Handle polynomial vf.
18638 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18640 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18643 Alan Hayward <alan.hayward@arm.com>
18644 David Sherwood <david.sherwood@arm.com>
18646 * match.pd: Handle bit operations involving three constants
18647 and try to fold one pair.
18649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18651 * tree-vect-loop-manip.c: Include gimple-fold.h.
18652 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18653 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18654 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18655 Add a path that uses a step of VF instead of 1, but disable it
18657 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18658 and niters_no_overflow parameters. Update calls to
18659 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18660 Create a new SSA name if the latter choses to use a ste other
18661 than zero, and return it via niters_vector_mult_vf_var.
18662 * tree-vect-loop.c (vect_transform_loop): Update calls to
18663 vect_do_peeling, vect_gen_vector_loop_niters and
18664 slpeel_make_loop_iterate_ntimes.
18665 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18666 (vect_gen_vector_loop_niters): Update declarations after above changes.
18668 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18670 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18671 128-bit round to integer instructions.
18672 (ceil<mode>2): Likewise.
18673 (btrunc<mode>2): Likewise.
18674 (round<mode>2): Likewise.
18676 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18678 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18679 unaligned VSX load/store on P8/P9.
18680 (expand_block_clear): Allow the use of unaligned VSX
18681 load/store on P8/P9.
18683 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18685 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18687 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18688 swap associated with both a load and a store.
18690 2018-01-02 Andrew Waterman <andrew@sifive.com>
18692 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18693 * config/riscv/riscv.md (clear_cache): Use it.
18695 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18697 * web.c: Remove out-of-date comment.
18699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18701 * expr.c (fixup_args_size_notes): Check that any existing
18702 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18703 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18704 (emit_single_push_insn): ...here.
18706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18708 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18709 (const_vector_encoded_nelts): New function.
18710 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18711 (const_vector_int_elt, const_vector_elt): Declare.
18712 * emit-rtl.c (const_vector_int_elt_1): New function.
18713 (const_vector_elt): Likewise.
18714 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18715 of CONST_VECTOR_ELT.
18717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18719 * expr.c: Include rtx-vector-builder.h.
18720 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18721 directly on the tree encoding.
18722 (const_vector_from_tree): Likewise.
18723 * optabs.c: Include rtx-vector-builder.h.
18724 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18725 sequence of "u" values.
18726 * vec-perm-indices.c: Include rtx-vector-builder.h.
18727 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18728 directly on the vec_perm_indices encoding.
18730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18732 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18733 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18734 * rtx-vector-builder.h: New file.
18735 * rtx-vector-builder.c: Likewise.
18736 * rtl.h (rtx_def::u2): Add a const_vector field.
18737 (CONST_VECTOR_NPATTERNS): New macro.
18738 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18739 (CONST_VECTOR_DUPLICATE_P): Likewise.
18740 (CONST_VECTOR_STEPPED_P): Likewise.
18741 (CONST_VECTOR_ENCODED_ELT): Likewise.
18742 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18743 (unwrap_const_vec_duplicate): Likewise.
18744 (const_vec_series_p): Check for a non-duplicated vector encoding.
18745 Say that the function only returns true for integer vectors.
18746 * emit-rtl.c: Include rtx-vector-builder.h.
18747 (gen_const_vec_duplicate_1): Delete.
18748 (gen_const_vector): Call gen_const_vec_duplicate instead of
18749 gen_const_vec_duplicate_1.
18750 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18751 (gen_const_vec_duplicate): Use rtx_vector_builder.
18752 (gen_const_vec_series): Likewise.
18753 (gen_rtx_CONST_VECTOR): Likewise.
18754 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18755 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18756 Build a new vector rather than modifying a CONST_VECTOR in-place.
18757 (handle_special_swappables): Update call accordingly.
18758 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18759 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18760 Build a new vector rather than modifying a CONST_VECTOR in-place.
18761 (handle_special_swappables): Update call accordingly.
18763 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18765 * simplify-rtx.c (simplify_const_binary_operation): Use
18766 CONST_VECTOR_ELT instead of XVECEXP.
18768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18770 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18771 the selector elements to be different from the data elements
18772 if the selector is a VECTOR_CST.
18773 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18774 ssizetype for the selector.
18776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18778 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18779 before testing each element individually.
18780 * tree-vect-generic.c (lower_vec_perm): Likewise.
18782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18784 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18785 * selftest-run-tests.c (selftest::run_tests): Call it.
18786 * vector-builder.h (vector_builder::operator ==): New function.
18787 (vector_builder::operator !=): Likewise.
18788 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18789 (vec_perm_indices::all_from_input_p): New function.
18790 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18791 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18792 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18793 instead of reading the VECTOR_CST directly. Detect whether both
18794 vector inputs are the same before constructing the vec_perm_indices,
18795 and update the number of inputs argument accordingly. Use the
18796 utility functions added above. Only construct sel2 if we need to.
18798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18800 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18801 the broadcast of the low byte.
18802 (expand_mult_highpart): Use an explicit encoding for the permutes.
18803 * optabs-query.c (can_mult_highpart_p): Likewise.
18804 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18805 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18806 (vectorizable_bswap): Likewise.
18807 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18808 explicit encoding for the power-of-2 permutes.
18809 (vect_permute_store_chain): Likewise.
18810 (vect_grouped_load_supported): Likewise.
18811 (vect_permute_load_chain): Likewise.
18813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18815 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18816 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18817 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18818 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18819 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18820 (vect_gen_perm_mask_any): Likewise.
18822 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18824 * int-vector-builder.h: New file.
18825 * vec-perm-indices.h: Include int-vector-builder.h.
18826 (vec_perm_indices): Redefine as an int_vector_builder.
18827 (auto_vec_perm_indices): Delete.
18828 (vec_perm_builder): Redefine as a stand-alone class.
18829 (vec_perm_indices::vec_perm_indices): New function.
18830 (vec_perm_indices::clamp): Likewise.
18831 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18832 (vec_perm_indices::new_vector): New function.
18833 (vec_perm_indices::new_expanded_vector): Update for new
18834 vec_perm_indices class.
18835 (vec_perm_indices::rotate_inputs): New function.
18836 (vec_perm_indices::all_in_range_p): Operate directly on the
18837 encoded form, without computing elided elements.
18838 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18839 encoding. Update for new vec_perm_indices class.
18840 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18841 the given vec_perm_builder.
18842 (expand_vec_perm_var): Update vec_perm_builder constructor.
18843 (expand_mult_highpart): Use vec_perm_builder instead of
18844 auto_vec_perm_indices.
18845 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18846 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18847 or double series encoding as appropriate.
18848 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18849 vec_perm_indices instead of auto_vec_perm_indices.
18850 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18851 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18852 (vect_permute_store_chain): Likewise.
18853 (vect_grouped_load_supported): Likewise.
18854 (vect_permute_load_chain): Likewise.
18855 (vect_shift_permute_load_chain): Likewise.
18856 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18857 (vect_transform_slp_perm_load): Likewise.
18858 (vect_schedule_slp_instance): Likewise.
18859 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18860 (vectorizable_mask_load_store): Likewise.
18861 (vectorizable_bswap): Likewise.
18862 (vectorizable_store): Likewise.
18863 (vectorizable_load): Likewise.
18864 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18865 vec_perm_indices instead of auto_vec_perm_indices. Use
18866 tree_to_vec_perm_builder to read the vector from a tree.
18867 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18868 vec_perm_builder instead of a vec_perm_indices.
18869 (have_whole_vector_shift): Use vec_perm_builder and
18870 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18871 truncation to calc_vec_perm_mask_for_shift.
18872 (vect_create_epilog_for_reduction): Likewise.
18873 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18874 from auto_vec_perm_indices to vec_perm_indices.
18875 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18876 instead of changing individual elements.
18877 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18878 the vector in d.perm.
18879 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18880 from auto_vec_perm_indices to vec_perm_indices.
18881 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18882 instead of changing individual elements.
18883 (arm_vectorize_vec_perm_const): Use new_vector to install
18884 the vector in d.perm.
18885 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18886 Update vec_perm_builder constructor.
18887 (rs6000_expand_interleave): Likewise.
18888 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18889 (rs6000_expand_interleave): Likewise.
18891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18893 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18894 to qimode could truncate the indices.
18895 * optabs.c (expand_vec_perm_var): Likewise.
18897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18899 * Makefile.in (OBJS): Add vec-perm-indices.o.
18900 * vec-perm-indices.h: New file.
18901 * vec-perm-indices.c: Likewise.
18902 * target.h (vec_perm_indices): Replace with a forward class
18904 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18905 * optabs.h: Include vec-perm-indices.h.
18906 (expand_vec_perm): Delete.
18907 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18908 (expand_vec_perm_const): Declare.
18909 * target.def (vec_perm_const_ok): Replace with...
18910 (vec_perm_const): ...this new hook.
18911 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18912 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18913 * doc/tm.texi: Regenerate.
18914 * optabs.def (vec_perm_const): Delete.
18915 * doc/md.texi (vec_perm_const): Likewise.
18916 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18917 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18918 expand_vec_perm for constant permutation vectors. Assert that
18919 the mode of variable permutation vectors is the integer equivalent
18920 of the mode that is being permuted.
18921 * optabs-query.h (selector_fits_mode_p): Declare.
18922 * optabs-query.c: Include vec-perm-indices.h.
18923 (selector_fits_mode_p): New function.
18924 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18925 is defined, instead of checking whether the vec_perm_const_optab
18926 exists. Use targetm.vectorize.vec_perm_const instead of
18927 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18928 fit in the vector mode before using a variable permute.
18929 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18930 vec_perm_indices instead of an rtx.
18931 (expand_vec_perm): Replace with...
18932 (expand_vec_perm_const): ...this new function. Take the selector
18933 as a vec_perm_indices rather than an rtx. Also take the mode of
18934 the selector. Update call to shift_amt_for_vec_perm_mask.
18935 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18936 Use vec_perm_indices::new_expanded_vector to expand the original
18937 selector into bytes. Check whether the indices fit in the vector
18938 mode before using a variable permute.
18939 (expand_vec_perm_var): Make global.
18940 (expand_mult_highpart): Use expand_vec_perm_const.
18941 * fold-const.c: Includes vec-perm-indices.h.
18942 * tree-ssa-forwprop.c: Likewise.
18943 * tree-vect-data-refs.c: Likewise.
18944 * tree-vect-generic.c: Likewise.
18945 * tree-vect-loop.c: Likewise.
18946 * tree-vect-slp.c: Likewise.
18947 * tree-vect-stmts.c: Likewise.
18948 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18950 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18951 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18952 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18953 (aarch64_vectorize_vec_perm_const): ...this new function.
18954 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18955 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18956 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18957 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18958 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18959 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18960 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18962 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18963 check for NEON modes.
18964 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18965 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18966 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18967 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18969 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18970 the old VEC_PERM_CONST conditions.
18971 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18972 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18973 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18974 (ia64_vectorize_vec_perm_const_ok): Merge into...
18975 (ia64_vectorize_vec_perm_const): ...this new function.
18976 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18977 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18978 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18979 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18980 * config/mips/mips.c (mips_expand_vec_perm_const)
18981 (mips_vectorize_vec_perm_const_ok): Merge into...
18982 (mips_vectorize_vec_perm_const): ...this new function.
18983 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18984 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18985 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18986 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18987 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18988 (rs6000_expand_vec_perm_const): Delete.
18989 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18991 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18992 (altivec_expand_vec_perm_const_le): Take each operand individually.
18993 Operate on constant selectors rather than rtxes.
18994 (altivec_expand_vec_perm_const): Likewise. Update call to
18995 altivec_expand_vec_perm_const_le.
18996 (rs6000_expand_vec_perm_const): Delete.
18997 (rs6000_vectorize_vec_perm_const_ok): Delete.
18998 (rs6000_vectorize_vec_perm_const): New function.
18999 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19000 an element count and rtx array.
19001 (rs6000_expand_extract_even): Update call accordingly.
19002 (rs6000_expand_interleave): Likewise.
19003 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19004 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19005 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19006 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19007 (rs6000_expand_vec_perm_const): Delete.
19008 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19009 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19010 (altivec_expand_vec_perm_const_le): Take each operand individually.
19011 Operate on constant selectors rather than rtxes.
19012 (altivec_expand_vec_perm_const): Likewise. Update call to
19013 altivec_expand_vec_perm_const_le.
19014 (rs6000_expand_vec_perm_const): Delete.
19015 (rs6000_vectorize_vec_perm_const_ok): Delete.
19016 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19017 reference to the SPE evmerge intructions.
19018 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19019 an element count and rtx array.
19020 (rs6000_expand_extract_even): Update call accordingly.
19021 (rs6000_expand_interleave): Likewise.
19022 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19023 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19025 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19029 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19030 vector mode and that that mode matches the mode of the data
19032 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19033 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19034 directly using expand_vec_perm_1 when forcing selectors into
19036 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19040 * optabs-query.h (can_vec_perm_p): Delete.
19041 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19042 * optabs-query.c (can_vec_perm_p): Split into...
19043 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19044 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19045 particular selector is valid.
19046 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19047 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19048 (vect_grouped_load_supported): Likewise.
19049 (vect_shift_permute_load_chain): Likewise.
19050 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19051 (vect_transform_slp_perm_load): Likewise.
19052 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19053 (vectorizable_bswap): Likewise.
19054 (vect_gen_perm_mask_checked): Likewise.
19055 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19056 implementations of variable permutation vectors into account
19057 when deciding which selector to use.
19058 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19059 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19060 with a false third argument.
19061 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19062 to test whether the constant selector is valid and can_vec_perm_var_p
19063 to test whether a variable selector is valid.
19065 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19067 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19068 * optabs-query.c (can_vec_perm_p): Likewise.
19069 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19070 instead of vec_perm_indices.
19071 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19072 (vect_gen_perm_mask_checked): Likewise,
19073 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19074 (vect_gen_perm_mask_checked): Likewise,
19076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19078 * optabs-query.h (qimode_for_vec_perm): Declare.
19079 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19080 (qimode_for_vec_perm): ...this new function.
19081 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19083 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19085 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19086 does not have a conditional at the top.
19088 2018-01-02 Richard Biener <rguenther@suse.de>
19090 * ipa-inline.c (big_speedup_p): Fix expression.
19092 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19095 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19098 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19102 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19103 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19104 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19105 cond_taken_branch_cost 3->4.
19107 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19109 PR tree-optimization/83581
19110 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19111 TODO_cleanup_cfg if any changes have been made.
19113 PR middle-end/83608
19114 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19115 convert_modes if target mode has the right side, but different mode
19118 PR middle-end/83609
19119 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19120 last argument when extracting from CONCAT. If either from_real or
19121 from_imag is NULL, use expansion through memory. If result is not
19122 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19123 the parts directly to inner mode, if even that fails, use expansion
19126 PR middle-end/83623
19127 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19128 check for bswap in mode rather than HImode and use that in expand_unop
19131 Copyright (C) 2018 Free Software Foundation, Inc.
19133 Copying and distribution of this file, with or without modification,
19134 are permitted in any medium without royalty provided the copyright
19135 notice and this notice are preserved.