1 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5 * config/arc/arc-cpus.def (hs4x): New cpu.
7 * config/arc/arc-tables.opt: Regenerate.
8 * config/arc/arc.c (arc_sched_issue_rate): New function.
9 (TARGET_SCHED_ISSUE_RATE): Define.
10 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
11 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
13 (attr tune): Add ARCHS4x tune values.
14 (attr tune_dspmpy): Define.
15 (*tst): Correct instruction type.
16 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
17 * config/arc/arcHS4x.md: New file.
18 * config/arc/fpu.md: Update instruction type attributes.
19 * config/arc/t-multilib: Regenerate.
21 2018-07-16 Tom de Vries <tdevries@suse.de>
24 * var-tracking.c (vt_initialize): Fix pre_dec handling.
26 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
28 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
31 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
33 PR tree-optimization/86514
34 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
35 conversion to a boolean type from a type with greater precision.
37 2018-07-16 Tom de Vries <tdevries@suse.de>
39 * var-tracking.c (vt_initialize): Print adjusted insn slim if
40 dump_flags request TDF_SLIM.
42 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
44 * fold-const.c (int_const_binop_1): Abstract...
45 (wide_int_binop): ...wide int code here.
46 (poly_int_binop): ...poly int code here.
47 Abstract the rest of int_const_binop_1 into int_const_binop.
48 * fold-const.h (wide_int_binop): New.
49 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
50 Remove useless PLUS/MINUS_EXPR case.
51 (zero_nonzero_bits_from_vr): Move wide int code...
52 (zero_nonzero_bits_from_bounds): ...here.
53 (extract_range_from_binary_expr_1): Move mask optimization code...
54 (range_easy_mask_min_max): ...here.
55 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
56 (range_easy_mask_min_max): New.
58 2018-07-15 Jeff Law <law@redhat.com>
61 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
64 2018-07-14 Jim Wilson <jimw@sifive.com>
66 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
68 2018-07-14 Paul Koning <ni1d@arrl.net>
70 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
72 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
74 * lto-streamer-out.c (copy_function_or_variable): Dump info about
77 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
78 Steve Munroe <munroesj52@gmail.com>
80 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
81 (_mm_andnot_si128): Likewise.
82 (_mm_or_si128): Likewise.
83 (_mm_xor_si128): Likewise.
85 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
88 * builtins.c (expand_builtin_memcmp): Inline the calls first
89 when result_eq is false.
90 (expand_builtin_strcmp): Inline the calls first.
91 (expand_builtin_strncmp): Likewise.
92 (inline_string_cmp): New routine. Expand a string compare
93 call by using a sequence of char comparison.
94 (inline_expand_builtin_string_cmp): New routine. Inline expansion
95 a call to str(n)cmp/memcmp.
96 * doc/invoke.texi (--param builtin-string-cmp-inline-length): New option.
97 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
99 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
101 * config/arm/driver-arm.c: Include arm-native.h.
102 (host_detect_local_cpu): Use auto-generated data tables.
103 (vendors, arm_cpu_table): Delete. Move part information to ...
104 * config/arm/arm-cpus.in: ... here.
105 * config/arm/parsecpu.awk (gen_native): New function.
106 (vendor, part): New CPU fields.
107 (END): Add support for building the native CPU detection tables.
108 * config/arm/t-arm (arm-native.h): Add build rule.
109 (driver-arm.o): Add dependency on arm-native.h.
111 2018-07-13 Richard Biener <rguenther@suse.de>
114 * match.pd (addr1 - addr2): Allow either of the operand to
117 2018-07-13 Tom de Vries <tdevries@suse.de>
119 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
122 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
124 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
125 arrays instead of numbered variables.
127 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
129 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
130 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
132 2018-07-13 Richard Biener <rguenther@suse.de>
135 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
136 instead of get_context_die.
138 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
139 Richard Biener <rguenther@suse.de>
142 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
143 that the loop latch destination where phi is defined.
145 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
147 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
148 (riscv_expand_epilogue): Add assertion to check interrupt mode.
149 (riscv_set_current_function): Extract getting interrupt type to new
151 (riscv_get_interrupt_type): New function.
152 (riscv_merge_decl_attributes): New function, checking interrupt type is
154 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
156 2018-07-12 Paul Koning <ni1d@arrl.net>
158 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
161 2018-07-12 Paul Koning <ni1d@arrl.net>
163 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
164 zero reference, add doloop_end instead.
165 * doc/md.texi (decrement_and_branch_until_zero): Remove.
166 (Looping patterns): Remove decrement_and_branch_until_zero. Add
167 detail for doloop_end.
169 2018-07-12 Martin Sebor <msebor@redhat.com>
172 * attribs.c (decl_attributes): Reject conflicting attributes before
173 calling attribute handlers.
175 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
177 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
179 (gcc::dump_manager::get_dump_file_name): likewise.
180 (dump_begin): Likewise.
181 * dumpfile.h (dump_begin): Update prototype.
182 (gcc::dump_manager::get_dump_file_name,
183 gcc::dump_manager::get_dump_file_name): Update prototype.
185 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
187 * internal-fn.h (vectorizable_internal_fn_p): New function.
188 * tree-vect-slp.c (compatible_calls_p): Likewise.
189 (vect_build_slp_tree_1): Remove nops argument. Handle calls
190 to internal functions.
191 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
193 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
195 * fold-const.h (inverse_conditions_p): Declare.
196 * fold-const.c (inverse_conditions_p): New function.
197 * match.pd: Use inverse_conditions_p. Add folds of view_converts
198 that test the inverse condition of a conditional internal function.
199 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
200 * internal-fn.c (internal_fn_mask_index): Handle conditional
202 (vectorized_internal_fn_supported_p): New function.
203 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
204 (any_pred_load_store): Replace with...
205 (need_to_predicate): ...this new variable.
206 (redundant_ssa_names): New variable.
207 (ifcvt_can_use_mask_load_store): Move initial checks to...
208 (ifcvt_can_predicate): ...this new function. Handle tree codes
209 for which a conditional internal function exists.
210 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
211 instead of ifcvt_can_use_mask_load_store. Update after variable
213 (predicate_load_or_store): New function, split out from
214 predicate_mem_writes.
215 (check_redundant_cond_expr): New function.
216 (value_available_p): Likewise.
217 (predicate_rhs_code): Likewise.
218 (predicate_mem_writes): Rename to...
219 (predicate_statements): ...this. Use predicate_load_or_store
220 and predicate_rhs_code.
221 (combine_blocks, tree_if_conversion): Update after above name changes.
222 (ifcvt_local_dce): Handle redundant_ssa_names.
223 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
224 general conditional functions.
225 * tree-vect-stmts.c (vectorizable_call): Likewise.
227 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
228 Alan Hayward <alan.hayward@arm.com>
229 David Sherwood <david.sherwood@arm.com>
231 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
232 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
233 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
234 plus and minus and convert them into IFN_COND_FMA-based sequences.
235 (convert_mult_to_fma): Handle conditional plus and minus.
237 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
239 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
240 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
241 (cond_fnms_optab): New optabs.
242 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
244 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
245 * internal-fn.h (get_conditional_internal_fn): Declare.
246 (get_unconditional_internal_fn): Likewise.
247 * internal-fn.c (cond_ternary_direct): New macro.
248 (expand_cond_ternary_optab_fn): Likewise.
249 (direct_cond_ternary_optab_supported_p): Likewise.
250 (FOR_EACH_COND_FN_PAIR): Likewise.
251 (get_conditional_internal_fn): New function.
252 (get_unconditional_internal_fn): Likewise.
253 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
254 (gimple_match_op::gimple_match_op): Add a new overload for 5
256 (gimple_match_op::set_op): Likewise.
257 (gimple_resimplify5): Declare.
258 * genmatch.c (decision_tree::gen): Generate simplifications for
260 * gimple-match-head.c (gimple_simplify): Define an overload for
261 5 operands. Handle calls with 5 arguments in the top-level overload.
262 (convert_conditional_op): Handle conversions from unconditional
263 internal functions to conditional ones.
264 (gimple_resimplify5): New function.
265 (build_call_internal): Pass a fifth operand.
266 (maybe_push_res_to_seq): Likewise.
267 (try_conditional_simplification): Try converting conditional
268 internal functions to unconditional internal functions.
269 Handle 3-operand unconditional forms.
270 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
271 Define ternary equivalents of the current rules for binary conditional
273 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
275 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
276 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
277 (optab): Handle them.
278 (SVE_COND_FP_TERNARY): New int iterator.
279 (sve_fmla_op, sve_fmad_op): New int attributes.
280 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
281 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
282 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
284 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
286 * target.def (preferred_else_value): New target hook.
287 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
288 * doc/tm.texi: Regenerate.
289 * targhooks.h (default_preferred_else_value): Declare.
290 * targhooks.c (default_preferred_else_value): New function.
291 * internal-fn.h (conditional_internal_fn_code): Declare.
292 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
293 (get_conditional_internal_fn): Use it.
294 (conditional_internal_fn_code): New function.
295 * gimple-match.h (gimple_match_cond): New struct.
296 (gimple_match_op): Add a cond member function.
297 (gimple_match_op::gimple_match_op): Update all forms to take a
299 * genmatch.c (expr::gen_transform): Use the same condition as res_op
300 for the suboperation, but don't specify a particular else_value.
301 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
302 (visit_nary_op, visit_reference_op_load): Pass
303 gimple_match_cond::UNCOND to the gimple_match_op constructor.
304 * gimple-match-head.c: Include tree-eh.h
305 (convert_conditional_op): New function.
306 (maybe_resimplify_conditional_op): Likewise.
307 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
308 (gimple_resimplify2): Likewise.
309 (gimple_resimplify3): Likewise.
310 (gimple_resimplify4): Likewise.
311 (maybe_push_res_to_seq): Return null for conditional operations.
312 (try_conditional_simplification): New function.
313 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
315 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
317 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
319 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
321 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
323 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
325 (hash_tree): Do not hash DECL_FCONTEXT
326 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
327 Do not stream DECL_FCONTEXT.
328 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
329 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
331 2018-07-12 Richard Biener <rguenther@suse.de>
334 * dwarf2out.c (gen_block_die): Only output blocks when they have
335 at least one !DECL_IGNORED_P variable.
337 2018-07-12 Richard Biener <rguenther@suse.de>
340 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
341 Remove -mieee-fp handling.
343 2018-07-12 Richard Biener <rguenther@suse.de>
345 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
346 left-over from last patch.
348 2018-07-12 Jakub Jelinek <jakub@redhat.com>
350 PR tree-optimization/86492
351 * gimple-ssa-store-merging.c
352 (imm_store_chain_info::coalesce_immediate_stores): Call
353 check_no_overlap even for the merge_overlapping case. Formatting fix.
355 2018-07-12 Richard Biener <rguenther@suse.de>
358 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
359 move possibly trapping operations into the conditional.
361 2018-07-12 Richard Biener <rguenther@suse.de>
363 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
364 (vn_lookup_simplify_result): Remove recursion limit applied
366 (vn_nary_build_or_lookup_1): Adjust.
367 (try_to_simplify): Likewise.
368 * gimple-match-head.c (gimple_resimplify1): Instead apply one
370 (gimple_resimplify2): Likewise.
371 (gimple_resimplify3): Likewise.
372 (gimple_resimplify4): Likewise.
374 2018-07-11 Jakub Jelinek <jakub@redhat.com>
376 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
377 Use __mmask64 type instead of __mmask8 for __M argument.
378 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
379 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
381 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
382 __mmask16 for __M argument.
383 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
384 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
385 to __mmask16 instead of __mmask8.
386 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
387 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
388 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
389 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
390 instead of __mmask16 for __U argument.
391 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
392 __mmask16 instead of __mmask8 for __U argument.
393 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
395 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
397 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
399 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
401 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
403 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
405 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
407 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
409 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
411 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
413 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
414 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
415 return type as well as __M argument type and all casts from __mmask8
417 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
418 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
419 return type as well as __M argument type and all casts from __mmask8
421 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
422 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
423 return type as well as __M argument type and all casts from __mmask8
425 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
426 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
427 return type as well as __M argument type and all casts from __mmask8
429 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
430 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
433 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
435 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
436 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
439 2018-07-11 Paul Koning <ni1d@arrl.net>
441 * doc/md.texi (define_subst): Document how multiple occurrences of
442 the same argument in the replacement pattern are handled.
444 2018-07-11 Paul Koning <ni1d@arrl.net>
446 * doc/extend.texi (Common Variable Attributes): Move "mode" into
448 (Common Type Attributes): Add "mode" attribute.
450 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
452 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
453 stream DECL_ORIGINAL_TYPE.
454 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
455 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
456 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
457 Do not walk original type.
458 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
460 (write_ts_decl_non_common_tree_pointers): Do not stream
462 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
463 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
465 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
467 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
468 thread twice from the same starting edge.
470 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
472 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
473 * gimple.c (gimple_call_nonnull_result_p): ...here...
474 (gimple_call_nonnull_arg): ...and here.
475 * gimple.h (gimple_call_nonnull_result_p): New.
476 (gimple_call_nonnull_arg): New.
478 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
480 * config/arm/arm-cpus.in: Move information from fpu field of each
481 cpu definition to the isa field.
482 * config/arm/parsecpu.awk (fpu): Delete match rule.
483 (gen_comm_data): Don't add bits from the CPU's FPU entry.
485 2018-07-11 Richard Biener <rguenther@suse.de>
488 * dwarf2out.c (init_sections_and_labels): Use
489 output_asm_line_debug_info consistently.
490 (dwarf2out_early_finish): Likewise.
491 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
494 2018-07-11 Richard Biener <rguenther@suse.de>
496 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
497 Rework father_bb setting in a way to avoid propagating constants
498 multiple times on a loop body.
500 2018-07-10 Mark Wielaard <mark@klomp.org>
503 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
505 2018-07-10 Richard Biener <rguenther@suse.de>
507 * hash-map.h (hash_map::iterator::operator*): Return
508 references to key and value.
510 2018-07-10 Jakub Jelinek <jakub@redhat.com>
513 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
514 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
515 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
516 (gimplify_omp_for): For composite loops, move outer
517 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
518 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
519 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
520 TREE_LIST for both the original class iterator and the "last" helper
521 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
522 loop, remember has_decl_expr from outer composite loops for the
523 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
525 2018-07-09 Martin Sebor <msebor@redhat.com>
529 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
530 accessing implicitly initialized array elements.
531 * expr.c (string_constant): Handle string initializers of
532 character arrays within aggregates.
533 * gimple-fold.c (fold_array_ctor_reference): Add argument.
534 Store element offset. As a special case, handle zero size.
535 (fold_nonarray_ctor_reference): Same.
536 (fold_ctor_reference): Add argument. Store subobject offset.
537 * gimple-fold.h (fold_ctor_reference): Add argument.
539 2018-07-09 Paul Koning <ni1d@arrl.net>
541 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
542 (pdp11_insn_cost): New function.
543 (pdp11_md_asm_adjust): New function.
544 (TARGET_INVALID_WITHIN_DOLOOP): Define.
545 (pdp11_rtx_costs): Update to match machine better.
546 (output_addr_const_pdp11): Correct format mismatch warnings.
547 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
548 * config/pdp11/pdp11.md: General change to add base_cost and/or
549 length attributes for use by new pdp11_insn_cost function.
550 (MIN_BRANCH): Correct definition.
552 (doloop_end): Use standard pattern name for looping pattern.
553 (doloop_end_nocc): New.
554 (movsf): Add another constraint alternative.
555 (zero_extendqihi2): Add constraint alternatives for not in place
557 (zero_extendhisi2): Remove.
558 (shift patterns): Add CC handling variants.
562 (define_peephole2): New peephole to recognize mov that sets CC for
565 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
567 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
568 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
571 2018-07-09 Jakub Jelinek <jakub@redhat.com>
574 * real.c (real_nextafter): Return true if result is denormal.
576 2018-07-09 Martin Liska <mliska@suse.cz>
578 * common.opt: Add back wrongly removed attribute.
580 2018-07-09 Richard Biener <rguenther@suse.de>
583 * dwarf2out.c (gen_block_die): For an early generated DIE
584 always output high/low PC attributes.
586 2018-07-09 Tom de Vries <tdevries@suse.de>
588 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
589 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
590 onto VAR_DECL with abstract origin.
592 2018-07-07 Jim Wilson <jimw@sifive.com>
594 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
596 2018-07-07 Tom de Vries <tdevries@suse.de>
598 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
600 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
602 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
604 (combine_bound): Use wide-int overflow calculation instead of
606 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
608 * fold-const.c (int_const_binop_2): Same.
609 (extract_muldiv_1): Same.
610 (fold_div_compare): Same.
611 (fold_abs_const): Same.
613 * poly-int.h (add): Same.
617 * predict.c (predict_iv_comparison): Same.
618 * profile-count.c (slow_safe_scale_64bit): Same.
619 * simplify-rtx.c (simplify_const_binary_operation): Same.
620 * tree-chrec.c (tree_fold_binomial): Same.
621 * tree-data-ref.c (split_constant_offset_1): Same.
622 * tree-if-conv.c (idx_within_array_bound): Same.
623 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
624 * tree-ssa-phiopt.c (minmax_replacement): Same.
625 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
626 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
627 * vr-values.c (vr_values::adjust_range_with_scev): Same.
628 * wide-int.cc (wi::add_large): Same.
629 (wi::mul_internal): Same.
630 (wi::sub_large): Same.
631 (wi::divmod_internal): Same.
632 * wide-int.h: Change overflow type to overflow_type for neg, add,
633 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
634 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
635 mul_internal, divmod_internal.
636 (overflow_type): New enum.
637 (accumulate_overflow): New.
639 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
641 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
642 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
644 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
646 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
647 argument is checked for zero before entering loop, avoid checking again.
649 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
651 * gimplify.h (generic_expr_could_trap_p): Set as global function.
652 * gimplify.h (generic_expr_could_trap_p): Likwise.
653 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
655 2018-07-06 Jakub Jelinek <jakub@redhat.com>
657 PR tree-optimization/86401
658 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
659 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
660 (fold_bit_and_mask): ... here. New helper function for match.pd.
661 * fold-const.h (fold_bit_and_mask): Declare.
662 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
664 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
667 * target.def (translate_mode_attribute): New hook.
668 * targhooks.h (default_translate_mode_attribute): Declare.
669 * targhooks.c (default_translate_mode_attribute): New function.
670 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
671 * doc/tm.texi: Regenerate.
672 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
673 (rs6000_translate_mode_attribute): New function.
675 2018-07-06 Paul Koning <ni1d@arrl.net>
677 * doc/md.texi (define_split): Document DONE and FAIL.
678 (define_peephole2): Ditto.
680 2018-07-05 Jeff Law <law@redhat.com>
682 PR tree-optimization/86010
683 * tree-ssa-dse.c (compute_trims): More aggressively trim at
684 both the head and tail of mem* and str* calls.
686 2018-07-05 Jim Wilson <jimw@sifive.com>
688 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
690 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
692 * config/aarch64/aarch64-simd.md: correct flags text for
693 MIN_EXPR replacement.
695 2018-07-05 James Clarke <jrtc27@jrtc27.com>
697 * configure: Regenerated.
699 2018-07-05 Carl Love <cel@us.ibm.com>
701 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
702 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
703 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
704 VSX_BUILTIN_DOUBLEL_V4SF.
706 2018-07-05 Martin Sebor <msebor@redhat.com>
709 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
710 than its domain to compute its the upper bound of a char array.
712 2018-07-05 Nathan Sidwell <nathan@acm.org>
714 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
715 * doc/cpp.texi: Update comment.
716 * doc/tm.texi: Rebuilt.
717 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
718 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
719 * doc/extend.texi (Backwards Compatibility): Clarify it is system
720 headers affected by extern "C".
721 * system.h: Poison NO_IMPLICIT_EXTERN_C.
722 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
723 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
724 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
725 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
726 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
727 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
728 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
729 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
730 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
731 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
732 config/sparc/sp64-elf.h, config/spu/spu.h,
733 config/stormy16/stormy16.h, config/v850/v850.h,
734 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
735 define NO_IMPLICIT_EXTERN_C.
736 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
738 2018-07-05 Tamar Christina <tamar.christina@arm.com>
741 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
742 instead of GET_MODE_SIZE when comparing Units.
744 2018-07-05 Tamar Christina <tamar.christina@arm.com>
747 * rtlanal.c (set_noop_p): Constrain on mode change,
748 include hard-reg-set.h
750 2018-07-05 Tamar Christina <tamar.christina@arm.com>
752 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
754 2018-07-05 Jakub Jelinek <jakub@redhat.com>
757 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
760 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
762 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
764 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
767 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
769 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
771 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
774 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
776 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
778 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
781 2018-07-04 Martin Liska <mliska@suse.cz>
783 * tree-switch-conversion.c: Define
784 max_ratio_for_speed and max_ratio_for_size constants.
786 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
787 Martin Liska <mliska@suse.cz>
792 * common.opt: Rename align options with 'str_' prefix.
793 * common/config/i386/i386-common.c (set_malign_value): New
795 (ix86_handle_option): Use it to set -falign-* options/
796 * config/aarch64/aarch64-protos.h (struct tune_params): Change
797 type from int to string.
798 * config/aarch64/aarch64.c: Update default values from int
800 * config/alpha/alpha.c (alpha_override_options_after_change):
802 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
803 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
804 max skip conditionally.
805 * config/i386/freebsd.h (SUBALIGN_LOG): New.
806 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
807 max skip conditionally.
808 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
809 max skip conditionally.
810 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
811 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
812 max skip conditionally.
813 * config/i386/i386.c (struct ptt): Change type from int to
815 (ix86_default_align): Set default values.
816 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
817 max skip conditionally.
818 * config/i386/iamcu.h (SUBALIGN_LOG): New.
819 (ASM_OUTPUT_MAX_SKIP_ALIGN):
820 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
821 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
822 max skip conditionally.
823 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
824 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
825 * config/i386/x86-64.h (SUBALIGN_LOG): New.
826 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
827 max skip conditionally.
828 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
829 * config/ia64/ia64.c (ia64_option_override): Set default values
830 for alignment options.
831 * config/m68k/m68k.c: Handle new str_align_* options.
832 * config/mips/mips.c (mips_set_compression_mode): Change
834 (mips_option_override): Set default values for options.
835 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
837 * config/rs6000/rs6000.c (rs6000_option_override_internal):
839 * config/rx/rx.c (rx_option_override): Likewise.
840 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
841 (LABEL_ALIGN): Use align_labels_log.
842 (LOOP_ALIGN): Use align_loops_align.
843 * config/s390/s390.c (s390_asm_output_function_label): Use new
845 * config/sh/sh.c (sh_override_options_after_change):
846 Change type of constants.
847 * config/spu/spu.c (spu_sched_init): Likewise.
848 * config/sparc/sparc.c (sparc_option_override): Set default
850 * config/visium/visium.c (visium_option_override): Likewise.
851 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
852 emit p2align format with last argument if it's not needed.
853 * doc/invoke.texi: Document extended format of -falign-*.
854 * final.c: Use align_labels alignment.
855 * flags.h (struct target_flag_state): Change type to use
857 (struct align_flags_tuple): New.
858 (struct align_flags): Likewise.
859 (align_loops_log): Redefine macro to use new types.
860 (align_loops_max_skip): Redefine macro to use new types.
861 (align_jumps_log): Redefine macro to use new types.
862 (align_jumps_max_skip): Redefine macro to use new types.
863 (align_labels_log): Redefine macro to use new types.
864 (align_labels_max_skip): Redefine macro to use new types.
865 (align_functions_log): Redefine macro to use new types.
866 (align_loops): Redefine macro to use new types.
867 (align_jumps): Redefine macro to use new types.
868 (align_labels): Redefine macro to use new types.
869 (align_functions): Redefine macro to use new types.
870 (align_functions_max_skip): Redefine macro to use new types.
871 (align_loops_value): New macro.
872 (align_jumps_value): New macro.
873 (align_labels_value): New macro.
874 (align_functions_value): New macro.
875 * function.c (invoke_set_current_function_hook): Propagate
876 alignment values from flags to global variables default in
878 * ipa-icf.c (sem_function::equals_wpa): Use
879 cl_optimization_option_eq instead of memcmp.
880 * lto-streamer.h (cl_optimization_stream_out): Support streaming
882 (cl_optimization_stream_in): Likewise.
883 * optc-save-gen.awk: Support strings in cl_optimization.
884 * opth-gen.awk: Likewise.
885 * opts.c (finish_options): Remove error checking of invalid
887 (MAX_CODE_ALIGN): Remove.
888 (MAX_CODE_ALIGN_VALUE): Likewise.
889 (parse_and_check_align_values): New function.
890 (check_alignment_argument): Likewise.
891 (common_handle_option): Use check_alignment_argument.
892 * opts.h (parse_and_check_align_values): Declare.
893 * toplev.c (init_alignments): Remove.
894 (read_log_maxskip): New.
895 (parse_N_M): Likewise.
896 (parse_alignment_opts): Likewise.
897 (backend_init_target): Remove usage of init_alignments.
898 * toplev.h (parse_alignment_opts): Declare.
899 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
901 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
902 * tree.c (cl_option_hasher::equal): New.
903 * varasm.c: Use new global macros.
905 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
907 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
908 Use a simpler align directive also if MAXSKIP = ALIGN-1.
909 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
910 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
911 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
912 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
913 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
914 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
915 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
916 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
917 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
918 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
919 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
921 2018-07-04 Martin Liska <mliska@suse.cz>
922 Jonathan Wakely <jwakely@redhat.com>
924 * coverage.c: Use correct type.
925 * doc/invoke.texi: Language correction.
927 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
930 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
931 ENDBRANCH for non-tail call which may return via indirect branch.
932 * doc/extend.texi: Document indirect_return attribute.
934 2018-07-03 Martin Sebor <msebor@redhat.com>
936 PR tree-optimization/86274
937 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
939 (format_floating): Correct handling of infinities and NaNs.
941 2018-07-03 Martin Sebor <msebor@redhat.com>
943 * print-tree.c (print_real_cst): New function.
944 (print_node_brief): Call it.
947 2018-07-03 Jeff Law <law@redhat.com>
949 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
950 into a single pattern.
952 * config/h8300/h8300.md (ors code_iterator): New.
953 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
954 a single pattern and single splitter.
955 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
956 (iorqi3_1, xorqi3_1): Likewise.
957 (iorqi3, xorqi3 expanders): Similarly.
959 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
960 (movmd_internal) into a single pattern using the P mode iterator.
961 (movmd splitters): Similarly.
962 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
963 (movsd splitters): Similarly.
965 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
966 ADDB, ADDW and ADDL into a single ADD attribute which selects the
967 right table based on the size of the operand.
968 * config/h8300/h8300.md (length_table): Corresponding changes. All
969 references to "addb", "addw" and "addl" changed to "add".
970 (btst patterns): Merge two variants into a single pattern.
971 (tstqi, tsthi): Likewise.
972 (addhi3_incdec, addsi3_incdec): Likewise.
973 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
974 (mulhi3, mulsi3): Likewise.
975 (udivhi3, udivsi3): Likewise.
976 (divhi3, divsi3): Likewise.
977 (andorqi3, andorhi3, andorsi3): Likewise.
979 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
982 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
983 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
985 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
987 PR tree-optimization/85694
988 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
989 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
991 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
992 (<u>avg<mode>3_ceil): New patterns.
994 2018-07-03 David Malcolm <dmalcolm@redhat.com>
996 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
997 scan-tree-dump directive.
998 * gcc.dg/vect/slp-perm-2.c: Likewise.
999 * gcc.dg/vect/slp-perm-3.c: Likewise.
1000 * gcc.dg/vect/slp-perm-5.c: Likewise.
1001 * gcc.dg/vect/slp-perm-6.c: Likewise.
1002 * gcc.dg/vect/slp-perm-7.c: Likewise.
1003 * gcc.dg/vect/slp-perm-8.c: Likewise.
1005 2018-07-03 Marek Polacek <polacek@redhat.com>
1008 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1010 2018-07-03 Richard Biener <rguenther@suse.de>
1013 * tree-ssa-structalias.c (find_func_clobbers): Properly
1014 handle indirect calls.
1016 2018-07-03 Jeff Law <law@redhat.com>
1018 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1019 (shifts): New code iterator.
1020 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1021 expander. Fix HImode handling on H8/SX.
1022 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1023 (subqi3, subhi3, subsi3 expanders): Likewise.
1024 (andqi3, andhi3, andsi3 expanders): Likewise.
1025 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1026 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1027 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1028 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1029 (zero_extendqihi2, zero_extendqisi2): Likewise.
1030 (extendqihi2, extendqisi2): Likewise.
1031 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1032 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1033 (rotlqi3_1, rotlhi3_1): Likewise.
1034 (logicalhi3_sn, logicalsi3_sn): Likewise.
1035 (logicalhi3, logicalsi3): Likewise.
1037 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1039 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1040 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1041 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1042 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1043 type to append_pattern_def_seq instead of creating a stmt_vec_info
1045 (build_mask_conversion): Likewise. Remove vinfo argument.
1046 (vect_add_conversion_to_patterm): Likewise, renaming to...
1047 (vect_add_conversion_to_pattern): ...this.
1048 (vect_recog_mask_conversion_pattern): Update call to
1049 build_mask_conversion. Pass the vector type to
1050 append_pattern_def_seq here too.
1051 (vect_recog_gather_scatter_pattern): Update call to
1052 vect_add_conversion_to_pattern.
1054 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1056 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1057 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1058 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1059 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1060 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1061 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1062 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1063 append_pattern_def_seq instead of new_pattern_def_seq.
1064 (vect_recog_divmod_pattern): Do both of the above.
1065 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1068 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1070 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1071 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1072 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1073 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1074 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1075 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1076 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1077 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1078 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1079 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1080 parameter with a single stmt_vec_info.
1081 (vect_recog_func_ptr): Likewise.
1082 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1083 (vect_try_gather_scatter_pattern): ...this.
1084 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1085 the stmt_vec_info of the statement to be matched. Don't clear
1086 STMT_VINFO_RELATED_STMT.
1087 (vect_pattern_recog): Update call accordingly.
1089 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1091 PR tree-optimization/85694
1092 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1093 (uavgM3_ceil): Document new optabs.
1094 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1095 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1097 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1098 (savg_ceil_optab): New optabs.
1099 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1100 (vect_vect_recog_func_ptrs): Add it.
1101 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1102 constant directly from the associated lhs.
1104 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1106 * tree-vect-patterns.c (vect_split_statement): New function.
1107 (vect_convert_input): Use it to try to split an existing cast.
1109 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1111 * poly-int.h (print_hex): New function.
1112 * dumpfile.h (dump_dec, dump_hex): Declare.
1113 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1114 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1115 min_input_precision, operation_precision and operation_sign.
1116 * tree-vect-patterns.c (vect_get_range_info): New function.
1117 (vect_same_loop_or_bb_p, vect_single_imm_use)
1118 (vect_operation_fits_smaller_type): Delete.
1119 (vect_look_through_possible_promotion): Add an optional
1120 single_use_p parameter.
1121 (vect_recog_over_widening_pattern): Rewrite to use new
1122 stmt_vec_info infomration. Handle one operation at a time.
1123 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1124 (vect_truncatable_operation_p, vect_set_operation_type)
1125 (vect_set_min_input_precision): New functions.
1126 (vect_determine_min_output_precision_1): Likewise.
1127 (vect_determine_min_output_precision): Likewise.
1128 (vect_determine_precisions_from_range): Likewise.
1129 (vect_determine_precisions_from_users): Likewise.
1130 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1131 (vect_vect_recog_func_ptrs): Put over_widening first.
1133 (vect_pattern_recog): Call vect_determine_precisions.
1135 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1137 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1138 statements that have been replaced by further pattern statements.
1139 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1141 2018-07-03 Richard Biener <rguenther@suse.de>
1143 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1144 always set *dt. Dump vectype in vectype overload.
1145 * dumpfile.h (dump_gimple_expr): New function.
1146 (dump_gimple_expr_loc): Likewise.
1147 * dumpfile.c (dump_gimple_expr): New function.
1148 (dump_gimple_expr_loc): Likewise.
1150 2018-07-02 Jeff Law <law@redhat.com>
1152 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1153 the H8/300, H8/300H and H8/S variants into a single pattern.
1154 (movhi_h8300, movqi_h8300hs): Similarly.
1155 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1156 (QHI mode iterator): New.
1158 * config/h8300/h8300.md: Remove trailing whitespace.
1160 2018-07-02 Jim Wilson <jimw@sifive.com>
1162 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1163 instead of emit_insn for interrupt returns.
1164 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1165 (riscv_sret, riscv_uret): Likewise.
1167 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1169 * pretty-print.c (selftest::test_pp_format): Move save and restore
1170 of quotes to class auto_fix_quotes, and add an instance.
1171 * selftest.c: Include "intl.h".
1172 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1173 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1174 * selftest.h (selftest::auto_fix_quotes): New class.
1176 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1178 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1179 (aarch64_sve_prepare_conditional_op): Remove.
1180 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1181 Allow aarch64_simd_reg_or_zero as select operand; remove
1182 the aarch64_sve_prepare_conditional_op call.
1183 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1184 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1185 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1186 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1187 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1188 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1189 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1190 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1191 and a splitters to match all of the *_any patterns.
1192 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1194 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1195 (SVE_COND_FP_BINARY_REV): Remove.
1196 (sve_int_op_rev, sve_fp_op_rev): New.
1197 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1198 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1199 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1200 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1201 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1202 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1203 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1204 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1205 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1207 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1208 Remove match_dup 1 from the inner unspec.
1209 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1211 * config/aarch64/aarch64.md (movprfx): New attr.
1212 (length): Default movprfx to 8.
1213 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1214 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1215 (*<su>mul<SVE_I>3_highpart): Likewise.
1216 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1217 (*v<ASHIFT><SVE_I>3): Likewise.
1218 (*<su><MAXMIN><SVE_I>3): Likewise.
1219 (*<su><MAXMIN><SVE_F>3): Likewise.
1220 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1221 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1222 (*div<SVE_F>4): Likewise.
1224 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1226 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1229 2018-07-02 Richard Biener <rguenther@suse.de>
1231 PR tree-optimization/86363
1232 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1233 memset argument refers to a non-variable address.
1235 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1237 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1238 {PLUS,MINUS}_EXPR code to...
1239 (adjust_symbolic_bound): ...here,
1240 (combine_bound): ...here,
1241 (set_value_range_with_overflow): ...and here.
1243 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1245 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1247 (extract_range_from_abs_expr): ...here.
1249 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1251 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1252 -fno-omit-frame-pointer when not optimizing.
1254 2018-07-02 Martin Liska <mliska@suse.cz>
1257 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1258 (propagate_nothrow): Likewise.
1260 2018-07-02 Martin Liska <mliska@suse.cz>
1263 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1265 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1267 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1268 function in r262149, changing "loc" param from source_location to
1269 const dump_location_t &.
1270 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1271 declaration, as above.
1273 2018-07-01 Paul Koning <ni1d@arrl.net>
1275 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1276 -munit-asm, -mgnu-asm, -mdec-asm.
1277 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1278 (pdp11_output_labelref): New.
1279 (pdp11_output_def): New.
1280 (pdp11_output_addr_vec_elt): New.
1281 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1282 %# and %@ format codes.
1283 (pdp11_option_override): New.
1284 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1285 (pdp11_output_ident): New.
1286 (pdp11_asm_named_section): New.
1287 (pdp11_asm_init_sections): New.
1288 (pdp11_file_start): New.
1289 (pdp11_file_end): New.
1290 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1291 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1292 %o, like %c but octal.
1293 (pdp11_option_override): New.
1294 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1296 (DATA_SECTION_ASM_OP): Ditto.
1297 (READONLY_DATA_SECTION_ASM_OP): New.
1298 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1299 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1300 (ASM_OUTPUT_LABELREF): Ditto.
1301 (ASM_OUTPUT_DEF): Ditto.
1302 (ASM_OUTPUT_EXTERNAL): New.
1303 (ASM_OUTPUT_SOURCE_FILENAME): New.
1304 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1305 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1306 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1307 %# and %@ format codes.
1308 * config/pdp11/pdp11.opt (mgnu-asm): New.
1309 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1310 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1311 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1313 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1315 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1316 dereferencing path[] beyond its length.
1318 (debug_all_paths): New.
1319 (rewire_first_differing_edge): New.
1320 (adjust_paths_after_duplication): New.
1321 (duplicate_thread_path): Call adjust_paths_after_duplication.
1323 (thread_through_all_blocks): Add new argument to
1324 duplicate_thread_path.
1326 2018-06-30 Jim Wilson <jimw@sifive.com>
1328 * config/riscv/predicates.md (p2m1_shift_operand): New.
1329 (high_mask_shift_operand): New.
1330 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1331 pattern using p2m1_shift_operand.
1332 (lshsi3_zero_extend_3+2): New combiner pattern using
1333 high_mask_shift_operand.
1335 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1337 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1339 (vect_recog_rotate_pattern): ...here.
1340 (vect_convert_input): Try to insert casts of invariants in the
1342 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1343 preheader to be empty.
1345 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1347 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1348 vector type. If given, install it in the new statement's
1350 (vect_element_precision): New function.
1351 (vect_unpromoted_value): New struct.
1352 (vect_unpromoted_value::vect_unpromoted_value): New function.
1353 (vect_unpromoted_value::set_op): Likewise.
1354 (vect_look_through_possible_promotion): Likewise.
1355 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1356 (vect_widened_op_tree, vect_convert_input): Likewise.
1357 (vect_convert_inputs, vect_convert_output): Likewise.
1358 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1359 to handle the optional cast of the multiplication result and
1360 vect_widened_op_tree to detect the widened multiplication itself.
1361 Do not require the input and output of promotion casts to have
1362 the same sign, but base the signedness of the operation on the
1363 input rather than the result. If the pattern includes two
1364 promotions, check that those promotions have the same sign.
1365 Do not restrict the MULT_EXPR handling to a double-width result;
1366 handle quadruple-width results and wider. Use vect_convert_inputs
1367 to convert the inputs to the common type.
1368 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1369 to handle the optional cast of the ABS result. Also allow a sign
1370 change or a sign extension between the ABS and MINUS.
1371 Use vect_widened_op_tree to detect the widened subtraction and use
1372 vect_convert_inputs to convert the inputs to the common type.
1373 (vect_handle_widen_op_by_const): Delete.
1374 (vect_recog_widen_op_pattern): New function.
1375 (vect_recog_widen_mult_pattern): Use it.
1376 (vect_recog_widen_shift_pattern): Likewise.
1377 (vect_recog_widen_sum_pattern): Use
1378 vect_look_through_possible_promotion to handle the promoted
1381 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1383 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1384 the containing gimple_seq *.
1385 * gimple-iterator.h (gsi_for_stmt): Declare it.
1386 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1387 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1388 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1389 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1390 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1392 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1394 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1395 statement being replaced is part of an existing pattern
1396 definition sequence, inserting the new pattern statements before
1398 (vect_pattern_recog_1): Don't return a bool. If the statement
1399 is already part of a pattern, instead apply pattern matching
1400 to the pattern definition statements. Don't clear the
1401 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1402 (vect_pattern_recog): Don't break after the first match;
1403 continue processing the pattern definition statements instead.
1404 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1406 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1408 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1409 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1410 (vect_recog_widen_sum_pattern): Use it.
1412 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1414 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1415 phi is not a pattern statement and has not been replaced by
1416 a pattern statement.
1417 * tree-vect-patterns.c (type_conversion_p): Don't check
1418 STMT_VINFO_IN_PATTERN_P.
1419 (vect_recog_vector_vector_shift_pattern): Likewise.
1420 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1421 the pattern statement rather than the original statement; check
1422 directly for a WIDEN_MULT_EXPR here.
1423 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1424 vect_is_simple_use to return the pattern statement rather
1425 than the original statement; use is_pattern_stmt_p to check
1426 for such a pattern statement.
1427 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1428 to return the pattern statement rather than the original statement;
1429 don't do the same transformation here.
1430 (vect_is_simple_use): If the defining statement has been replaced
1431 by a pattern statement, return the pattern statement instead.
1432 Remove the corresponding (local) transformation from the vectype
1435 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1437 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1438 end and default to null.
1439 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1440 (vectorizable_reduction): Update calls accordingly, dropping the
1441 gimple ** argument if the passed-back statement isn't needed.
1442 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1443 (vect_recog_rotate_pattern): Likewise.
1444 (vect_recog_mask_conversion_pattern): Likewise.
1445 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1446 (vect_mask_constant_operand_p): Likewise.
1447 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1448 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1449 (get_group_load_store_type, get_load_store_type): Likewise.
1450 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1451 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1452 (vectorizable_conversion, vectorizable_assignment): Likewise.
1453 (vectorizable_shift, vectorizable_operation): Likewise.
1454 (vectorizable_store, vect_is_simple_cond): Likewise.
1455 (vectorizable_condition, vectorizable_comparison): Likewise.
1456 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1457 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1458 and move it to the end. Cope with null def_stmt_outs.
1460 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1462 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1464 2018-06-29 Jeff Law <law@redhat.com>
1466 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1467 displacements for TARGET_V850E2V3 and newer.
1468 (TARGET_LRA_P): Remove. Defaults to LRA now.
1469 * config/v850/v850.md (sign23byte_load): Remove.
1470 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1471 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1473 2018-06-29 Martin Liska <mliska@suse.cz>
1476 * coverage.c (coverage_init): Mangle full path name.
1477 * doc/invoke.texi: Document the change.
1478 * gcov-io.c (mangle_path): New.
1479 * gcov-io.h (mangle_path): Likewise.
1480 * gcov.c (mangle_name): Use mangle_path for path mangling.
1482 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1484 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1485 if starting source register is not even.
1487 2018-06-29 Martin Liska <mliska@suse.cz>
1489 PR tree-optimization/86263
1490 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1491 Make edge redirection.
1493 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1495 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1496 (dump_scope_depth): New variable.
1497 (get_dump_scope_depth): New function.
1498 (dump_begin_scope): New function.
1499 (dump_end_scope): New function.
1500 * dumpfile.h (get_dump_scope_depth): New declaration.
1501 (dump_begin_scope): New declaration.
1502 (dump_end_scope): New declaration.
1503 (class auto_dump_scope): New class.
1504 (AUTO_DUMP_SCOPE): New macro.
1505 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1508 2018-06-29 Richard Biener <rguenther@suse.de>
1510 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1511 compute_all_dependences succeeds.
1512 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1513 exceed --param loop-max-datarefs-for-datadeps.
1515 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1517 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1519 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1522 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1523 alternative 0 in preferred_for_speed attribute.
1525 2018-06-28 Paul Koning <ni1d@arrl.net>
1527 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1528 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1529 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1530 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1531 for shift insn patterns.
1533 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1535 * cgraph.c (cgraph_node::get_body): Replace assignments to
1536 "dump_file" with calls to set_dump_file.
1537 * dumpfile.c (alt_dump_file): Make static, and group with...
1538 (alt_flags): ...this definition.
1539 (dumps_are_enabled): New variable.
1540 (refresh_dumps_are_enabled): New function.
1541 (set_dump_file): New function.
1542 (set_alt_dump_file): New function.
1543 (gcc::dump_manager::dump_start): Replace assignments to
1544 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1546 (gcc::dump_manager::dump_finish): Likewise.
1547 * dumpfile.h (alt_dump_file): Delete decl.
1548 (dumps_are_enabled): New variable decl.
1549 (set_dump_file): New function decl.
1550 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1552 * tree-nested.c (lower_nested_functions): Replace assignments to
1553 "dump_file" with calls to set_dump_file.
1555 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1557 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1558 goto_locus of each outgoing edge of each basic block.
1560 2018-06-28 Richard Biener <rguenther@suse.de>
1562 * dwarf2out.c (decl_scope_table): Remove.
1563 (push_decl_scope): Likewise.
1564 (pop_decl_scope): Likewise.
1565 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1566 (gen_struct_or_union_type_die): Likewise.
1567 (gen_tagged_type_die): Likewise.
1568 (dwarf2out_init): Do not initialize decl_scope_table.
1569 (dwarf2out_c_finalize): Do not free it.
1571 2018-06-28 Richard Biener <rguenther@suse.de>
1573 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1574 deciding whether to not re-use a DIE.
1576 2018-06-28 Richard Biener <rguenther@suse.de>
1578 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1579 DW_AT_abstract_origin attribute.
1581 2018-06-28 Martin Liska <mliska@suse.cz>
1583 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1584 Use newly introduced constants.
1585 * tree-switch-conversion.h (struct jump_table_cluster):
1586 Define max_ratio_for_size and max_ratio_for_speed.
1588 2018-06-28 Martin Liska <mliska@suse.cz>
1590 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1591 Add new checking assert to catch invalid state.
1592 (jump_table_cluster::can_be_handled): Handle single case
1594 (jump_table_cluster::is_beneficial): Bail out for such case.
1595 (bit_test_cluster::find_bit_tests):
1596 Add new checking assert to catch invalid state.
1597 (bit_test_cluster::can_be_handled): Handle single case
1599 (bit_test_cluster::is_beneficial): Bail out for such case.
1600 (switch_decision_tree::analyze_switch_statement):
1603 2018-06-28 Martin Liska <mliska@suse.cz>
1605 * common.opt: Introduce -completion option.
1606 * gcc.c (driver_handle_option): Handle it.
1607 (driver::main): Print completions if completion
1609 * opt-suggestions.c (option_proposer::get_completions):
1611 (option_proposer::suggest_completion): Likewise.
1612 (option_proposer::find_param_completions): Likewise.
1613 (verify_autocompletions): Likewise.
1614 (test_completion_valid_options): Likewise.
1615 (test_completion_valid_params): Likewise.
1616 (in_completion_p): Likewise.
1617 (empty_completion_p): Likewise.
1618 (test_completion_partial_match): Likewise.
1619 (test_completion_garbage): Likewise.
1620 (opt_proposer_c_tests): Likewise.
1621 * opt-suggestions.h: Declare new functions.
1622 * opts.c (common_handle_option): Handle OPT__completion_.
1623 * selftest-run-tests.c (selftest::run_tests): Add
1624 opt_proposer_c_tests.
1625 * selftest.c (assert_str_startswith): New.
1626 * selftest.h (assert_str_startswith): Likewise.
1627 (opt_proposer_c_tests): New.
1628 (ASSERT_STR_STARTSWITH): Likewise.
1630 2018-06-28 Martin Liska <mliska@suse.cz>
1632 * Makefile.in: Add opt-suggestions.o.
1633 * gcc-main.c: Include opt-suggestions.h.
1634 * gcc.c (driver::driver): Likewise.
1635 (driver::~driver): Remove m_option_suggestions.
1636 (driver::build_option_suggestions): Moved to option_proposer.
1637 (driver::suggest_option): Likewise.
1638 (driver::handle_unrecognized_options): Use option_proposer.
1639 * gcc.h (class driver): Add new memver m_option_proposer.
1640 * opt-suggestions.c: New file.
1641 * opt-suggestions.h: New file.
1643 2018-06-28 Martin Liska <mliska@suse.cz>
1645 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1646 (auto_string_vec::~auto_string_vec): Likewise.
1648 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1650 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1651 prevent_decl_creation_for_types fields up and add reset_location field.
1652 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1653 statement if id->reset_location is true.
1654 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1655 id->reset_location is true.
1656 (copy_phis_for_bb): Force input_location on the arguments if
1657 id->reset_location is true.
1658 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1659 is set on the function to be inlined.
1661 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1663 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1665 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1667 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1668 registers for Pmode.
1669 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1670 hard registers for the clobbered pseudo.
1672 2018-06-27 Paul Koning <ni1d@arrl.net>
1674 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1675 mutually exclusive options.
1676 * config/pdp11/constraints.md (h): New constraint.
1677 (O): Update definition to match shift code generation.
1678 (D): New constraint.
1679 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1681 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1683 (output_jump): Change arguments.
1684 (pdp11_fixed_cc_regs): New function.
1685 (pdp11_cc_mode): Ditto.
1686 (pdp11_expand_shift): Ditto.
1687 (pdp11_assemble_shift): Ditto.
1688 (pdp11_small_shift): Ditto.
1689 (pdp11_branch_cost): Remove.
1690 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1692 (pdp11_register_move_cost): Update for CC registers.
1693 (pdp11_rtx_costs): Add case for LSHIFTRT.
1694 (pdp11_output_jump): Add CCNZ mode conditional branches.
1695 (notice_update_cc_on_set): Remove.
1696 (pdp11_cc_mode): New function.
1697 (simple_memory_operand): Correct pre/post decrement case.
1698 (no_side_effect_operand): New function.
1699 (pdp11_regno_reg_class): Add CC_REGS class.
1700 (pdp11_fixed_cc_regs): New function.
1701 (pdp11_small_shift): New function.
1702 (pdp11_expand_shift): New function to expand shift insns.
1703 (pdp11_assemble_shift): New function to output shifts.
1704 (pdp11_branch_cost): Remove.
1705 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1706 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1707 (WCHAR_TYPE): Ditto.
1708 (PTRDIFF_TYPE): Ditto.
1709 (ADJUST_INSN_LENGTH): New macro.
1710 (FIXED_REGISTERS): Add CC registers.
1711 (CALL_USED_REGISTERS): Ditto.
1713 (REG_CLASS_NAMES): Ditto.
1714 (REG_CLASS_CONTENTS): Ditto.
1715 (SELECT_CC_MODE): Use new function.
1716 (TARGET_FLAGS_REGNUM): New macro.
1717 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1718 (cc0_reg_rtx): Remove.
1719 (CC_STATUS_MDEP): Remove.
1720 (CC_STATUS_MDEFP_INIT): Remove.
1721 (CC_IN_FPU): Remove.
1722 (NOTICE_UPDATE_CC): Remove.
1723 (REGISTER_NAMES): Add CC registers.
1724 (BRANCH_COST): Change to constant 1.
1725 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1727 * config/pdp11/pdp11.opt (mbcopy): Remove.
1728 (mbcopy-builtin): Remove.
1729 (mbranch-cheap): Remove.
1730 (mbranch-expensive): Remove.
1731 * config/pdp11/predicates.md (expand_shift_operand): Update to
1732 match shift code generation.
1733 (ccnz_operator): New predicate.
1734 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1735 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1736 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1738 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1739 description of O constraint.
1741 2018-06-27 Jeff Law <law@redhat.com>
1742 Austin Law <austinklaw@gmail.com>
1744 * config/v850/v850.md (addsi3_set_flags): New pattern.
1745 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1746 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1747 (zero_extendhisi2_v850_set_flags): Likewise.
1748 (zero_extendqisi2_v850_set_flags): Likewise.
1749 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1750 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1751 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1753 * config/v850/v850-protos.h (notice_update_cc): Remove.
1754 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1755 (v850_print_operand): Handle 'D' and "d".
1756 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1757 Add handling of arithmetic/logical operations compared against zero.
1758 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1759 Do not look at v850_compare_op, instead get mode from last argument.
1760 (v850_gen_compare): Remove
1761 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1762 after reload for prologue insns.
1763 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1765 (construct_save_jarl): Likewise.
1766 (TARGET_FLAGS_REGNUM): Define.
1767 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1768 (NOTICE_UPDATE_CC): Remove.
1769 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1770 than cc0. Conditionalize on reload_completed.
1771 (cmpsi_insn, setfcc_insn): Likewise.
1772 (tst1 splitter): Turn into define_and_split which sets the flags
1774 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1775 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1776 (cstoresf4, cstoredf4): Clobber the flags.
1777 (cmpsi, cmpsf, cmpdf): Remove expanders.
1778 (setf_insn): Remove pattern.
1779 (addsi3): Turn into define_and_split which clobbers the flags after
1780 reload and a suitable pattern (addsi3_clobber_flags) for use after
1782 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1783 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1784 (ashrsi3, ashrsi3_v850e2): Likewise.
1785 (bins): Clobber the flags.
1786 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1787 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1788 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1789 (call_value_internal_short, call_value_internal_long): Likewise.
1790 (callt_save_interrupt, callt_return_interrupt): Likewise.
1791 (save_interrupt, return_interrupt): Likewise.
1792 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1793 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1794 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1795 (All FP comparisons): Only allow after reload has completed.
1797 (divh, divhu): Tweak output template.
1798 (branch_z_normal, branch_z_invert): Remove
1799 (branch_nz_normal, branch_nz_invert): Likewise.
1800 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1802 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1803 * config/v850/v850.c (notice_update_cc): Remove.
1804 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1805 (CC_NO_CARRY): Likewise.
1806 (NOTICE_UPDATE_CC): Define to nothing.
1807 * config/v850/v850.md: Remove block comment on cc0 handling
1808 Remove "cc" attribute from all patterns. Remove cc_status handling
1809 from all patterns. Minor formatting fixes.
1811 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1813 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1814 (cortex-a76.cortex-a55): Likewise.
1815 * config/aarch64/aarch64-tune.md: Regenerate.
1816 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1817 cortex-a76.cortex-a55.
1819 2018-06-27 Jeff Law <law@redhat.com>
1821 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1822 (MULTILIB_DIRNAMES): Similarly.
1824 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1826 * gimple.h (gimple_return_retbnd): Delete.
1827 (gimple_return_set_retbnd): Likewise.
1828 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1829 gimple_return_set_retbnd.
1830 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1831 gimple_return_retbnd and adjust.
1832 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1833 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1834 Explicitly return NULL in a couple more cases. Move assertion
1835 on debug statements and remove unreachable code.
1836 (reset_debug_binding): Do not test id->retbnd.
1837 (expand_call_inline): Do not set it.
1839 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1841 * configure.ac: Add --disable-gcov option.
1842 * configure: Regenerate.
1843 * Makefile.in: Honour @enable_gcov@.
1844 * doc/install.texi: Document --disable-gcov.
1846 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1848 * config/arm/arm-cpus.in (cortex-a76): New entry.
1849 (cortex-a76.cortex-a55): Likewise.
1850 * config/arm/arm-tables.opt: Regenerate.
1851 * config/arm/arm-tune.md: Likewise.
1852 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1853 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1854 cortex-a76.cortex-a55.
1856 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1859 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1861 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1863 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1865 (EPILOGUE_USES): Likewise.
1867 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1869 * tree-inline.c (remap_location): New function extracted from...
1870 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1871 (copy_phis_for_bb): ...here. Call remap_location.
1872 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1874 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1876 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1877 unaligned vsx for 16B memset.
1879 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1882 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1883 ieee128_float_type_node to long_double_type_node unless
1884 TARGET_LONG_DOUBLE_128 is set.
1886 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1888 * cfgloop.c (get_loop_location): Convert return type from
1889 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1890 by implicit construction from rtx_insn *, and using
1891 dump_user_location_t::from_function_decl for the fallback case.
1892 * cfgloop.h (get_loop_location): Convert return type from
1893 location_t to dump_user_location_t.
1894 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1895 dump_printf_loc to pass in a dump_location_t rather than a
1896 location_t, via the gimple stmt.
1897 * coverage.c (get_coverage_counts): Update calls to
1898 dump_printf_loc to pass in dump_location_t rather than a
1900 * doc/optinfo.texi (Dump types): Convert example of
1901 dump_printf_loc from taking "locus" to taking "insn". Update
1902 description of the "_loc" calls to cover dump_location_t.
1903 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1905 (dump_user_location_t::dump_user_location_t): New constructors,
1906 from gimple *stmt and rtx_insn *.
1907 (dump_user_location_t::from_function_decl): New function.
1908 (dump_loc): Make static.
1909 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1910 const dump_location_t &.
1911 (dump_generic_expr_loc): Delete.
1912 (dump_printf_loc): Convert param "loc" from location_t to
1913 const dump_location_t &.
1914 (selftest::test_impl_location): New function.
1915 (selftest::dumpfile_c_tests): New function.
1916 * dumpfile.h: Include "profile-count.h".
1917 (class dump_user_location_t): New class.
1918 (struct dump_impl_location_t): New struct.
1919 (class dump_location_t): New class.
1920 (dump_printf_loc): Convert 2nd param from source_location to
1921 const dump_location_t &.
1922 (dump_generic_expr_loc): Delete.
1923 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1924 const dump_location_t &.
1925 * gimple-fold.c (fold_gimple_assign): Update call to
1926 dump_printf_loc to pass in a dump_location_t rather than a
1927 location_t, via the gimple stmt.
1928 (gimple_fold_call): Likewise.
1929 * gimple-loop-interchange.cc
1930 (loop_cand::analyze_iloop_reduction_var): Update for change to
1931 check_reduction_path.
1932 (tree_loop_interchange::interchange): Update for change to
1934 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1935 change in return-type of find_loop_location.
1936 (graphite_regenerate_ast_isl): Likewise.
1937 * graphite-optimize-isl.c (optimize_isl): Likewise.
1938 * graphite.c (graphite_transform_loops): Likewise.
1939 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1940 pass in a dump_location_t rather than a location_t, via the
1942 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1943 * ipa.c (walk_polymorphic_call_targets): Likewise.
1944 * loop-unroll.c (report_unroll): Convert "locus" param from
1945 location_t to dump_location_t.
1946 (decide_unrolling): Update for change to get_loop_location's
1948 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1949 location_t to dump_user_location_t.
1950 (grid_find_single_omp_among_assignments_1): Updates calls to
1951 dump_printf_loc to pass in a dump_location_t rather than a
1952 location_t, via the gimple stmt.
1953 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1954 location_t to dump_location_t. Updates calls to dump_printf_loc
1955 to pass in a dump_location_t rather than a location_t, via the
1957 (grid_inner_loop_gridifiable_p): Likewise.
1958 (grid_dist_follows_simple_pattern): Likewise.
1959 (grid_gfor_follows_tiling_pattern): Likewise.
1960 (grid_target_follows_gridifiable_pattern): Likewise.
1961 (grid_attempt_target_gridification): Convert initialization
1962 of local "grid" from memset to zero-initialization; FIXME: does
1963 this require C++11? Update call to dump_printf_loc to pass in a
1964 optinfo_location rather than a location_t, via the gimple stmt.
1965 * profile.c (read_profile_edge_counts): Updates call to
1966 dump_printf_loc to pass in a dump_location_t rather than a
1968 (compute_branch_probabilities): Likewise.
1969 * selftest-run-tests.c (selftest::run_tests): Call
1971 * selftest.h (dumpfile_c_tests): New decl.
1972 * tree-loop-distribution.c (pass_loop_distribution::execute):
1973 Update for change in return type of find_loop_location.
1974 * tree-parloops.c (parallelize_loops): Likewise.
1975 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1976 "locus" from location_t to dump_user_location_t.
1977 (canonicalize_loop_induction_variables): Likewise.
1978 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1979 for change in return type of find_loop_location.
1980 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1981 to dump_printf_loc to pass in a dump_location_t rather than a
1982 location_t, via the stmt.
1983 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1985 * tree-vect-loop-manip.c (find_loop_location): Convert return
1986 type from source_location to dump_user_location_t.
1987 (vect_do_peeling): Update for above change.
1988 (vect_loop_versioning): Update for change in type of
1990 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1991 from location_t to dump_user_location_t.
1992 (vect_estimate_min_profitable_iters): Update for change in type
1994 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1995 location_t to dump_location_t.
1996 (vect_slp_bb): Update for change in type of vect_location.
1997 * tree-vectorizer.c (vect_location): Convert from source_location
1998 to dump_user_location_t.
1999 (try_vectorize_loop_1): Update for change in vect_location's type.
2000 (vectorize_loops): Likewise.
2001 (increase_alignment): Likewise.
2002 * tree-vectorizer.h (vect_location): Convert from source_location
2003 to dump_user_location_t.
2004 (find_loop_location): Convert return type from source_location to
2005 dump_user_location_t.
2006 (check_reduction_path): Convert 1st param from location_t to
2007 dump_user_location_t.
2008 * value-prof.c (check_counter): Update call to dump_printf_loc to
2009 pass in a dump_user_location_t rather than a location_t; update
2010 call to error_at for change in type of "locus".
2011 (check_ic_target): Update call to dump_printf_loc to
2012 pass in a dump_user_location_t rather than a location_t, via the
2015 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2017 * config/s390/s390.h (enum processor_flags): Do not use
2018 default tune parameter when -march was specified.
2020 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2023 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2024 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2027 2018-06-26 Richard Biener <rguenther@suse.de>
2029 PR tree-optimization/86287
2031 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2032 (vect_analyze_loop): Initialize n_stmts.
2034 2018-06-26 Richard Biener <rguenther@suse.de>
2037 * fold-const.c (fold_convertible_p): Pointer extension
2040 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2043 * dwarf2out.c (loc_list_has_views): Adjust comments.
2044 (dw_loc_list): Split single cross-partition range with
2047 2018-06-25 Jeff Law <law@redhat.com>
2049 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2050 on -mbig-switch by default.
2052 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2054 (const_float_0_operand): Remove unused predicate.
2055 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2056 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2057 (recipsf2): New expander. Original pattern now called
2059 (recipdf2, recipdf2_insn): Similarly.
2060 (rsqrtsf2, rsqrtsf2_insn): Similarly
2061 (rsqrtdf2, rsqrtdf2_insn): Similarly
2063 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2065 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2066 Simplify logic for FreeBSD (twice).
2068 2018-06-25 Martin Sebor <msebor@redhat.com>
2070 PR tree-optimization/86204
2071 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2072 a strnlen result if it's less than the length of the string.
2074 2018-06-25 Martin Sebor <msebor@redhat.com>
2076 PR tree-optimization/85700
2077 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2078 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2079 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2081 2018-06-25 Martin Sebor <msebor@redhat.com>
2083 * doc/extend.texi (Zero-length arrays): Update and clarify.
2085 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2087 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2088 added IEEE/IBM long double multilib support on PowerPC little
2089 endian Linux systems.
2090 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2091 (MULTILIB_DEFAULTS): Likewise.
2092 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2094 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2095 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2096 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2098 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2101 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2102 (REORDER_45): Likewise.
2104 2018-06-25 Jeff Law <law@redhat.com>
2106 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2107 dividend to 32 bits. Adjust length.
2108 (udivmodhi4): Cleanup output template. Fix length.
2110 2018-06-25 Carl Love <cel@us.ibm.com>
2112 * config/rs6000/vsx.md: Change word selector to prefered location.
2114 2018-06-25 Richard Biener <rguenther@suse.de>
2116 PR tree-optimization/86304
2117 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2118 epilogue-if-converted loops as well.
2120 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2122 * lto-section-out.c (lto_begin_section): Do not print section
2123 name for noaddr and unnumbered dumps.
2125 2018-06-25 Richard Biener <rguenther@suse.de>
2127 * tree-vectorizer.h (struct vec_info_shared): New structure
2128 with parts split out from struct vec_info and loop_nest from
2129 struct _loop_vec_info.
2130 (struct vec_info): Adjust accordingly.
2131 (struct _loop_vec_info): Likewise.
2132 (LOOP_VINFO_LOOP_NEST): Adjust.
2133 (LOOP_VINFO_DATAREFS): Likewise.
2134 (LOOP_VINFO_DDRS): Likewise.
2135 (struct _bb_vec_info): Likewise.
2136 (BB_VINFO_DATAREFS): Likewise.
2137 (BB_VINFO_DDRS): Likewise.
2138 (struct _stmt_vec_info): Add dr_aux member.
2139 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2140 (DR_MISALIGNMENT_UNINITIALIZED): New.
2141 (set_dr_misalignment): Adjust.
2142 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2143 (vect_analyze_loop): Adjust prototype.
2144 (vect_analyze_loop_form): Likewise.
2145 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2146 Compute dependences lazily.
2147 (vect_record_base_alignments): Use shared datarefs/ddrs.
2148 (vect_verify_datarefs_alignment): Likewise.
2149 (vect_analyze_data_refs_alignment): Likewise.
2150 (vect_analyze_data_ref_accesses): Likewise.
2151 (vect_analyze_data_refs): Likewise.
2152 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2153 constructor parameter for shared part.
2154 (vect_analyze_loop_form): Pass in shared part and adjust.
2155 (vect_analyze_loop_2): Pass in storage for the number of
2156 stmts. Move loop nest finding to the caller. Compute
2158 (vect_analyze_loop): Pass in shared part.
2159 (vect_transform_loop): Verify shared datarefs are unchanged.
2160 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2161 constructor parameter for shared part.
2162 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2163 (vect_slp_bb): Verify shared datarefs are unchanged before
2165 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2167 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2168 DR_MISALIGNMENT_UNINITIALIZED.
2169 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2170 parameter for shared part.
2171 (vec_info::~vec_info): Adjust.
2172 (vec_info_shared::vec_info_shared): New.
2173 (vec_info_shared::~vec_info_shared): Likewise.
2174 (vec_info_shared::save_datarefs): Likewise.
2175 (vec_info_shared::check_datarefs): Likewise.
2176 (try_vectorize_loop_1): Construct shared part live for analyses
2177 of a single loop for multiple vector sizes.
2178 * tree-parloops.c (gather_scalar_reductions): Adjust.
2180 2018-06-25 Richard Biener <rguenther@suse.de>
2182 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2183 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2184 (vect_analyze_data_refs): Remove similar code from here and
2185 simplify accordingly.
2187 2018-06-25 Richard Biener <rguenther@suse.de>
2189 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2190 for reverse storage order accesses rather than asserting
2191 they cannot happen here.
2193 2018-06-25 Tom de Vries <tdevries@suse.de>
2196 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2197 Use data16 instead of .byte for insn prefix.
2199 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2202 * parser.c (make_char_string_pack): Pass this literal chars
2203 through cpp_interpret_string.
2204 (cp_parser_userdef_numeric_literal): Check the result of
2205 make_char_string_pack.
2207 2018-06-24 Maya Rashish <coypu@sdf.org>
2209 * ginclude/stddef.h: Simplify conditions around avoiding
2210 re-definition of __size_t.
2212 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2214 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2215 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2217 2018-06-22 Maya Rashish <coypu@sdf.org>
2219 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2220 made to OpenBSD and NetBSD through the years.
2222 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2224 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2225 behavior of vec_pack (vector double, vector double) to match
2226 behavior of vec_float2 (vector double, vector double).
2228 2018-06-22 Olivier Hainque <hainque@adacore.com>
2230 * gimplify.c (gimplify_function_tree): Prevent creation
2231 of a trampoline for the address of the current function
2232 passed to entry/exit instrumentation hooks.
2234 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2237 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2240 2018-06-22 Martin Liska <mliska@suse.cz>
2242 PR tree-optimization/86263
2243 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2244 Bail out if is_enabled is false.
2245 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2247 (jump_table_cluster::is_enabled): New function.
2249 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2251 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2252 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2253 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2254 (lto_input_ts_binfo_tree_pointers): Likewise.
2255 * tree-streamer-out.c (streamer_write_tree_bitfields,
2256 write_ts_binfo_tree_pointers): Likewise.
2257 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2259 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2261 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2263 2018-06-22 Martin Liska <mliska@suse.cz>
2265 * symbol-summary.h (get): Make it pure and inline move
2266 functionality from ::get function.
2267 (get): Remove and inline into ::get and ::get_create.
2268 (get_create): Move code from ::get function.
2270 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2273 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2274 -x assembler-with-cpp.
2276 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2278 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2279 _FILE_OFFSET_BITS=64 for C++.
2281 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2283 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2284 conversion insn that shows up when pr85657-3.c is compiled using
2285 IEEE 128-bit long double.
2286 (neg<mode>2_internal): Use the correct mode to check whether the
2287 mode is IBM extended.
2288 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2289 multiply and divide external functions from being created more
2292 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2294 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2296 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2297 the edge can be forwarded.
2298 (cfg_layout_merge_blocks): Likewise.
2300 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2302 * except.c (finish_eh_generation): Commit edge insertions only after
2303 the EH edges have been redirected from post-landing to landing pads.
2305 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2307 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2308 create_tmp_var_for to create the FRAME decl.
2309 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2311 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2313 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2314 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2315 debug statement when resetting its value.
2316 (expand_call_inline): Copy the locus of the call onto the assignment
2317 of the return value, if any. Use local variable in more cases.
2319 2018-06-21 Martin Liska <mliska@suse.cz>
2321 * ipa-pure-const.c (propagate_nothrow): Use
2322 funct_state_summaries->get.
2323 (dump_malloc_lattice): Likewise.
2324 (propagate_malloc): Likewise.
2326 2018-06-21 Richard Biener <rguenther@suse.de>
2328 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2329 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2330 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2332 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2333 BLOCK_ABSTRACT_ORIGIN unconditionally.
2335 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2337 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2339 * ipa-reference.c (ipa_reference_c_finalize): Delete
2340 ipa_ref_opt_sum_summaries and set it to NULL.
2342 2018-06-21 Tom de Vries <tdevries@suse.de>
2344 PR tree-optimization/85859
2345 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2346 test with comment from bb_no_side_effects_p.
2348 2018-06-21 Richard Biener <rguenther@suse.de>
2350 PR tree-optimization/86232
2351 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2352 max for constant niter.
2354 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2356 * config/aarch64/aarch64-simd.md
2357 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2359 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2361 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2362 Make opernads of the unspec commutative.
2364 2018-06-21 Richard Biener <rguenther@suse.de>
2366 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2367 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2368 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2369 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2371 (vect_record_base_alignments): Do not record base alignment
2372 for gathers or scatters.
2373 (vect_compute_data_ref_alignment): Drop return value that is always
2374 true. Bail out early for gathers or scatters.
2375 (vect_enhance_data_refs_alignment): Bail out early for gathers
2377 (vect_find_same_alignment_drs): Likewise.
2378 (vect_analyze_data_refs_alignment): Remove dead code.
2379 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2380 (vect_analyze_data_refs): For possible gathers or scatters do
2381 not create an alternate DR, just check their possible validity
2382 and mark them. Adjust DECL_NONALIASED handling to not rely
2384 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2385 update inits of gathers or scatters.
2386 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2387 Also copy gather/scatter flag to pattern vinfo.
2389 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2391 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2392 behavior of vec_packsu (vector unsigned long long, vector unsigned
2393 long long) to match behavior of vec_packs with same signature.
2395 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2396 Thomas Schwinge <thomas@codesourcery.com>
2397 Cesar Philippidis <cesar@codesourcery.com>
2399 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2400 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2401 (gimplify_adjust_omp_clauses): Likewise.
2402 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2403 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2404 (gimplify_omp_target_update): Update handling of acc update and
2406 * omp-low.c (install_var_field): Remove unused parameter
2407 base_pointers_restrict.
2408 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2409 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2411 (omp_target_base_pointers_restrict_p): Delete.
2412 (scan_omp_target): Update call to scan_sharing_clauses.
2413 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2415 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2416 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2417 (convert_local_omp_clauses): Likewise.
2418 * tree-pretty-print.c (dump_omp_clause): Likewise.
2419 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2421 (omp_clause_code_name): Likewise.
2423 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2426 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2429 PR tree-optimization/86231
2430 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2431 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2433 2018-06-20 Tom de Vries <tdevries@suse.de>
2435 PR tree-optimization/86097
2436 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2437 iv type if signedness of iv type is not the same as that of *nit.
2439 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2441 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2442 EDGE_EH edges, verify they are all EDGE_EH.
2444 2018-06-20 Maya Rashish <coypu@sdf.org>
2446 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2448 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2450 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2451 * config/aarch64/aarch64.c (xgene1_tunings): Add
2452 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2453 (aarch64_mode_valid_for_sched_fusion_p):
2454 Allow 16-byte modes.
2455 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2456 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2458 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2460 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2461 * config/aarch64/iterators.md (VQ2): New mode iterator.
2463 2018-06-20 Martin Liska <mliska@suse.cz>
2465 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2466 Change default ratio from 10 to 8.
2468 2018-06-20 Martin Liska <mliska@suse.cz>
2470 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2472 (bit_test_cluster::find_bit_tests): Likewise.
2473 (switch_decision_tree::analyze_switch_statement): Find clusters.
2474 * tree-switch-conversion.h (struct jump_table_cluster): Document
2477 2018-06-20 Martin Liska <mliska@suse.cz>
2479 * tree-switch-conversion.c (switch_conversion::collect):
2480 Record m_uniq property.
2481 (switch_conversion::expand): Bail out for special conditions.
2482 (group_cluster::~group_cluster): New.
2483 (group_cluster::group_cluster): Likewise.
2484 (group_cluster::dump): Likewise.
2485 (jump_table_cluster::emit): New.
2486 (switch_decision_tree::fix_phi_operands_for_edges): New.
2487 (struct case_node): Remove struct.
2488 (jump_table_cluster::can_be_handled): New.
2489 (case_values_threshold): Moved to header.
2490 (reset_out_edges_aux): Likewise.
2491 (jump_table_cluster::is_beneficial): New.
2492 (bit_test_cluster::can_be_handled): Likewise.
2493 (add_case_node): Remove.
2494 (bit_test_cluster::is_beneficial): New.
2495 (case_bit_test::cmp): New.
2496 (bit_test_cluster::emit): New.
2497 (expand_switch_as_decision_tree_p): Remove.
2498 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2499 (fix_phi_operands_for_edge): Likewise.
2500 (switch_decision_tree::analyze_switch_statement): New.
2501 (compute_cases_per_edge): Move ...
2502 (switch_decision_tree::compute_cases_per_edge): ... here.
2503 (try_switch_expansion): Likewise.
2504 (switch_decision_tree::try_switch_expansion): Likewise.
2505 (record_phi_operand_mapping): Likewise.
2506 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2507 (emit_case_decision_tree): Likewise.
2508 (switch_decision_tree::emit): Likewise.
2509 (balance_case_nodes): Likewise.
2510 (switch_decision_tree::balance_case_nodes): Likewise.
2511 (dump_case_nodes): Likewise.
2512 (switch_decision_tree::dump_case_nodes): Likewise.
2513 (emit_jump): Likewise.
2514 (switch_decision_tree::emit_jump): Likewise.
2515 (emit_cmp_and_jump_insns): Likewise.
2516 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2517 (emit_case_nodes): Likewise.
2518 (switch_decision_tree::emit_case_nodes): Likewise.
2519 (conditional_probability): Remove.
2520 * tree-switch-conversion.h (enum cluster_type): New.
2522 (struct cluster): Likewise.
2523 (cluster::cluster): Likewise.
2524 (struct simple_cluster): Likewise.
2525 (simple_cluster::simple_cluster): Likewise.
2526 (struct group_cluster): Likewise.
2527 (struct jump_table_cluster): Likewise.
2528 (struct bit_test_cluster): Likewise.
2529 (struct min_cluster_item): Likewise.
2530 (struct case_tree_node): Likewise.
2531 (case_tree_node::case_tree_node): Likewise.
2532 (jump_table_cluster::case_values_threshold): Likewise.
2533 (struct case_bit_test): Likewise.
2534 (struct switch_decision_tree): Likewise.
2535 (struct switch_conversion): Likewise.
2536 (switch_decision_tree::reset_out_edges_aux): Likewise.
2538 2018-06-20 Martin Liska <mliska@suse.cz>
2540 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2541 (hoist_edge_and_branch_if_true): Likewise.
2542 (expand_switch_using_bit_tests_p): Likewise.
2543 (struct case_bit_test): Likewise.
2544 (case_bit_test_cmp): Likewise.
2545 (emit_case_bit_tests): Likewise.
2546 (switch_conversion::switch_conversion): New class.
2547 (struct switch_conv_info): Remove old struct.
2548 (collect_switch_conv_info): More to ...
2549 (switch_conversion::collect): ... this.
2550 (check_range): Likewise.
2551 (switch_conversion::check_range): Likewise.
2552 (check_all_empty_except_final): Likewise.
2553 (switch_conversion::check_all_empty_except_final): Likewise.
2554 (check_final_bb): Likewise.
2555 (switch_conversion::check_final_bb): Likewise.
2556 (create_temp_arrays): Likewise.
2557 (switch_conversion::create_temp_arrays): Likewise.
2558 (free_temp_arrays): Likewise.
2559 (gather_default_values): Likewise.
2560 (switch_conversion::gather_default_values): Likewise.
2561 (build_constructors): Likewise.
2562 (switch_conversion::build_constructors): Likewise.
2563 (constructor_contains_same_values_p): Likewise.
2564 (switch_conversion::contains_same_values_p): Likewise.
2565 (array_value_type): Likewise.
2566 (switch_conversion::array_value_type): Likewise.
2567 (build_one_array): Likewise.
2568 (switch_conversion::build_one_array): Likewise.
2569 (build_arrays): Likewise.
2570 (switch_conversion::build_arrays): Likewise.
2571 (gen_def_assigns): Likewise.
2572 (switch_conversion::gen_def_assigns): Likewise.
2573 (prune_bbs): Likewise.
2574 (switch_conversion::prune_bbs): Likewise.
2575 (fix_phi_nodes): Likewise.
2576 (switch_conversion::fix_phi_nodes): Likewise.
2577 (gen_inbound_check): Likewise.
2578 (switch_conversion::gen_inbound_check): Likewise.
2579 (process_switch): Use the newly created class.
2580 (switch_conversion::expand): New.
2581 (switch_conversion::~switch_conversion): New.
2582 * tree-switch-conversion.h: New file.
2584 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2586 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2587 tree-vect-patterns.c.
2588 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2589 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2590 (vect_recog_sad_pattern): Likewise.
2591 (vect_recog_widen_sum_pattern): Likewise.
2592 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2593 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2594 (vect_recog_rotate_pattern): Likewise.
2595 (vect_recog_mult_pattern): Likewise.
2596 (vect_recog_vector_vector_shift_pattern): Likewise.
2597 (vect_recog_divmod_pattern): Likewise.
2598 (vect_recog_mixed_size_cond_pattern): Likewise.
2599 (vect_recog_bool_pattern): Likewise.
2600 (vect_recog_mask_conversion_pattern): Likewise.
2601 (vect_try_gather_scatter_pattern): Likewise.
2602 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2603 (vect_recog_over_widening_pattern): Likewise.
2604 (vect_recog_gather_scatter_pattern): Likewise.
2605 (vect_recog_func_ptr): Move from tree-vectorizer.h
2606 (vect_vect_recog_func_ptrs): Move further down the file.
2607 (vect_recog_func): Likewise. Remove the third argument.
2608 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2609 (vect_pattern_recog_1): Expect the pattern function to do any
2610 necessary target tests. Also expect it to provide a vector type.
2611 Remove the type_in handling.
2613 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2615 * tree-vect-patterns.c (vect_pattern_detected): New function.
2616 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2617 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2618 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2619 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2620 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2621 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2622 (vect_recog_mask_conversion_pattern)
2623 (vect_try_gather_scatter_pattern): Likewise.
2625 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2627 * tree-vect-patterns.c (vect_get_internal_def): New function.
2628 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2629 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2630 (search_type_for_mask_1): Use it.
2632 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2634 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2635 redundant WIDEN_SUM_EXPR handling.
2636 (vect_recog_sad_pattern): Likewise.
2638 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2640 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2641 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2642 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2643 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2644 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2646 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2648 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2649 use the stmt_vec_info of the original bb statement for the
2650 new zero assignment, even if the call is part of a pattern.
2652 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2654 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2655 that the sequence is attached to the original statement rather
2656 than the pattern statement.
2657 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2658 PATTERN_DEF_SEQ from the original statement rather than
2659 the main pattern statement.
2660 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2661 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2662 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2664 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2666 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2667 definition statements before the early exit for statements that aren't
2669 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2671 (vect_transform_loop): ...here. Process pattern definition
2672 statements without first checking whether the main pattern
2673 statement is live or relevant.
2675 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2677 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2678 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2680 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2682 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2683 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2684 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2685 (expand_block_compare): Change select_block_compare_mode call.
2686 (expand_strncmp_align_check): Use new functions, fix comment.
2687 (emit_final_str_compare_gpr): New function.
2688 (expand_strn_compare): Refactor and clean up code.
2689 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2691 2018-06-19 Tony Reix <tony.reix@atos.com>
2692 Damien Bergamini <damien.bergamini@atos.com>
2693 David Edelsohn <dje.gcc@gmail.com>
2695 * collect2.c (static_obj): New variable.
2696 (static_libs): New variable.
2697 (is_in_list): Uncomment declaration.
2698 (main): Track AIX libraries linked statically.
2699 (is_in_list): Uncomment definition.
2700 (scan_prog_file): Don't add AIX shared libraries initializer
2701 to constructor list if linking statically.
2703 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2705 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2707 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2709 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2711 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2714 2018-06-19 Martin Liska <mliska@suse.cz>
2716 * config/i386/i386.c (ix86_can_inline_p): Do not use
2717 ipa_fn_summaries::get_create.
2718 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2720 (devirtualization_time_bonus): Likewise.
2721 (ipcp_propagate_stage): Likewise.
2722 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2723 (edge_set_predicate): Likewise.
2724 (evaluate_conditions_for_known_args): Likewise.
2725 (evaluate_properties_for_edge): Likewise.
2726 (ipa_call_summary::reset): Tranform to ...
2727 (ipa_call_summary::~ipa_call_summary): ... this.
2728 (ipa_fn_summary::reset): Transform to ...
2729 (ipa_fn_summary::~ipa_fn_summary): ... this.
2730 (ipa_fn_summary_t::remove): Rename to ...
2731 (ipa_fn_summary_t::remove_callees): ... this.
2732 (ipa_fn_summary_t::duplicate): Use placement new
2733 instead of memory copy.
2734 (ipa_call_summary_t::duplicate): Likewise.
2735 (ipa_call_summary_t::remove): Remove.
2736 (dump_ipa_call_summary): Change get_create to get.
2737 (ipa_dump_fn_summary): Dump only when summary exists.
2738 (analyze_function_body): Use symbol_summary::get instead
2740 (compute_fn_summary): Likewise.
2741 (estimate_edge_devirt_benefit): Likewise.
2742 (estimate_edge_size_and_time): Likewise.
2743 (inline_update_callee_summaries): Likewise.
2744 (remap_edge_change_prob): Likewise.
2745 (remap_edge_summaries): Likewise.
2746 (ipa_merge_fn_summary_after_inlining): Likewise.
2747 (write_ipa_call_summary): Likewise.
2748 (ipa_fn_summary_write): Likewise.
2749 (ipa_free_fn_summary): Likewise.
2750 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2751 (struct ipa_call_summary): Likewise.
2752 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2754 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2755 (estimate_size_after_inlining): Likewise.
2756 (estimate_growth): Likewise.
2757 (growth_likely_positive): Likewise.
2758 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2759 (inline_call): Likewise.
2760 * ipa-inline.c (caller_growth_limits): Likewise.
2761 (can_inline_edge_p): Likewise.
2762 (can_inline_edge_by_limits_p): Likewise.
2763 (compute_uninlined_call_time): Likewise.
2764 (compute_inlined_call_time): Likewise.
2765 (want_inline_small_function_p): Likewise.
2766 (edge_badness): Likewise.
2767 (update_caller_keys): Likewise.
2768 (update_callee_keys): Likewise.
2769 (inline_small_functions): Likewise.
2770 (inline_to_all_callers_1): Likewise.
2771 (dump_overall_stats): Likewise.
2772 (early_inline_small_functions): Likewise.
2773 (early_inliner): Likewise.
2774 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2775 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2776 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2777 * ipa-split.c (execute_split_functions): Likewise.
2778 * symbol-summary.h: Likewise.
2779 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2781 2018-06-19 Richard Biener <rguenther@suse.de>
2783 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2784 (vectorize_loops): ... here. Fix dbgcnt handling.
2785 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2787 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2790 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2791 ieee128 argument takes up only one (vector) register, not two (floating
2794 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2796 * gimplify.c (gimplify_init_constructor): Really never clear for an
2797 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2799 2018-06-19 Richard Biener <rguenther@suse.de>
2801 PR tree-optimization/86179
2802 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2803 after failed recognition.
2805 2018-06-18 Martin Sebor <msebor@redhat.com>
2808 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2809 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2810 Handle integer subtraction.
2811 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2812 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2814 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2816 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2817 param from rtx to rtx_insn *.
2818 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2820 (frv_ifcvt_modify_insn): Likwise.
2821 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2822 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2823 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2824 as_a <rtx_insn *> cast to local "unprotected_region" once
2825 it's been established that it's not NULL or pc_rtx.
2826 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2827 param "sethi" from rtx to rtx_insn *.
2828 (nds32_group_float_insns): Likewise for param "insn".
2829 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2831 (vax_output_int_subtract): Likewise.
2832 * config/vax/vax.c (vax_output_int_add): Likewise for param
2834 (vax_output_int_subtract): Likewise.
2835 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2836 (emit_pattern_after): Likewise for param "after".
2837 (emit_insn_after): Likewise.
2838 (emit_jump_insn_after): Likewise.
2839 (emit_call_insn_after): Likewise.
2840 (emit_debug_insn_after): Likewise.
2841 (emit_pattern_before): Likewise for param "before".
2842 (emit_insn_before): Likewise.
2843 (emit_jump_insn_before): Likewise.
2844 * final.c (get_insn_template): Likewise for param "insn", removing
2846 * output.h (get_insn_template): Likewise for 2nd param.
2847 * rtl.h (emit_insn_before): Likewise.
2848 (emit_jump_insn_before): Likewise.
2849 (emit_debug_insn_before_noloc): Likewise.
2850 (emit_insn_after): Likewise.
2851 (emit_jump_insn_after): Likewise.
2852 (emit_call_insn_after): Likewise.
2853 (emit_debug_insn_after): Likewise.
2854 (set_insn_deleted): Likewise for param.
2856 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2859 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2860 floating point modes, so that IFmode is numerically greater than
2861 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2862 to declare the ordering. This prevents IFmode from being
2863 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2864 machine. Include rs6000-modes.h to share the fractional values
2865 between genmodes* and the rest of the compiler.
2869 * config/rs6000/rs6000-modes.h: New file.
2870 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2871 meaning of rs6000_long_double_size so that 126..128 selects an
2872 appropriate 128-bit floating point type.
2873 (rs6000_option_override_internal): Likewise.
2874 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2875 (TARGET_LONG_DOUBLE_128): Change the meaning of
2876 rs6000_long_double_size so that 126..128 selects an appropriate
2877 128-bit floating point type.
2878 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2879 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2880 source and destination to match the standard usage.
2881 (truncifkf2): Likewise.
2882 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2883 ISA 2.07 to use an explicit clobber, instead of passing in a
2885 (copysign<mode>3_soft): Likewise.
2887 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2889 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2890 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2891 (vect_slp_analyze_instance_dependence): Likewise.
2892 (vect_enhance_data_refs_alignment): Likewise.
2893 (vect_analyze_data_refs_alignment): Likewise.
2894 (vect_slp_analyze_and_verify_instance_alignment
2895 (vect_analyze_data_ref_accesses): Likewise.
2896 (vect_prune_runtime_alias_test_list): Likewise.
2897 (vect_analyze_data_refs): Likewise.
2898 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2899 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2900 (vect_analyze_scalar_cycles_1): Likewise.
2901 (vect_get_loop_niters): Likewise.
2902 (vect_analyze_loop_form_1): Likewise.
2903 (vect_update_vf_for_slp): Likewise.
2904 (vect_analyze_loop_operations): Likewise.
2905 (vect_analyze_loop): Likewise.
2906 (vectorizable_induction): Likewise.
2907 (vect_transform_loop): Likewise.
2908 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2909 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2910 (vect_make_slp_decision): Likewise.
2911 (vect_detect_hybrid_slp): Likewise.
2912 (vect_slp_analyze_operations): Likewise.
2913 (vect_slp_bb): Likewise.
2914 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2915 (vectorizable_bswap): Likewise.
2916 (vectorizable_call): Likewise.
2917 (vectorizable_simd_clone_call): Likewise.
2918 (vectorizable_conversion): Likewise.
2919 (vectorizable_assignment): Likewise.
2920 (vectorizable_shift): Likewise.
2921 (vectorizable_operation): Likewise.
2922 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2924 2018-06-18 Martin Sebor <msebor@redhat.com>
2926 PR tree-optimization/81384
2927 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2928 * builtins.c (expand_builtin_strnlen): New function.
2929 (expand_builtin): Call it.
2930 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2931 * builtins.def (BUILT_IN_STRNLEN): New.
2932 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2933 Warn for bounds in excess of maximum object size.
2934 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2935 single-value ranges. Handle strnlen.
2936 (handle_builtin_strlen): Handle strnlen.
2937 (strlen_check_and_optimize_stmt): Same.
2938 * doc/extend.texi (Other Builtins): Document strnlen.
2940 2018-06-18 Maya Rashish <coypu@sdf.org>
2942 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2943 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2944 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2946 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2948 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2950 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2952 * tree.c (escaped_string::escape): Replace cast to char * by
2953 const_cast<char *> (unescaped).
2955 2018-06-18 Nick Clifton <nickc@redhat.com>
2958 * tree.c (escaped_string): New class. Converts an unescaped
2959 string into its escaped equivalent.
2960 (warn_deprecated_use): Use the new class to convert the
2961 deprecation message, if present.
2962 (test_escaped_strings): New self test.
2963 (test_c_tests): Add test_escaped_strings.
2964 * doc/extend.texi (deprecated): Add a note that the
2965 deprecation message is affected by the -fmessage-length
2966 option, and that control characters will be escaped.
2967 (#pragma GCC error): Document this pragma.
2968 (#pragma GCC warning): Likewise.
2969 * doc/invoke.texi (-fmessage-length): Document this option's
2970 effect on the #warning and #error preprocessor directives and
2971 the deprecated attribute.
2973 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2975 * tree.c (decl_value_expr_lookup): Revert latest change.
2976 (decl_value_expr_insert): Likewise.
2978 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2980 * gimplify.c (nonlocal_vlas): Delete.
2981 (nonlocal_vla_vars): Likewise.
2982 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2984 (gimplify_body): Do not create and destroy nonlocal_vlas.
2985 * tree-nested.c: Include diagnostic.h.
2986 (use_pointer_in_frame): Tweak.
2987 (lookup_field_for_decl): Add assertion and declare the transformation.
2988 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2989 internal error when the reference is in a wrong context. Do not
2990 create a debug decl by default.
2991 (note_nonlocal_block_vlas): Delete.
2992 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2993 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2994 create a debug decl by default.
2995 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2996 call is in a wrong context.
2997 (fixup_vla_decls): New function.
2998 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2999 debug variables were created.
3000 * tree.c (decl_value_expr_lookup): Add checking assertion.
3001 (decl_value_expr_insert): Likewise.
3003 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3006 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3007 * tree-scalar-evolution.c (interpret_expr): Likewise.
3008 (expression_expensive_p): Likewise.
3009 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3010 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3011 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3012 (ssa_defined_by_minus_one_stmt_p): New.
3014 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3017 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3018 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3019 * dojump.c (do_jump): Likewise.
3020 * expr.c (expand_expr_real_2): Check operand type's sign.
3021 * fold-const.c (const_unop): Handle ABSU_EXPR.
3022 (fold_abs_const): Likewise.
3023 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3024 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3025 (strip_sign_op_1): Likesise.
3026 * match.pd: Add new pattern to generate ABSU_EXPR.
3027 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3028 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3029 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3030 * tree-inline.c (estimate_operator_cost): Likewise.
3031 * tree-pretty-print.c (dump_generic_node): Likewise.
3032 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3033 * tree.def (ABSU_EXPR): New.
3035 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3038 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3039 documented as preserved for backward compatibility only.
3040 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3042 PR rtl-optimization/86108
3043 * bb-reorder.c (create_forwarder_block): Renamed to ...
3044 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3045 jump from new landing pad to the second part.
3046 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3049 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3052 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3053 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3054 Only call store_expr for halves if the mode is the same.
3057 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3058 Fix up comment formatting.
3060 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3062 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3063 (typed_splay_tree::closure,
3064 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3065 (typed_splay_tree::typed_splay_tree,
3066 typed_splay_tree::operator =): Declared private.
3067 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3068 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3069 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3070 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3071 typed_splay_tree::splay_tree_splay,
3072 typed_splay_tree::splay_tree_foreach_helper,
3073 typed_splay_tree::splay_tree_insert,
3074 typed_splay_tree::splay_tree_remove,
3075 typed_splay_tree::splay_tree_lookup,
3076 typed_splay_tree::splay_tree_predecessor,
3077 typed_splay_tree::splay_tree_successor,
3078 typed_splay_tree::splay_tree_min,
3079 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3080 (typed_splay_tree::root, typed_splay_tree::comp,
3081 typed_splay_tree::delete_key,
3082 typed_splay_tree::delete_value): New data members.
3083 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3084 typed_splay_tree::remove.
3086 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3088 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3089 -mginv and -mno-ginv to the assembler.
3090 * config/mips/mips.opt (-mcrc): New option.
3092 * doc/invoke.text (-mcrc): Document.
3095 2018-06-15 Nick Clifton <nickc@redhat.com>
3098 * tree.c (escaped_string): New class. Converts an unescaped
3099 string into its escaped equivalent.
3100 (warn_deprecated_use): Use the new class to convert the
3101 deprecation message, if present.
3102 (test_escaped_strings): New self test.
3103 (test_c_tests): Add test_escaped_strings.
3104 * doc/extend.texi (deprecated): Add a note that the
3105 deprecation message is affected by the -fmessage-length
3106 option, and that control characters will be escaped.
3107 (#pragma GCC error): Document this pragma.
3108 (#pragma GCC warning): Likewise.
3109 * doc/invoke.texi (-fmessage-length): Document this option's
3110 effect on the #warning and #error preprocessor directives and
3111 the deprecated attribute.
3113 2018-06-15 Richard Biener <rguenther@suse.de>
3115 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3116 here, also noting vector size used.
3117 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3118 size used in MSG_OPTIMIZED_LOCATIONS dump.
3119 (pass_slp_vectorize::execute): Adjust.
3121 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3124 * config/arc/arc.c (arc_return_address_register): Fix
3127 2018-06-15 Richard Biener <rguenther@suse.de>
3130 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3131 leave useless conversion stripping to force_gimple_operand_gsi.
3132 (gimplify_build2): Likewise.
3133 (gimplify_build1): Likewise.
3135 2018-06-15 Richard Biener <rguenther@suse.de>
3138 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3139 before adjusting their block.
3141 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3143 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3144 multilibs for *-*-rtems*.
3145 * config/riscv/t-rtems: New file.
3147 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3150 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3151 unsigned_type_for returns NULL.
3154 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3155 subregs of multi-word pseudos unless the float mode has word size.
3157 2018-06-14 Richard Biener <rguenther@suse.de>
3160 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3161 duplicate and harmful type_hash_canon.
3162 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3164 2018-06-14 Richard Biener <rguenther@suse.de>
3167 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3170 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3172 * config/rtems.h (STDINT_LONG32): Define.
3174 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3175 Prachi Godbole <prachi.godbole@imgtec.com>
3177 * config/mips/mips-cpus.def: Define P6600.
3178 * config/mips/mips-tables.opt: Regenerate.
3179 * config/mips/mips.c (mips_ucbranch_type): New enum.
3180 (mips_rtx_cost_data): Add support for P6600.
3181 (mips_issue_rate): Likewise.
3182 (mips_multipass_dfa_lookahead): Likewise.
3183 (mips_avoid_hazard): Likewise.
3184 (mips_reorg_process_insns): Likewise.
3185 (mips_classify_branch_p6600): New function.
3186 * config/mips/mips.h (TUNE_P6600): New define.
3187 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3188 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3189 * config/mips/mips.md: Include p6600.md.
3190 (processor): Add p6600.
3191 * config/mips/p6600.md: New file.
3192 * doc/invoke.texi: Add p6600 to supported architectures.
3194 2018-06-13 Martin Sebor <msebor@redhat.com>
3196 PR tree-optimization/86114
3197 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3199 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3201 2018-06-13 Richard Biener <rguenther@suse.de>
3203 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3204 Properly set vector type of the intermediate stmt.
3205 * tree-vect-stmts.c (vectorizable_operation): The destination
3206 var always has vectype_out type.
3208 2018-06-13 Jeff Law <law@redhat.com>
3210 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3211 integer 0 for argument to print_rtl_with_bb.
3212 (rl78_reorg): Likewise.
3214 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3216 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3217 from rtx to rtx_insn *.
3218 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3220 (add_sched_insns_for_speculation): Likewise for local "target",
3221 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3222 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3223 from rtx_insn ** to rtx_code_label **.
3224 (reorg_emit_nops): Likewise.
3225 (c6x_reorg): Likewise for local "call_labels".
3226 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3228 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3229 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3230 the loops over LABEL_REFS.
3231 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3233 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3234 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3235 (split_branches): Strengthen local "olabel" from rtx to
3236 rtx_insn *, adding a safe_as_a cast.
3237 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3239 (add_insn_after): Likewise for first two params.
3240 (add_insn_before): Likewise.
3241 (remove_insn): Likewise for param.
3242 (emit_pattern_before_noloc): Likewise for second and third params.
3243 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3244 (emit_call_insn_before_noloc): Likewise.
3245 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3247 (emit_barrier_before): Likewise.
3248 (emit_label_before): Strengthen "label" param from "rtx" to
3249 "rtx_code_label *". Strengthen "before" param from "rtx" to
3251 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3253 (emit_pattern_after_noloc): Likewise.
3254 (emit_insn_after_noloc): Likewise.
3255 (emit_jump_insn_after_noloc): Likewise.
3256 (emit_call_insn_after_noloc): Likewise.
3257 (emit_debug_insn_after_noloc): Likewise.
3258 (emit_barrier_after): Likewise.
3259 (emit_label_after): Likewise for both params.
3260 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3261 "loc" param from "int" to "location_t".
3262 (emit_insn_after_setloc): Likewise.
3263 (emit_jump_insn_after_setloc): Likewise.
3264 (emit_call_insn_after_setloc): Likewise.
3265 (emit_debug_insn_after_setloc): Likewise.
3266 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3267 "loc" param from "int" to "location_t".
3268 (emit_pattern_before): Convert NULL_RTX to NULL.
3269 (emit_insn_before_setloc): Convert "loc" param from "int" to
3271 (emit_jump_insn_before_setloc): Likewise.
3272 (emit_call_insn_before_setloc): Likewise.
3273 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3274 rtx_insn *. Convert "loc" param from "int" to "location_t".
3275 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3276 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3277 Convert 3rd param from "int" to "location_t".
3278 (emit_barrier_before, emit_barrier_after, next_real_insn):
3279 Strengthen param from rtx to rtx_insn *.
3280 (emit_label_before): Strengthen 1st param from "rtx" to
3281 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3283 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3284 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3285 Strengthen 2nd param from "rtx" to "rtx_insn *".
3286 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3287 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3288 Likewise. Convert 3rd param from "int" to "location_t".
3289 (emit_label_after): Strengthen 1st param from "rtx" to
3291 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3293 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3294 from "rtx" to "rtx_insn *".
3296 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3298 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3299 bodies streamed in with -Q.
3300 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3301 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3302 * gimple-streamer-out.c: Include gimple-pretty-print.h
3303 (output_bb): Dump stmts streamed.
3304 * lto-section-out.c: Include print-tree.h
3305 (lto_begin_section): Dump sections created.
3306 (lto_output_decl_index): Dump decl encoded.
3307 * lto-streamer-out.c: Include print-tree.h
3308 (create_output_block): Dump output block created.
3309 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3310 (output_function): Dump function output.
3311 (output_constructor): Dump constructor streamed.
3312 (write_global_stream): Output indexes encoded.
3313 (produce_asm_for_decls): Dump streams encoded.
3314 * lto-streamer.c (streamer_dump_file): New global var.
3315 * lto-streamer.h (streamer_dump_file): Declare.
3316 * passes.c (ipa_write_summaries): Initialize streamer dump.
3317 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3320 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3323 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3324 offsets for register save directives. Emit a second batch of save
3325 directives, if need be, when the function accesses prior frames.
3327 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3329 * config/arc/fpu.md (fmasf4): Force operand to register.
3330 (fnmasf4): Likewise.
3332 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3334 * config/arc/arc-protos.h (arc_pad_return): Remove.
3335 * config/arc/arc.c (machine_function): Remove force_short_suffix
3337 (arc_print_operand): Adjust printing of '&'.
3338 (arc_verify_short): Remove conditional printing of short suffix.
3339 (arc_final_prescan_insn): Remove reference to size_reason.
3340 (pad_return): New function.
3341 (arc_reorg): Call pad_return.
3342 (arc_pad_return): Remove.
3343 (arc_init_machine_status): Remove reference to force_short_suffix.
3344 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3345 (attr length): When attribute iscompact is true force to 2
3346 regardless; in the case of maybe check if we want to force the
3347 instruction to have 4 bytes length.
3348 (nopv): Change it to generate 4 byte long nop as well.
3349 (blockage): New pattern.
3350 (simple_return): Remove call to arc_pad_return.
3351 (p_return_i): Likewise.
3353 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3355 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3357 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3359 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3362 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3364 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3365 for ARC700 and ARCv2.
3367 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3370 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3371 operands[2] instead of operands[1].
3374 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3376 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3377 case, check whether the outer register overlaps an unallocatable
3378 register, not just whether it fits the required class.
3380 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3382 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3384 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3385 (memrefs_conflict_p): Likewise.
3386 (init_alias_analysis): Likewise.
3387 * cfgexpand.c (expand_debug_expr): Likewise.
3388 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3389 * cse.c (fold_rtx): Likewise.
3390 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3391 * expr.c (emit_block_move_hints): Likewise.
3392 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3393 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3394 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3395 (emit_group_store): Likewise.
3396 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3397 to read the PRE/POST_MODIFY increment.
3398 * calls.c (store_one_arg): Use strip_offset.
3399 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3401 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3403 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3404 (simplify_binary_operation_1): Extend CONST_INT handling to
3406 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3407 than a HOST_WIDE_INT.
3408 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3410 (adjust_mems, add_stores): Update accodingly.
3411 (vt_canonicalize_addr): Track polynomial offsets.
3412 (emit_note_insn_var_location): Likewise.
3413 (vt_add_function_parameter): Likewise.
3414 (vt_initialize): Likewise.
3416 2018-06-12 Jeff Law <law@redhat.com>
3418 * config.gcc (alpha*-*-freebsd*): Remove.
3419 * config/alpha/freebsd.h: Remove.
3421 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3424 * spellcheck-tree.c (levenshtein_distance): Rename to...
3425 (get_edit_distance): ...this, and update for underlying renaming.
3426 * spellcheck-tree.h (levenshtein_distance): Rename to...
3427 (get_edit_distance): ...this.
3428 * spellcheck.c (levenshtein_distance): Rename to...
3429 (get_edit_distance): ...this. Convert from Levenshtein distance
3430 to Damerau-Levenshtein distance by supporting transpositions of
3431 adjacent characters. Rename "v1" to "v_next" and "v0" to
3433 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3434 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3435 update for underlying renaming.
3436 (selftest::levenshtein_distance_unit_test): Rename to...
3437 (selftest::test_get_edit_distance_unit): ...this, and update for
3438 underlying renaming.
3439 (selftest::test_find_closest_string): Add example from PR 69968
3440 where transposition helps
3441 (selftest::test_metric_conditions): Update for renaming.
3442 (selftest::test_metric_conditions): Likewise.
3443 (selftest::spellcheck_c_tests): Likewise.
3444 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3445 (get_edit_distance): ...this.
3446 (best_match::consider): Update for renaming.
3448 2018-06-12 Martin Sebor <msebor@redhat.com>
3450 PR tree-optimization/85259
3451 * builtins.c (compute_objsize): Handle constant offsets.
3452 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3453 true iff a warning has been issued.
3454 * gimple.h (gimple_nonartificial_location): New function.
3455 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3456 gimple_nonartificial_location and handle -Wno-system-headers.
3457 (handle_builtin_stxncpy): Same.
3459 2018-06-12 Martin Sebor <msebor@redhat.com>
3462 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3464 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3466 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3467 BUILTIN_VEC_XST entries for pointer to double and long long.
3469 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3472 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3474 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3477 2018-06-12 Martin Liska <mliska@suse.cz>
3479 * doc/options.texi: Document IntegerRange.
3481 2018-06-12 Martin Liska <mliska@suse.cz>
3483 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3484 * opt-functions.awk: Handle Deprecated flag.
3485 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3487 (read_cmdline_option): Report warning for a deprecated option.
3488 * opts.h (struct cl_option): Add new field cl_deprecated.
3489 (CL_ERR_DEPRECATED): New.
3491 2018-06-12 Martin Liska <mliska@suse.cz>
3493 * doc/options.texi: Document Deprecated option flag.
3495 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3497 * config/arc/arc-arch.h (arc_extras): New enum.
3498 (arc_cpu_t):Add field extra.
3499 (arc_cpu_types): Consider the extras.
3500 * config/arc/arc-cpus.def: Add extras info.
3501 * config/arc/arc-opts.h (processor_type): Consider extra field.
3502 * config/arc/arc.c (arc_override_options): Handle extra field.
3504 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3506 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3507 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3509 * config/arc/arc.c (arc_init): Update pic warning.
3510 (irq_range): Update irq range parsing warnings.
3511 (arc_override_options): Update various warning messages.
3512 (arc_handle_aux_attribute): Likewise.
3514 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3516 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3518 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3520 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3521 compiler messages directives.
3523 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3525 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3526 * config/mips/mips-tables.opt: Regenerate.
3527 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3529 * doc/invoke.texi: Document -march=i6500.
3531 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3533 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3534 (i6400_gpmul): Add cpu_unit.
3535 (i6400_gpdiv): Likewise.
3536 (i6400_msa_add_d): Update reservations.
3537 (i6400_msa_int_add) Likewise.
3538 (i6400_msa_short_logic3) Likewise.
3539 (i6400_msa_short_logic2) Likewise.
3540 (i6400_msa_short_logic) Likewise.
3541 (i6400_msa_move) Likewise.
3542 (i6400_msa_cmp) Likewise.
3543 (i6400_msa_short_float2) Likewise.
3544 (i6400_msa_div_d) Likewise.
3545 (i6400_msa_long_logic1) Likewise.
3546 (i6400_msa_long_logic2) Likewise.
3547 (i6400_msa_mult) Likewise.
3548 (i6400_msa_long_float2) Likewise.
3549 (i6400_msa_long_float4) Likewise.
3550 (i6400_msa_long_float5) Likewise.
3551 (i6400_msa_long_float8) Likewise.
3552 (i6400_fpu_fadd): Include frint type.
3553 (i6400_fpu_store): New define_insn_reservation.
3554 (i6400_fpu_load): Likewise.
3555 (i6400_fpu_move): Likewise.
3556 (i6400_fpu_fcmp): Likewise.
3557 (i6400_fpu_fmadd): Likewise.
3558 (i6400_int_mult): Include imul3nc type and update reservation.
3559 (i6400_int_div): Include idiv3 type and update reservation.
3560 (i6400_int_load): Update to check type not move_type.
3561 (i6400_int_store): Likewise.
3562 (i6400_int_prefetch): Set zero latency.
3564 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3566 * gcc.c: Document new %@{...} sequence.
3567 (LINK_COMMAND_SPEC): Use it for the -L switches.
3568 (cpp_unique_options): Use it for the -I switches.
3569 (at_file_argbuf): New global variable.
3570 (in_at_file): Likewise.
3571 (alloc_args): Create at_file_argbuf.
3572 (clear_args): Truncate at_file_argbuf.
3573 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3574 (open_at_file): New function.
3575 (close_at_file): Likewise.
3576 (create_at_file): Delete.
3577 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3580 (validate_switches_from_spec): Deal with %@{...} sequence.
3581 (validate_switches): Likewise.
3582 (driver::finalize): Call clear_args.
3584 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3586 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3588 2018-06-11 Martin Sebor <msebor@redhat.com>
3590 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3591 (Wno-class-memaccess): Add @opindex.
3592 (Wno-templates, Wno-multiple-inheritance): Same.
3593 (Wno-virtual-inheritance, Wno-namespaces): Same.
3594 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3595 (Wno-format-overflow, Wno-format-truncation): Same.
3596 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3597 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3598 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3599 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3600 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3601 (Wno-misspelled-isr): Same.
3603 2018-06-11 Martin Sebor <msebor@redhat.com>
3605 * PR tree-optimization/86083
3606 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3608 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3610 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3612 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3615 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3616 on the correct operand.
3617 (*movdi_internal64): Ditto.
3619 2018-06-11 Martin Liska <mliska@suse.cz>
3621 PR tree-optimization/86089
3622 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3624 2018-06-11 Julia Koval <julia.koval@intel.com>
3626 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3627 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3628 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3630 2018-06-11 Olivier Hainque <hainque@adacore.com>
3632 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3633 for Ada with strict dwarf2.
3635 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3638 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3641 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3643 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3645 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3647 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3650 2018-06-08 Martin Liska <mliska@suse.cz>
3652 * tree-cfg.h (debug_function): Fix argument type to match
3655 2018-06-08 Martin Liska <mliska@suse.cz>
3657 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3658 Remove usage of MPX-related (and removed) fields.
3659 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3661 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3663 * cfg.c (debug): Use TDF_NONE rather than 0.
3664 * cfghooks.c (debug): Likewise.
3665 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3666 (struct dump_option_value_info): Convert to...
3667 (struct kv_pair): ...this template type.
3668 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3670 (optinfo_verbosity_options): Likewise.
3671 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3673 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3674 than int for "optgroup_flags" param.
3675 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3677 (dump_dec): Likewise.
3678 (dump_finish): Use TDF_NONE rather than 0.
3679 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3680 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3681 than 0. Update for change to option_ptr.
3682 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3683 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3684 0. Update for changes to optinfo_verbosity_options and
3686 (opt_info_switch_p): Convert optgroup_flags from int to
3688 (dump_basic_block): Use dump_flags_t rather than int
3689 for "dump_kind" param.
3690 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3691 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3692 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3693 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3694 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3695 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3696 TDF_NONE): Convert from macros to...
3697 (enum dump_flag): ...this new enum.
3698 (dump_flags_t): Update to use enum.
3699 (operator|, operator&, operator~, operator|=, operator&=):
3700 Implement for dump_flags_t.
3701 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3702 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3703 Convert from macros to...
3704 (enum optgroup_flag): ...this new enum.
3705 (optgroup_flags_t): New typedef.
3706 (operator|, operator|=): Implement for optgroup_flags_t.
3707 (struct dump_file_info): Convert field "alt_flags" to
3708 dump_flags_t. Convert field "optgroup_flags" to
3710 (dump_basic_block): Use dump_flags_t rather than int for param.
3711 (dump_generic_expr_loc): Likewise.
3712 (dump_dec): Likewise.
3713 (dump_register): Convert param "optgroup_flags" to
3715 (opt_info_enable_passes): Likewise.
3716 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3718 * gimple-pretty-print.c (debug): Likewise.
3719 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3720 (merged_store_group::apply_stores): Likewise.
3721 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3722 * gimple.c (verify_gimple_pp): Likewise.
3723 * graphite-poly.c (print_pbb_body): Likewise.
3724 * passes.c (pass_manager::register_one_dump_file): Convert
3725 local "optgroup_flags" to optgroup_flags_t.
3726 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3728 (debug_body): Likewise.
3729 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3730 to optgroup_flags_t.
3731 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3733 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3734 (convert_mult_to_fma): Likewise.
3735 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3736 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3737 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3738 "dump_kind" to dump_flags_t.
3740 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3742 * config/rs6000/rs6000.c (min, max): Delete.
3744 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3746 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3747 -mabi=spe and -mabi=no-spe.
3749 2018-06-08 Martin Liska <mliska@suse.cz>
3751 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3752 where we expect an existing summary.
3754 2018-06-08 Martin Liska <mliska@suse.cz>
3756 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3757 * ipa-inline.h (estimate_edge_growth): Likewise.
3759 2018-06-08 Martin Liska <mliska@suse.cz>
3761 * cgraph.c (function_version_hasher::hash): Use
3762 cgraph_node::get_uid ().
3763 (function_version_hasher::equal):
3764 * cgraph.h (cgraph_node::get_uid): New method.
3765 * ipa-inline.c (update_caller_keys): Use
3766 cgraph_node::get_uid ().
3767 (update_callee_keys): Likewise.
3768 * ipa-utils.c (searchc): Likewise.
3769 (ipa_reduced_postorder): Likewise.
3770 * lto-cgraph.c (input_node): Likewise.
3771 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3772 * symbol-summary.h (symtab_insertion): Likewise.
3773 (symtab_removal): Likewise.
3774 (symtab_duplication): Likewise.
3775 * tree-pretty-print.c (dump_function_header): Likewise.
3776 * tree-sra.c (convert_callers_for_node): Likewise.
3778 2018-06-08 Martin Liska <mliska@suse.cz>
3780 * cgraph.c (symbol_table::create_edge): Always assign a new
3782 (symbol_table::free_edge): Do not recycle numbers.
3783 * cgraph.h (cgraph_edge::get): New method.
3784 * symbol-summary.h (symtab_removal): Use it.
3785 (symtab_duplication): Likewise.
3786 (call_summary::hashable_uid): Remove.
3788 2018-06-08 Martin Liska <mliska@suse.cz>
3790 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3791 (initialize_growth_caches): Remove.
3792 (free_growth_caches): Likewise.
3793 (do_estimate_edge_time): Use edge_growth_cache.
3794 (do_estimate_edge_size): Likewise.
3795 (do_estimate_edge_hints): Likewise.
3796 * ipa-inline.c (reset_edge_caches): Likewise.
3797 (recursive_inlining): Likewise.
3798 (inline_small_functions): Likewise.
3799 * ipa-inline.h (initialize_growth_caches): Remove.
3800 (estimate_edge_size): Likewise.
3801 (estimate_edge_time): Likewise.
3802 (estimate_edge_hints): Likewise.
3803 (reset_edge_growth_cache): Likewise.
3804 * symbol-summary.h (call_summary::remove): New method.
3806 2018-06-08 Martin Liska <mliska@suse.cz>
3808 * ipa-cp.c (class edge_clone_summary): New summary.
3809 (grow_edge_clone_vectors): Remove.
3810 (ipcp_edge_duplication_hook): Remove.
3811 (class edge_clone_summary_t): New call_summary class.
3812 (ipcp_edge_removal_hook): Remove.
3813 (edge_clone_summary_t::duplicate): New function.
3814 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3815 (create_specialized_node): Likewise.
3816 (ipcp_driver): Initialize edge_clone_summaries and do not
3819 2018-06-08 Martin Liska <mliska@suse.cz>
3821 * symbol-summary.h (get): New function.
3822 (call_summary::m_initialize_when_cloning): New class member.
3824 2018-06-08 Martin Liska <mliska@suse.cz>
3826 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3827 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3828 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3829 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3830 (remove_cgraph_node_from_order): Use the removed_nodes set.
3831 (do_per_function_toporder): Likwise.
3832 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3833 instead of summary_uid.
3834 (symtab_removal): Likewise.
3835 (symtab_duplication): Likewise.
3837 2018-06-08 Martin Liska <mliska@suse.cz>
3839 * ipa-cp.c (ipcp_store_bits_results): Use
3840 ipcp_transformation_sum.
3841 (ipcp_store_vr_results): Likewise.
3842 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3844 (ipcp_transformation_initialize): ... this.
3845 (ipa_set_node_agg_value_chain):
3846 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3847 (write_ipcp_transformation_info): Likewise.
3848 (read_ipcp_transformation_info): Likewise.
3849 (ipcp_update_bits): Likewise.
3850 (ipcp_update_vr): Likewise.
3851 (ipcp_transform_function): Likewise.
3852 * ipa-prop.h: Rename ipcp_transformation_summary to
3853 ipcp_transformation.
3854 (class ipcp_transformation_t): New function summary.
3855 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3856 (ipa_get_agg_replacements_for_node): Likewise.
3858 2018-06-08 Martin Liska <mliska@suse.cz>
3860 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3862 (class funct_state_summary_t): New function_summary class.
3863 (has_function_state): Remove.
3864 (get_function_state): Likewise.
3865 (set_function_state): Likewise.
3866 (add_new_function): Likewise.
3867 (funct_state_summary_t::insert): New function.
3868 (duplicate_node_data): Remove.
3869 (remove_node_data): Remove.
3870 (funct_state_summary_t::duplicate): New function.
3871 (register_hooks): Create new funct_state_summaries.
3872 (pure_const_generate_summary): Use it.
3873 (pure_const_write_summary): Likewise.
3874 (pure_const_read_summary): Likewise.
3875 (propagate_pure_const): Likewise.
3876 (propagate_nothrow): Likewise.
3877 (dump_malloc_lattice): Likewise.
3878 (propagate_malloc): Likewise.
3879 (execute): Do not register hooks, just remove summary
3881 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3884 2018-06-08 Martin Liska <mliska@suse.cz>
3886 * ipa-reference.c (remove_node_data): Remove.
3887 (duplicate_node_data): Likewise.
3888 (class ipa_ref_var_info_summary_t): New class.
3889 (class ipa_ref_opt_summary_t): Likewise.
3890 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3891 (get_reference_optimization_summary): Use
3892 ipa_ref_opt_sum_summaries.
3893 (set_reference_vars_info): Remove.
3894 (set_reference_optimization_summary): Likewise.
3895 (ipa_init): Create summaries.
3896 (init_function_info): Use function summary.
3897 (ipa_ref_opt_summary_t::duplicate): New function.
3898 (ipa_ref_opt_summary_t::remove): New function.
3899 (get_read_write_all_from_node): Fix GNU coding style.
3900 (propagate): Use function summary.
3901 (write_node_summary_p): Fix GNU coding style.
3902 (stream_out_bitmap): Likewise.
3903 (ipa_reference_read_optimization_summary): Use function summary.
3904 (ipa_reference_c_finalize): Do not release hooks.
3906 2018-06-08 Martin Liska <mliska@suse.cz>
3908 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3909 (analyze_function_body): Extract multiple calls of get_create.
3910 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3911 * ipa-inline.c (recursive_inlining): Use ::get method.
3912 * ipa-inline.h (estimate_edge_growth): Likewise.
3914 2018-06-08 Martin Liska <mliska@suse.cz>
3916 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3918 (hsa_function_summary::hsa_function_summary): Use the new enum
3920 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3921 * hsa-gen.c (hsa_get_host_function): Likewise.
3922 (get_brig_function_name): Likewise.
3923 * ipa-hsa.c (process_hsa_functions): Likewise.
3924 (ipa_hsa_write_summary): Likewise.
3925 * symbol-summary.h (symtab_duplication): Use ::get function/
3926 (get): New function.
3928 2018-06-08 Martin Liska <mliska@suse.cz>
3930 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3932 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3933 (hsa_register_kernel): Likewise.
3934 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3935 * hsa-gen.c (hsa_get_host_function): Likewise.
3936 (get_brig_function_name): Likewise.
3937 (generate_hsa): Likewise.
3938 (pass_gen_hsail::execute): Likewise.
3939 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3940 (devirtualization_time_bonus): Likewise.
3941 (ipcp_propagate_stage): Likewise.
3942 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3943 (edge_set_predicate): Likewise.
3944 (evaluate_conditions_for_known_args): Likewise.
3945 (evaluate_properties_for_edge): Likewise.
3946 (ipa_fn_summary::reset): Likewise.
3947 (ipa_fn_summary_t::duplicate): Likewise.
3948 (dump_ipa_call_summary): Likewise.
3949 (ipa_dump_fn_summary): Likewise.
3950 (analyze_function_body): Likewise.
3951 (compute_fn_summary): Likewise.
3952 (estimate_edge_devirt_benefit): Likewise.
3953 (estimate_edge_size_and_time): Likewise.
3954 (estimate_calls_size_and_time): Likewise.
3955 (estimate_node_size_and_time): Likewise.
3956 (inline_update_callee_summaries): Likewise.
3957 (remap_edge_change_prob): Likewise.
3958 (remap_edge_summaries): Likewise.
3959 (ipa_merge_fn_summary_after_inlining): Likewise.
3960 (ipa_update_overall_fn_summary): Likewise.
3961 (read_ipa_call_summary): Likewise.
3962 (inline_read_section): Likewise.
3963 (write_ipa_call_summary): Likewise.
3964 (ipa_fn_summary_write): Likewise.
3965 (ipa_free_fn_summary): Likewise.
3966 * ipa-hsa.c (process_hsa_functions): Likewise.
3967 (ipa_hsa_write_summary): Likewise.
3968 (ipa_hsa_read_section): Likewise.
3969 * ipa-icf.c (sem_function::merge): Likewise.
3970 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3971 (do_estimate_edge_time): Likewise.
3972 (estimate_size_after_inlining): Likewise.
3973 (estimate_growth): Likewise.
3974 (growth_likely_positive): Likewise.
3975 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3976 (inline_call): Likewise.
3977 * ipa-inline.c (caller_growth_limits): Likewise.
3978 (can_inline_edge_p): Likewise.
3979 (can_inline_edge_by_limits_p): Likewise.
3980 (compute_uninlined_call_time): Likewise.
3981 (compute_inlined_call_time): Likewise.
3982 (want_inline_small_function_p): Likewise.
3983 (edge_badness): Likewise.
3984 (update_caller_keys): Likewise.
3985 (update_callee_keys): Likewise.
3986 (recursive_inlining): Likewise.
3987 (inline_small_functions): Likewise.
3988 (inline_to_all_callers_1): Likewise.
3989 (dump_overall_stats): Likewise.
3990 (early_inline_small_functions): Likewise.
3991 (early_inliner): Likewise.
3992 * ipa-inline.h (estimate_edge_growth): Likewise.
3993 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3994 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3995 * ipa-prop.h (IPA_NODE_REF): Likewise.
3996 (IPA_EDGE_REF): Likewise.
3997 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3998 (propagate_malloc): Likewise.
3999 * ipa-split.c (execute_split_functions): Likewise.
4000 * symbol-summary.h: Rename get to get_create.
4002 (get_create): Likewise.
4003 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4005 2018-06-08 Martin Liska <mliska@suse.cz>
4007 * symbol-summary.h (release): Move definition out of class
4009 (symtab_removal): Likewise.
4010 (symtab_duplication): Likewise.
4012 2018-06-08 Martin Liska <mliska@suse.cz>
4014 * symbol-summary.h (function_summary): Move constructor
4015 implementation out of class declaration.
4016 (release): Likewise.
4017 (symtab_insertion): Likewise.
4018 (symtab_removal): Likewise.
4019 (symtab_duplication): Likewise.
4022 2018-06-08 Martin Liska <mliska@suse.cz>
4024 * Makefile.in: Remove support for MPX (macros, related functions,
4025 fields in cgraph_node, ...).
4026 * builtin-types.def (BT_BND): Likewise.
4027 (BT_FN_BND_CONST_PTR): Likewise.
4028 (BT_FN_CONST_PTR_BND): Likewise.
4029 (BT_FN_VOID_PTR_BND): Likewise.
4030 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4031 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4032 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4033 (expand_builtin_mempcpy_with_bounds): Likewise.
4034 (expand_builtin_memset_with_bounds): Likewise.
4035 (expand_builtin_memset_args): Likewise.
4036 (std_expand_builtin_va_start): Likewise.
4037 (expand_builtin): Likewise.
4038 (expand_builtin_with_bounds): Likewise.
4039 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4040 (DEF_LIB_BUILTIN_CHKP): Likewise.
4041 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4042 (DEF_CHKP_BUILTIN): Likewise.
4043 (BUILT_IN_MEMCPY): Likewise.
4044 (BUILT_IN_MEMMOVE): Likewise.
4045 (BUILT_IN_MEMPCPY): Likewise.
4046 (BUILT_IN_MEMSET): Likewise.
4047 (BUILT_IN_STPCPY): Likewise.
4048 (BUILT_IN_STRCAT): Likewise.
4049 (BUILT_IN_STRCHR): Likewise.
4050 (BUILT_IN_STRCPY): Likewise.
4051 (BUILT_IN_STRLEN): Likewise.
4052 (BUILT_IN_MEMCPY_CHK): Likewise.
4053 (BUILT_IN_MEMMOVE_CHK): Likewise.
4054 (BUILT_IN_MEMPCPY_CHK): Likewise.
4055 (BUILT_IN_MEMSET_CHK): Likewise.
4056 (BUILT_IN_STPCPY_CHK): Likewise.
4057 (BUILT_IN_STRCAT_CHK): Likewise.
4058 (BUILT_IN_STRCPY_CHK): Likewise.
4059 * calls.c (store_bounds): Likewise.
4060 (emit_call_1): Likewise.
4061 (special_function_p): Likewise.
4062 (maybe_warn_nonstring_arg): Likewise.
4063 (initialize_argument_information): Likewise.
4064 (finalize_must_preallocate): Likewise.
4065 (compute_argument_addresses): Likewise.
4066 (expand_call): Likewise.
4067 * cfgexpand.c (expand_call_stmt): Likewise.
4068 (expand_return): Likewise.
4069 (expand_gimple_stmt_1): Likewise.
4070 (pass_expand::execute): Likewise.
4071 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4072 (cgraph_node::remove): Likewise.
4073 (cgraph_node::dump): Likewise.
4074 (cgraph_node::verify_node): Likewise.
4075 * cgraph.h (chkp_function_instrumented_p): Likewise.
4076 (symtab_node::get_alias_target): Likewise.
4077 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4078 (cgraph_local_p): Likewise.
4079 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4080 (cgraph_edge::rebuild_references): Likewise.
4081 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4082 (walk_polymorphic_call_targets): Likewise.
4083 (cgraph_node::expand_thunk): Likewise.
4084 (symbol_table::output_weakrefs): Likewise.
4085 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4086 (ix86_handle_option): Likewise.
4087 * config/i386/constraints.md: Likewise.
4088 * config/i386/i386-builtin-types.def (BND): Likewise.
4092 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4093 (BDESC_FIRST): Likewise.
4095 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4096 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4097 * config/i386/i386.c (enum reg_class): Likewise.
4098 (ix86_target_string): Likewise.
4099 (ix86_option_override_internal): Likewise.
4100 (ix86_conditional_register_usage): Likewise.
4101 (ix86_valid_target_attribute_inner_p): Likewise.
4102 (ix86_set_indirect_branch_type): Likewise.
4103 (ix86_set_current_function): Likewise.
4104 (ix86_function_arg_regno_p): Likewise.
4105 (init_cumulative_args): Likewise.
4106 (ix86_function_arg_advance): Likewise.
4107 (ix86_function_arg): Likewise.
4108 (ix86_pass_by_reference): Likewise.
4109 (ix86_function_value_regno_p): Likewise.
4110 (ix86_function_value_1): Likewise.
4111 (ix86_function_value_bounds): Likewise.
4112 (ix86_return_in_memory): Likewise.
4113 (ix86_setup_incoming_vararg_bounds): Likewise.
4114 (ix86_va_start): Likewise.
4115 (indirect_thunk_need_prefix): Likewise.
4116 (print_reg): Likewise.
4117 (ix86_print_operand): Likewise.
4118 (ix86_expand_call): Likewise.
4119 (ix86_output_function_return): Likewise.
4120 (reg_encoded_number): Likewise.
4121 (BDESC_VERIFYS): Likewise.
4122 (ix86_init_mpx_builtins): Likewise.
4123 (ix86_init_builtins): Likewise.
4124 (ix86_emit_cmove): Likewise.
4125 (ix86_emit_move_max): Likewise.
4126 (ix86_expand_builtin): Likewise.
4127 (ix86_builtin_mpx_function): Likewise.
4128 (ix86_get_arg_address_for_bt): Likewise.
4129 (ix86_load_bounds): Likewise.
4130 (ix86_store_bounds): Likewise.
4131 (ix86_load_returned_bounds): Likewise.
4132 (ix86_store_returned_bounds): Likewise.
4133 (ix86_class_likely_spilled_p): Likewise.
4134 (ix86_hard_regno_mode_ok): Likewise.
4135 (x86_order_regs_for_local_alloc): Likewise.
4136 (ix86_mitigate_rop): Likewise.
4137 (ix86_bnd_prefixed_insn_p): Likewise.
4138 (ix86_mpx_bound_mode): Likewise.
4139 (ix86_make_bounds_constant): Likewise.
4140 (ix86_initialize_bounds): Likewise.
4141 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4142 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4143 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4144 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4145 (TARGET_CHKP_BOUND_MODE): Likewise.
4146 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4147 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4148 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4149 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4150 * config/i386/i386.h (TARGET_MPX): Likewise.
4151 (TARGET_MPX_P): Likewise.
4152 (VALID_BND_REG_MODE): Likewise.
4153 (FIRST_BND_REG): Likewise.
4154 (LAST_BND_REG): Likewise.
4155 (enum reg_class): Likewise.
4156 (BND_REG_P): Likewise.
4157 (BND_REGNO_P): Likewise.
4158 (BNDmode): Likewise.
4159 (ADJUST_INSN_LENGTH): Likewise.
4160 * config/i386/i386.md: Likewise.
4161 * config/i386/i386.opt: Likewise.
4162 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4163 (defined): Likewise.
4164 (LINK_MPX): Likewise.
4165 (MPX_SPEC): Likewise.
4166 (LIBMPX_SPEC): Likewise.
4167 (LIBMPXWRAPPERS_SPEC): Likewise.
4168 (CHKP_SPEC): Likewise.
4169 * config/i386/predicates.md: Likewise.
4170 * dbxout.c (dbxout_type): Likewise.
4171 * doc/extend.texi: Likewise.
4172 * doc/invoke.texi: Likewise.
4173 * doc/md.texi: Likewise.
4174 * doc/tm.texi: Likewise.
4175 * doc/tm.texi.in: Likewise.
4176 * dwarf2out.c (is_base_type): Likewise.
4177 (gen_formal_types_die): Likewise.
4178 (gen_subprogram_die): Likewise.
4179 (gen_type_die_with_usage): Likewise.
4180 (gen_decl_die): Likewise.
4181 (dwarf2out_late_global_decl): Likewise.
4182 * expr.c (expand_assignment): Likewise.
4183 (emit_storent_insn): Likewise.
4184 (store_expr_with_bounds): Likewise.
4185 (store_expr): Likewise.
4186 (expand_expr_real_1): Likewise.
4187 * expr.h (store_expr_with_bounds): Likewise.
4188 * function.c (use_register_for_decl): Likewise.
4189 (struct bounds_parm_data): Likewise.
4190 (assign_parms_augmented_arg_list): Likewise.
4191 (assign_parm_find_entry_rtl): Likewise.
4192 (assign_parm_is_stack_parm): Likewise.
4193 (assign_parm_load_bounds): Likewise.
4194 (assign_bounds): Likewise.
4195 (assign_parms): Likewise.
4196 (expand_function_start): Likewise.
4197 * gcc.c (CHKP_SPEC): Likewise.
4198 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4199 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4200 (wrestrict_dom_walker::check_call): Likewise.
4201 * gimple.c (gimple_build_call_from_tree): Likewise.
4202 * gimple.h (enum gf_mask): Likewise.
4203 (gimple_call_with_bounds_p): Likewise.
4204 (gimple_call_set_with_bounds): Likewise.
4205 * gimplify.c (gimplify_init_constructor): Likewise.
4206 * ipa-cp.c (initialize_node_lattices): Likewise.
4207 (propagate_constants_across_call): Likewise.
4208 (find_more_scalar_values_for_callers_subset): Likewise.
4209 * ipa-hsa.c (process_hsa_functions): Likewise.
4210 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4211 * ipa-icf.c (sem_function::merge): Likewise.
4212 * ipa-inline.c (early_inliner): Likewise.
4213 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4214 (warn_function_cold): Likewise.
4215 (propagate_pure_const): Likewise.
4216 * ipa-ref.h (enum GTY): Likewise.
4217 * ipa-split.c (find_retbnd): Likewise.
4218 (consider_split): Likewise.
4219 (split_function): Likewise.
4220 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4221 * ipa.c (walk_polymorphic_call_targets): Likewise.
4222 (symbol_table::remove_unreachable_nodes): Likewise.
4223 (process_references): Likewise.
4224 (cgraph_build_static_cdtor_1): Likewise.
4225 * lto-cgraph.c (lto_output_node): Likewise.
4226 (output_refs): Likewise.
4227 (compute_ltrans_boundary): Likewise.
4228 (input_overwrite_node): Likewise.
4229 (input_node): Likewise.
4230 (input_cgraph_1): Likewise.
4231 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4232 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4233 (class pass_chkp_instrumentation_passes): Likewise.
4234 (make_pass_chkp_instrumentation_passes): Likewise.
4235 * passes.def: Likewise.
4236 * rtl.h (struct GTY): Likewise.
4237 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4238 * stor-layout.c (layout_type): Likewise.
4239 * symtab.c: Likewise.
4240 * target.def: Likewise.
4241 * targhooks.c (default_chkp_bound_type): Likewise.
4242 (default_chkp_bound_mode): Likewise.
4243 (default_builtin_chkp_function): Likewise.
4244 (default_chkp_function_value_bounds): Likewise.
4245 (default_chkp_make_bounds_constant): Likewise.
4246 (default_chkp_initialize_bounds): Likewise.
4247 * targhooks.h (default_chkp_bound_type): Likewise.
4248 (default_chkp_bound_mode): Likewise.
4249 (default_builtin_chkp_function): Likewise.
4250 (default_chkp_function_value_bounds): Likewise.
4251 (default_chkp_make_bounds_constant): Likewise.
4252 (default_chkp_initialize_bounds): Likewise.
4253 * toplev.c (compile_file): Likewise.
4254 (process_options): Likewise.
4255 * tree-core.h (DEF_BUILTIN): Likewise.
4256 (DEF_BUILTIN_CHKP): Likewise.
4257 * tree-inline.c (declare_return_variable): Likewise.
4258 (remap_gimple_stmt): Likewise.
4259 (copy_bb): Likewise.
4260 (initialize_inlined_parameters): Likewise.
4261 (expand_call_inline): Likewise.
4262 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4263 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4264 (make_pass_ipa_chkp_produce_thunks): Likewise.
4265 (make_pass_chkp): Likewise.
4266 (make_pass_chkp_opt): Likewise.
4267 (make_pass_chkp_instrumentation_passes): Likewise.
4268 * tree-pretty-print.c (dump_generic_node): Likewise.
4269 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4270 * tree-ssa-dce.c (propagate_necessity): Likewise.
4271 (eliminate_unnecessary_stmts): Likewise.
4272 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4273 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4274 * tree-ssa-sccvn.h: Likewise.
4275 * tree-ssa-strlen.c (get_string_length): Likewise.
4276 (valid_builtin_call): Likewise.
4277 (adjust_last_stmt): Likewise.
4278 (handle_builtin_strchr): Likewise.
4279 (handle_builtin_strcpy): Likewise.
4280 (handle_builtin_stxncpy): Likewise.
4281 (handle_builtin_memcpy): Likewise.
4282 (handle_builtin_strcat): Likewise.
4283 (strlen_check_and_optimize_stmt): Likewise.
4284 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4285 * tree-streamer-in.c: Likewise.
4286 * tree-streamer.c (record_common_node): Likewise.
4287 * tree.c (tree_code_size): Likewise.
4288 (wide_int_to_tree_1): Likewise.
4289 (type_contains_placeholder_1): Likewise.
4290 (build_common_tree_nodes): Likewise.
4291 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4292 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4293 (POINTER_BOUNDS_P): Likewise.
4294 (BOUNDED_TYPE_P): Likewise.
4295 (BOUNDED_P): Likewise.
4296 (CALL_WITH_BOUNDS_P): Likewise.
4297 (pointer_bounds_type_node): Likewise.
4298 * value-prof.c (gimple_ic): Likewise.
4299 * var-tracking.c (vt_add_function_parameters): Likewise.
4300 * varasm.c (make_decl_rtl): Likewise.
4301 (assemble_start_function): Likewise.
4302 (output_constant): Likewise.
4303 (maybe_assemble_visibility): Likewise.
4304 * varpool.c (ctor_for_folding): Likewise.
4305 * chkp-builtins.def: Remove.
4306 * ipa-chkp.c: Remove.
4307 * ipa-chkp.h: Remove.
4308 * rtl-chkp.c: Remove.
4309 * rtl-chkp.h: Remove.
4310 * tree-chkp-opt.c: Remove.
4311 * tree-chkp.c: Remove.
4312 * tree-chkp.h: Remove.
4314 2018-06-07 Carl Love <cel@us.ibm.com>
4316 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4317 vextract_fp_from_shortl): Add BE support.
4319 2018-06-07 Paul Koning <ni1d@arrl.net>
4321 * compare-elim.c (try_merge_compare): Don't merge compare if
4322 address contains a side effect.
4323 (try_eliminate_compare): Likewise.
4325 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4327 * config.gcc: Support "tremont".
4328 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4329 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4331 * config/i386/i386.c (m_TREMONT): Define.
4332 (processor_target_table): Add "tremont".
4333 (PTA_TREMONT): Define.
4334 (ix86_lea_outperforms): Add TARGET_TREMONT.
4335 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4336 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4337 and M_INTEL_GOLDMONT_PLUS.
4338 (fold_builtin_cpu): Add "tremont".
4339 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4340 (ix86_option_override_internal): Add "tremont".
4341 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4342 (processor_type): Add PROCESSOR_TREMONT.
4343 * config/i386/x86-tune.def: Add m_TREMONT.
4344 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4346 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4348 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4349 symbol defined for msp430i* devices to be lower case.
4351 gcc/testsuite/gcc.target/msp430/
4352 * msp430i-device-symbol.c: New test.
4353 * msp430f-device-symbol.c: New test.
4354 * msp430.h: New test header file.
4356 2018-06-07 Richard Biener <rguenther@suse.de>
4358 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4359 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4360 Properly wrap signed arithmetic if overflow wraps.
4362 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4364 PR tree-optimization/69615
4365 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4366 of a cast from a same precision integral SSA_NAME in a bb dominated
4367 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4368 cast to utype if rhs2 has already a compatible type.
4370 2018-06-07 Richard Biener <rguenther@suse.de>
4372 PR tree-optimization/85935
4373 * graphite-scop-detection.c (find_params_in_bb): Analyze
4374 condition operands with respect to the correct loop. Assert
4375 the analysis doesn't fail.
4377 2018-06-04 Carl Love <cel@us.ibm.com>
4379 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4380 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4381 as it is slightly cheaper.
4382 (first_match_or_eos_index_<mode>):
4383 Calculate index using natural element order.
4384 (first_match_index_<mode>):
4385 Calculate index using natural element order.
4386 (first_match_or_eos_index_<mode>):
4387 Calculate index using natural order.
4388 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4389 for BE and LE modes.
4390 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4391 P9V_BUILTIN_VCLZLSBB_V16QI.
4392 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4395 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4397 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4398 indentation and line wrap for many prototypes. Add missing
4399 @smallexample directives around block of prototypes for vec_xl and
4402 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4404 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4405 track if we pass or return IEEE 128-bit floating point.
4406 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4407 C++ mangling that is compatible with GCC 8.1.
4408 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4409 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4410 floating point types.
4411 (rs6000_function_arg_advance_1): Likewise.
4412 (rs6000_mangle_type): Optionally generate mangled names that match
4413 what GCC 8.1 generated for IEEE 128-bit floating point types.
4414 (rs6000_globalize_decl_name): If we have an external function that
4415 passes or returns IEEE 128-bit types, generate a weak reference
4416 from the mangled name used in GCC 8.1 to the current mangled
4418 (rs6000_init_builtins): Make __ibm128 use the long double type if
4419 long double is IBM extended double. Make __float128 use the long
4420 double type if long double is IEEE 128-bit.
4423 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4424 macro for __ibm128 built-in functions.
4425 (PACK_IF): Add __ibm128 pack/unpack functions.
4426 (UNPACK_IF): Likewise.
4427 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4428 enable long double built-in functions if long double is IEEE
4429 128-bit floating point.
4430 (rs6000_invalid_builtin): Update long double built-in function
4432 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4433 functions, adjust the built-in function to use the long double
4434 built-in function if __ibm128 and long double are the same type.
4435 * doc/extend.texi (PowerPC builtins): Update documention for
4436 __builtin_{,un}pack_longdouble. Add documentation for
4437 __builtin_{,un}pack_ibm128.
4439 2018-06-06 Jim Wilson <jimw@sifive.com>
4441 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4442 (struct machine_function): New field interrupt_mode.
4443 (riscv_handle_type_attribute): New function. Add forward declaration.
4444 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4445 (riscv_expand_epilogue): Check interrupt_mode field.
4446 (riscv_set_current_function): Check interrupt attribute args and
4447 set interrupt_mode field.
4448 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4449 (riscv_sret, riscv_uret): New.
4450 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4451 new arguments to interrupt attribute.
4453 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4456 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4457 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4459 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4461 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4462 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4463 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4464 several redundant entries.
4466 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4468 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4469 type from "rtx" to "rtx_insn *".
4470 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4471 for local "call_insn", removing cast.
4472 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4474 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4476 PR tree-optimization/86066
4477 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4478 for BIT_INSERT_EXPR stores.
4480 2018-06-06 Richard Biener <rguenther@suse.de>
4482 PR tree-optimization/86062
4483 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4484 component refs ontop
4485 of to be offsetted base.
4487 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4489 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4490 to be static and remove check on interrupt attribute name.
4492 gcc/testsuite/gcc.target/msp430/
4493 * function-attributes-4.c: New test.
4494 * static-interrupts.c: New test.
4496 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4498 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4499 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4501 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4504 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4506 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4507 Remove second argument, change how error is called.
4508 (aarch64_layout_arg): Remove second argument from
4509 aarch64_err_no_fpadvsimd call.
4510 (aarch64_init_cumulative_args): Ditto.
4511 (aarch64_gimplify_va_arg_expr): Ditto.
4512 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4514 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4516 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4517 (*simple_return_indirect_internal<mode>): Rename from
4518 simple_return_indirect_internal. Use W mode iterator.
4519 (rstorssp): New expander.
4520 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4521 (clrssbsy): New expander.
4522 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4524 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4526 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4528 (cmse_check_pointed_object): Likewise.
4530 2018-06-05 Martin Liska <mliska@suse.cz>
4532 PR gcov-profile/47618
4533 * doc/invoke.texi: Document how -fprofile-dir format
4536 2018-06-05 Richard Biener <rguenther@suse.de>
4538 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4539 removal pretend DOM info isn't available so we do not update
4540 it and only remove edges, not dominated blocks. Actually free
4541 DOM info in case we removed something. Remove unreachable blocks.
4542 (mfb_keep_latches): Work with either DOM info or marked backedges.
4543 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4544 first. Mark backedges if DOM info isn't available.
4545 (Re-)compute DOM info after cleanup_control_flow_pre.
4547 2018-06-05 Richard Biener <rguenther@suse.de>
4549 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4550 (locus_discrim_hasher::hash): Adjust.
4551 (locus_discrim_hasher::equal): Likewise.
4552 (next_discriminator_for_locus): Work on line directly.
4553 (same_line_p): Pass in expanded locus1 as well.
4554 (assign_discriminators): Avoid redundant location expansions.
4556 2018-06-05 Richard Biener <rguenther@suse.de>
4558 PR tree-optimization/86046
4559 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4560 if required after clearing TREE_ADDRESSABLE.
4562 2018-06-05 Richard Biener <rguenther@suse.de>
4564 PR tree-optimization/86047
4565 * tree-ssa-loop.c (for_each_index): Glob handling of all
4566 decls and constants and really handle all of them.
4568 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4571 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4572 qualifier_void_pointer and qualifier_const_void_pointer.
4573 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4574 (arm_init_builtins): Handle the above.
4575 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4576 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4577 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4580 2018-06-05 Martin Liska <mliska@suse.cz>
4582 * auto-profile.c (read_autofdo_file): Do not use
4583 gcov_ctr_summary struct.
4584 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4585 * coverage.c (struct counts_entry): Likewise.
4586 (read_counts_file): Read just single summary entry.
4587 (get_coverage_counts): Use gcov_summary struct.
4588 * coverage.h (get_coverage_counts): Likewise.
4589 * gcov-dump.c (dump_working_sets): Likewise.
4590 (tag_summary): Dump just single summary.
4591 * gcov-io.c (gcov_write_summary): Write just histogram
4593 (gcov_read_summary): Read just single summary.
4594 (compute_working_sets): Use gcov_summary struct.
4595 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4596 of GCOV_COUNTERS_SUMMABLE.
4597 (GCOV_COUNTERS_SUMMABLE): Remove.
4598 (GCOV_FIRST_VALUE_COUNTER): Replace with
4599 GCOV_COUNTER_V_INTERVAL.
4600 (struct gcov_ctr_summary): Remove.
4601 (struct gcov_summary): Directly use fields of former
4603 (compute_working_sets): Use gcov_summary struct.
4604 * gcov.c (read_count_file): Do not use ctrs fields.
4605 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4607 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4609 * profile.c: Likewise.
4610 * profile.h: Likewise.
4612 2018-06-05 Martin Liska <mliska@suse.cz>
4614 PR gcov-profile/84846
4615 * gcov.c (output_lines): Print working directory only
4616 in intermediate format.
4618 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4620 * config/s390/s390-builtin-types.def: Add void function type.
4621 * config/s390/s390-builtins.def: Use the function type for the
4624 2018-06-04 Jim Wilson <jimw@sifive.com>
4626 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4628 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4629 and maybe_eh_return. Change regno to unsigned int. Use new args to
4630 handle EH_RETURN_DATA_REGNO registers properly.
4631 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4632 (riscv_expand_epilogue): Update comment. Change argument name and
4633 type. Update code to use new name and type. Pass new args to
4634 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4636 * config/riscv/riscv.md (NORMAL_RETURN): New.
4637 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4638 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4639 (eh_return): Call gen_eh_return_internal and emit barrier.
4640 (eh_return_internal): Call riscv_expand_epilogue.
4642 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4644 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4645 bit_insertion field and declare can_be_merged_into method.
4646 (merged_store_group::can_be_merged_into): New method.
4647 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4648 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4649 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4651 2018-06-04 Richard Biener <rguenther@suse.de>
4653 PR tree-optimization/85955
4654 * builtins.c (fold_builtin_sincos): Convert pointers to
4655 destination to appropriate type before dereferencing.
4657 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4659 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4661 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4663 * expr.c (expand_expr_real_1): Force the operand into memory if
4664 its TYPE_MODE is BLKmode and if there is no integer mode for
4665 the number of bits being extracted.
4667 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4671 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4672 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4674 2018-06-04 Richard Biener <rguenther@suse.de>
4676 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4677 (cleanup_tree_cfg_noloop): ... single caller. Do
4678 start_recording_case_labels later.
4680 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4682 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4683 to _IMMINTRIN_H_INCLUDED.
4684 * config/i386/pconfigintrin.h: Ditto.
4685 * config/i386/waitpkgintrin.h: Ditto.
4686 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4687 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4688 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4689 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4690 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4691 waitpkgintrin.h and cldemoteintrin.h.
4693 2018-06-04 Richard Biener <rguenther@suse.de>
4695 PR tree-optimization/86038
4696 * tracer.c (find_best_successor): Check probability for
4697 being initialized, bail out if not.
4699 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4702 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4703 of bits to ignore when comparing architectures.
4705 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4707 PR tree-optimization/69615
4708 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4709 maximum or minimum of the type, try to merge it also as if
4710 range1 is + [-, x - 1] or + [x + 1, -].
4713 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4715 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4717 PR tree-optimization/86034
4718 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4719 the unsigned bitfield type in a bit insertion sequence if it does not
4720 have a larger precision than the bitfield size.
4721 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4723 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4725 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4727 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4729 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4730 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4731 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4732 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4734 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4736 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4737 Disable -fdelete-null-pointer-checks for ELF toolchain.
4739 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4740 Kito Cheng <kito.cheng@gmail.com>
4742 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4743 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4744 (nds32*-*-*): Add glibc and uclibc conditions.
4745 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4746 (TARGET_EXCEPT_UNWIND_INFO): Define.
4747 * config/nds32/elf.h: New file.
4748 * config/nds32/linux.h: New file.
4749 * config/nds32/nds32-elf.opt: New file.
4750 * config/nds32/nds32-linux.opt: New file.
4751 * config/nds32/nds32-fp-as-gp.c
4752 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4753 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4755 (nds32_asm_file_end): Ditto.
4756 (nds32_print_operand): Ditto.
4757 (nds32_insert_attributes): Ditto.
4758 (nds32_init_libfuncs): New function.
4759 (TARGET_HAVE_TLS): Define.
4760 (TARGET_INIT_LIBFUNCS): Define.
4761 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4763 (TARGET_ELF): Apply different mcmodel setting.
4764 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4765 been migrated into elf.h and linux.h files.
4766 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4767 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4768 (mcmodel): The content has been migrated into nds32-elf.opt and
4769 nds32-linux.opt files.
4770 * config/nds32/t-elf: New file.
4771 * config/nds32/t-linux: New file.
4773 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4774 Shiva Chen <shiva0217@gmail.com>
4776 * config/nds32/constants.md (unspec_volatile_element): Add
4777 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4778 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4780 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4781 (make_pass_nds32_fp_as_gp): Declare.
4782 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4784 (nds32_asm_function_end_prologue): Remove unused asm output.
4785 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4786 (nds32_asm_file_start): Output necessary fp_as_gp information.
4787 (nds32_option_override): Adjust register usage.
4788 (nds32_expand_prologue): Consider fp_as_gp situation.
4789 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4790 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4793 (simple_return): Ditto.
4794 (omit_fp_begin): Output special directive for fp_as_gp.
4795 (omit_fp_end): Output special directive for fp_as_gp.
4796 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4797 mforbid-fp-as-gp): New options.
4799 2018-06-01 Mark Wielaard <mark@klomp.org>
4801 * dwarf2out.c (dwarf2out_finish): Remove generation of
4802 DW_AT_loclists_base.
4804 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4806 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4807 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4808 (struct merged_store_group): Add bit_insertion field.
4809 (dump_char_array): Use standard hexadecimal format.
4810 (merged_store_group::merged_store_group): Set bit_insertion to false.
4811 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4812 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4813 also print the mask in the dump file.
4814 (pass_store_merging::gate): Minor tweak.
4815 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4816 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4817 stores with INTEGER_CST stores.
4818 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4819 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4820 and use it throughout. Generate bit insertion sequences if need be.
4821 (pass_store_merging::process_store): Remove redundant condition.
4822 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4824 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4826 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4827 the 128-bit floating point types. Fix function comment.
4829 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4831 * config/aarch64/aarch64-simd.md
4832 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4834 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4837 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4839 PR tree-optimization/85989
4840 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4842 (backprop::intersect_uses): Check it when deciding whether this
4843 is a backedge reference.
4844 (backprop::process_block): Add each phi to m_visited_phis
4845 after visiting it, then clear it at the end.
4847 2018-06-01 Richard Biener <rguenther@suse.de>
4849 * tree-vectorizer.h (vect_dr_stmt): New function.
4850 (vect_get_load_cost): Adjust.
4851 (vect_get_store_cost): Likewise.
4852 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4853 Use vect_dr_stmt instead of DR_SMTT.
4854 (vect_record_base_alignments): Likewise.
4855 (vect_calculate_target_alignment): Likewise.
4856 (vect_compute_data_ref_alignment): Likewise and make static.
4857 (vect_update_misalignment_for_peel): Likewise.
4858 (vect_verify_datarefs_alignment): Likewise.
4859 (vector_alignment_reachable_p): Likewise.
4860 (vect_get_data_access_cost): Likewise. Pass down
4861 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4862 (vect_get_peeling_costs_all_drs): Likewise.
4863 (vect_peeling_hash_get_lowest_cost): Likewise.
4864 (vect_enhance_data_refs_alignment): Likewise.
4865 (vect_find_same_alignment_drs): Likewise.
4866 (vect_analyze_data_refs_alignment): Likewise.
4867 (vect_analyze_group_access_1): Likewise.
4868 (vect_analyze_group_access): Likewise.
4869 (vect_analyze_data_ref_access): Likewise.
4870 (vect_analyze_data_ref_accesses): Likewise.
4871 (vect_vfa_segment_size): Likewise.
4872 (vect_small_gap_p): Likewise.
4873 (vectorizable_with_step_bound_p): Likewise.
4874 (vect_prune_runtime_alias_test_list): Likewise.
4875 (vect_analyze_data_refs): Likewise.
4876 (vect_supportable_dr_alignment): Likewise.
4877 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4878 (vect_gen_prolog_loop_niters): Likewise.
4879 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4880 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4882 (vect_recog_mask_conversion_pattern): Likewise.
4883 (vect_try_gather_scatter_pattern): Likewise.
4884 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4885 to vect_get_store_cost.
4886 (vect_get_store_cost): Get stmt_info instead of DR.
4887 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4888 (vect_get_load_cost): Get stmt_info instead of DR.
4890 2018-06-01 Richard Biener <rguenther@suse.de>
4893 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4894 vars wrapped in MEM_REFs.
4896 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4898 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4899 Fix subreg tests so that we only return a choice between
4900 GENERAL_REGS and FP_REGS if the original classes included both.
4902 2018-06-01 Richard Biener <rguenther@suse.de>
4905 * tree-ssa-structalias.c (get_function_part_constraint):
4906 Handle NULL fi->decl.
4907 (find_func_aliases_for_call): Properly handle indirect
4908 fi from direct call.
4909 (find_func_clobbers): Likewise.
4910 (ipa_pta_execute): Likewise.
4911 (create_variable_info_for): For functions that are ifunc_resolver
4912 resolve to a varinfo that contains the result of the resolver call.
4913 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4916 2018-05-31 Michael Collison <michael.collison@arm.com>
4918 * config/aarch64/aarch64.md:
4919 (*fix_to_zero_extenddfdi2): New pattern.
4920 * gcc.target/aarch64/fix_extend1.c: New testcase.
4922 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4926 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4927 and BUILT_IN_STRNCMP_EQ.
4928 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4929 BUILT_IN_STRNCMP_EQ.
4930 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4931 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4932 (gimple_fold_builtin): Likewise.
4933 * tree-ssa-strlen.c (compute_string_length): New function.
4934 (determine_min_obsize): New function.
4935 (handle_builtin_string_cmp): New function to handle calls to
4936 string compare functions.
4937 (strlen_optimize_stmt): Add handling to builtin string compare
4939 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4940 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4941 * tree.c (build_common_builtin_nodes): Add new defines of
4942 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4944 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4947 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4948 functions with naked attribute.
4950 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4952 * config/i386/sse.md (avx_vec_concat<mode>):
4953 Substitute concat_tg_mode mode attribute with xtg_mode.
4954 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4955 (concat_tg_mode): Remove mode attribute.
4957 2018-05-31 Martin Sebor <msebor@redhat.com>
4960 * calls.c (alloc_max_size): Correct a logic error/typo.
4961 Treat excessive arguments as infinite. Warn for invalid arguments.
4962 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4964 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4967 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4968 and movx for Haswell.
4970 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4971 Cesar Philippidis <cesar@codesourcery.com>
4974 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4975 when emitting error on private/firstprivate reductions.
4976 * omp-low.c (lower_omp_target): Avoid reference-type processing
4977 on pointers for firstprivate clause.
4979 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4981 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4984 * config/aarch64/aarch64-simd.md
4985 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4986 (aarch64_ld1_x3_<mode>): Likewise
4987 (aarch64_st1x2<VALLDIF:mode>): Likewise
4988 (aarch64_st1_x2_<mode>): Likewise
4989 (aarch64_st1x3<VALLDIF:mode>): Likewise
4990 (aarch64_st1_x3_<mode>): Likewise
4991 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4992 (vld1_s8_x3): Likewise.
4993 (vld1_u16_x3): Likewise.
4994 (vld1_s16_x3): Likewise.
4995 (vld1_u32_x3): Likewise.
4996 (vld1_s32_x3): Likewise.
4997 (vld1_u64_x3): Likewise.
4998 (vld1_s64_x3): Likewise.
4999 (vld1_f16_x3): Likewise.
5000 (vld1_f32_x3): Likewise.
5001 (vld1_f64_x3): Likewise.
5002 (vld1_p8_x3): Likewise.
5003 (vld1_p16_x3): Likewise.
5004 (vld1_p64_x3): Likewise.
5005 (vld1q_u8_x3): Likewise.
5006 (vld1q_s8_x3): Likewise.
5007 (vld1q_u16_x3): Likewise.
5008 (vld1q_s16_x3): Likewise.
5009 (vld1q_u32_x3): Likewise.
5010 (vld1q_s32_x3): Likewise.
5011 (vld1q_u64_x3): Likewise.
5012 (vld1q_s64_x3): Likewise.
5013 (vld1q_f16_x3): Likewise.
5014 (vld1q_f32_x3): Likewise.
5015 (vld1q_f64_x3): Likewise.
5016 (vld1q_p8_x3): Likewise.
5017 (vld1q_p16_x3): Likewise.
5018 (vld1q_p64_x3): Likewise.
5019 (vst1_s64_x2): Likewise.
5020 (vst1_u64_x2): Likewise.
5021 (vst1_f64_x2): Likewise.
5022 (vst1_s8_x2): Likewise.
5023 (vst1_p8_x2): Likewise.
5024 (vst1_s16_x2): Likewise.
5025 (vst1_p16_x2): Likewise.
5026 (vst1_s32_x2): Likewise.
5027 (vst1_u8_x2): Likewise.
5028 (vst1_u16_x2): Likewise.
5029 (vst1_u32_x2): Likewise.
5030 (vst1_f16_x2): Likewise.
5031 (vst1_f32_x2): Likewise.
5032 (vst1_p64_x2): Likewise.
5033 (vst1q_s8_x2): Likewise.
5034 (vst1q_p8_x2): Likewise.
5035 (vst1q_s16_x2): Likewise.
5036 (vst1q_p16_x2): Likewise.
5037 (vst1q_s32_x2): Likewise.
5038 (vst1q_s64_x2): Likewise.
5039 (vst1q_u8_x2): Likewise.
5040 (vst1q_u16_x2): Likewise.
5041 (vst1q_u32_x2): Likewise.
5042 (vst1q_u64_x2): Likewise.
5043 (vst1q_f16_x2): Likewise.
5044 (vst1q_f32_x2): Likewise.
5045 (vst1q_f64_x2): Likewise.
5046 (vst1q_p64_x2): Likewise.
5047 (vst1_s64_x3): Likewise.
5048 (vst1_u64_x3): Likewise.
5049 (vst1_f64_x3): Likewise.
5050 (vst1_s8_x3): Likewise.
5051 (vst1_p8_x3): Likewise.
5052 (vst1_s16_x3): Likewise.
5053 (vst1_p16_x3): Likewise.
5054 (vst1_s32_x3): Likewise.
5055 (vst1_u8_x3): Likewise.
5056 (vst1_u16_x3): Likewise.
5057 (vst1_u32_x3): Likewise.
5058 (vst1_f16_x3): Likewise.
5059 (vst1_f32_x3): Likewise.
5060 (vst1_p64_x3): Likewise.
5061 (vst1q_s8_x3): Likewise.
5062 (vst1q_p8_x3): Likewise.
5063 (vst1q_s16_x3): Likewise.
5064 (vst1q_p16_x3): Likewise.
5065 (vst1q_s32_x3): Likewise.
5066 (vst1q_s64_x3): Likewise.
5067 (vst1q_u8_x3): Likewise.
5068 (vst1q_u16_x3): Likewise.
5069 (vst1q_u32_x3): Likewise.
5070 (vst1q_u64_x3): Likewise.
5071 (vst1q_f16_x3): Likewise.
5072 (vst1q_f32_x3): Likewise.
5073 (vst1q_f64_x3): Likewise.
5074 (vst1q_p64_x3): Likewise.
5076 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5078 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5079 user_label_prefix to name.
5081 * tree-core.h: Update comment about the format of NAME string
5082 passed to handler in attribute_spec.
5084 * config/msp430/msp430.md: Remove erroneous subreg expression from
5085 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5086 zero_extend{q,h}isi2.
5088 2018-05-30 Borislav Petkov <bp@suse.de>
5090 * doc/extend.texi: Document some architecture specific
5091 constraints and sort entries.
5093 2018-05-30 Martin Sebor <msebor@redhat.com>
5096 * builtins.c (expand_builtin_stpcpy_1): New function.
5097 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5098 only if the former succeeds.
5100 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5102 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5105 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5107 * doc/invoke.texi (-flinker-output): Document
5109 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5111 * passes.c (ipa_write_summaries): Only modify statements if body
5113 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5114 incrementally linking.
5115 (ipa_passes): Likewise.
5116 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5117 pass down resolution info.
5118 * common.opt (flag_incremental_link): Update info.
5119 * gcc.c (plugin specs): Turn flinker-output=* to
5120 -plugin-opt=-linker-output-known
5121 * toplev.c (compile_file): Also cut compilation when doing incremental
5123 * flag-types. (enum lto_partition_model): Add
5124 LTO_LINKER_OUTPUT_NOLTOREL.
5125 (invoke.texi): Add -flinker-output docs.
5126 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5127 link same way as WPA; do not stream in dead initializers.
5129 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5130 darf2out_register_external_decl): Support incremental link.
5132 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5134 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5136 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5138 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5139 it down to simple_object_copy_lto_debug_sections.
5140 (run_gcc): Determine incremental LTO link time and configure
5141 lto1 into non-wpa mode, disable renaming of debug sections.
5143 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5145 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5146 descriptions of various incorrectly documented functions.
5148 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5151 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5152 address check not strict.
5154 2018-05-30 Richard Biener <rguenther@suse.de>
5156 PR tree-optimization/85964
5157 * tracer.c (better_p): Drop initialized count check, we only
5158 call the function with initialized counts now.
5159 (find_best_successor): Do find a best edge if one
5160 has uninitialized count.
5161 (find_best_predecessor): Likewise. Do BB frequency check only
5162 if count is initialized.
5164 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5166 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5167 (aarch64_ldrstr_offset_compare): New.
5168 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5169 load/store orderings.
5170 (aarch64_gen_adjusted_ldpstp): Likewise.
5172 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5174 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5175 Check for subset of GENERAL_REGS and FP_REGS.
5176 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5179 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5181 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5182 and wi::to_poly_offset. Add the current offset and then check
5183 whether the sum fits, rather than using an unchecked addition of
5184 a checked term. Check for a shwi rather than a uhwi.
5185 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5186 (store_constructor): Use poly_int_tree_p.
5187 (expand_expr_real_1): Likewise.
5188 * function.c (assign_temp): Likewise.
5189 * fold-const.c (const_binop): Use poly_int_tree_p and
5191 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5193 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5194 to_poly_offset for MEM offsets.
5195 * ipa-icf.c (sem_variable::equals): Likewise.
5196 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5197 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5198 wi::to_poly_offset for BIT_FIELD_REF offsets.
5199 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5201 * var-tracking.c (emit_note_insn_var_location): Use
5202 tree_to_poly_uint64.
5204 2018-05-29 Jim Wilson <jimw@sifive.com>
5206 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5208 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5211 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5212 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5214 (sse4_1_round<mode>2): Use nonimmediate_operand
5215 for operand 1 predicate.
5217 2018-05-29 Martin Sebor <msebor@redhat.com>
5218 Richard Biener <rguenther@suse.de>
5221 * calls.c (get_size_range): Call determine_value_range instead
5222 of get_value_range..
5223 * tree-vrp.h (determine_value_range): Declared new function.
5224 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5226 2018-05-29 Richard Biener <rguenther@suse.de>
5228 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5229 sure to use non-pattern stmts for get_earlier_stmt arguments.
5230 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5231 called on pattern stmts.
5232 (get_later_stmt): Likewise.
5234 2018-05-29 Martin Liska <mliska@suse.cz>
5236 PR gcov-profile/85759
5237 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5240 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5242 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5244 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5248 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5249 VEC_PACK_FLOAT_EXPR): New tree codes.
5250 * tree-pretty-print.c (op_code_prio): Handle
5251 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5252 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5253 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5254 * tree-inline.c (estimate_operator_cost): Likewise.
5255 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5256 * fold-const.c (const_binop): Likewise.
5257 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5258 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5259 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5260 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5261 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5262 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5263 * expr.c (expand_expr_real_2): Likewise.
5264 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5265 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5266 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5268 * optabs.c (expand_widen_pattern_expr): For
5269 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5270 sign from result type rather than operand's type.
5271 (expand_binop_directly): For vec_packu_float_optab and
5272 vec_packs_float_optab allow result type to be different from operand's
5274 * optabs-tree.c (optab_for_tree_code): Handle
5275 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5276 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5277 * tree-vect-generic.c (expand_vector_operations_1): Handle
5278 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5279 VEC_PACK_FLOAT_EXPR.
5280 * tree-vect-stmts.c (supportable_widening_operation): Handle
5282 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5283 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5284 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5285 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5286 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5288 (vec_pack<floatprefix>_float_<mode>): New expander.
5289 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5291 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5292 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5293 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5294 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5295 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5297 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5298 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5299 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5300 VEC_PACK_FLOAT_EXPR): Document.
5302 2018-05-29 Richard Biener <rguenther@suse.de>
5304 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5306 (stmt_vec_info_vec): Make pointer.
5307 (init_stmt_vec_info_vec): Remove.
5308 (free_stmt_vec_info_vec): Likewise.
5309 (set_stmt_vec_info_vec): New function.
5310 (free_stmt_vec_infos): Likewise.
5311 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5312 (set_vinfo_for_stmt): Likewise.
5313 (get_earlier_stmt): Likewise.
5314 (get_later_stmt): Likewise.
5315 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5316 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5317 (vec_info::~vec_info): Free stmt_vec_infos.
5318 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5319 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5320 (pass_slp_vectorize::execute): Likewise.
5321 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5322 (free_stmt_vec_info_vec): Likewise.
5323 (set_stmt_vec_info_vec): New function.
5324 (free_stmt_vec_infos): Likewise.
5325 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5326 the global stmt_vec_info_vec.
5327 * tree-parloops.c (gather_scalar_reductions): Use
5328 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5331 2018-05-29 Richard Biener <rguenther@suse.de>
5333 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5335 2018-05-29 Martin Liska <mliska@suse.cz>
5336 David Malcolm <dmalcolm@redhat.com>
5338 * vec.c (test_reverse): New.
5339 (vec_c_tests): Add new test.
5340 * vec.h (vl_ptr>::reverse): New function.
5342 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5344 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5346 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5349 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5351 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5353 2018-05-28 Richard Biener <rguenther@suse.de>
5355 PR tree-optimization/85933
5356 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5357 look at stmts marked as vectorizable.
5359 2018-05-28 Richard Biener <rguenther@suse.de>
5361 PR tree-optimization/85934
5362 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5363 vector boolean check before scalar optimization.
5365 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5367 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5370 2018-05-28 Mark Wielaard <mark@klomp.org>
5372 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5373 if it is an expression containing a minus sign.
5375 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5377 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5379 2018-05-27 Paul Koning <ni1d@arrl.net>
5381 * config/pdp11/pdp11.md (truncsihi2): Remove.
5383 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5384 Chung-Ju Wu <jasonwucj@gmail.com>
5386 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5388 (unaligned_store_dw): Ditto.
5389 * config/nds32/nds32-memory-manipulation.c
5390 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5391 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5392 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5393 (emit_setmem_word_loop): Rename to ...
5394 (emit_setmem_doubleword_loop): ... this.
5395 (nds32_gen_dup_4_byte_to_word_value): New function.
5396 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5397 (nds32_expand_setmem_loop): Refine implementation.
5398 (nds32_expand_setmem_loop_v3m): Ditto.
5399 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5402 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5404 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5406 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5408 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5409 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5410 machine->attr_no_prologue_p.
5411 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5412 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5413 (nds32_expand_epilogue): Consider attr_naked_p.
5414 (nds32_expand_epilogue_v3pop): Likewise.
5415 (nds32_can_use_return_insn): Likewise.
5416 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5417 attr_no_prologue_p fields.
5418 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5420 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5423 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5425 * config/i386/sse.md
5426 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5428 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5430 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5432 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5434 (*<floatsuffix>floatv2div2sf2): Rename to ...
5435 (*float<floatunssuffix>v2div2sf2): ... this.
5436 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5437 (float<floatunssuffix>v2div2sf2_mask): ... this.
5438 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5439 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5440 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5442 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5444 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5446 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5448 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5450 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5452 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5453 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5454 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5455 gen_ufix_truncv8dfv8si2.
5456 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5457 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5458 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5459 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5460 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5461 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5462 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5463 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5465 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5469 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5471 2018-05-25 Jim Wilson <jimw@sifive.com>
5473 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5474 * config/riscv/riscv.c (struct machine_function): Add
5475 interrupt_handler_p and attribute_checked_p fields.
5476 (riscv_attribute_table): Add interrupt.
5477 (riscv_interrupt_type_p): New.
5478 (riscv_save_reg_p): Save extra regs for interrupt handler.
5479 (riscv_use_save_libcall): Return false for interrupt handler.
5480 (riscv_first_stack_step): Add forward declaration.
5481 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5482 for interrupt handler with large frame. Use it for saved reg list.
5483 (riscv_expand_prologue): Move flag_stack_usage_info support to
5484 eliminate duplication.
5485 (riscv_expand_epilogue): Generate mret for interrupt handler.
5486 (riscv_epilogue_uses): New.
5487 (riscv_can_use_return_insn): Return false for interrupt handler.
5488 (riscv_function_ok_for_sibcall): Likewise.
5489 (riscv_set_current_function): Add interrupt handler support.
5490 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5491 * config/riscv/riscv.md (UNSPECV_MRET): New.
5493 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5495 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5497 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5499 PR tree-optimization/85712
5500 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5501 this candidate has already been replaced in-situ by a copy.
5503 2018-05-25 Jason Merrill <jason@redhat.com>
5505 PR c++/80485 - inline function non-zero address.
5506 * symtab.c (nonzero_address): Check DECL_COMDAT.
5508 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5511 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5512 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5513 extension of SImode operation. Use const123_operand predicate.
5514 (*saddsi_1): Remove.
5515 (*saddl_se_1): Ditto.
5517 (*ssubl_se_1): Ditto.
5518 * config/alpha/predicates.md (const123_operand): New predicate.
5519 * config/alpha/constraints.md (P): Use IN_RANGE.
5521 2018-05-25 Richard Biener <rguenther@suse.de>
5523 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5525 (ref_maybe_used_by_stmt_p): Likewise.
5526 (stmt_may_clobber_ref_p): Likewise.
5527 (stmt_may_clobber_ref_p_1): Likewise.
5528 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5530 (ref_maybe_used_by_stmt_p): Likewise.
5531 (stmt_may_clobber_ref_p): Likewise.
5532 (stmt_may_clobber_ref_p_1): Likewise.
5533 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5534 the alias oracle to disambiguate DRs with stmts DR analysis
5536 (vect_analyze_data_refs): Do not give up on not analyzable
5537 DRs for BB vectorization. Remove code truncating the dataref
5540 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5543 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5544 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5545 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5547 2018-05-25 Richard Biener <rguenther@suse.de>
5549 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5550 function, combining stmt data ref gathering and fatal analysis
5552 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5553 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5554 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5555 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5556 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5558 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5560 PR tree-optimization/85720
5561 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5562 SCC if all partitions are builtins.
5563 (version_loop_by_alias_check): New parameter. Generate cancelable
5564 runtime alias check if all partitions are builtins.
5565 (distribute_loop): Update call to above function.
5567 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5569 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5570 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5571 (parm_default_def_partition_arg): Ditto.
5572 (set_parm_default_def_partition): Ditto.
5573 (get_parm_default_def_partitions): Ditto and make it static.
5574 (get_undefined_value_partitions): Ditto and make it static.
5575 (remove_ssa_form): Refactor call to init_var_map here.
5576 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5577 computation for loop region.
5578 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5579 (register_default_def): Delete.
5580 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5581 (parm_default_def_partition_arg): Ditto.
5582 (set_parm_default_def_partition): Ditto.
5583 (get_parm_default_def_partitions): Ditto and make it static.
5584 (get_undefined_value_partitions): Ditto and make it static.
5585 (coalesce_with_default, coalesce_with_default): Update comment.
5586 (create_coalesce_list_for_region): New func factored out from
5587 create_outofssa_var_map.
5588 (populate_coalesce_list_for_outofssa): New func factored out from
5589 create_outofssa_var_map and coalesce_ssa_name.
5590 (create_outofssa_var_map): Delete.
5591 (coalesce_ssa_name): Refactor to support live range computation.
5592 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5593 (get_parm_default_def_partitions): Delete.
5594 (get_undefined_value_partitions): Ditto.
5595 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5596 computation for loop region.
5597 (new_tree_live_info, loe_visit_block): Ditto.
5598 (live_worklist, set_var_live_on_entry): Ditto.
5599 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5600 * tree-ssa-live.h (struct _var_map): New fields.
5601 (init_var_map): Change decl.
5602 (region_contains_p): New.
5604 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5606 * tree-ssa-live.h (live_merge_and_clear): Delete.
5608 2018-05-25 Richard Biener <rguenther@suse.de>
5611 * tree-dump.c (dequeue_and_dump): Remove access to removed
5612 operand 2 of a SWITCH_EXPR.
5614 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5616 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5617 multiplication and division.
5618 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5619 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5620 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5621 (cond_udiv_optab, cond_umod_optab): New optabs.
5622 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5623 (IFN_COND_RDIV): New internal functions.
5624 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5625 TRUNC_MOD_EXPR and RDIV_EXPR.
5626 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5627 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5629 (SVE_INT_BINARY): Include mult.
5630 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5631 (optab, sve_int_op): Handle mult.
5632 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5634 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5635 for SVE_INT_BINARY_SD.
5637 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5639 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5640 (optab, sve_int_op): Handle div and udiv.
5641 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5642 for SVE_INT_BINARY_SD.
5643 (*<optab><mode>3): New insn for the same.
5645 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5647 * tree-vect-patterns.c: Include predict.h.
5648 (vect_recog_divmod_pattern): Restrict check for division support
5649 to when optimizing for size.
5651 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5653 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5654 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5655 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5656 (gimple_match_op::set_op): Likewise.
5657 (gimple_resimplify4): Declare.
5658 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5659 (expr::gen_transform): Likewise.
5660 (decision_tree::gen): Generate a simplification routine for 4 operands.
5661 * gimple-match-head.c (gimple_simplify): Add an overload for
5662 4 operands. In the top-level function, handle up to 4 call
5663 arguments and call gimple_resimplify4.
5664 (gimple_resimplify4): New function.
5665 (build_call_internal): Pass a fourth operand.
5666 (maybe_push_to_seq): Likewise.
5667 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5668 Fold VEC_COND_EXPRs of an operation and a default value into
5669 an IFN_COND_* function if possible.
5670 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5672 (SVE_COND_FP_BINARY): Include them.
5673 (optab, sve_fp_op): Handle them.
5674 (SVE_INT_BINARY_REV): New code iterator.
5675 (SVE_COND_FP_BINARY_REV): New int iterator.
5676 (commutative): New int attribute.
5677 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5679 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5681 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5682 (*cond_<optab><mode>): New patterns for reversed operands.
5684 2018-05-25 Richard Biener <rguenther@suse.de>
5686 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5687 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5688 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5689 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5690 * tree-vect-data-refs.c (everywhere): Adjust users.
5691 * tree-vect-loop.c (everywhere): Likewise.
5692 * tree-vect-slp.c (everywhere): Likewise.
5693 * tree-vect-stmts.c (everywhere): Likewise.
5694 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5696 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5698 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5700 (gcc_cv_as_section_exclude): ... this.
5701 Try Solaris as #exclude syntax.
5702 * configure: Regenerate.
5703 * config.in: Regenerate.
5704 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5706 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5707 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5709 * varasm.c (default_elf_asm_named_section): Don't check if
5710 HAVE_GAS_SECTION_EXCLUDE is defined.
5712 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5714 * doc/md.texi: Update the documentation of the cond_* optabs
5715 to mention the new final operand. Fix GET_MODE_NUNITS call.
5716 Describe the scalar case too.
5717 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5718 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5720 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5721 (get_conditional_internal_fn): Update comment.
5722 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5723 accumulator value as a final argument to conditional functions.
5724 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5725 a define_expand and add an "else" operand. Assert for now that
5726 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5727 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5728 (*cond_<optab><mode>): New patterns.
5729 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5730 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5731 (UNSPEC_COND_EOR): Delete.
5732 (optab): Remove associated mappings.
5733 (SVE_INT_BINARY): New code iterator.
5734 (sve_int_op): Remove int attribute and add "minus" to the code
5736 (SVE_COND_INT_OP): Delete.
5737 (SVE_COND_FP_OP): Rename to...
5738 (SVE_COND_FP_BINARY): ...this.
5740 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5742 * optabs.c (can_reuse_operands_p): New function.
5743 (maybe_legitimize_operands): Try to reuse the results for
5746 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5748 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5749 Add {q} suffix to insn mnemonic.
5751 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5753 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5754 (msp430_warn_func_return): New.
5756 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5758 * fold-const.c (tree_nonzero_bits): New function.
5759 * fold-const.h (tree_nonzero_bits): Likewise.
5760 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5761 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5763 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5767 * varasm.c (assemble_alias): Check ifunc_resolver only on
5770 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5773 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5774 when memory input operand is handled.
5776 2018-05-24 Luis Machado <luis.machado@linaro.org>
5778 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5780 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5782 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5784 * match.pd: Delay FMA folds until after vectorization.
5786 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5789 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5790 address check not strict.
5792 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5794 * gimple-match.h (gimple_match_op): New class.
5795 (mprts_hook): Replace parameters with a gimple_match_op *.
5796 (maybe_build_generic_op): Likewise.
5797 (gimple_simplified_result_is_gimple_val): Replace parameters with
5798 a const gimple_match_op *.
5799 (gimple_simplify): Replace code_helper * and tree * parameters with
5800 a gimple_match_op * parameter.
5801 (gimple_resimplify1): Replace code_helper *, tree and tree *
5802 parameters with a gimple_match_op * parameter.
5803 (gimple_resimplify2): Likewise.
5804 (gimple_resimplify3): Likewise.
5805 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5806 parameters with a gimple_match_op * parameter.
5807 * gimple-match-head.c (gimple_simplify): Change prototypes of
5808 auto-generated functions to take a gimple_match_op * instead of
5809 separate code_helper * and tree * parameters. Make the same
5810 change in the top-level overload and update calls to the
5811 gimple_resimplify routines. Update calls to the auto-generated
5812 functions and to maybe_push_res_to_seq in the publicly-facing
5813 operation-specific gimple_simplify overloads.
5814 (gimple_match_op::MAX_NUM_OPS): Define.
5815 (gimple_resimplify1): Replace rcode and ops with a single res_op
5816 parameter. Update call to gimple_simplify.
5817 (gimple_resimplify2): Likewise.
5818 (gimple_resimplify3): Likewise.
5819 (mprts_hook): Replace parameters with a gimple_match_op *.
5820 (maybe_build_generic_op): Likewise.
5821 (build_call_internal): Replace type, nargs and ops with
5822 a gimple_match_op *.
5823 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5824 with a single gimple_match_op *. Update calls to mprts_hook,
5825 build_call_internal and gimple_simplified_result_is_gimple_val.
5826 Factor out code that is common to the tree_code and combined_fn cases.
5827 * genmatch.c (expr::gen_transform): Replace tem_code and
5828 tem_ops with a gimple_match_op called tem_op. Update calls
5829 to the gimple_resimplify functions and maybe_push_res_to_seq.
5830 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5831 res_ops. Update call to the gimple_resimplify functions.
5832 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5833 (decision_tree::gen): Make the functions take a gimple_match_op *
5834 called res_op instead of separate res_code and res_ops parameters.
5835 Update call accordingly.
5836 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5837 and ops with a single res_op parameter. Update calls to
5838 maybe_build_generic_op and maybe_push_res_to_seq.
5839 (fold_stmt_1): Update calls to gimple_simplify and
5840 replace_stmt_with_simplification.
5841 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5842 and gimple_simplified_result_is_gimple_val.
5843 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5845 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5846 with a gimple_match_op *.
5847 (vn_nary_build_or_lookup): Likewise. Update call to
5848 vn_nary_build_or_lookup_1.
5849 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5850 gimple_match_op *. Update calls to the gimple_resimplify routines
5851 and to gimple_simplified_result_is_gimple_val.
5852 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5853 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5854 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5855 (visit_nary_op): Likewise.
5856 (visit_reference_op_load): Likewise.
5858 2018-05-23 Luis Machado <luis.machado@linaro.org>
5860 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5861 modifier for printing the step amount.
5863 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5866 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5869 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5871 * doc/sourcebuild.texi (Endianness): New subsubsection.
5873 2018-05-23 Luis Machado <luis.machado@linaro.org>
5875 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5876 <prefetch_dynamic_strides>: New const bool field.
5877 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5878 prefetch_dynamic_strides.
5879 (exynosm1_prefetch_tune): Likewise.
5880 (thunderxt88_prefetch_tune): Likewise.
5881 (thunderx_prefetch_tune): Likewise.
5882 (thunderx2t99_prefetch_tune): Likewise.
5883 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5885 (aarch64_override_options_internal): Update to set
5886 PARAM_PREFETCH_DYNAMIC_STRIDES.
5887 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5888 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5889 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5890 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5891 prefetch-dynamic-strides setting.
5893 2018-05-23 Luis Machado <luis.machado@linaro.org>
5895 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5896 <minimum_stride>: New const int field.
5897 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5898 minimum_stride field defaulting to -1.
5899 (exynosm1_prefetch_tune): Likewise.
5900 (thunderxt88_prefetch_tune): Likewise.
5901 (thunderx_prefetch_tune): Likewise.
5902 (thunderx2t99_prefetch_tune): Likewise.
5903 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5904 <default_opt_level>: Set to 3.
5905 (aarch64_override_options_internal): Update to set
5906 PARAM_PREFETCH_MINIMUM_STRIDE.
5907 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5908 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5909 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5910 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5911 stride is constant and is below the minimum stride threshold.
5913 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5915 * config/arm/arm-cpus.in (mode26): Delete.
5916 (armv4): Delete mode26 reference.
5917 * config/arm/arm.c (arm_configure_build_target): Delete use of
5920 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5922 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5924 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5925 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5927 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5928 (floatunsdidf2): Ditto.
5930 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5932 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5933 (fixuns_trunc<mode>si2_avx512f): Ditto.
5934 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5935 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5936 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5938 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5940 PR rtl-optimization/79985
5941 * df-scan.c (df_insn_refs_collect): Remove special case for
5942 global registers and asm statements.
5944 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5946 * extend.texi (Global Register Variables): Rewrite the bullet list.
5947 Note that the register is available for allocation. Note that access
5948 via inline asm must use constraints. Add note about async-signal
5949 handlers. Remove paragraph about automagic register selection.
5951 2018-05-23 Richard Biener <rguenther@suse.de>
5953 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5954 of fixed offset from memset VN.
5956 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5958 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5960 (alloc_cand_and_find_basis): Initialize first_interp field.
5961 (slsr_process_mul): Modify first_interp field.
5962 (slsr_process_add): Likewise.
5963 (slsr_process_cast): Modify first_interp field for each new
5965 (slsr_process_copy): Likewise.
5966 (dump_candidate): Dump first_interp field.
5967 (replace_mult_candidate): Process all interpretations, not just
5969 (replace_rhs_if_not_dup): Likewise.
5970 (replace_one_candidate): Likewise.
5972 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5974 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5976 (aarch64_needs_frame_chain): New function.
5977 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5979 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5982 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5983 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5984 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5985 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5986 as true for strict-align.
5987 (aarch64_can_inline_p): Perform checks even when callee has no
5988 attributes to check for strict alignment.
5989 * doc/extend.texi (AArch64 Function Attributes): Document
5991 * doc/invoke.texi: (AArch64 Options): Likewise.
5993 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5995 PR tree-optimization/85853
5996 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5997 the handling of the root of the node to...
5998 (vect_slp_analyze_node_operations_1): ...this new function,
5999 and run the whole thing with the child nodes' def types
6000 set according to their SLP node's def type.
6002 2018-05-23 Richard Biener <rguenther@suse.de>
6005 * tree-data-ref.c (create_runtime_alias_checks): Defer
6006 and ignore overflow warnings.
6008 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6010 PR tree-optimization/85822
6011 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6014 2018-05-23 Richard Biener <rguenther@suse.de>
6016 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6017 memset constants via native_interpret_expr.
6019 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6022 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6024 (cgraph_node::create_alias): Likewise.
6025 (cgraph_node::get_availability): Check ifunc_resolver instead
6026 of looking up ifunc attribute.
6027 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6028 * varasm.c (do_assemble_alias): Likewise.
6029 (assemble_alias): Likewise.
6030 (default_binds_local_p_3): Likewise.
6031 * cgraph.h (cgraph_node): Add ifunc_resolver.
6032 (cgraph_node::only_called_directly_or_aliased_p): Return false
6034 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6036 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6037 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6038 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6039 instead of looking up ifunc attribute.
6041 2018-05-22 Luis Machado <luis.machado@linaro.org>
6043 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6045 2018-05-22 Martin Sebor <msebor@redhat.com>
6048 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6049 only when expasion succeeds.
6050 (expand_builtin_strcmp): Same.
6051 (expand_builtin_strncmp): Same.
6053 2018-05-22 Martin Sebor <msebor@redhat.com>
6055 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6057 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6058 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6060 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6061 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6063 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6064 Add check for legitimate_address.
6065 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6066 (aarch64_swap_ldrstr_operands): New.
6067 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6070 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6071 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6073 * config/aarch64/aarch64.md: New patterns to generate stp
6075 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6076 single words and double words.
6077 (load_pair_sw, load_pair_dw): Likewise.
6078 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6080 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6082 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6083 for different mode ldpstp and add peephole for merged zero stores.
6085 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6087 (aarch64_gen_store_pair): Rename calls to match new patterns.
6088 (aarch64_gen_load_pair): Rename calls to match new patterns.
6089 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6090 (load_pair<DREG:mode><DREG2:mode>): ... This.
6091 (store_pair<mode>): Rename to...
6092 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6093 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6095 (V_INT_EQUIV): Handle SImode.
6096 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6099 2018-05-22 Martin Sebor <msebor@redhat.com>
6102 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6103 or ajust the presumed bound on an operation to avoid unnecessary
6106 2018-05-22 Martin Sebor <msebor@redhat.com>
6108 PR tree-optimization/85826
6109 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6110 assuming that a DECL necesarily has a constant size.
6112 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6115 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6117 2018-05-22 Richard Biener <rguenther@suse.de>
6119 PR tree-optimization/85834
6120 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6121 non-constant and non-zero memset arguments.
6123 2018-05-22 Martin Liska <mliska@suse.cz>
6126 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6128 2018-05-22 Richard Biener <rguenther@suse.de>
6130 PR tree-optimization/85863
6131 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6132 comparisons when vectype is specified.
6133 (vectorizable_condition): Do not specify vectype for
6134 vect_is_simple_cond when SLP vectorizing.
6136 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6139 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6140 define __ibm128 as long double.
6141 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6142 as a distinct type when IEEE 128-bit support is enabled.
6143 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6144 128-bit types to use the correct functions.
6145 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6146 convert between 128-bit floating point types that have different
6147 modes but the same representation, instead of using gen_lowpart to
6149 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6151 (IFKF_reg): New attributes to give the register constraints for
6153 (extend<mode>tf2_internal): New insns to mark an explicit
6154 conversion between 128-bit floating point types that have a
6155 different mode but share the same representation.
6157 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6159 PR tree-optimization/85814
6160 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6161 a null return from get_strinfo when unsharing the next
6162 strinfo in the chain.
6164 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6167 * varasm.c (weak_finish): Clean up weak_decls.
6169 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6171 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6172 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6173 UNSPEC_UADALP values.
6174 * config/aarch64/iterators.md (ABAL): New int iterator.
6177 (sur): Add mappings for the above.
6178 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6180 (aarch64_<sur>abal<mode>_4): Likewise.
6181 (aarch64_<sur>adalp<mode>_3): Likewise.
6182 (<sur>sadv16qi): New define_expand.
6184 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6186 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6187 (*movdf_internal): Ditto.
6188 (*rcpsf2_sse): Ditto.
6189 (*rsqrtsf2_sse): Ditto.
6190 (*sqrt<mode>2_sse): Ditto.
6192 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6194 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6196 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6197 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6198 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6199 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6201 * config/aarch64/arm_neon.h: Likewise.
6202 * config/aarch64/iterators.md (VQ_I): New.
6204 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6206 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6208 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6211 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6213 * config/nds32/constraints.md (S): New constraint.
6214 * config/nds32/nds32.md (call_internal): Use constraint S.
6215 (call_value_internal): Likewise.
6216 (sibcall_internal): Likewise.
6217 (sibcall_value_internal): Likewise.
6219 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6220 Chung-Ju Wu <jasonwucj@gmail.com>
6222 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6225 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6226 Chung-Ju Wu <jasonwucj@gmail.com>
6228 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6229 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6230 (nds32_rtx_costs_impl): Simplify.
6231 (nds32_address_cost_impl): Simplify.
6232 (nds32_init_rtx_costs): New function.
6233 (nds32_rtx_costs_speed_prefer): Likewise.
6234 (nds32_rtx_costs_size_prefer): Likewise.
6235 (nds32_address_cost_speed_prefer): Likewise.
6236 (nds32_address_cost_speed_fwprop): Likewise.
6237 (nds32_address_cost_size_prefer): Likewise.
6238 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6239 * config/nds32/nds32.c (nds32_option_override): Use
6240 nds32_init_rtx_costs function.
6242 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6244 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6245 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6246 (TARGET_PIPELINE_N8): Likewise.
6247 (TARGET_PIPELINE_N10): Likewise.
6248 (TARGET_PIPELINE_N13): Likewise.
6249 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6251 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6253 * config/nds32/nds32-fpu.md: Update copyright year.
6255 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6257 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6259 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6261 * config/nds32/nds32.c
6262 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6263 * config/nds32/nds32.opt (minline-asm-r15): New option.
6265 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6267 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6269 * config/nds32/nds32.md (abssi2): New pattern.
6271 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6273 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6274 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6275 Merge insn pattern from sse_cvtsi2ss<round_name> and
6276 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6277 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6278 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6279 using SWI48 mode iterator.
6280 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6281 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6282 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6283 pattern from sse_cvttss2si<round_saeonly_name>
6284 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6285 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6286 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6287 using SWI48 mode iterator.
6288 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6289 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6290 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6291 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6292 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6293 using SWI48 mode iterator.
6294 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6295 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6296 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6297 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6298 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6299 SWI48 mode iterator.
6300 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6301 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6302 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6303 pattern from sse_cvttsd2si<round_saeonly_name>
6304 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6306 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6308 * config/nds32/nds32-md-auxiliary.c
6309 (nds32_valid_smw_lwm_base_p): Refine.
6310 (nds32_output_smw_single_word): Refine.
6311 (nds32_output_smw_double_word): New.
6312 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6314 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6316 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6317 (nds32_output_stack_pop): Refine.
6318 (nds32_expand_unaligned_load): Refine.
6319 (nds32_expand_unaligned_store): Refine.
6321 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6322 Chung-Ju Wu <jasonwucj@gmail.com>
6324 * config/nds32/constants.md: Add TP_REGNUM constant.
6325 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6326 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6328 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6329 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6330 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6331 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6332 and PIC code generation.
6333 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6335 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6337 * config/nds32/nds32.md: Support TLS and PIC.
6338 * config/nds32/nds32.c: Support TLS and PIC.
6339 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6340 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6343 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6345 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6346 mode with E_ prefix.
6348 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6349 Chung-Ju Wu <jasonwucj@gmail.com>
6351 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6352 * config/nds32/nds32-md-auxiliary.c
6353 (symbolic_reference_mentioned_p): New.
6354 (nds32_legitimize_ict_address): New.
6355 (nds32_expand_ict_move): New.
6356 (nds32_indirect_call_referenced_p): New.
6357 (nds32_symbol_binds_local_p): Delete.
6358 (nds32_long_call_p): Modify.
6359 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6360 * config/nds32/nds32-protos.h
6361 (symbolic_reference_mentioned_p): Declare.
6362 (nds32_legitimize_ict_address): Declare.
6363 (nds32_expand_ict_move): Declare.
6364 (nds32_indirect_call_referenced_p): Declare.
6365 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6366 (nds32_relax_group): Use nds32_ict_const_p as condition.
6367 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6368 (nds32_asm_file_start): Output ict_model directive in asm code.
6369 (nds32_legitimate_address_p): Consider indirect call.
6370 (nds32_print_operand): Consider indirect call.
6371 (nds32_print_operand_address): Consider indirect call.
6372 (nds32_insert_attributes): Handle "indirect_call" attribute.
6373 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6374 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6375 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6376 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6377 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6378 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6379 (TARGET_ICT_MODEL_SMALL): Define.
6380 (TARGET_ICT_MODEL_LARGE): Define.
6381 * config/nds32/nds32.md (movsi): Consider ict model.
6382 (call, call_value): Consider ict model.
6383 (sibcall, sibcall_value): Consider ict model.
6384 * config/nds32/nds32.opt (mict-model): New option.
6385 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6388 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6389 Monk Chiang <sh.chiang04@gmail.com>
6390 Jim Wilson <jimw@sifive.com>
6392 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6393 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6394 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6395 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6396 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6397 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6398 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6399 compute save_libcall_adjustment properly.
6400 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6401 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6402 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6403 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6404 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6405 (ABI_SPEC): Handle mabi=ilp32e.
6406 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6407 (RVE): Add RVE mask.
6408 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6409 <-march>: Add rv32e as an example.
6411 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6414 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6415 (intra_create_variable_infos): Handle C++ constructors.
6417 2018-05-18 Martin Liska <mliska@suse.cz>
6419 * passes.def: Remove a redundant pass.
6421 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6424 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6426 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6428 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6430 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6432 (armv2, armv2a, armv3, armv3m): Delete architectures.
6433 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6434 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6435 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6437 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6438 (*mulsidi3adddi): Likewise.
6439 (mulsidi3): Likewise.
6440 (*mulsidi3_nov6): Likewise.
6441 (umulsidi3): Likewise.
6442 (umulsidi3_nov6): Likewise.
6443 (umaddsidi4): Likewise.
6444 (*umulsidi3adddi): Likewise.
6445 (smulsi3_highpart): Likewise.
6446 (*smulsi3_highpart_nov6): Likewise.
6447 (umulsi3_highpart): Likewise.
6448 (*umulsi3_highpart_nov6): Likewise.
6449 * config/arm/arm.h (arm_arch3m): Delete.
6450 * config/arm/arm.c (arm_arch3m): Delete.
6451 (arm_option_override_internal): Update armv3-related comment.
6452 (arm_configure_build_target): Delete use of isa_bit_mode32.
6453 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6454 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6455 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6457 (mulusa3): Likewise.
6458 * config/arm/arm-protos.h (arm_arch3m): Delete.
6459 * config/arm/arm-tables.opt: Regenerate.
6460 * config/arm/arm-tune.md: Likewise.
6461 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6462 deleted architectures.
6464 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6466 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6467 (armv5t, armv5te): New features.
6468 (ARMv5, ARMv5e): Delete fgroups.
6469 (ARMv5t, ARMv5te): Adjust for above changes.
6471 (armv5, armv5e): Delete arches.
6472 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6474 (*call_reg_arm): Likewise.
6475 (*call_value_reg_armv5): Likewise.
6476 (*call_value_reg_arm): Likewise.
6477 (*call_symbol): Likewise.
6478 (*call_value_symbol): Likewise.
6479 (*sibcall_insn): Likewise.
6480 (*sibcall_value_insn): Likewise.
6482 (prefetch): Likewise.
6483 (define_split and define_peephole2 dependent on arm_arch5):
6485 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6487 (TARGET_ARM_QBIT): Likewise.
6488 (TARGET_DSP_MULTIPLY): Likewise.
6489 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6490 (arm_arch5, arm_arch5e): Delete.
6491 (arm_arch5t, arm_arch5te): Declare.
6492 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6493 (arm_arch5t): Declare.
6494 (arm_option_reconfigure_globals): Update for the above.
6495 (arm_options_perform_arch_sanity_checks): Update comment, replace
6496 use of arm_arch5 with arm_arch5t.
6497 (use_return_insn): Likewise.
6498 (arm_emit_call_insn): Likewise.
6499 (output_return_instruction): Likewise.
6500 (arm_final_prescan_insn): Likewise.
6501 (arm_coproc_builtin_available): Likewise.
6502 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6503 arm_arch5e with arm_arch5t and arm_arch5te.
6504 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6505 (arm_arch5t, arm_arch5te): Declare.
6506 * config/arm/arm-tables.opt: Regenerate.
6507 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6508 * config/arm/t-multilib: Likewise.
6509 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6510 instead of arm_arch5.
6511 (*call_reg_thumb1): Likewise.
6512 (*call_value_reg_thumb1_v5): Likewise.
6513 (*call_value_reg_thumb1): Likewise.
6514 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6516 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6518 2018-05-18 Martin Liska <mliska@suse.cz>
6520 PR gcov-profile/84846
6521 * doc/gcov.texi: Document -t option of gcov tool.
6523 2018-05-18 Martin Liska <mliska@suse.cz>
6525 PR gcov-profile/84846
6526 * gcov.c (print_usage): Add new -t option.
6527 (process_args): Handle the option.
6528 (generate_results): Use stdout as output when requested by
6531 2018-05-18 Martin Liska <mliska@suse.cz>
6533 PR gcov-profile/84846
6534 * coverage.c (coverage_init): Write PWD to .gcno file.
6535 * doc/gcov.texi: Document how working directory is printed.
6536 * gcov-dump.c (dump_gcov_file): Print PWD.
6537 * gcov.c (output_intermediate_file): Likewise.
6538 (read_graph_file): Read PWD string.
6539 (output_lines): Print PWD.
6541 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6544 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6545 for retval and return false if all args to phi are zero.
6547 2018-05-18 Richard Biener <rguenther@suse.de>
6549 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6551 (evrp_dom_walker::before_dom_children): Call it.
6553 2018-05-18 Richard Biener <rguenther@suse.de>
6555 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6556 results when processing array refs with variable index.
6558 2018-05-18 Toon Moene <toon@moene.org>
6560 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6561 directly after that of -floop-interchange. Indicate that both
6562 options are enabled by default when specifying -O3.
6564 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6566 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6567 iterator. Delete separate integer-mode vec_set<mode> expander.
6568 (aarch64_simd_vec_setv2di): Delete.
6569 (vec_setv2di): Delete.
6570 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6571 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6572 the "w, r" alternative.
6574 2018-05-18 Martin Liska <mliska@suse.cz>
6576 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6577 * tree-pass.h (make_pass_lower_switch_O0): New function.
6578 * tree-switch-conversion.c (node_has_low_bound): Remove.
6579 (node_has_high_bound): Likewise.
6580 (node_is_bounded): Likewise.
6581 (class pass_lower_switch): Make it a template type and create
6583 (pass_lower_switch::execute): Add template argument.
6584 (make_pass_lower_switch): New function.
6585 (make_pass_lower_switch_O0): New function.
6586 (do_jump_if_equal): Remove.
6587 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6588 all the hard work to tree optimization passes.
6590 2018-05-18 Martin Liska <mliska@suse.cz>
6592 * dbgcnt.c (limit_low): Renamed from limit.
6593 (limit_high): New variable.
6594 (dbg_cnt_is_enabled): Check for upper limit.
6595 (dbg_cnt): Adjust dumping.
6596 (dbg_cnt_set_limit_by_index): Add new argument for high
6598 (dbg_cnt_set_limit_by_name): Likewise.
6599 (dbg_cnt_process_single_pair): Parse new format.
6600 (dbg_cnt_process_opt): Use strtok.
6601 (dbg_cnt_list_all_counters): Remove 'value' and add
6603 * doc/invoke.texi: Document changes.
6605 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6607 * doc/sourcebuild.texi (scalar_all_fma): Document.
6608 * tree.def (FMA_EXPR): Delete.
6609 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6610 * internal-fn.c (ternary_direct): New macro.
6611 (expand_ternary_optab_fn): Likewise.
6612 (direct_ternary_optab_supported_p): Likewise.
6613 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6614 * builtins.c (fold_builtin_fma): Delete.
6615 (fold_builtin_3): Don't call it.
6616 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6617 * expr.c (expand_expr_real_2): Likewise.
6618 * fold-const.c (operand_equal_p): Likewise.
6619 (fold_ternary_loc): Likewise.
6620 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6621 * gimple.c (DEFTREECODE): Likewise.
6622 * gimplify.c (gimplify_expr): Likewise.
6623 * optabs-tree.c (optab_for_tree_code): Likewise.
6624 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6625 * tree-eh.c (operation_could_trap_p): Likewise.
6626 (stmt_could_throw_1_p): Likewise.
6627 * tree-inline.c (estimate_operator_cost): Likewise.
6628 * tree-pretty-print.c (dump_generic_node): Likewise.
6629 (op_code_prio): Likewise.
6630 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6631 * tree-ssa-operands.c (get_expr_operands): Likewise.
6632 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6633 * fold-const-call.h (fold_fma): Delete.
6634 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6635 CFN_FNMA and CFN_FNMS.
6637 * genmatch.c (combined_fn): New enum.
6638 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6639 (commutative_op): New function.
6640 (commutate): Use it. Handle more than 2 operands.
6641 (dt_operand::gen_gimple_expr): Use commutative_op.
6642 (parser::parse_expr): Allow :c to be used with non-binary
6643 operators if the commutative operand is known.
6644 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6645 CFN_FMS, CFN_FNMA and CFN_FNMS.
6646 (backprop::process_assign_use): Remove FMA_EXPR handling.
6647 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6648 (gen_hsa_fma): New function.
6649 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6650 IFN_FNMA and IFN_FNMS.
6651 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6652 * gimple-fold.h (follow_all_ssa_edges): Declare.
6653 * gimple-fold.c (follow_all_ssa_edges): New function.
6654 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6655 gimple_build interface and use follow_all_ssa_edges to fold the result.
6656 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6657 instead of checking for optabs directly.
6658 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6659 rather than FMA_EXPRs.
6660 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6661 call to IFN_FMA instead of an FMA_EXPR.
6663 2018-05-17 Jim Wilson <jimw@sifive.com>
6665 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6666 sign extended promoted subreg, and the range does not have the sign bit
6667 set, then do a sign extend.
6669 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6670 test, check for sign extended subreg and/or constant operands, and
6671 do a sign extend in that case.
6673 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6675 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6676 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6678 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6679 Change logics_shift_reg to logics_shift_imm.
6680 (thunderx2t99_fp_loadpair_basic): Delete.
6681 (thunderx2t99_fp_storepair_basic): Delete.
6682 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6683 (thunderx2t99_asimd_polynomial): Delete.
6684 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6685 and neon_fp_mul_d_scalar_q.
6686 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6687 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6688 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6689 (thunderx2t99_asimd_lut): Add missing tbl types.
6690 (thunderx2t99_asimd_ext): Delete.
6691 (thunderx2t99_asimd_load1_1_mult): Delete.
6692 (thunderx2t99_asimd_load1_2_mult): Delete.
6693 (thunderx2t99_asimd_load1_ldp): New.
6694 (thunderx2t99_asimd_load1): New.
6695 (thunderx2t99_asimd_load2): Add missing *load2* types.
6696 (thunderx2t99_asimd_load3): New.
6697 (thunderx2t99_asimd_load4): New.
6698 (thunderx2t99_asimd_store1_1_mult): Delete.
6699 (thunderx2t99_asimd_store1_2_mult): Delete.
6700 (thunderx2t99_asimd_store2_mult): Delete.
6701 (thunderx2t99_asimd_store2_onelane): Delete.
6702 (thunderx2t99_asimd_store_stp): New.
6703 (thunderx2t99_asimd_store1): New.
6704 (thunderx2t99_asimd_store2): New.
6705 (thunderx2t99_asimd_store3): New.
6706 (thunderx2t99_asimd_store4): New.
6708 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6710 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6711 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6713 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6714 Segher Boessenkool <segher@kernel.crashing.org>
6717 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6720 2018-05-17 Richard Biener <rguenther@suse.de>
6722 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6723 for pruning loop and prune defs feeding only already visited PHIs.
6725 2018-05-17 Richard Biener <rguenther@suse.de>
6727 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6729 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6730 Richard Biener <rguenther@suse.de>
6732 PR tree-optimization/85793
6733 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6734 for VMAT_ELEMENTWISE.
6736 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6738 * internal-fn.h (lookup_internal_fn): Declare
6739 * internal-fn.c (lookup_internal_fn): New function.
6740 * gimple.c (gimple_build_call_from_tree): Handle calls to
6742 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6743 internal function names.
6744 * tree-pretty-print.c (dump_generic_node): Likewise.
6745 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6747 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6749 * gimple-fold.h (gimple_build): Make the function forms take
6750 combined_fn rather than built_in_function.
6751 (gimple_simplify): Likewise.
6752 * gimple-match-head.c (gimple_simplify): Likewise.
6753 * gimple-fold.c (gimple_build): Likewise.
6754 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6755 rather than gimple_build_call_internal.
6756 (get_initial_defs_for_reduction): Likewise.
6757 (vect_create_epilog_for_reduction): Likewise.
6758 (vectorizable_live_operation): Likewise.
6760 2018-05-17 Martin Liska <mliska@suse.cz>
6762 * gimple-ssa-sprintf.c (format_directive): Do not use
6763 space in between 'G_' and '('.
6765 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6768 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6769 even if the mask is not all ones.
6772 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6774 (ix86_gimple_fold_builtin): Likewise.
6777 * config/i386/i386.c: Include tree-vector-builder.h.
6778 (ix86_vector_shift_count): New function.
6779 (ix86_fold_builtin): Fold shift builtins by scalar count.
6780 (ix86_gimple_fold_builtin): Likewise.
6782 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6783 _mm512_setzero): New intrinsics.
6785 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6786 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6788 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6789 code generation for cases where splatting a value is not useful.
6790 * simplify-rtx.c (simplify_ternary_operation): Simplify
6791 vec_merge across a vec_duplicate and a paradoxical subreg forming
6792 a vector mode to a vec_concat.
6794 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6796 * config.gcc: Support "goldmont-plus".
6797 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6799 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6800 PROCESSOR_GOLDMONT_PLUS.
6801 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6802 (processor_target_table): Add "goldmont-plus".
6803 (PTA_GOLDMONT_PLUS): Define.
6804 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6805 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6806 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6807 (fold_builtin_cpu): Add "goldmont-plus".
6808 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6809 (ix86_option_override_internal): Add "goldmont-plus".
6810 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6811 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6812 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6813 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6815 2018-05-17 Richard Biener <rguenther@suse.de>
6817 PR tree-optimization/85757
6818 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6819 remove defs that only feed that PHI from further processing.
6821 2018-05-16 Jim Wilson <jimw@sifive.com>
6823 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6825 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6827 2018-05-16 Mark Wielaard <mark@klomp.org>
6829 * dwarf2out.c (count_index_strings): New function.
6830 (output_indirect_strings): Call count_index_strings and generate
6831 header for dwarf_version >= 5.
6833 2018-05-16 Mark Wielaard <mark@klomp.org>
6835 * dwarf2out.c (dwarf_FORM): New function.
6836 (set_indirect_string): Use dwarf_FORM.
6837 (reset_indirect_string): Likewise.
6838 (size_of_die): Likewise.
6839 (value_format): Likewise.
6840 (output_die): Likewise.
6841 (add_skeleton_AT_string): Likewise.
6842 (output_macinfo_op): Likewise.
6843 (index_string): Likewise.
6844 (output_index_string_offset): Likewise.
6845 (output_index_string): Likewise.
6846 (count_index_strings): Likewise.
6848 2018-05-16 Carl Love <cel@us.ibm.com>
6850 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6851 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6853 2018-05-16 Martin Jambor <mjambor@suse.cz>
6855 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6856 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6858 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6860 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6861 (fnma<mode>4): Likewise.
6862 (fms<mode>4): Likewise.
6863 (fnms<mode>4): Likewise.
6864 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6865 (aarch64_fnma<mode>4): Likewise.
6866 (aarch64_fms<mode>4): Likewise.
6867 (aarch64_fnms<mode>4): Likewise.
6868 (aarch64_fnmadd<mode>4): Likewise.
6870 2018-05-16 Jason Merrill <jason@redhat.com>
6872 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6874 2018-05-16 Richard Biener <rguenther@suse.de>
6876 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6877 (dump_stmt_cost): Declare.
6878 (add_stmt_cost): Dump cost we add.
6879 (add_stmt_costs): New function.
6880 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6882 (vect_analyze_stmt): Adjust prototype.
6883 (vectorizable_condition): Likewise.
6884 (vectorizable_live_operation): Likewise.
6885 (vectorizable_reduction): Likewise.
6886 (vectorizable_induction): Likewise.
6887 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6888 cost vector to pass to vectorizable_ and record afterwards.
6889 (vect_model_reduction_cost): Take cost vector argument and adjust.
6890 (vect_model_induction_cost): Likewise.
6891 (vectorizable_reduction): Likewise.
6892 (vectorizable_induction): Likewise.
6893 (vectorizable_live_operation): Likewise.
6894 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6895 SLP_TREE_NUMBER_OF_VEC_STMTS.
6896 (vect_analyze_slp_cost_1): Remove.
6897 (vect_analyze_slp_cost): Likewise.
6898 (vect_slp_analyze_node_operations): Take visited args and
6899 a target cost vector. Avoid processing already visited stmt sets.
6900 (vect_slp_analyze_operations): Use a local cost vector to gather
6901 costs and register those of non-discarded instances.
6902 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6903 (vect_schedule_slp_instance): Remove copying of
6904 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6906 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6907 adding cost. Record cost entry location.
6908 (vect_prologue_cost_for_slp_op): Function to compute cost of
6909 a constant or invariant generated for SLP vect in the prologue,
6910 split out from vect_analyze_slp_cost_1.
6911 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6912 (vect_model_promotion_demotion_cost): Likewise.
6913 (vect_model_store_cost): Likewise, make static.
6914 (vect_model_load_cost): Likewise.
6915 (vectorizable_bswap): Add cost vector arg and adjust.
6916 (vectorizable_call): Likewise.
6917 (vectorizable_simd_clone_call): Likewise.
6918 (vectorizable_conversion): Likewise.
6919 (vectorizable_assignment): Likewise.
6920 (vectorizable_shift): Likewise.
6921 (vectorizable_operation): Likewise.
6922 (vectorizable_store): Likewise.
6923 (vectorizable_load): Likewise.
6924 (vectorizable_condition): Likewise.
6925 (vectorizable_comparison): Likewise.
6926 (can_vectorize_live_stmts): Likewise.
6927 (vect_analyze_stmt): Likewise.
6928 (vect_transform_stmt): Adjust calls to vectorizable_*.
6929 * tree-vectorizer.c: Include gimple-pretty-print.h.
6930 (dump_stmt_cost): New function.
6932 2018-05-16 Richard Biener <rguenther@suse.de>
6934 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6935 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6936 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6937 (check_name): New callback.
6938 (dse_classify_store): Track cycles via a visited bitmap of PHI
6939 defs and simplify handling of in-loop and across loop dead stores
6940 and properly fail for loop-variant refs. Handle byte-tracking with
6941 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6944 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6946 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6947 (vect_get_mask_type_for_stmt): Likewise.
6948 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6950 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6951 to determine the statement's vector type and the vector type that
6952 should be used for calculating nunits. Deal with cases in which
6953 the type has to be deferred.
6954 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6955 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6956 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6957 (vect_determine_vf_for_stmt): New functions, split out from...
6958 (vect_determine_vectorization_factor): ...here.
6959 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6960 (vect_get_mask_type_for_stmt): New functions, split out from
6961 vect_determine_vectorization_factor.
6963 2018-05-16 Richard Biener <rguenther@suse.de>
6965 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6966 verify the [VEC_]COND_EXPR embedded comparison.
6968 2018-05-15 Martin Sebor <msebor@redhat.com>
6970 PR tree-optimization/85753
6971 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6972 RECORD_TYPE in addition to ARRAY_TYPE.
6974 2018-05-15 Martin Sebor <msebor@redhat.com>
6977 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6979 2018-05-15 Richard Biener <rguenther@suse.de>
6981 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6982 add by_clobber_p one. Change algorithm to collect all defs
6983 representing uses we need to walk and try reducing them to
6984 a single one before failing.
6985 (dse_dom_walker::dse_optimize_stmt): Adjust.
6987 2018-05-13 Mark Wielaard <mark@klomp.org>
6989 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6990 (size_of_loc_descr): Likewise.
6991 (output_loc_operands): Likewise.
6992 (output_loc_operands_raw): Likewise.
6993 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6994 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6995 (hash_loc_operands): Likewise.
6996 (compare_loc_operands): Likewise.
6998 2018-05-14 Mark Wielaard <mark@klomp.org>
7000 * dwarf2out.c (count_index_addrs): New function.
7001 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7003 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7005 PR tree-optimization/83648
7006 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7007 return value as malloc candidate.
7009 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7012 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7013 param as true in call to suggest_attribute.
7015 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7017 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7018 -mreadonly-in-sdata.
7020 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7022 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7024 (aarch64_crypto_aesd_fused): Likewise.
7026 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7028 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7029 (movsi_aarch64): Likewise.
7030 (load_pairsi): Likewise.
7031 (load_pairdi): Likewise.
7032 (store_pairsi): Likewise.
7033 (store_pairdi): Likewise.
7034 (load_pairsf): Likewise.
7035 (load_pairdf): Likewise.
7036 (store_pairsf): Likewise.
7037 (store_pairdf): Likewise.
7038 (zero_extend): Likewise.
7039 (trunc): Swap alternatives.
7040 (fcvt_target): Add '?' to prefer w over r.
7042 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7045 * config/i386/i386.md: Disallow non-commutative arithmetics in
7046 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7047 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7048 in the peephole2 before it.
7050 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7052 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7053 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7054 (ix86_handle_option): Handle -mcldemote.
7055 * config.gcc: New header.
7056 * config/i386/cldemoteintrin.h: New file.
7057 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7058 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7060 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7061 OPTION_MASK_ISA_CLDEMOTE.
7062 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7063 (ix86_valid_target_attribute_inner_p): Ditto.
7064 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7065 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7066 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7067 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7068 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7070 * config/i386/i386.opt: Add -mcldemote.
7071 * config/i386/x86intrin.h: New header.
7072 * doc/invoke.texi: Add -mcldemote.
7074 2018-05-14 Richard Biener <rguenther@suse.de>
7076 * doc/match-and-simplify.texi: Adjust :s documentation.
7078 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7080 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7081 intended memcpy size.
7082 (REORDER_45): Likewise.
7084 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7086 * sort.cc: New file.
7087 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7088 * vec.c (qsort_chk): Use gcc_qsort.
7089 * Makefile.in (OBJS-libcommon): Add sort.o.
7090 (build/sort.o): New target. Use it...
7091 (BUILD_RTL): ... here, and...
7092 (build/gencfn-macros): ... here, and...
7093 (build/genmatch): ... here.
7095 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7096 Chung-Ju Wu <jasonwucj@gmail.com>
7098 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7099 * config/nds32/nds32-graywolf.md: New file.
7100 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7101 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7103 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7104 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7105 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7106 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7107 * config/nds32/pipelines.md: Include n15 settings.
7109 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7110 Chung-Ju Wu <jasonwucj@gmail.com>
7112 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7113 * config/nds32/nds32-n13.md: New file.
7114 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7115 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7117 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7118 * config/nds32/nds32.md (pipeline_model): Add n13.
7119 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7120 * config/nds32/pipelines.md: Include n13 settings.
7122 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7123 Chung-Ju Wu <jasonwucj@gmail.com>
7125 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7126 * config/nds32/nds32-n10.md: New file.
7127 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7128 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7130 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7131 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7132 * config/nds32/nds32.md (pipeline_model): Add n10.
7133 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7134 * config/nds32/pipelines.md: Include n10 settings.
7136 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7137 Kito Cheng <kito.cheng@gmail.com>
7138 Chung-Ju Wu <jasonwucj@gmail.com>
7140 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7141 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7142 Add enum values for DSP extension instructions.
7143 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7145 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7146 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7148 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7149 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7150 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7151 * config/nds32/nds32-intrinsic.md: Likewise.
7152 * config/nds32/nds32_intrinsic.h: Likewise.
7153 * config/nds32/nds32-md-auxiliary.c: Likewise.
7154 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7155 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7156 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7157 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7158 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7159 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7160 TYPE_DMAC in switch statement.
7161 * config/nds32/nds32.c: New checking and implementation for DSP
7162 extension instructions.
7163 * config/nds32/nds32.h: Likewise.
7164 * config/nds32/nds32.md: Likewise.
7165 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7166 * config/nds32/predicates.md: Implement new predicates for DSP
7169 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7171 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7172 Reformat alternatives and attributes so it is easier to identify
7173 which constraints/attributes go with which instruction.
7174 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7175 (mov<mode>_softfloat32, FMOVE64): Likewise.
7176 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7177 (mov<mode>_softfloat64, FMOVE64): Likewise.
7179 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7181 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7183 (Basic PowerPC Built-in Functions): The new name of the
7184 subsection previously known as "PowerPC Built-in Functions".
7185 (Basic PowerPC Built-in Functions Available on all Configurations):
7187 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7188 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7189 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7190 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7192 2018-05-11 Martin Jambor <mjambor@suse.cz>
7195 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7198 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7201 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7203 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7205 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7206 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7207 (ix86_handle_option): Handle -mwaitpkg.
7208 * config.gcc: New header.
7209 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7210 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7211 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7213 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7214 OPTION_MASK_ISA_WAITPKG.
7215 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7216 (ix86_option_override_internal): Add PTA_WAITPKG.
7217 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7218 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7219 IX86_BUILTIN_TPAUSE.
7220 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7221 __builtin_ia32_umwait and __builtin_ia32_tpause.
7222 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7223 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7224 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7225 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7226 UNSPECV_TPAUSE): New.
7227 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7228 * config/i386/i386.opt: Add -mwaitpkg.
7229 * config/i386/waitpkgintrin.h: New file.
7230 * config/i386/x86intrin.h: New header.
7231 * doc/invoke.texi: Add -mwaitpkg.
7233 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7236 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7238 (cortex-m0): Use armv6s-m isa.
7239 (cortex-m0plus): Likewise.
7240 (cortex-m1): Likewise.
7241 (cortex-m0.small-multiply): Likewise.
7242 (cortex-m0plus.small-multiply): Likewise.
7243 (cortex-m1.small-multiply): Likewise.
7245 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7246 Jakub Jelinek <jakub@redhat.com>
7248 PR tree-optimization/85692
7249 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7250 source permute as well.
7252 2018-05-11 Martin Liska <mliska@suse.cz>
7255 * doc/extend.texi: Document LLVM style format for no_sanitize
7258 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7260 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7261 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7262 (mode_supports_vsx_dform_quad): Likewise.
7263 (mode_supports_vmx_dform): Move these functions to be next to the
7264 other mode_supports functions.
7265 (mode_supports_dq_form): Likewise.
7266 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7267 mode_supports_dq_form.
7268 (reg_offset_addressing_ok_p): Likewise.
7269 (offsettable_ok_by_alignment): Likewise.
7270 (rs6000_legitimate_offset_address_p): Likewise.
7271 (legitimate_lo_sum_address_p): Likewise.
7272 (rs6000_legitimize_address): Likewise.
7273 (rs6000_legitimize_reload_address): Likewise.
7274 (rs6000_secondary_reload_inner): Likewise.
7275 (rs6000_preferred_reload_class): Likewise.
7276 (rs6000_output_move_128bit): Likewise.
7278 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7280 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7281 Generate SImode target register for null target.
7282 <case IX86_BUILTIN_XGETBV>: Ditto.
7283 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7284 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7286 2018-05-10 Carl Love <cel@us.ibm.com>
7288 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7289 dcbtt and dcbtstt if operands[2] is 0.
7291 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7294 * config/i386/sse.md (usadv64qi): New expander.
7296 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7298 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7299 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7300 -maltivec=be support.
7301 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7302 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7303 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7304 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7305 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7306 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7307 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7308 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7309 altivec_vsumsws): Adjust.
7310 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7311 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7312 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7314 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7315 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7316 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7317 (altivec_lve<VI_char>x): Delete expand.
7318 (*altivec_lve<VI_char>x_internal): Rename to...
7319 (altivec_lve<VI_char>x): ... this.
7320 (altivec_lvxl_<mode>): Delete expand.
7321 (*altivec_lvxl_<mode>_internal): Rename to ...
7322 (altivec_lvxl_<mode>): ... this.
7323 (altivec_stvxl_<mode>): Delete expand.
7324 (*altivec_stvxl_<mode>_internal): Rename to ...
7325 (altivec_stvxl_<mode>): ... this.
7326 (altivec_stve<VI_char>x): Delete expand.
7327 (*altivec_stve<VI_char>x_internal): Rename to ...
7328 (altivec_stve<VI_char>x): ... this.
7329 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7330 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7331 reduc_plus_scal_<mode>): Adjust.
7332 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7334 (rs6000_cpu_cpp_builtins): Adjust.
7335 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7336 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7337 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7338 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7339 -maltivec=be support.
7340 (rs6000_split_vec_extract_var): Adjust.
7341 (rs6000_split_v4si_init): Adjust.
7342 (swap_selector_for_mode): Delete.
7343 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7344 altivec_expand_stvex_be): Delete.
7345 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7346 -maltivec=be support.
7347 (rs6000_gimple_fold_builtin): Ditto.
7348 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7350 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7351 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7352 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7353 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7354 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7355 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7356 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7357 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7358 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7359 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7360 anonymous split): Adjust.
7361 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7362 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7364 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7366 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7367 when --with-gxx-include-dir is also specified.
7368 * configure: Regenerate.
7370 2018-05-09 Jim Wilson <jimw@sifive.com>
7373 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7374 * config/riscv/t-withmultilib: New.
7375 * config/riscv/withmultilib.h: New.
7376 * doc/install.texi: Document RISC-V --with-multilib-list support.
7378 2018-05-09 Richard Biener <rguenther@suse.de>
7380 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7382 (vect_bb_vectorization_profitable_p): Adjust. Compute
7383 actual scalar cost using the cost vector and the add_stmt_cost
7386 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7388 PR rtl-optimization/85645
7389 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7390 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7392 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7394 PR rtl-optimization/85645
7395 * shrink-wrap.c (spread_components): Return a boolean saying if
7396 anything was changed.
7397 (try_shrink_wrapping_separate): Iterate spread_components until
7398 nothing changes anymore.
7400 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7402 PR rtl-optimization/85645
7403 * regrename.c (build_def_use): Also kill the chains that include the
7404 destination of a REG_CFA_REGISTER note.
7406 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7408 PR rtl-optimization/85645
7409 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7410 insn that has a REG_CFA_REGISTER note.
7412 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7414 * cfgexpand.c (expand_clobber): New function.
7415 (expand_gimple_stmt_1): Use it.
7416 * tree-vect-stmts.c (vect_clobber_variable): New function,
7418 (vectorizable_simd_clone_call): ...here.
7419 (vectorizable_store): Emit a clobber either side of an
7420 IFN_STORE_LANES sequence.
7421 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7423 2018-05-09 Tom de Vries <tom@codesourcery.com>
7426 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7427 (define_insn "trap_if_false"): Add exit after trap.
7429 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7431 PR rtl-optimization/85638
7432 * bb-reorder.c: Include common/common-target.h.
7433 (create_forwarder_block): New function extracted from...
7434 (fix_up_crossing_landing_pad): ...here. Rename into...
7435 (dw2_fix_up_crossing_landing_pad): ...this.
7436 (sjlj_fix_up_crossing_landing_pad): New function.
7437 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7438 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7439 from both partitions and exit the loop after one iteration.
7441 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7444 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7446 (Basic PowerPC Built-in Functions): The new name of the
7447 subsection previously known as "PowerPC Built-in Functions".
7448 (Basic PowerPC Built-in Functions Available on all Configurations):
7450 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7452 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7454 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7456 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7459 2018-05-08 Jim Wilson <jimw@sifive.com>
7461 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7462 (LD_EMUL_SUFFIX): New.
7463 (LINK_SPEC): Use it.
7465 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7467 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7469 (Basic PowerPC Built-in Functions): The new name of the
7470 subsection previously known as "PowerPC Built-in Functions".
7471 (Basic PowerPC Built-in Functions Available on all Configurations):
7473 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7475 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7477 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7479 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7482 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7485 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7486 after cmpelim optimization.
7488 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7490 * config.gcc: Support "goldmont".
7491 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7492 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7494 * config/i386/i386.c (m_GOLDMONT): Define.
7495 (processor_target_table): Add "goldmont".
7496 (PTA_GOLDMONT): Define.
7497 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7498 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7499 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7500 (fold_builtin_cpu): Add "goldmont".
7501 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7502 (ix86_option_override_internal): Add "goldmont".
7503 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7504 (processor_type): Add PROCESSOR_GOLDMONT.
7505 * config/i386/i386.md: Add CPU "glm".
7506 * config/i386/glm.md: New file.
7507 * config/i386/x86-tune.def: Add m_GOLDMONT.
7508 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7510 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7513 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7515 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7516 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7517 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7520 * config/i386/i386.c (ix86_fold_builtin): Handle
7521 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7524 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7525 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7527 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7530 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7531 (check_arch): Likewise.
7532 (check_fpu): Return the result rather than printing it.
7533 (end arch): Fix operator precedence.
7534 (end cpu): Likewise.
7535 (END): Print the result from check_fpu.
7537 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7538 Alan Hayward <alan.hayward@arm.com>
7539 David Sherwood <david.sherwood@arm.com>
7541 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7542 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7543 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7544 (*fcmuo<mode>_and): New patterns.
7546 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7548 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7549 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7550 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7551 (cmp_op, sve_imm_con): New code attributes.
7552 (SVE_COND_INT_CMP, imm_con): Delete.
7553 (cmp_op): Remove above unspecs from int attribute.
7554 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7556 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7557 comparison-specific unspecs.
7558 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7559 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7560 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7561 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7562 (*vec_fcm<cmp_op><mode>): Rename to...
7563 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7564 (*vec_fcmuo<mode>): Rename to...
7565 (*fcmuo<mode>): ...this and adjust likewise.
7566 (*pred_fcm<cmp_op><mode>): New pattern.
7567 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7568 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7570 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7572 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7573 (aarch64_emit_sve_predicated_cond): New function.
7574 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7575 (aarch64_emit_unspec_cond_or): Replace with...
7576 (aarch64_emit_sve_or_conds): ...this new function. Use
7577 aarch64_emit_sve_ptrue_op for the individual comparisons and
7578 aarch64_emit_binop to OR them together.
7579 (aarch64_emit_inverted_unspec_cond): Replace with...
7580 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7581 aarch64_emit_sve_ptrue_op for the comparison and
7582 aarch64_emit_unop to invert the result.
7583 (aarch64_expand_sve_vec_cmp_float): Update after the above
7584 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7586 2018-05-07 Nathan Sidwell <nathan@acm.org>
7588 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7589 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7590 (Backwards Compatibility): Likewise.
7592 2018-05-07 Luis Machado <luis.machado@linaro.org>
7596 2018-05-07 Luis Machado <luis.machado@linaro.org>
7598 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7599 <prefetch_dynamic_strides>: New const bool field.
7600 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7601 prefetch_dynamic_strides.
7602 (exynosm1_prefetch_tune): Likewise.
7603 (thunderxt88_prefetch_tune): Likewise.
7604 (thunderx_prefetch_tune): Likewise.
7605 (thunderx2t99_prefetch_tune): Likewise.
7606 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7608 (aarch64_override_options_internal): Update to set
7609 PARAM_PREFETCH_DYNAMIC_STRIDES.
7610 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7611 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7612 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7613 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7614 prefetch-dynamic-strides setting.
7616 2018-05-07 Luis Machado <luis.machado@linaro.org>
7618 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7619 <minimum_stride>: New const int field.
7620 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7621 minimum_stride field.
7622 (exynosm1_prefetch_tune): Likewise.
7623 (thunderxt88_prefetch_tune): Likewise.
7624 (thunderx_prefetch_tune): Likewise.
7625 (thunderx2t99_prefetch_tune): Likewise.
7626 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7627 (aarch64_override_options_internal): Update to set
7628 PARAM_PREFETCH_MINIMUM_STRIDE.
7629 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7630 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7631 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7632 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7633 stride is constant and is below the minimum stride threshold.
7635 2018-05-07 Luis Machado <luis.machado@linaro.org>
7637 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7640 2018-05-07 Luis Machado <luis.machado@linaro.org>
7642 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7643 <prefetch_dynamic_strides>: New const bool field.
7644 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7645 prefetch_dynamic_strides.
7646 (exynosm1_prefetch_tune): Likewise.
7647 (thunderxt88_prefetch_tune): Likewise.
7648 (thunderx_prefetch_tune): Likewise.
7649 (thunderx2t99_prefetch_tune): Likewise.
7650 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7652 (aarch64_override_options_internal): Update to set
7653 PARAM_PREFETCH_DYNAMIC_STRIDES.
7654 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7655 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7656 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7657 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7658 prefetch-dynamic-strides setting.
7660 2018-05-07 Luis Machado <luis.machado@linaro.org>
7662 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7663 <minimum_stride>: New const int field.
7664 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7665 minimum_stride field.
7666 (exynosm1_prefetch_tune): Likewise.
7667 (thunderxt88_prefetch_tune): Likewise.
7668 (thunderx_prefetch_tune): Likewise.
7669 (thunderx2t99_prefetch_tune): Likewise.
7670 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7671 (aarch64_override_options_internal): Update to set
7672 PARAM_PREFETCH_MINIMUM_STRIDE.
7673 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7674 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7675 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7676 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7677 stride is constant and is below the minimum stride threshold.
7679 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7682 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7683 the type is addressable. Don't force op into register if it has
7686 2018-05-05 Roland McGrath <mcgrathr@google.com>
7689 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7690 any section for which we don't know a specific type it should have,
7691 regardless of name. Previously this was done only for the exact
7692 names ".init_array", ".fini_array", and ".preinit_array".
7693 (default_elf_asm_named_section): Add comment about
7694 relationship with default_section_type_flags and SECTION_NOTYPE.
7695 (get_section): Don't consider it a type conflict if one side has
7696 SECTION_NOTYPE and the other doesn't, as long as neither has the
7697 SECTION_BSS et al used in the default_section_type_flags logic.
7699 2018-05-05 Tom de Vries <tom@codesourcery.com>
7702 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7703 (workaround_barsyncs): New function.
7704 (nvptx_reorg): Use workaround_barsyncs.
7705 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7706 (define_expand "nvptx_membar_cta"): New define_expand.
7707 (define_insn "*nvptx_membar_cta"): New insn.
7709 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7711 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7712 To improve optimization opportunities.
7713 * builtin-types.def: The new needed builtin types for the above.
7715 2018-05-04 Richard Biener <rguenther@suse.de>
7717 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7718 * gimple-ssa-store-merging.c
7719 (imm_store_chain_info::output_merged_store): Remove redundant create,
7720 release split_store vector contents on failure.
7721 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7722 scalar stmt vector on cache hit.
7724 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7726 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7728 * config.gcc (powerpc-xilinx-eabi*): Remove.
7729 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7731 (fusion_addis_mem_combo_load): Ditto.
7732 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7734 (rs6000_cpu_cpp_builtins): Ditto.
7735 * config/rs6000/rs6000-linux.c
7736 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7737 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7738 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7740 (rs6000_setup_reg_addr_masks): Ditto.
7741 (rs6000_init_hard_regno_mode_ok): Ditto.
7742 (rs6000_option_override_internal): Ditto.
7743 (legitimate_lo_sum_address_p): Ditto.
7744 (rs6000_legitimize_address): Ditto.
7745 (rs6000_legitimize_reload_address): Ditto.
7746 (rs6000_legitimate_address_p): Ditto.
7747 (abi_v4_pass_in_fpr): Ditto.
7748 (setup_incoming_varargs): Ditto.
7749 (rs6000_gimplify_va_arg): Ditto.
7750 (rs6000_split_multireg_move): Ditto.
7751 (rs6000_savres_strategy): Ditto.
7752 (rs6000_emit_prologue_components): Ditto.
7753 (rs6000_emit_epilogue_components): Ditto.
7754 (rs6000_emit_prologue): Ditto.
7755 (rs6000_emit_epilogue): Ditto.
7756 (rs6000_elf_file_end): Ditto.
7757 (rs6000_function_value): Ditto.
7758 (rs6000_libcall_value): Ditto.
7759 * config/rs6000/rs6000.h: Ditto.
7760 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7761 (TARGET_MINMAX): ... this. New.
7762 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7763 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7764 (*movsi_internal1_single): Delete.
7765 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7766 mfpu=, mxilinx-fpu): Delete.
7767 * config/rs6000/singlefp.h: Delete.
7768 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7769 * config/rs6000/t-rs6000: Ditto.
7770 * config/rs6000/t-xilinx: Delete.
7771 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7772 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7773 (VStype_simple): Delete.
7774 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7775 * config/rs6000/xfpu.h: Delete.
7776 * config/rs6000/xfpu.md: Delete.
7777 * config/rs6000/xilinx.h: Delete.
7778 * config/rs6000/xilinx.opt: Delete.
7779 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7780 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7782 2018-05-04 Tom de Vries <tom@codesourcery.com>
7785 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7788 2018-05-04 Richard Biener <rguenther@suse.de>
7791 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7792 (expand_complex_div_wide): Likewise.
7793 (expand_complex_operations_1): Likewise.
7794 (expand_complex_libcall): Preserve EH info of the original stmt.
7795 (tree_lower_complex): Handle removed blocks.
7796 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7797 on complex multiplication and division libcall builtins.
7799 2018-05-04 Richard Biener <rguenther@suse.de>
7802 * fold-const.c (negate_expr_p): Restrict negation of operand
7803 zero of a division to when we know that can happen without
7805 (fold_negate_expr_1): Likewise.
7807 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7810 * real.h (real_nextafter): Declare.
7811 * real.c (real_nextafter): New function.
7812 * fold-const-call.c (fold_const_nextafter): New function.
7813 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7814 CASE_CFN_NEXTTOWARD.
7815 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7816 even when arg1_mode is different from arg0_mode.
7818 2018-05-03 Nathan Sidwell <nathan@acm.org>
7820 * doc/extend.texi (Deprecated Features): Remove
7822 (Backwards Compatibility): Likewise.
7823 * doc/invoke.texi (C++ Language Options): Likewise.
7824 (C++ Dialect Options): Likewise.
7826 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7829 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7830 _mm512_mask_mullox_epi64): New intrinsics.
7832 2018-05-03 Tom de Vries <tom@codesourcery.com>
7835 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7836 dump files): Add offload-tree.
7838 2018-05-03 Richard Biener <rguenther@suse.de>
7840 PR tree-optimization/85615
7841 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7842 to loops not nested in BBs loop father to avoid creating multi-entry
7845 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7847 PR tree-optimization/70291
7848 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7849 arguments. Change return type to tree. Emit libcall as a new
7850 statement rather than replacing existing one when inplace_p is true.
7851 (expand_complex_multiplication_components): New function.
7852 (expand_complex_multiplication): Expand floating-point complex
7853 multiplication using the above.
7854 (expand_complex_division): Rename inner_type parameter to type.
7855 Update expand_complex_libcall call-site.
7856 (expand_complex_operations_1): Update expand_complex_multiplication
7857 and expand_complex_division call-sites.
7859 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7862 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7863 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7864 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7865 the highest significant bit of the shift count mask is clear. In
7866 check whether and[sq]i3 is needed verify that all significant bits
7867 of the shift count other than the highest are set.
7869 2018-05-02 Tom de Vries <tom@codesourcery.com>
7872 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7873 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7874 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7875 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7876 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7877 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7878 __builtin_goacc_parlevel_size.
7880 2018-05-02 Richard Biener <rguenther@suse.de>
7882 PR tree-optimization/85597
7883 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7884 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7887 2018-05-02 Tom de Vries <tom@codesourcery.com>
7890 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7891 dump files): Add ltrans-tree.
7893 2018-05-02 Tom de Vries <tom@codesourcery.com>
7896 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7897 dump files): Add wpa-ipa.
7899 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7901 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7902 powerpc*-*-linux*paired* target.
7903 * config/rs6000/750cl.h: Delete.
7904 * config/rs6000/paired.h: Delete.
7905 * config/rs6000/paired.md: Delete.
7906 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7908 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7909 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7910 comment. Remove paired float support.
7911 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7912 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7914 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7915 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7917 * config/rs6000/rs6000.c: Remove paired float support.
7918 (paired_expand_vector_init, paired_expand_vector_move,
7919 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7920 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7921 paired_expand_builtin, paired_expand_predicate_builtin,
7922 paired_init_builtins): Delete.
7923 * config/rs6000/rs6000.h: Remove paired float support.
7924 * config/rs6000/rs6000.md: Remove paired float support.
7925 (move_from_CR_ov_bit): Delete.
7926 * config/rs6000/rs6000.opt (mpaired): Delete.
7927 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7928 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7930 2018-05-02 Richard Biener <rguenther@suse.de>
7933 * gimplify.c (gimplify_save_expr): When in SSA form allow
7934 SAVE_EXPRs to compute to SSA vars.
7936 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7939 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7940 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7941 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7942 clobber operands[2], instead use a new pseudo. Formatting fixes.
7944 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7946 PR tree-optimization/85586
7947 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7948 exit early for statements in the same group if the accesses are
7951 2018-05-02 Tom de Vries <tom@codesourcery.com>
7954 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7957 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7959 PR tree-optimization/85143
7960 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7962 2018-05-01 Tom de Vries <tom@codesourcery.com>
7965 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7966 not found" error message.
7968 2018-05-01 Tom de Vries <tom@codesourcery.com>
7971 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7972 * vec.c (test_ordered_remove_if): New function.
7973 (vec_c_tests): Call test_ordered_remove_if.
7974 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7975 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7976 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7977 VEC_ORDERED_REMOVE_IF.
7979 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7981 PR tree-optimization/82665
7982 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7983 pointer subtraction where arguments come from a memchr call.
7985 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7987 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7988 --push-state --as-needed and --pop-state instead of --as-needed and
7989 --no-as-needed if ld supports it.
7990 * configure: Regenerated.
7993 * doc/install.texi2html: Replace _002d with - and _002a with * in
7994 generated html files using sed.
7996 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7999 * gcc-rich-location.c (blank_line_before_p): New function.
8000 (use_new_line): New function.
8001 (gcc_rich_location::add_fixit_insert_formatted): New function.
8002 * gcc-rich-location.h
8003 (gcc_rich_location::add_fixit_insert_formatted): New function.
8005 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8007 * selftest.c (assert_streq): Rename "expected" and "actual" to
8008 "val1" and "val2". Extend NULL-handling to cover both inputs
8009 symmetrically, while still requiring both to be non-NULL for a pass.
8010 * selftest.h (assert_streq): Rename "expected" and "actual" to
8012 (ASSERT_EQ): Likewise.
8013 (ASSERT_EQ_AT): Likewise.
8014 (ASSERT_KNOWN_EQ): Likewise.
8015 (ASSERT_KNOWN_EQ_AT): Likewise.
8016 (ASSERT_NE): Likewise.
8017 (ASSERT_MAYBE_NE): Likewise.
8018 (ASSERT_MAYBE_NE_AT): Likewise.
8019 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8020 the assertion to pass.
8021 (ASSERT_STREQ_AT): Likewise.
8023 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8025 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8026 interaction with -pie.
8028 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8030 * selftest.h: Fix alphabetization of per-source-file selftest
8033 2018-04-30 Jason Merrill <jason@redhat.com>
8035 PR c++/61982 - dead stores to destroyed objects.
8036 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8039 2018-04-30 Jason Merrill <jason@redhat.com>
8041 * tree.c (build_clobber): New.
8042 * tree.h: Declare it.
8043 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8045 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8047 * diagnostic-show-locus.c (layout::layout): Update for
8048 location_get_source_line returning a char_span.
8049 (struct char_span): Move to input.h.
8050 (struct correction): Update for fields in char_span becoming
8052 (struct source_line): Update for location_get_source_line
8053 returning a char_span.
8054 (layout::print_line): Likewise.
8055 * edit-context.c (edited_file::print_content): Likewise.
8056 (edited_file::print_diff_hunk): Likewise.
8057 (edited_file::print_run_of_changed_lines): Likewise.
8058 (edited_file::get_num_lines): Likewise.
8059 (edited_line::edited_line): Likewise.
8060 * final.c (asm_show_source): Likewise.
8061 * input.c (location_get_source_line): Convert return type
8062 from const char * to char_span, losing the final "line_len"
8064 (dump_location_info): Update for the above.
8065 (get_substring_ranges_for_loc): Likewise. Use a char_span
8066 when handling the literal within the line.
8067 (test_reading_source_line): Update for location_get_source_line
8068 returning a char_span.
8069 * input.h (class char_span): Move here from
8070 diagnostic-show-locus.c, converting from a struct to a class.
8071 Make data members private.
8072 (char_span::operator bool): New.
8073 (char_span::length): New.
8074 (char_span::get_buffer): New.
8075 (char_span::operator[]): New.
8076 (char_span::subspan): Make const.
8077 (char_span::xstrdup): New.
8078 (location_get_source_line): Convert return type from const char *
8079 to char_span, losing the final "line_size" param.
8081 2018-04-30 Jan Hubicka <jh@suse.cz>
8083 * lto-wrapper.c (ltrans_priorities): New static var.
8084 (cmp_priority): New.
8085 (run_gcc): Read priorities and if doing parallel build order
8086 the Makefile by them.
8088 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8090 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8092 2018-04-30 Richard Biener <rguenther@suse.de>
8094 * tree-cfg.c (verify_address): Remove base argument, add
8095 flag whether to check TREE_ADDRESSABLE and do that.
8096 (verify_expr): Remove.
8097 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8098 (verify_gimple_assign_single): Likewise.
8099 (verify_gimple_switch): Likewise.
8100 (verify_expr_location_1): Dereference tp once. Add (disabled)
8101 piece from verify_expr.
8102 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8104 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8106 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8108 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8110 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8111 (small_data_pattern): Likewise.
8112 (arc_rewrite_small_data): Likewise.
8113 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8114 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8115 (get_symbol_alignment): New function.
8116 (legitimate_small_data_address_p): Likewise.
8117 (legitimate_scaled_address): Update, call
8118 legitimate_small_data_address_p.
8119 (output_sdata): New static variable.
8120 (arc_print_operand): Update how we handle small data operands.
8121 (arc_print_operand_address): Likewise.
8122 (arc_legitimate_address_p): Update, use
8123 legitimate_small_data_address_p.
8124 (arc_rewrite_small_data_p): Remove.
8125 (arc_rewrite_small_data_1): Likewise.
8126 (arc_rewrite_small_data): Likewise.
8127 (small_data_pattern): Likewise.
8128 (compact_sda_memory_operand): Update to use
8129 legitimate_small_data_address_p and get_symbol_alignment.
8130 (prepare_move_operands): Don't rewite sdata pattern.
8131 (prepare_extend_operands): Remove.
8132 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8134 (zero_extendqisi2): Likewise.
8135 (zero_extendhisi2): Likewise.
8136 (extendqihi2): Likewise.
8137 (extendqisi2): Likewise.
8138 (extendhisi2): Likewise.
8142 * config/arc/constraints.md (Usd): Change it to memory constraint.
8144 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8146 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8147 as source of std instructions.
8148 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8149 allow 6-bit constants as source for store instructions.
8150 (movdi_insn): Update instruction pattern to allow 6-bit constants
8151 as source for store instructions.
8153 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8155 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8157 2018-04-30 Nathan Sidwell <nathan@acm.org>
8158 Sandra Loosemore <sandra@codesourcery.com>
8160 * dumpfile.c (dump_open): Allow '-' for stdout.
8161 * doc/invoke.texi (Developer Options): Document dump filename
8162 determination early. Document stdin/stdout selection.
8164 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8166 Microblaze Target: PIC data text relative
8168 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8169 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8171 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8172 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8173 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8174 New addressing mode for data-text relative position indepenedent code.
8175 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8176 'ADDRESS_SYMBOLIC_TXT_REL'.
8177 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8178 (microblaze_legitimate_pic_operand): Exclude function calls from
8179 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8180 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8182 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8183 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8184 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8185 for 'address + offset'.
8186 (microblaze_expand_prologue): Add new function prologue call for
8188 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8189 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8190 table in case of TARGET_PIC_DATA_TEXT_REL.
8191 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8192 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8193 Add new macros 'UNSPEC_TEXT',
8194 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8195 + exclude function calls from 'UNSPEC_PLT' in case of data text
8197 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8198 new target hook for generating address diff vector tables in case of
8200 * doc/tm.texi : Regenerate.
8201 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8202 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8203 of addr diff vector generation.
8204 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8205 target hook definition.
8206 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8207 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8208 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8209 option for data text relative.
8211 2018-04-30 Richard Biener <rguenther@suse.de>
8213 * tree-chrec.h (evolution_function_is_constant_p): Remove
8215 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8217 2018-04-30 Richard Biener <rguenther@suse.de>
8220 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8222 2018-04-30 Richard Biener <rguenther@suse.de>
8224 PR tree-optimization/28364
8225 PR tree-optimization/85275
8226 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8227 copying first exit test.
8229 2018-04-28 Mark Wielaard <mark@klomp.org>
8231 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8233 (dwarf_AT): Handle DW_AT_addr_base.
8234 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8236 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8239 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8240 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8241 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8242 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8244 2018-04-28 Richard Biener <rguenther@suse.de>
8246 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8247 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8248 to reflect use. Only add interesting stmts.
8250 2018-04-27 Martin Jambor <mjambor@suse.cz>
8253 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8254 the jump function allows for passing through aggregate values.
8256 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8258 * input.h (in_system_header_at): Convert from macro to inline
8260 (from_macro_expansion_at): Likewise.
8261 (from_macro_definition_at): Likewise.
8263 2018-04-27 Jeff Law <law@redhat.com>
8265 * config.gcc: Mark tile* targets as deprecated/obsolete.
8267 2018-04-27 Richard Biener <rguenther@suse.de>
8269 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8272 2018-04-27 Richard Biener <rguenther@suse.de>
8274 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8276 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8278 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8279 with Yd constraint. Set "preferred_for_speed" attribute from
8280 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8282 (*movdi_internal): Ditto.
8283 (movti_interunit splitters): Remove
8284 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8285 (movdi_interunit splitters): Ditto.
8286 * config/i386/constraints.md (Ye): Remove.
8287 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8289 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8292 * config/aarch64/constraints.md (Usg): Limit to 31.
8295 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8297 PR tree-optimization/85529
8298 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8299 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8300 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8301 zero extension or masking of the MSB bit.
8302 (optimize_range_tests): Add FIRST_BB argument, pass it through
8303 to optimize_range_tests_var_bound.
8304 (maybe_optimize_range_tests, reassociate_bb): Adjust
8305 optimize_range_tests callers.
8307 2018-04-26 Richard Biener <rguenther@suse.de>
8308 Jakub Jelinek <jakub@redhat.com>
8310 * cgraph.h (symbol_table): Just declare debug method here.
8311 * symtab.c (symbol_table::debug): Define.
8313 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8315 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8317 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8319 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8320 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8321 (*movdi_internal): Substitute Yi and Yj constraint with x
8322 and Ym and Yn constraint with y constraint. Update "isa"
8323 attribute and set "preferred_for_speed" attribute from
8324 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8325 (*movsi_internal): Ditto.
8326 (*movdf_internal): Ditto.
8327 (*movsf_internal): Ditto.
8328 (*zero_extendsidi2): Ditto.
8329 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8330 (sse2_loadld): Ditto.
8331 (*vec_extract<ssevecmodelower>_0): Ditto.
8332 (*vec_extractv4si_0_zext_sse4): Ditto.
8333 (vec_concatv2di): Ditto.
8334 (*vec_dup<mode>): Ditto.
8335 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8336 * config/i386/constraints.md (Yi): Remove.
8341 2018-04-26 Nathan Sidwell <nathan@acm.org>
8343 * dumpfile.c (dump_open): New.
8344 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8345 (dump_finish): Detect stdio/stderr by value not name.
8347 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8349 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8351 2018-04-26 Tom de Vries <tom@codesourcery.com>
8354 * config/nvptx/nvptx.c (verify_neutering_jumps)
8355 (verify_neutering_labels): New function
8356 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8358 2018-04-26 Tom de Vries <tom@codesourcery.com>
8361 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8362 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8365 2018-04-26 Richard Biener <rguenther@suse.de>
8366 Tom de Vries <tom@codesourcery.com>
8369 * lto-streamer-out.c (output_function): Fixup loops if required to match
8370 discovery done in the reader.
8372 2018-04-26 Richard Biener <rguenther@suse.de>
8374 PR tree-optimization/85116
8375 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8376 have a loop exit from the single latch predecessor. Remove
8377 case of header with just condition.
8378 (ch_base::copy_headers): Exclude infinite loops from any
8380 (pass_ch::execute): Record exits.
8382 2018-04-26 Richard Biener <rguenther@suse.de>
8384 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8385 prologue cost vector and pass it to vect_get_load_cost.
8386 (vect_get_peeling_costs_all_drs): Likewise.
8387 (vect_peeling_hash_get_lowest_cost): Likewise.
8388 (vect_enhance_data_refs_alignment): Likewise.
8390 2018-04-26 Richard Biener <rguenther@suse.de>
8393 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8394 checking of integer<->pointer conversions.
8395 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8396 sign-/zero-extending pointer types.
8397 (expand_omp_for_static_chunk): Likewise.
8399 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8400 Jean Lee <xiaoyur347@gmail.com>
8402 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8403 (TARGET_ASAN_SHADOW_OFFSET): Define.
8404 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8405 true for -fsanitize=address.
8407 2018-04-25 Mark Wielaard <mark@klomp.org>
8409 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8412 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8414 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8415 than "alu", remove explicit "memory" and "imm_disp" attributes.
8416 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8419 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8420 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8421 gen_lowpart_no_emit.
8423 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8426 * config/i386/i386.c (ix86_expand_builtin): Change memory
8427 operand to XI, extend p0 to Pmode.
8428 * config/i386/i386.md: Change unspec volatile and operand
8429 1 mode to XI, change operand 0 mode to P.
8431 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8433 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8434 GET_MODE_MASK before any checking.
8435 (nds32_can_use_bset_p): Likewise.
8436 (nds32_can_use_btgl_p): Likewise.
8438 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8440 * config/nds32/nds32-doubleword.md: New define_split pattern for
8441 illegal register number.
8443 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8445 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8447 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8449 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8451 2018-04-25 Richard Biener <rguenther@suse.de>
8453 * lto-streamer.h (LTO_major_version): Bump to 8.
8455 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8457 * BASE-VER: Set to 9.0.0.
8459 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8461 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8462 in __abskf2 and __powikf2.
8464 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8467 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8468 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8469 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8470 Use the above on operand 2. Reindent.
8471 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8473 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8476 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8478 * config/i386/i386.opt (mcet): Removed.
8479 * doc/install.texi: Remove -mcet documentation.
8480 * doc/invoke.texi: Likewise.
8482 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8485 * doc/install.texi: Remove -mcet from bootstrap-cet.
8487 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8490 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8491 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8495 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8496 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8497 containing a CONST_VECTOR.
8499 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8501 * doc/install.texi: Update newlib dependency for nvptx.
8503 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8506 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8507 instead of INTVAL when shifting x left.
8509 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8511 PR tree-optimization/85478
8512 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8513 vect_grouped_store_supported for single element vectors.
8515 2018-04-24 Richard Biener <rguenther@suse.de>
8518 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8519 load cost increase to the case of non-constant step.
8521 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8524 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8525 destination if any_malformed_asm.
8527 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8530 * expr.c (store_field): In the bitfield case, if the value comes from
8531 a function call and is returned in registers by means of a PARALLEL,
8532 do not change the mode of the temporary unless BLKmode and VOIDmode.
8534 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8536 PR rtl-optimization/85423
8537 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8538 dependencies to debug insns when the previous insn is non-debug.
8540 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8542 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8543 enums into a single definition.
8544 (fls): Fix predicates and printing.
8547 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8549 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8550 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8551 and short u6 immediate.
8552 (check_if_valid_sleep_operand): Remove.
8553 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8555 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8557 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8558 flag_always_save_lp condition.
8559 * config/nds32/nds32.opt (malways-save-lp): New option.
8561 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8563 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8564 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8565 * config/nds32/nds32.h
8566 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8567 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8569 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8571 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8572 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8574 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8575 Chung-Ju Wu <jasonwucj@gmail.com>
8577 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8579 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8580 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8582 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8584 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8586 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8588 * config/nds32/nds32-protos.h (nds32_data_alignment,
8589 nds32_local_alignment): Declare.
8590 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8591 nds32_local_alignment): New functions.
8592 (TARGET_CONSTANT_ALIGNMENT): Define.
8593 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8595 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8597 * config/nds32/nds32.c
8598 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8599 (TARGET_MODES_TIEABLE_P): Likewise.
8601 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8603 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8606 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8607 Chung-Ju Wu <jasonwucj@gmail.com>
8609 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8610 for unaligned access.
8611 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8613 * config/nds32/nds32-intrinsic.md: Likewise.
8614 * config/nds32/nds32_intrinsic.h: Likewise.
8615 * config/nds32/nds32.h (nds32_builtins): Likewise.
8616 * config/nds32/nds32.opt (munaligned-access): New option.
8617 * config/nds32/nds32.c (nds32_asm_file_start): Display
8618 flag_unaligned_access status.
8620 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8622 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8623 -mno-relax is present.
8624 * config/riscv/linux.h (LINK_SPEC): Ditto.
8626 2018-04-20 Martin Sebor <msebor@redhat.com>
8629 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8631 (gimple_fold_builtin_stxcpy_chk): Same.
8632 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8634 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8637 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8638 __powikf2 when long double is IEEE 128-bit.
8640 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8642 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8643 step to make sure stack always aligned.
8645 2018-04-20 Carl Love <cel@us.ibm.com>
8648 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8649 size check for arg0.
8651 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8652 Tom de Vries <tom@codesourcery.com>
8655 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8656 Emit insns for calls too.
8657 (nvptx_find_par): Always look for worker-level predecessor insn.
8658 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8660 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8661 (nvptx_process_pars): Propagate frames for calls.
8663 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8666 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8668 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8669 (ix86_handle_option): Don't handle OPT_mibt.
8670 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8672 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8674 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8675 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8676 (ix86_target_macros): Define __CET__ with flag_cf_protection
8677 for -fcf-protection.
8678 * config/i386/i386.c (isa2_opts): Remove -mibt.
8679 * config/i386/i386.h (TARGET_IBT): Removed.
8680 (TARGET_IBT_P): Likewise.
8681 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8682 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8683 * config/i386/i386.opt (mcet): Update help message.
8686 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8687 -mcet as an alias for -mshstk.
8689 2018-04-20 Richard Biener <rguenther@suse.de>
8692 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8693 complexity by forcing a single use of the multiply operand.
8695 2018-04-20 Martin Jambor <mjambor@suse.cz>
8698 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8699 recursion dependency to only apply to non-clones.
8701 2018-04-20 Martin Jambor <mjambor@suse.cz>
8704 * ipa-cp.c (create_specialized_node): Check that clones of
8705 self-recursive edges exist during IPA-CP.
8707 2018-04-19 Toon Moene <toon@moene.org>
8709 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8712 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8714 PR tree-optimization/85467
8715 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8716 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8717 VECTOR_CST element to type.
8719 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8722 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8723 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8724 (builtin_longjmp): Likewise.
8725 (save_stack_nonlocal): New pattern.
8726 (restore_stack_nonlocal): Likewise.
8728 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8731 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8732 Replace ASM_OUTPUT_LABEL with fprintf.
8734 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8737 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8738 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8739 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8740 define __IBT__ and __SHSTK__ for -fcf-protection.
8741 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8743 (ix86_trampoline_init): Likewise.
8744 (x86_output_mi_thunk): Likewise.
8745 (ix86_notrack_prefixed_insn_p): Likewise.
8746 (ix86_option_override_internal): Don't disallow -fcf-protection.
8747 * config/i386/i386.md (rdssp<mode>): Also enable for
8749 (incssp<mode>): Likewise.
8750 (nop_endbr): Likewise.
8751 * config/i386/i386.opt (mcet): Change help message to built-in
8755 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8756 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8757 enable CET built-in functions.
8759 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8761 * common/config/i386/i386-common.c
8762 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8763 OPTION_MASK_ISA_MOVDIRI_UNSET,
8764 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8765 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8766 * config.gcc (movdirintrin.h): New header.
8767 * config/i386/cpuid.h (bit_MOVDIRI,
8768 bit_MOVDIR64B): New bits.
8769 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8771 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8772 (VOID, PVOID, PCVOID)): New function types.
8773 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8774 __builtin_ia32_directstoreu_u64,
8775 __builtin_ia32_movdir64b): New builtins.
8776 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8777 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8779 (ix86_valid_target_attribute_inner_p): Ditto.
8780 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8781 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8782 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8783 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8784 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8785 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8786 (movdiri<mode>, movdir64b_<mode>): New.
8787 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8788 * config/i386/immintrin.h: Include movdirintrin.h.
8789 * config/i386/movdirintrin.h: New file.
8790 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8792 2018-04-19 Richard Biener <rguenther@suse.de>
8795 * cfg.c (clear_bb_flags): When loop state says we have
8796 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8798 2018-04-19 Richard Biener <rguenther@suse.de>
8800 PR tree-optimization/84737
8801 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8802 copying restrict info.
8803 (vect_setup_realignment): Use it.
8804 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8805 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8806 the first DR to all generated stores.
8807 (vectorizable_load): Likewise for loads.
8809 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8811 PR tree-optimization/85446
8812 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8813 the integral and pointer types to have the same precision.
8815 * doc/install.texi: Document --disable-cet being the default and
8818 2018-04-18 Martin Liska <mliska@suse.cz>
8820 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8823 2018-04-18 Martin Liska <mliska@suse.cz>
8826 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8829 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8830 arguments if they are comparable.
8832 2018-04-18 Martin Liska <mliska@suse.cz>
8835 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8838 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8841 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8844 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8845 ENDBR after calling __morestack.
8847 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8850 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8851 by using gcc_base_ver to generate a gcc_driver_version, and use
8852 it when generating GCC_DRIVER_NAME.
8853 * configure: Regenerate.
8855 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8858 * config.gcc: Obsolete powerpc*-*-*spe*.
8860 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8863 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8864 (stabstr_D): Change type of unum from unsigned int to
8865 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8868 2018-04-17 Jim Wilson <jimw@sifive.com>
8871 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8872 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8873 Set arg_pointer_offset after using pretend_args_size.
8875 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8877 PR rtl-optimization/85431
8878 * dse.c (record_store): Ignore zero width stores.
8881 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8882 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8883 __builtin_stack_restore rather than after it.
8884 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8885 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8886 argument instead of virtual_dynamic_stack_rtx.
8888 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8890 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8892 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8893 Add note to error message to explain internal mapping of overloaded
8894 built-in function name to non-overloaded built-in function name.
8895 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8898 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8901 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8902 where the inputs overlap with the output.
8904 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8906 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8907 (=v, v) alternative and explicit "memory" attribute.
8908 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8909 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8911 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8912 "sselog1" type instead of "sselog".
8913 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8914 "sselog". Remove explicit "memory" attribute.
8915 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8916 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8918 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8919 "sselog1" type instead of "sselog". Remove explicit "memory"
8921 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8922 use "sselog1" type instead of "sselog". Remove explicit "memory"
8924 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8925 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8927 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8928 "sselog1" type instead of "sselog". Remove explicit "memory"
8930 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8931 use "sselog1" type instead of "sselog". Remove explicit "memory"
8935 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8938 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8941 2018-04-17 Martin Jambor <mjambor@suse.cz>
8944 * ipa-cp.c (create_specialized_node): Call
8945 expand_all_artificial_thunks if necessary.
8947 2018-04-17 Martin Liska <mliska@suse.cz>
8950 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8951 in message, remote space in between '_G' and '('.
8953 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8956 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8957 avx512f_vmcmp<mode>3<round_saeonly_name>,
8958 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8959 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8960 avx512f_rndscale<mode><round_saeonly_name>,
8961 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8962 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8963 Use %<iptr>2 instead of %2 for -masm=intel.
8964 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8965 avx512f_vcvttss2usi<round_saeonly_name>,
8966 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8968 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8969 avx512f_vcvttsd2usi<round_saeonly_name>,
8970 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8971 Use %q1 instead of %1 for -masm=intel.
8972 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8973 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8974 of %3 for -masm=intel.
8975 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8977 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8979 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8981 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8982 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8984 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8986 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8987 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8988 %g1 and one with %0 and %1.
8989 (avx512er_vmrcp28<mode><round_saeonly_name>,
8990 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8992 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8993 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8994 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8995 of %0 and %{%4%} for -masm=intel.
8996 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8997 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8998 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8999 order of %0 and %{%5%}%{z%} for -masm=intel.
9001 2018-04-17 Jan Hubicka <jh@suse.cz>
9004 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9006 2018-04-17 Martin Liska <mliska@suse.cz>
9009 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9010 for target_clone error message. Make default implementation
9011 clone to be a local declaration.
9012 (separate_attrs): Add new argument and check for an empty
9014 (expand_target_clones): Handle it.
9015 (ipa_target_clone): Make redirection just for target_clones
9018 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9019 Tom de Vries <tom@codesourcery.com>
9022 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9023 tiled basic blocks without omp continue statements.
9025 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9028 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9029 vec_extract expression as having side effects to make sure it gets
9032 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9035 * config/i386/i386.c (get_builtin_code_for_version): Check
9038 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9041 * gcc/config.gcc: Support "skylake".
9042 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9044 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9045 (processor_target_table): Add "skylake".
9046 (ix86_option_override_internal): Add "skylake".
9047 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9048 PROCESSOR_CANNONLAKE.
9049 (get_builtin_code_for_version): Fix priority for
9050 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9051 PROCESSOR_SKYLAKE-AVX512.
9052 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9053 (processor_type): Add PROCESSOR_SKYLAKE.
9055 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9056 Jason Merrill <jason@redhat.com>
9059 * convert.c (convert_to_integer_1): Use direct recursion for
9060 enumeral types and types with a precision less than the number
9061 of bits in their mode.
9063 2018-04-16 Julia Koval <julia.koval@intel.com>
9066 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9067 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9069 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9072 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9073 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9074 and -mno-direct-move.
9076 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9079 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9080 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9081 Ensure negative shifts result in {0}.
9083 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9085 PR rtl-optimization/79916
9086 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9087 regs (if any) to define how to gnerate SD moves when LRA is in
9090 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9092 PR rtl-optimization/85393
9093 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9094 * except.c (expand_dw2_landing_pad_for_region): Make static.
9095 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9096 a label and unconditional jump to old_bb, rather than
9097 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9100 PR rtl-optimization/85376
9101 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9102 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9103 instead of a specific value.
9105 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9106 Bin Cheng <bin.cheng@arm.com>
9108 PR tree-optimization/82965
9109 PR tree-optimization/83991
9110 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9111 by_profile_only parameter.
9112 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9113 information if the loop was predicted to iterate too many times.
9114 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9116 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9119 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9122 2018-04-13 Martin Liska <mliska@suse.cz>
9123 Jakub Jelinek <jakub@redhat.com>
9126 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9127 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9128 * builtins.c (expand_builtin_memory_copy_args): Use
9129 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9130 handle dest_addr == pc_rtx.
9132 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9135 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9136 asked to not generate direct moves.
9137 (fix_trunc<mode>si2_stfiwx): Similar.
9138 (fix_trunc<mode>si2_internal): Similar.
9140 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9143 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9144 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9145 lookup if dest in some wider mode is known to be const0_rtx and
9146 if so, record permanent equivalence for it to be ZERO_EXTEND of
9147 the narrower mode destination.
9149 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9151 * lto-streamer-out.c (output_function): Revert 259346.
9152 * omp-expand.c (expand_oacc_for): Likewise.
9154 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9156 PR rtl-optimization/85354
9157 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9158 * sel-sched.c (sel_global_init): ... here.
9160 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9163 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9164 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9165 mode for PE-COFF targets.
9166 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9167 (i386_pe_asm_lto_end): Likewise.
9168 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9169 (TARGET_ASM_LTO_END): Likewise.
9170 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9171 (i386_pe_asm_lto_start): New function.
9172 (i386_pe_asm_lto_end): Likewise.
9174 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9175 Richard Biener <rguenther@suse.de>
9178 * lto-streamer-out.c (output_function): Fix CFG loop state before
9180 * omp-expand.c (expand_oacc_for): Handle calls to internal
9181 functions like regular functions.
9183 2018-04-12 Richard Biener <rguenther@suse.de>
9186 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9187 for the early LTO debug to properly generate references to it
9188 during DIE emission. Do not re-use that for the skeleton for
9190 (dwarf2out_early_finish): Likewise.
9192 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9195 * config/i386/sse.md
9196 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9197 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9198 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9199 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9200 and output is a reg, avoid creating invalid lowpart subreg, but
9201 instead split into a 512-bit move. Don't split if not AVX512VL,
9202 input is xmm16+ reg and output is a mem.
9203 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9204 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9205 xmm16+ reg and output is a mem.
9207 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9209 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9210 also for flag_dwarf2_cfi_asm.
9212 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9214 PR rtl-optimization/85342
9215 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9216 a bool scalar var inside of the loop instead. Don't try to update
9217 recog_data.operand after failed apply_change_group.
9219 2018-04-12 Tom de Vries <tom@codesourcery.com>
9222 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9223 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9224 array with flexible array member as array without given dimension.
9225 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9226 argument for undefined param to true.
9228 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9231 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9232 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9233 from PowerPC section.
9234 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9235 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9236 help text that is too long.
9237 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9238 help text that is too long.
9239 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9240 help text that is too long.
9242 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9244 * config/alpha/alpha.md (stack_probe_internal): Rename
9245 from "probe_stack". Update all callers.
9247 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9249 PR rtl-optimization/84566
9250 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9251 sched_macro_fuse_insns.
9253 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9256 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9257 (compute_block_dependences): ... from here.
9259 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9261 PR tree-optimization/85331
9262 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9263 from int to HOST_WIDE_INT.
9265 2018-04-11 Martin Jambor <mjambor@suse.cz>
9268 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9269 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9270 not the same as the source val.
9271 (cgraph_edge_brings_value_p): New parameter.
9272 (gather_edges_for_value): Pass destination value to
9273 cgraph_edge_brings_value_p.
9274 (perhaps_add_new_callers): Likewise.
9275 (get_info_about_necessary_edges): Likewise and exclude values brought
9276 only by self-recursive edges.
9277 (create_specialized_node): Redirect only clones of self-calling edges.
9278 (+self_recursive_pass_through_p): New function.
9279 (find_more_scalar_values_for_callers_subset): Use it.
9280 (find_aggregate_values_for_callers_subset): Likewise.
9281 (known_aggs_to_agg_replacement_list): Removed.
9282 (decide_whether_version_node): Re-calculate known constants for all
9283 remaining context clones.
9285 2018-04-11 Richard Biener <rguenther@suse.de>
9288 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9289 from early DWARF output.
9290 (dwarf2out_early_finish): Output line info unconditionally into
9291 early DWARF and add reference to it.
9293 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9296 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9297 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9298 other than V2DFmode using iptr mode attribute.
9299 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9301 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9303 PR rtl-optimization/84659
9304 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9306 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9309 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9311 (output_loc_list): Pass address of a dummy size variable even in the
9312 locview handling loop.
9313 (index_location_lists): Add comment on why skip_loc_list_entry can't
9316 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9319 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9322 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9325 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9326 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9327 and -mstring-compare-inline-limit.
9329 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9332 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9333 for stack clash protection in a register whenever we need it to be in
9336 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9338 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9339 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9341 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9344 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9346 (mlong-double-): Ditto.
9347 * config/rs6000/sysv4.opt (msdata=): Ditto.
9348 (mtls-size=): Ditto.
9350 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9352 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9353 erroneous entries for
9354 "vector int vec_ldl (int, long int *)", and
9355 "vector unsigned int vec_ldl (int, unsigned long int *)".
9356 Add comments and entries for
9357 "vector bool char vec_ldl (int, bool char *)",
9358 "vector bool short vec_ldl (int, bool short *)",
9359 "vector bool int vec_ldl (int, bool int *)",
9360 "vector bool long long vec_ldl (int, bool long long *)",
9361 "vector pixel vec_ldl (int, pixel *)",
9362 "vector long long vec_ldl (int, long long *)",
9363 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9364 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9365 type tree bool_long_long_type_node and correct definition of
9366 bool_V2DI_type_node to make reference to this new type tree.
9367 (rs6000_mangle_type): Replace erroneous reference to
9368 bool_long_type_node with bool_long_long_type_node.
9369 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9370 comments to emphasize sign distinctions for char and int types and
9371 replace RS6000_BTI_bool_long constant with
9372 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9373 use of RS6000_BTI_pixel.
9374 (bool_long_type_node): Remove this macro definition.
9375 (bool_long_long_type_node): New macro definition
9377 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9379 PR rtl-optimization/85300
9380 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9381 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9382 simplify_unary_operation fails.
9384 2018-04-10 Martin Liska <mliska@suse.cz>
9386 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9387 cgraph_edge and ipa_ref.
9389 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9393 * config/i386/sse.md
9394 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9395 computation of the VEC_MERGE selector from mask.
9396 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9397 Fix decoding of the VEC_MERGE selector into mask.
9399 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9401 PR tree-optimization/85286
9402 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9404 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9406 * final.c (final_1): Set insn_last_address as well as
9407 insn_current_address.
9409 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9412 * explow.c (emit_stack_probe): Call validize_mem on memory location
9413 before passing it to gen_probe_stack. Create address operand and
9414 legitimize it for the probe_stack_address case.
9416 2018-04-09 Jan Hubicka <jh@suse.cz>
9419 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9420 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9421 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9423 (free_lang_data): Rebuild type inheritance graph.
9425 2018-04-09 Martin Sebor <msebor@redhat.com>
9427 * invoke.texi (-finline-small-functions): Mention other optimization
9429 (-findirect-inlining, -fpartial-inlining): Same.
9430 (-finline-functions-called-once): Same.
9431 (-freorder-blocks-and-partition): Same.
9433 2018-04-09 Jan Hubicka <jh@suse.cz>
9436 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9437 jumps; choose last target that matches the criteria (i.e.
9438 no partition changes for non-crossing jumps).
9439 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9440 support for redirecting crossing jumps to non-crossing.
9442 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9444 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9445 also for naked functions.
9447 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9449 * config/arc/arc.md (add_shift): New pattern.
9450 (add_shift2): Likewise.
9451 (sub_shift): Likewise.
9452 (sub_shift_cmp0_noout): Likewise.
9453 (compare_si_ashiftsi): Likewise.
9454 (xbfu_cmp0_noout): New combine pattern.
9455 (xbfu_cmp0"): Likewise.
9456 (movsi_set_cc_insn): Place the predicable variant first.
9457 (commutative_binary_cmp0_noout): Remove clobber.
9458 (commutative_binary_cmp0): New pattern.
9459 (noncommutative_binary_cmp0): Likewise.
9460 (noncommutative_binary_cmp0_noout): Likewise.
9461 (noncommutative_binary_comparison_result_used): Removed.
9462 (rsub_cmp0): New pattern.
9463 (rsub_cmp0_noout): Likewise.
9464 (extzvsi): Changed, keep only meaningful variants.
9465 (SQH, SEZ): New iterators.
9466 (SQH_postfix): New mode attribute.
9467 (SEZ_prefix): New code attribute.
9468 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9469 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9470 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9472 (noncommutative_operator): Check the availability of barrel
9475 2018-04-09 Richard Biener <rguenther@suse.de>
9477 PR tree-optimization/85284
9478 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9479 Only use the niter constraining form of simple_iv when the exit
9482 2018-04-09 Tom de Vries <tom@codesourcery.com>
9485 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9486 (define_expand "*memory_barrier"): New define_expand.
9487 (define_insn "memory_barrier"): New insn.
9489 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9491 PR rtl-optimization/80463
9492 PR rtl-optimization/83972
9493 PR rtl-optimization/83480
9495 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9496 correct producer for the insn.
9497 (tidy_control_flow): Fixup seqnos in case of debug insns.
9499 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9501 PR rtl-optimization/83913
9503 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9504 different sched-times when merging exprs.
9506 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9508 PR rtl-optimization/83962
9510 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9511 tidy_fallthru_edge and tidy_control_flow.
9513 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9515 PR rtl-optimization/83530
9517 * sel-sched.c (force_next_insn): New global variable.
9518 (remove_insn_for_debug): When force_next_insn is true, also leave only
9519 next insn in the ready list.
9520 (sel_sched_region): When the region wasn't scheduled, make another pass
9521 over it with force_next_insn set to 1.
9523 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9525 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9527 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9528 for interrupt control.
9529 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9530 functions for interrupt control.
9531 * config/nds32/nds32-intrinsic.md: Likewise.
9532 * config/nds32/nds32_intrinsic.h: Likewise.
9533 * config/nds32/nds32.h (nds32_builtins): Likewise.
9535 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9537 * config/nds32/nds32.c (nds32_init_machine_status,
9538 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9539 strict_aligned_p field.
9540 (nds32_expand_to_rtl_hook): New function.
9541 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9542 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9544 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9545 Chung-Ju Wu <jasonwucj@gmail.com>
9547 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9548 * config/nds32/nds32-n7.md: New file.
9549 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9550 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9552 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9553 * config/nds32/nds32.md (pipeline_model): Add n7.
9554 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9555 * config/nds32/pipelines.md: Include n7 settings.
9557 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9558 Chung-Ju Wu <jasonwucj@gmail.com>
9560 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9561 * config/nds32/nds32-e8.md: New file.
9562 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9563 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9565 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9566 * config/nds32/nds32.md (pipeline_model): Add e8.
9567 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9568 * config/nds32/pipelines.md: Include e8 settings.
9570 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9571 Chung-Ju Wu <jasonwucj@gmail.com>
9573 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9574 * config/nds32/nds32-n8.md: New file.
9575 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9576 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9578 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9579 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9580 * config/nds32/nds32.md (pipeline_model): Add n8.
9581 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9582 * config/nds32/pipelines.md: Include n8 settings.
9584 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9585 Chung-Ju Wu <jasonwucj@gmail.com>
9587 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9588 * config/nds32/nds32-n9-2r1w.md: New file.
9589 * config/nds32/nds32-n9-3r2w.md: New file.
9590 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9591 nds32_register_ports): New or modify for cpu n9.
9592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9594 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9595 * config/nds32/nds32-utils.c: New file.
9596 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9597 TARGET_MUL_SLOW): Define.
9598 * config/nds32/nds32.md (pipeline_model): New attribute.
9599 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9600 New options that support cpu n9.
9601 * config/nds32/pipelines.md: Include n9 settings.
9602 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9604 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9606 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9607 information if necessary.
9608 (output_cond_branch_compare_zero): Likewise.
9609 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9610 (nds32_target_alignment): Refine for alignment.
9611 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9612 (FUNCTION_BOUNDARY): Modify.
9613 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9615 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9617 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9619 * config/nds32/constants.md (unspec_volatile_element): Add values for
9620 TLB operation and data prefetch.
9621 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9622 functions for TLB operation and data prefetch.
9623 * config/nds32/nds32-intrinsic.md: Likewise.
9624 * config/nds32/nds32_intrinsic.h: Likewise.
9625 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9626 (nds32_print_operand): Likewise.
9627 * config/nds32/nds32.h (nds32_builtins): Likewise.
9629 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9630 Andrew Pinski <pinsika@gcc.gnu.org>
9633 * match.pd: Use constant_boolean_node of correct type instead of
9634 boolean_true_node or boolean_false_node for simplifying
9635 pointer comparisons to zero.
9637 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9639 PR tree-optimization/80021
9640 * tree.c (verify_type_variant): Make error call in verify_variant_match
9641 translatable and remove final full stop.
9643 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9645 * config/nds32/constants.md (unspec_volatile_element): Add
9646 UNSPEC_VOLATILE_EH_RETURN.
9647 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9648 nds32_output_stack_pop): Support dwarf exception handling process.
9649 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9650 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9651 exception handling process.
9652 (nds32_compute_stack_frame): Likewise.
9653 (nds32_return_addr_rtx): Likewise.
9654 (nds32_initial_elimination_offset): Likewise.
9655 (nds32_expand_prologue): Likewise.
9656 (nds32_expand_epilogue): Likewise.
9657 (nds32_dynamic_chain_address): New function.
9658 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9660 (DYNAMIC_CHAIN_ADDRESS): Define.
9661 (EH_RETURN_DATA_REGNO): Define.
9662 (EH_RETURN_STACKADJ_RTX): Define.
9663 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9664 patterns for dwarf exception handling.
9666 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9668 * config/nds32/nds32.h: Clean up obsolete macros.
9670 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9672 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9673 Add enum values for particular instructions.
9674 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9675 particular intrinsic functions.
9676 * config/nds32/nds32-intrinsic.md: Likewise.
9677 * config/nds32/nds32_intrinsic.h: Likewise.
9678 * config/nds32/nds32.h (nds32_builtins): Likewise.
9679 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9680 (btst, ave): New patterns for particular instructions.
9682 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9684 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9685 Add enum values for atomic load/store and memory sync.
9686 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9688 * config/nds32/nds32-intrinsic.md: Likewise.
9689 * config/nds32/nds32_intrinsic.h: Likewise.
9690 * config/nds32/nds32.h (nds32_builtins): Likewise.
9692 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9694 PR tree-optimization/85257
9695 * fold-const.c (native_encode_vector): If not all elts could fit
9696 and off is -1, return 0 rather than offset.
9697 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9698 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9699 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9700 adjust buffer in native_interpret_expr call.
9702 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9704 * config/nds32/constants.md (unspec_volatile_element): Add cache
9705 control enum values.
9706 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9707 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9708 * config/nds32/nds32.c (nds32_cctl_names): New.
9709 (nds32_print_operand): Handle cache control register names.
9710 * config/nds32/nds32.h (nds32_builtins): New enum values.
9711 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9713 * config/nds32/nds32.md (type): Add mmu.
9714 * config/nds32/pipelines.md (simple_insn): Add mmu.
9716 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9718 * config/nds32/nds32.md (type): Remove call.
9719 * config/nds32/pipelines.md (simple_insn): Likewise.
9721 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9723 * config/nds32/constants.md (unspec_volatile_element): Add
9724 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9725 UNSPEC_VOLATILE_FMFCFG.
9726 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9727 description for fmfcfg and fmfcsr.
9728 (bdesc_1arg): Add fmtcsr.
9729 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9730 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9731 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9732 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9733 unspec_fmfcfg): New patterns.
9734 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9735 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9736 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9737 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9738 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9739 __nds32__fmfcfg): Define.
9741 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9743 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9744 intrinsic register names.
9745 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9746 intrinsic register enum values and macros.
9748 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9750 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9751 for load/store addressing form.
9752 (nds32_print_operand_address): Likewise.
9754 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9757 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9758 based on LABEL_REF. Remove useless assertion.
9759 (pic_address_needs_scratch): Fix formatting.
9760 (sparc_legitimize_pic_address): Minor tweaks.
9761 (sparc_delegitimize_address): Adjust assertion accordingly.
9762 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9763 into symbolic_operand.
9764 (movsi_high_pic_label_ref): Likewise.
9765 (movsi_lo_sum_pic_label_ref): Likewise.
9766 (movdi_pic_label_ref): Likewise.
9767 (movdi_high_pic_label_ref): Likewise.
9768 (movdi_lo_sum_pic_label_ref): Likewise.
9770 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9772 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9773 custom LIB_SPEC setup.
9775 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9776 Kito Cheng <kito.cheng@gmail.com>
9778 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9779 * config/riscv/freebsd.h: New.
9781 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9783 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9784 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9787 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9788 Kito Cheng <kito.cheng@gmail.com>
9790 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9791 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9792 * config/nds32/nds32-protos.h (nds32_output_call,
9793 nds32_output_return): Declare.
9794 * config/nds32/nds32.md: Refine all the call and return patterns.
9796 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9799 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9800 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9802 PR rtl-optimization/84872
9803 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9804 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9807 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9809 * expr.c (copy_blkmode_to_reg): Revert 254862.
9810 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9812 2018-04-06 Richard Biener <rguenther@suse.de>
9815 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9816 after seeing a component reference with an adjacent field. Treat
9817 refs to arrays at struct end of external decls similar to
9818 refs to unconstrained commons.
9820 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9823 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9824 look through SAVE_EXPRs with non-side-effects argument. Adjust
9826 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9829 2018-04-06 Richard Biener <rguenther@suse.de>
9832 * alias.c (find_base_term): New wrapper around find_base_term
9833 unwinding CSELIB_VAL_PTR changes.
9834 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9837 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9839 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9841 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9842 constant definitions.
9843 ("nop"): lr 0,0 -> nopr r0
9844 ("nop_lr0", "nop_lr1"): New insn definitions.
9846 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9848 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9849 NDS32_V3PUSH_AVAILABLE_P macro.
9851 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9852 Chung-Ju Wu <jasonwucj@gmail.com>
9854 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9855 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9856 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9857 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9858 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9859 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9860 * config/nds32/constraints.md: New constraints and checking for hard
9861 float configuration.
9862 * config/nds32/iterators.md: New mode iterator and attribute for hard
9863 float configuration.
9864 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9866 * config/nds32/nds32-fpu.md: New file.
9867 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9868 deal with hard float code generation.
9869 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9871 (abi_type, float_reg_number): New enum type.
9872 * config/nds32/nds32-predicates.c: New predicates for hard float.
9873 * config/nds32/nds32-protos.h: Declare functions for hard float.
9874 * config/nds32/nds32.c: Implementation for hard float configuration.
9875 * config/nds32/nds32.h: Definitions for hard float configuration.
9876 * config/nds32/nds32.md: Include hard float machine description and
9877 modify patterns for hard float configuration.
9878 * config/nds32/nds32.opt: New options for hard float configuration.
9879 * config/nds32/predicates.md: New predicates for hard float
9882 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9884 * common/config/nds32/nds32-common.c
9885 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9887 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9890 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9891 CONSTRUCTOR_ELT (ctor, ...)->value.
9893 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9896 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9898 2018-04-05 Tom de Vries <tom@codesourcery.com>
9901 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9904 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9905 Kito Cheng <kito.cheng@gmail.com>
9907 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9908 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9909 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9910 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9912 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9913 Kito Cheng <kito.cheng@gmail.com>
9915 * config/nds32/constraints.md (Ufe): New memory constraint.
9916 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9917 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9918 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9920 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9921 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9923 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9925 * config/nds32/nds32.md: Use optimize_size in the condition for
9926 alu-shift instructions.
9928 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9930 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9932 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9934 * config/nds32/nds32.md (negsi2): Refine pattern.
9936 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9937 Chung-Ju Wu <jasonwucj@gmail.com>
9939 * config/nds32/iterators.md (shift_rotate): New code iterator.
9940 (shift): New code attribute.
9941 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9942 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9943 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9944 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9945 bit-wise operations.
9946 (andsi3, *andsi3): Ditto.
9947 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9948 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9949 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9950 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9951 nds32_ior_operand, nds32_xor_operand): New predicates.
9953 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9955 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9956 (addsi3, subsi3): ... this.
9958 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9960 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9962 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9964 * config/nds32/nds32.md: Adjust indention.
9966 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9968 * config/nds32/nds32.md (feature): New attribute.
9970 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9972 * config/nds32/nds32.md (subtype): New attribute.
9974 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9977 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9978 expansion to perform a bitwise AND of the argument followed by a
9979 boolean negation of the result.
9981 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9983 PR rtl-optimization/84878
9984 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9985 the basic block. Assert the use reference is not artificial and that
9986 it has an associated insn.
9988 2018-04-04 Michael Matz <matz@suse.de>
9990 * builtins.c (compute_objsize): Pass correct operand
9991 to array_at_struct_end_p.
9993 2018-04-04 Richard Biener <rguenther@suse.de>
9996 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9997 from contexts for DINFO_LEVEL_TERSE and below.
9999 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10001 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10002 resiter_operand condition.
10003 * config/nds32/nds32.md (*move<mode>): Ditto.
10005 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10006 Monk Chiang <sh.chiang04@gmail.com>
10008 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10010 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10012 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10014 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10015 Kito Cheng <kito.cheng@gmail.com>
10017 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10018 nds32_cond_code_str, output_cond_branch,
10019 output_cond_branch_compare_zero, nds32_expand_cbranch,
10020 nds32_expand_cstore, nds32_expand_movcc,
10021 nds32_output_cbranchsi4_equality_zero,
10022 nds32_output_cbranchsi4_equality_reg,
10023 nds32_output_cbranchsi4_equality_reg_or_const_int,
10024 nds32_output_cbranchsi4_greater_less_zero: New functions.
10025 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10026 nds32_expand_cstore, nds32_expand_movcc,
10027 nds32_output_cbranchsi4_equality_zero,
10028 nds32_output_cbranchsi4_equality_reg,
10029 nds32_output_cbranchsi4_equality_reg_or_const_int,
10030 nds32_output_cbranchsi4_greater_less_zero): Declare.
10031 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10032 nds32_rimm11s_operand): New predicates.
10033 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10034 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10037 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10039 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10040 * config/nds32/nds32.md: Ditto.
10041 * config/nds32/pipelines.md: Ditto.
10043 2018-04-04 Richard Biener <rguenther@suse.de>
10045 PR tree-optimization/85168
10046 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10047 propagating abnormals.
10049 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10051 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10053 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10054 Kito Cheng <kito.cheng@gmail.com>
10056 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10057 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10058 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10059 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10060 * config/nds32/nds32.md (sibcall_internal): New.
10061 (sibcall_register): Remove.
10062 (sibcall_immediate): Remove.
10063 (sibcall_value_internal): New.
10064 (sibcall_value_register): Remove.
10065 (sibcall_value_immediate): Remove.
10066 * config/nds32/predicates.md (nds32_general_register_operand): New.
10067 (nds32_call_address_operand): New.
10069 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10071 PR rtl-optimization/85167
10072 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10073 bb_defs if *split_p, instead preinitialize it to NULL.
10075 PR tree-optimization/85156
10076 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10077 evaluating the argument multiple times.
10079 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10081 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10083 (_mm_cvtpd_ps): Likewise.
10084 (_mm_cvttpd_epi32): Likewise.
10085 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10086 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10087 vector, pixel, and bool following altivec.h include.
10089 2018-04-03 Martin Sebor <msebor@redhat.com>
10091 * doc/extend.texi (Common Function Attributes): Clarify.
10092 (const attribute): Likewise.
10093 (pure attribute): Likewise.
10095 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10098 * config/i386/i386.c (ix86_expand_vector_set): Use
10099 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10101 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10103 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10104 instructions when changing rounding bits to preserve precision bits
10105 in the x87 control word.
10107 2018-04-03 Martin Liska <mliska@suse.cz>
10109 PR tree-optimization/82491
10110 * rtl.h (strip_offset_and_add): Replace += suboffset with
10111 poly_uint64 () + suboffset.
10113 2018-03-29 Martin Liska <mliska@suse.cz>
10114 Martin Jambor <mjambor@suse.cz>
10117 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10118 param_type is not an integral or pointer type.
10120 2018-04-03 Richard Biener <rguenther@suse.de>
10122 * sese.h (recompute_all_dominators): Remove.
10124 2018-04-02 Martin Sebor <msebor@redhat.com>
10126 * doc/invoke.texi (-Wrestrict): Fix typos.
10128 2018-04-02 Jim Wilson <jimw@sifive.com>
10130 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10131 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10132 (<optab>di3, <optab>si3_extend): Likewise.
10133 (<optab>si3_mask, <optab>si3_mask_1): New.
10134 (<optab>di3_mask, <optab>di3_mask_1): New.
10135 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10136 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10137 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10139 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10141 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10144 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10146 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10147 (nds32_canonicalize_comparison): New function.
10149 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10150 Kito Cheng <kito.cheng@gmail.com>
10151 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10153 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10154 * config/nds32/constants.md (unspec_volatile_element): Add
10155 UNSPEC_VOLATILE_RELAX_GROUP.
10156 * config/nds32/nds32-relax-opt.c: New file.
10157 * config/nds32/nds32-predicates.c
10158 (nds32_symbol_load_store_p): New function.
10159 * config/nds32/nds32-protos.h
10160 (nds32_symbol_load_store_p): Declare function.
10161 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10162 * config/nds32/nds32.c
10163 (nds32_register_pass): New function to register pass.
10164 (nds32_register_passes): New function to register passes.
10165 * config/nds32/nds32.md (relax_group): New pattern.
10166 * config/nds32/nds32.opt (mrelax-hint): New option.
10167 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10169 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10171 * config/nds32/t-nds32: Modify files dependency.
10173 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10175 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10176 (PROFILE_HOOK): Define its implementation.
10178 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10180 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10181 type and 32-bit size.
10183 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10185 PR middle-end/85090
10186 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10187 (V_128_256): New mode iterator.
10188 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10189 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10190 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10192 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10193 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10195 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10198 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10199 NaN inputs correctly.
10201 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10204 * config/rs6000/vsx.md (??r): New mode attribute.
10205 (*vsx_mov<mode>_64bit): Use it.
10206 (*vsx_mov<mode>_32bit): Likewise.
10208 2018-03-30 Martin Sebor <msebor@redhat.com>
10210 PR tree-optimization/84818
10211 * builtins.c (check_access): Use warning_n.
10213 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10216 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10218 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10221 2018-03-30 Julia Koval <julia.koval@intel.com>
10224 * x86-tune.def (movx, partial_reg_dependency): Enable for
10227 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10229 PR inline-asm/84985
10230 * lra-constraints.c (process_alt_operands): Move setting
10231 this_alternative_matches below.
10233 2018-03-29 Martin Liska <mliska@suse.cz>
10236 * doc/invoke.texi: Document how LTO works with debug info.
10237 Describe auto-load support of binutils. Mention 'x86-64'
10238 as valid option value of -march option.
10240 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10242 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10245 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10246 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10247 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10250 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10253 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10254 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10255 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10256 for RS6000_BTM_POWERPC64.
10257 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10258 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10259 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10262 (DIVDEU): Likewise.
10264 2018-03-28 Carl Love <cel@us.ibm.com>
10267 2017-09-27 Carl Love <cel@us.ibm.com>
10269 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10270 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10271 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10274 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10276 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10277 instead of __vector bool.
10278 (_mm_max_pu8): Likewise.
10279 (_mm_min_pi16): Likewise.
10281 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10284 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10285 (DIVWEUO): Likewise.
10286 (DIVDEO): Likewise.
10287 (DIVDEUO): Likewise.
10288 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10289 DIVWEUO and DIVDEUO.
10290 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10291 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10292 (div_extend): Likewise.
10293 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10295 (__builtin_divweuo): Likewise.
10296 (__builtin_divdeo): Likewise.
10297 (__builtin_divdeuo): Likewise.
10299 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10302 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10303 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10305 PR tree-optimization/82004
10306 * gimple-match-head.c (optimize_pow_to_exp): New function.
10307 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10308 Don't fold to exp if optimize_pow_to_exp is false.
10310 2018-03-28 Martin Liska <mliska@suse.cz>
10313 * calls.c (initialize_argument_information): Fix trailing space.
10314 * common.opt: Fix typo and provide better explanation for
10315 -fsanitize-coverage option.
10316 * config/i386/i386.opt: Fix typo.
10318 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10319 Martin Liska <mliska@suse.cz>
10322 * gimplify.c (asan_poison_variable): Don't do the check for
10323 gimplify_omp_ctxp here.
10324 (gimplify_decl_expr): Do it here.
10325 (gimplify_target_expr): Likewise.
10327 2018-03-28 Martin Liska <mliska@suse.cz>
10330 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10331 chkp_type_bounds_count if MPX is not enabled.
10333 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10335 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10337 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10340 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10341 function to create the function decl for complex long double
10342 multiply and divide for -mabi=ieeelongdouble.
10343 (init_float128_ieee): Call it.
10345 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10348 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10349 -fcf-protection=branch -mibt.
10350 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10352 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10355 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10356 arm_word_relocations.
10358 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10361 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10362 extern array declarations.
10364 2018-03-27 Richard Biener <rguenther@suse.de>
10366 PR middle-end/84067
10367 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10368 explicit single_use checks.
10370 2018-03-27 Richard Biener <rguenther@suse.de>
10372 PR tree-optimization/85082
10373 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10376 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10378 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10379 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10380 Turn on fasynchronous-unwind-tables and funwind-tables.
10382 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10385 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10386 (*bmi_blsr_<mode>_ccz): Ditto.
10388 2018-03-26 Tom de Vries <tom@codesourcery.com>
10390 PR tree-optimization/85063
10391 * omp-general.c (offloading_function_p): New function. Factor out
10393 * omp-offload.c (pass_omp_target_link::gate): ... here.
10394 * omp-general.h (offloading_function_p): Declare.
10395 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10396 with attribute omp declare target for offloading functions.
10398 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10400 PR tree-optimization/84005
10401 * tree-data-ref.h (get_base_for_alignment): Declare.
10402 * tree-data-ref.c (get_base_for_alignment_1): New function.
10403 (get_base_for_alignment): Likewise.
10404 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10405 get_base_for_alignment to find a suitable base object, instead
10406 of always using drb->base_address.
10408 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10410 PR inline-asm/85022
10411 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10412 known size by default.
10414 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10416 PR inline-asm/85030
10417 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10418 and non BLKmode operands.
10420 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10423 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10424 Clean up attributes.
10426 2018-03-23 Richard Biener <rguenther@suse.de>
10429 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10430 we are going to emit early debug for LTO.
10432 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10434 PR inline-asm/85034
10435 * function.c (match_asm_constraints_1): Don't optimize if input
10436 doesn't satisfy general_operand predicate for output's mode.
10438 PR inline-asm/85022
10439 * alias.c (write_dependence_p): Don't require for x_canonicalized
10440 non-VOIDmode if x has VOIDmode.
10443 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10444 just don't try to optimize it rather than assert it never happens.
10446 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10448 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10449 macro expansions for definition of ST_INTERNAL_<mode> and
10450 LD_INTERNAL_<mode> builtins.
10451 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10453 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10455 (altivec_expand_st_builtin): Likewise.
10456 (altivec_expand_builtin): Remove calls to deleted functions.
10457 (rs6000_address_for_altivec): Delete this function.
10458 * config/rs6000/vector.md: Remove expands for
10459 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10461 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10464 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10465 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10466 re-computing once computed.
10467 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10468 (arm_init_machine_status): Initialize
10469 machine->static_chain_stack_bytes.
10471 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10474 * doc/extend.texi: Add four new prototypes for vec_ld.
10475 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10476 definitions for more logical presentation.
10477 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10478 entries for V1TI variants of __builtin_altivec_ld builtin.
10479 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10480 handling of V1TI variant of LVX icode pattern.
10481 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10482 (rs6000_gimple_fold_builtin): Likewise.
10483 (altivec_init_builtins): Add code to define
10484 __builtin_altivec_lvx_v1ti function.
10486 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10488 PR inline-asm/84941
10489 * function.c (match_asm_constraints_1): Don't do the optimization
10490 if input isn't a REG, SUBREG, MEM or constant.
10492 2018-03-22 Tom de Vries <tom@codesourcery.com>
10494 PR tree-optimization/84956
10495 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10496 bb_has_abnormal_pred.
10498 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10501 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10502 DECL_INITIAL (decl) to decl at the end.
10503 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10504 adjust the comment.
10506 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10508 * doc/extend.texi (__builtin_tgmath): Document when complex
10509 integer types are treated as _Complex _Float64.
10511 2018-03-21 Tom de Vries <tom@codesourcery.com>
10513 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10515 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10517 PR tree-optimization/84960
10518 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10519 if it is ENTRY block, move them into single succ of ENTRY in that case.
10521 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10523 PR tree-optimization/84811
10524 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10525 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10526 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10527 (known_subrange_p): Update accordingly. Cast each value involved
10528 in the size comparison, rather than casting the result of the
10531 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10533 PR tree-optimization/84982
10534 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10535 by flipping the least significant bit rather than all bits from
10536 bitpos to bitpos + bitsize - 1.
10538 2018-03-21 Nathan Sidwell <nathan@acm.org>
10540 * doc/extend.texi (Deprecated Features): Remove mention of
10541 long-deleted deprecations.
10543 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10546 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10547 * configure: Regenerate.
10549 2018-03-21 Tom de Vries <tom@codesourcery.com>
10551 PR tree-optimization/83126
10552 * tree-parloops.c (num_phis): New function.
10553 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10555 2018-03-21 Nathan Sidwell <nathan@acm.org>
10557 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10558 mention anon-struct/union members and trailing attributes.
10560 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10562 PR tree-optimization/84969
10563 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10564 builtin memset partitions if they set different rhs values.
10566 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10568 PR rtl-optimization/84989
10569 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10570 VEC_DUPLICATE with scalar result mode.
10572 2018-03-21 Martin Liska <mliska@suse.cz>
10575 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10576 not intended return statement.
10578 2018-03-21 Martin Liska <mliska@suse.cz>
10581 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10582 (chkp_find_bound_slots_1): Limit number of iterations.
10584 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10587 * Minor grammar fixes for x86 options.
10589 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10592 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10593 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10595 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10598 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10599 (altivec_lvx_<mode>_1op): Likewise.
10600 (altivec_stvx_<mode>_2op): Likewise.
10601 (altivec_stvx_<mode>_1op): Likewise.
10602 (altivec_lvx_<VM2:mode>): New define_expand.
10603 (altivec_stvx_<VM2:mode>): Likewise.
10604 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10605 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10606 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10607 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10608 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10609 (rs6000_gen_lvx): Likewise.
10610 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10611 (altivec_expand_stv_builtin): Likewise.
10612 (altivec_expand_builtin): Likewise.
10613 * config/rs6000/vector.md: Likewise.
10615 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10618 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10621 2018-03-20 Richard Biener <rguenther@suse.de>
10624 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10625 sign-conversions as zero, fall back to standard scalar_stmt
10628 2018-03-20 Martin Liska <mliska@suse.cz>
10631 * predict.c (rebuild_frequencies): Handle case when we have
10632 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10634 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10637 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10638 flag_section_anchors.
10639 * varasm.c (use_blocks_for_decl_p): Remove hack for
10640 dw2_force_const_mem.
10643 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10645 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10646 be created, use lowpart_subreg of operands[0] rather than operands[0]
10648 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10649 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10650 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10651 and n constraint instead of aarch64_shift_imm_di and Usd.
10652 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10653 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10655 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10658 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10659 to favor GPR over NEON registers.
10660 (<shift>di3_neon): Likewise.
10662 2018-03-20 Tom de Vries <tom@codesourcery.com>
10665 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10666 (nvptx_process_pars): Emit bar.sync asap and alap.
10668 2018-03-20 Tom de Vries <tom@codesourcery.com>
10671 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10672 seen_label if seen_label is already set.
10674 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10677 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10678 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10679 Use 1U instead of 1. Formatting fixes.
10682 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10683 instead of TREE_TYPE (s1) for the return value.
10685 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10687 PR tree-optimization/84946
10688 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10689 bitsize + bitsize in poly_uint64 rather than poly_int64.
10692 * dwarf2asm.c: Include fold-const.c.
10693 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10694 of decl rather than decl itself.
10696 PR rtl-optimization/84643
10697 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10699 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10702 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10703 calling assemble_variable.
10705 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10708 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10709 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10711 2018-03-19 Jim Wilson <jimw@sifive.com>
10714 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10715 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10716 (riscv_first_stack_step): Likewise.
10717 (riscv_option_override): Use STACK_BOUNDARY instead of
10718 MIN_STACK_BOUNDARY.
10719 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10720 MIN_STACK_BOUNDARY.
10721 (BIGGEST_ALIGNMENT): Set to 128.
10722 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10723 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10726 2018-03-19 Richard Biener <rguenther@suse.de>
10728 PR tree-optimization/84933
10729 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10730 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10732 2018-03-19 Richard Biener <rguenther@suse.de>
10734 PR tree-optimization/84859
10735 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10736 (cond_if_else_store_replacement): Perform sinking operation on
10737 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10738 Generalize what a BB with a single eligible store is.
10740 2018-03-19 Richard Biener <rguenther@suse.de>
10742 PR tree-optimization/84929
10743 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10744 chrec_is_positive against non-chrec arg.
10746 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10749 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10751 2018-03-18 Martin Liska <mliska@suse.cz>
10753 PR rtl-optimization/84635
10754 * regrename.c (build_def_use): Use matches_mode only when
10757 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10759 PR tree-optimization/84913
10760 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10761 vectorize chains of COND_EXPRs.
10763 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10765 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10767 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10769 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10771 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10773 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10775 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10776 Kito Cheng <kito.cheng@gmail.com>
10778 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10779 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10780 (nds32_adjust_reg_alloc_order): New function.
10781 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10783 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10785 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10786 nds32_print_operand, nds32_print_operand_address): Use
10787 HOST_WIDE_INT_PRINT_DEC instead.
10789 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10791 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10793 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10796 * config/i386/i386.c (initial_ix86_tune_features,
10797 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10798 unsigned long long.
10799 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10800 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10801 rather than 1u << ix86_tune. Formatting fix.
10802 (ix86_option_override_internal): Change ix86_arch_mask from
10803 unsigned int to unsigned HOST_WIDE_INT, initialize to
10804 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10805 (ix86_function_specific_restore): Likewise.
10807 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10810 * postreload.c (reload_combine_recognize_pattern): Perform
10811 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10812 truncate_int_for_mode the result for the destination's mode.
10815 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10816 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10818 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10821 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10824 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10825 regno_allocno_class_array and sorted_pseudos.
10826 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10827 insns where regno is used.
10829 2018-03-16 Martin Liska <mliska@suse.cz>
10832 * multiple_target.c (create_dispatcher_calls): Redirect
10833 reference in the symbol table.
10835 2018-03-16 Martin Liska <mliska@suse.cz>
10838 * multiple_target.c (create_dispatcher_calls): Redirect also
10841 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10845 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10846 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10847 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10849 2018-03-16 Julia Koval <julia.koval@intel.com>
10851 * doc/invoke.texi (Skylake Server): Add CLWB.
10852 Cannonlake): Remove CLWB.
10854 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10856 PR tree-optimization/84841
10857 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10859 (FLOAT_ONE_CONST_TYPE): Define.
10860 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10861 (sort_by_operand_rank): Put entries with higher constant_type last
10862 rather than first to match comments.
10864 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10866 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10869 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10872 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10873 check and use address of target always.
10875 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10878 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10879 (indirect_thunk_bnd_needed): Likewise.
10880 (indirect_thunks_used): Likewise.
10881 (indirect_thunks_bnd_used): Likewise.
10882 (indirect_return_needed): New.
10883 (indirect_return_bnd_needed): Likewise.
10884 (output_indirect_thunk_function): Add a bool argument for
10886 (output_indirect_thunk_function): Don't generate alias for
10887 function return thunk.
10888 (ix86_code_end): Call output_indirect_thunk_function to generate
10889 function return thunks.
10890 (ix86_output_function_return): Set indirect_return_bnd_needed
10891 and indirect_return_needed instead of indirect_thunk_bnd_needed
10892 and indirect_thunk_needed.
10894 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10896 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10897 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10898 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10900 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10901 Paul Hua <paul.hua.gm@gmail.com>
10904 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10906 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10908 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10909 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10910 resp. SFmode cases.
10912 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10915 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10916 instead of GET_MODE_SIZE when comparing Units.
10918 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10921 * varasm.c (hash_section): Return an unchangeble hash value
10922 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10923 Return !aarch64_can_use_per_function_literal_pools_p ().
10925 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10928 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10929 rather than address of cmode as last argument to prepare_cmp_insn.
10931 2018-03-15 Julia Koval <julia.koval@intel.com>
10933 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10934 F_AVX512VNNI, F_AVX512BITALG): New.
10936 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10939 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10940 insn for floating-point loads and stores.
10942 2018-03-14 Carl Love <cel@us.ibm.com>
10944 * config/rs6000/rs6000-c.c: Add macro definitions for
10945 ALTIVEC_BUILTIN_VEC_PERMXOR.
10946 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10947 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10948 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10949 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10951 * config/doc/extend.texi: Add prototypes for vec_permxor.
10953 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10956 * diagnostic-show-locus.c (class layout_point): Convert m_line
10957 from int to linenum_type.
10958 (line_span::comparator): Use linenum "compare" function when
10959 comparing line numbers.
10960 (test_line_span): New function.
10961 (layout_range::contains_point): Convert param "row" from int to
10963 (layout_range::intersects_line_p): Likewise.
10964 (layout::will_show_line_p): Likewise.
10965 (layout::print_source_line): Likewise.
10966 (layout::should_print_annotation_line_p): Likewise.
10967 (layout::print_annotation_line): Likewise.
10968 (layout::print_leading_fixits): Likewise.
10969 (layout::annotation_line_showed_range_p): Likewise.
10970 (struct line_corrections): Likewise for field m_row.
10971 (line_corrections::line_corrections): Likewise for param "row".
10972 (layout::print_trailing_fixits): Likewise.
10973 (layout::get_state_at_point): Likewise.
10974 (layout::get_x_bound_for_row): Likewise.
10975 (layout::print_line): Likewise.
10976 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10977 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10978 * input.c (selftest::test_linenum_comparisons): New function.
10979 (selftest::input_c_tests): Call it.
10980 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10981 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10982 * selftest.h (ASSERT_GT): New macro.
10983 (ASSERT_GT_AT): New macro.
10984 (ASSERT_LT): New macro.
10985 (ASSERT_LT_AT): New macro.
10987 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10989 PR rtl-optimization/84780
10990 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10992 2018-03-14 Martin Liska <mliska@suse.cz>
10994 * tree.c (record_node_allocation_statistics): Use
10995 get_stats_node_kind.
10996 (get_stats_node_kind): New function extracted from
10997 record_node_allocation_statistics.
10998 (free_node): Use get_stats_node_kind.
11000 2018-03-14 Richard Biener <rguenther@suse.de>
11002 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11003 that the value-set of ANTIC_IN doesn't grow.
11006 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11008 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11009 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11011 2018-03-14 Julia Koval <julia.koval@intel.com>
11013 * config.gcc (icelake-client, icelake-server): New.
11015 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11016 (initial_ix86_arch_features): Ditto.
11017 (PTA_SKYLAKE): Add SGX.
11018 (PTA_ICELAKE): Remove.
11019 (PTA_ICELAKE_CLIENT): New.
11020 (PTA_ICELAKE_SERVER): New.
11021 (ix86_option_override_internal): Split up icelake on icelake client and
11023 (get_builtin_code_for_version): Ditto.
11024 (fold_builtin_cpu): Ditto.
11025 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11026 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11027 * config/i386/i386.h (processor_type): Ditto.
11028 * doc/invoke.texi: Ditto.
11030 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11033 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11034 INTEGER_CST offset, add it together with bitpos / 8 and
11035 sign extend based on POINTER_SIZE.
11039 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11042 * config/i386/constraints.md (Yc): New register constraint.
11043 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11044 Use Yc constraint for alternative 2 of operand 0. Remove
11045 preferred_for_speed attribute.
11047 2018-03-14 Richard Biener <rguenther@suse.de>
11049 PR tree-optimization/84830
11050 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11051 with the old one to avoid oscillations.
11053 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11056 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11058 (assign_by_spills): Return a flag of reload assignment failure.
11059 Do not process the reload assignment failures. Do not spill other
11060 reload pseudos if they has the same reg class. Update n if
11062 (lra_assign): Add a return arg. Set up from the result of
11063 assign_by_spills call.
11064 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11065 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11066 usage_insns if it is not NULL.
11067 (spill_hard_reg_in_range): New function.
11068 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11069 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11070 function prototypes.
11071 (lra_assign): Change prototype.
11072 * lra.c (lra): Add code to deal with fails by splitting hard reg
11075 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11077 * config/riscv/riscv.opt (mrelax): New option.
11078 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11079 "norelax" when riscv_mrelax is disabled.
11080 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11082 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11085 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11086 reassociation for int modes.
11088 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11090 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11091 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11093 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11094 * config/aarch64/aarch64-sve.md
11095 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11096 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11097 (*extend<mode><Vwide>2): Rename to...
11098 (aarch64_sve_extend<mode><Vwide>2): ...this.
11099 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11100 renaming the old pattern to...
11101 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11103 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11104 define_expand, renaming the old pattern to...
11105 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11106 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11107 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11108 account when deciding which SVE instruction the optab should use.
11109 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11111 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11113 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11114 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11115 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11116 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11117 (tlsdesc_small_<mode>): Turn a define_expand and use
11118 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11119 (tlsdesc_small_advsimd_<mode>): ...this.
11120 (tlsdesc_small_sve_<mode>): New pattern.
11122 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11124 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11125 (UNSPEC_UMUL_HIGHPART): New constants.
11126 (MUL_HIGHPART): New int iteraor.
11127 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11128 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11130 (*<su>mul<mode>3_highpart): New define_insn.
11132 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11135 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11138 2018-03-13 Martin Liska <mliska@suse.cz>
11141 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11143 (sem_item_optimizer::~sem_item_optimizer): Release it.
11144 (sem_item_optimizer::merge_classes): Register variable aliases.
11145 (sem_item_optimizer::fixup_pt_set): New function.
11146 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11147 * ipa-icf.h: Declare new variables and functions.
11149 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11151 PR middle-end/84834
11152 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11153 integer_pow2p@2 and test integer_pow2p in condition.
11154 (A < 0 ? C : 0): Similarly for @1.
11156 PR middle-end/84831
11157 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11158 characters starting at p contain '\0' character, don't look beyond
11162 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11163 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11166 * reg-stack.c (change_stack): Change update_end var from int to
11167 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11168 also call set_block_for_insn on the newly added insns and rescan.
11171 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11172 on the last operand.
11175 * tree.c (stabilize_reference_1): Return save_expr (e) for
11176 STATEMENT_LIST even if it doesn't have side-effects.
11178 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11180 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11182 2018-03-12 Renlin Li <renlin.li@arm.com>
11184 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11185 aarch64_output_scalar_simd_mov_immediate.
11187 2018-03-12 Martin Sebor <msebor@redhat.com>
11189 PR tree-optimization/83456
11190 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11191 for perfectly overlapping calls to memcpy.
11192 (gimple_fold_builtin_memory_chk): Same.
11193 (gimple_fold_builtin_strcpy): Handle no-warning.
11194 (gimple_fold_builtin_stxcpy_chk): Same.
11195 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11197 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11199 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11200 parameter. Use it for SFmode.
11201 (rs6000_function_arg_advance_1): Adjust.
11202 (rs6000_function_arg): Adjust.
11203 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11205 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11207 PR rtl-optimization/84169
11208 PR rtl-optimization/84780
11209 * combine.c (can_combine_p): Check for a 2-insn combination whether
11210 the destination register is used between the two insns, too.
11212 2018-03-12 Richard Biener <rguenther@suse.de>
11214 PR tree-optimization/84803
11215 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11216 for refs DR analysis didn't process.
11218 2018-03-12 Richard Biener <rguenther@suse.de>
11220 PR tree-optimization/84777
11221 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11222 force-vectorize loops ignore whether we are optimizing for size.
11224 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11226 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11227 (TARGET_MD_ASM_ADJUST): Define.
11229 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11230 Kito Cheng <kito.cheng@gmail.com>
11231 Chung-Ju Wu <jasonwucj@gmail.com>
11233 * config/nds32/nds32.c (nds32_compute_stack_frame,
11234 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11235 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11236 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11237 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11238 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11239 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11240 * config/nds32/nds32.md (prologue, epilogue): Use macro
11241 NDS32_V3PUSH_AVAILABLE_P to do checking.
11243 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11246 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11247 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11248 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11249 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11250 addition of most attributes on !orig_type_die or the attribute not
11251 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11253 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11254 Chung-Ju Wu <jasonwucj@gmail.com>
11256 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11257 __NDS32_VH__ macro.
11258 * config/nds32/nds32.opt (mvh): New option.
11260 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11261 Chung-Ju Wu <jasonwucj@gmail.com>
11263 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11265 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11266 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11269 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11270 Chung-Ju Wu <jasonwucj@gmail.com>
11272 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11274 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11275 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11277 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11278 Kito Cheng <kito.cheng@gmail.com>
11279 Chung-Ju Wu <jasonwucj@gmail.com>
11281 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11282 UNSPEC_FFMISM and UNSPEC_FLMISM.
11283 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11284 for ffb, ffmism and flmism.
11285 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11286 (unspec_ffmism): Ditto.
11287 (unspec_flmism): Ditto.
11288 (nds32_expand_builtin_impl): Check if string extension is available.
11289 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11290 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11292 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11295 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11298 * lra-assigns.c (assign_by_spills): Return a flag of reload
11299 assignment failure. Do not process the reload assignment
11300 failures. Do not spill other reload pseudos if they has the same
11302 (lra_assign): Add a return arg. Set up from the result of
11303 assign_by_spills call.
11304 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11305 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11306 usage_insns if it is not NULL.
11307 (spill_hard_reg_in_range): New function.
11308 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11309 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11310 function prototypes.
11311 (lra_assign): Change prototype.
11312 * lra.c (lra): Add code to deal with fails by splitting hard reg
11315 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11318 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11320 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11323 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11324 (dw_val_node): Add val_symbolic_view.
11325 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11326 (symview_upper_bound): New.
11327 (new_line_info_table): Initialize symviews_since_reset.
11328 (dwarf2out_source_line): Count symviews_since_reset and set
11329 symview_upper_bound.
11330 (dw_val_equal_p): Handle symview.
11331 (add_AT_symview): New.
11332 (print_dw_val): Handle symview.
11333 (attr_checksum, attr_checksum_ordered): Likewise.
11334 (same_dw_val_p, size_of_die): Likewise.
11335 (value_format, output_die): Likewise.
11336 (add_high_low_attributes): Use add_AT_symview for entry_view.
11337 (dwarf2out_finish): Reset symview_upper_bound, clear
11340 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11343 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11344 Add strict argument and use it.
11345 (rs6000_split_multireg_move): Update for new strict argument.
11346 (mem_operand_gpr): Disallow all non-offsettable addresses.
11347 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11349 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11352 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11353 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11354 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11357 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11358 decl, use remap_type if we want to use the type.
11360 2018-03-09 Martin Sebor <msebor@redhat.com>
11362 PR tree-optimization/84526
11363 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11365 (builtin_access::generic_overlap): Be prepared to handle non-array
11368 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11370 PR rtl-optimization/84682
11371 * lra-constraints.c (process_address_1): Check is_address flag
11372 for address constraints.
11373 (process_alt_operands): Likewise.
11374 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11375 preprocess_constraints.
11376 * recog.h (preprocess_constraints): Add oploc parameter.
11378 * recog.c (preprocess_constraints): Test address_operand for
11379 CT_ADDRESS constraints.
11381 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11384 * lra-assigns.c (assign_by_spills): Return a flag of reload
11385 assignment failure. Do not process the reload assignment
11386 failures. Do not spill other reload pseudos if they has the same
11388 (lra_assign): Add a return arg. Set up from the result of
11389 assign_by_spills call.
11390 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11391 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11392 usage_insns if it is not NULL.
11393 (spill_hard_reg_in_range): New function.
11394 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11395 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11396 function prototypes.
11397 (lra_assign): Change prototype.
11398 * lra.c (lra): Add code to deal with fails by splitting hard reg
11401 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11404 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11405 Accept complain bool parameter. Only emit errors if it is true.
11406 (arm_parse_cpu_option_name): Likewise.
11407 (arm_target_thumb_only): Adjust callers of the above.
11408 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11409 prototype to take a default true bool parameter.
11410 (arm_parse_arch_option_name): Likewise.
11412 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11413 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11417 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11418 * configure: Regenerate.
11419 * configure.ac ("linker --version-script option"): New.
11420 ("linker soname option"): New.
11422 2018-03-09 Richard Biener <rguenther@suse.de>
11424 PR tree-optimization/84775
11425 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11426 immediate uses of predicate stmts and mark them modified.
11429 PR tree-optimization/84178
11430 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11432 (version_loop_for_if_conversion): Delay update_ssa call.
11433 (tree_if_conversion): Delay update_ssa until after predicate
11436 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11439 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11440 when the function accesses prior frames.
11442 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11445 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11446 gen_llsym, otherwise call maybe_gen_llsym.
11448 PR inline-asm/84742
11449 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11450 has ',' character inside of it.
11452 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11455 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11456 as clobbering CC_REGNUM.
11458 2018-03-08 Richard Biener <rguenther@suse.de>
11460 PR middle-end/84552
11461 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11462 (follow_copies_to_constant): Do not follow SSA names registered
11465 2018-03-08 Richard Biener <rguenther@suse.de>
11467 PR tree-optimization/84178
11468 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11470 (version_loop_for_if_conversion): Delay update_ssa call.
11471 (tree_if_conversion): Delay update_ssa until after predicate
11474 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11476 PR tree-optimization/84178
11477 * tree-if-conv.c (release_bb_predicate): Remove the
11478 the assertion that the stmts have NULL use_ops.
11479 Discard the statements, asserting that they haven't
11480 yet been added to a BB.
11482 2018-03-08 Richard Biener <rguenther@suse.de>
11484 PR tree-optimization/84746
11485 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11486 (phi_translate): Pass in destination ANTIC_OUT set.
11487 (phi_translate_1): Likewise. For a simplified result lookup
11488 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11489 (phi_translate_set): Adjust.
11490 (do_pre_regular_insertion): Likewise.
11491 (do_pre_partial_partial_insertion): Likewise.
11493 2018-03-08 Martin Liska <mliska@suse.cz>
11495 PR gcov-profile/84735
11496 * doc/gcov.texi: Document usage of profile files.
11497 * gcov-io.h: Document changes in the format.
11499 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11503 * dwarf2out.c (struct dw_line_info_table): Update comments for
11505 (FORCE_RESET_NEXT_VIEW): New.
11506 (FORCE_RESETTING_VIEW_P): New.
11507 (RESETTING_VIEW_P): Check for -1 too.
11508 (ZERO_VIEW_P): Likewise.
11509 (new_line_info_table): Force-reset next view.
11510 (dwarf2out_begin_function): Likewise.
11511 (dwarf2out_source_line): Simplify zero_view_p initialization.
11512 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11513 view directly. Omit view when omitting .loc at line 0.
11515 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11517 PR tree-optimization/84740
11518 * tree-switch-conversion.c (process_switch): Call build_constructors
11519 only if info.phi_count is non-zero.
11521 PR tree-optimization/84739
11522 * tree-tailcall.c (find_tail_calls): Check call arguments against
11523 DECL_ARGUMENTS (current_function_decl) rather than
11524 DECL_ARGUMENTS (func) when checking for tail recursion.
11526 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11528 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11529 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11530 Volker Reichelt's entry and add entries for people that perform
11531 GCC fuzzy testing and report numerous bugs.
11533 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11536 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11537 readonly data in sdata, if that is disabled.
11538 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11539 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11540 -mreadonly-in-sdata option.
11542 2018-03-07 Martin Sebor <msebor@redhat.com>
11544 PR tree-optimization/84468
11545 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11546 basic block when looking for nul assignment.
11548 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11551 * except.h (output_function_exception_table): Adjust prototype.
11552 * except.c (output_function_exception_table): Remove FNNAME parameter
11553 and add SECTION parameter. Ouput one part of the table at a time.
11554 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11555 the first part of the exception table and emit unwind directives.
11556 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11557 (i386_pe_seh_cold_init): Likewise.
11558 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11559 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11560 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11561 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11562 * config/i386/winnt.c: Include except.h.
11563 (struct seh_frame_state): Add reg_offset, after_prologue and
11564 in_cold_section fields.
11565 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11566 (i386_pe_seh_cold_init): New function.
11567 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11568 to seh->in_cold_section.
11569 (seh_emit_push): Record the offset of the push.
11570 (seh_emit_save): Record the offet of the save.
11571 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11572 Test seh->after_prologue to disregard the epilogue.
11573 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11574 (i386_pe_end_cold_function): New function.
11576 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11579 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11580 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11583 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11585 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11586 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11588 PR middle-end/84723
11589 * multiple_target.c: Include tree-inline.h and intl.h.
11590 (expand_target_clones): Diagnose and fail if node->definition and
11591 !tree_versionable_function_p (node->decl).
11593 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11595 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11597 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11598 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11599 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11601 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11604 * combine.c (try_combine): Use reg_or_subregno instead of handling
11605 just paradoxical SUBREGs and REGs.
11607 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11609 * config/arc/arc.c (arc_finalize_pic): Remove function.
11610 (arc_must_save_register): We use single base PIC register, remove
11611 checks to save/restore the PIC register.
11612 (arc_expand_prologue): Likewise.
11613 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11615 (arc_verify_short): Likewise.
11616 (arc_attr_type): Likewise.
11617 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11618 (walk_stores): Likewise.
11619 (arc_address_cost): Make it static.
11620 (arc_verify_short): Likewise.
11621 (branch_dest): Likewise.
11622 (arc_attr_type): Likewise.
11623 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11624 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11625 (arc_final_prescan_insn): Remove inserting the nops due to
11626 hardware hazards. It is done in reorg step.
11627 (insn_length_variant_t): Remove.
11628 (insn_length_parameters_t): Likewise.
11629 (arc_insn_length_parameters): Likewise.
11630 (arc_get_insn_variants): Likewise.
11631 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11633 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11635 PR inline-asm/84683
11636 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11639 PR tree-optimization/84687
11640 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11642 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11644 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11646 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11647 Rename to ppc_speculation_barrier.
11648 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11649 __builtin_ppc_speculation_barrier.
11651 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11654 * combine.c (combine_simplify_rtx): Don't try to simplify if
11655 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11658 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11660 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11661 to 32 bytes when compiling for POWER9.
11663 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11666 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11667 regparm >= 3 with no arg reg available also for calls with
11668 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11671 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11673 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11675 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11678 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11680 2018-03-05 Richard Biener <rguenther@suse.de>
11682 PR tree-optimization/84486
11683 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11684 When inserting a __builtin_assume_aligned call set the LHS
11685 SSA name alignment info accordingly.
11687 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11689 PR tree-optimization/84114
11690 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11691 Avoid reassociation of FLOAT_MODE addition.
11693 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11695 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11696 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11697 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11698 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11699 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11700 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11701 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11703 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11704 __builtin_ia32_wbinvd): New builtins.
11705 (SPECIAL_ARGS2): New.
11706 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11707 (SPECIAL_ARGS2): New.
11708 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11709 (ix86_valid_target_attribute_inner_p): Ditto.
11710 (ix86_init_mmx_sse_builtins): Add special_args2.
11711 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11712 TARGET_WBNOINVD_P): New.
11713 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11714 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11715 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11716 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11717 * config/i386/pconfigintrin.h: New file.
11718 * config/i386/wbnoinvdintrin.h: Ditto.
11719 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11721 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11723 2018-03-05 Richard Biener <rguenther@suse.de>
11725 PR tree-optimization/84670
11726 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11728 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11729 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11730 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11731 doesn't grow if all successors (recursively) were visited at least
11734 2018-03-05 Richard Biener <rguenther@suse.de>
11736 PR tree-optimization/84650
11737 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11738 if executed in the loop pipeline.
11740 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11742 * doc/configfiles.texi (Configuration Files): Move info about
11743 conditionalizing $target-protos.h to...
11744 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11745 differs from $target-protos.h.
11747 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11748 Chung-Ju Wu <jasonwucj@gmail.com>
11750 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11751 * config/nds32/nds32-multiple.md (setmemsi): Define.
11752 * config/nds32/nds32-memory-manipulation.c
11753 (nds32_gen_dup_4_byte_to_word_value): New.
11754 (emit_setmem_word_loop): New.
11755 (emit_setmem_byte_loop): New.
11756 (nds32_expand_setmem_loop): New.
11757 (nds32_expand_setmem_loop_v3m): New.
11758 (nds32_expand_setmem_unroll): New.
11759 (nds32_expand_setmem): New.
11761 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11762 Chung-Ju Wu <jasonwucj@gmail.com>
11764 * config/nds32/nds32-memory-manipulation.c
11765 (nds32_emit_load_store): New.
11766 (nds32_emit_post_inc_load_store): New.
11767 (nds32_emit_mem_move): New.
11768 (nds32_emit_mem_move_block): New.
11769 (nds32_expand_movmemsi_loop_unknown_size): New.
11770 (nds32_expand_movmemsi_loop_known_size): New.
11771 (nds32_expand_movmemsi_loop): New.
11772 (nds32_expand_movmemsi_unroll): New.
11773 (nds32_expand_movmemqi): Rename ...
11774 (nds32_expand_movmemsi): ... to this.
11775 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11776 (movmemsi): ... to this.
11777 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11778 (nds32_expand_movmemsi): ... to this.
11780 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11781 Monk Chiang <sh.chiang04@gmail.com>
11782 Chung-Ju Wu <jasonwucj@gmail.com>
11784 * config/nds32/nds32-protos.h
11785 (nds32_expand_load_multiple): New arguments.
11786 (nds32_expand_store_multiple): Ditto.
11787 (nds32_valid_multiple_load_store): Rename ...
11788 (nds32_valid_multiple_load_store_p): ... to this.
11789 * config/nds32/nds32-memory-manipulation.c
11790 (nds32_expand_load_multiple): Refine implementation.
11791 (nds32_expand_store_multiple): Ditto.
11792 * config/nds32/nds32-multiple.md
11793 (load_multiple): Update nds32_expand_load_multiple interface.
11794 (store_multiple): Update nds32_expand_store_multiple interface.
11795 * config/nds32/nds32-predicates.c
11796 (nds32_valid_multiple_load_store): Rename ...
11797 (nds32_valid_multiple_load_store_p): ... to this and refine
11799 * config/nds32/predicates.md
11800 (nds32_load_multiple_and_update_address_operation): New predicate.
11801 (nds32_store_multiple_and_update_address_operation): New predicate.
11803 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11804 Chung-Ju Wu <jasonwucj@gmail.com>
11806 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11807 (combo): New attribute.
11808 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11810 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11812 * config/nds32/nds32.opt: Change -mcmodel= default value.
11814 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11815 Monk Chiang <sh.chiang04@gmail.com>
11816 Chung-Ju Wu <jasonwucj@gmail.com>
11818 * config/nds32/constants.md (unspec_element): New enum.
11819 * config/nds32/constraints.md (Umw): New constraint.
11820 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11821 * config/nds32/nds32-intrinsic.md: Likewise.
11822 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11823 (nds32_valid_smw_lwm_base_p): New.
11824 (nds32_output_smw_single_word): New.
11825 (nds32_output_lmw_single_word): New.
11826 (nds32_expand_unaligned_load): New.
11827 (nds32_expand_unaligned_store): New.
11828 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11829 (nds32_output_smw_single_word): Declare.
11830 (nds32_output_lmw_single_word): Declare.
11831 (nds32_expand_unaligned_load): Declare.
11832 (nds32_expand_unaligned_store): Declare.
11833 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11834 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11835 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11836 NDS32_BUILTIN_UASTORE_DW.
11837 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11840 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11841 Kito Cheng <kito.cheng@gmail.com>
11842 Chung-Ju Wu <jasonwucj@gmail.com>
11844 * config/nds32/nds32-intrinsic.c
11845 (nds32_expand_builtin_null_ftype_reg): Delete.
11846 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11847 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11848 (nds32_read_argument): New.
11849 (nds32_legitimize_target): Ditto.
11850 (nds32_legitimize_argument): Ditto.
11851 (nds32_check_constant_argument): Ditto.
11852 (nds32_expand_unop_builtin): Ditto.
11853 (nds32_expand_unopimm_builtin): Ditto.
11854 (nds32_expand_binop_builtin): Ditto.
11855 (nds32_builtin_decl_impl): Ditto.
11856 (builtin_description): Ditto.
11857 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11858 (nds32_init_builtins_impl): Ditto.
11859 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11860 (nds32_builtin_decl): New.
11861 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11862 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11864 2018-03-02 Jeff Law <law@redhat.com>
11866 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11867 (redundant_insn, fill_simple_delay_slots): Likewise.
11868 (fill_slots_from_thread): Likewise.
11869 * resource.c (mark_referenced_resources): Likewise.
11870 (mark_set_resources, find_dead_or_set_registers): Likewise.
11872 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11874 * substring-locations.h (format_warning_va): Formatting fix for
11875 ATTRIBUTE_GCC_DIAG.
11876 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11878 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11879 * substring-locations.c: Include intl.h.
11880 (format_warning_va): Turned into small wrapper around
11881 format_warning_n_va, renamed to ...
11882 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11883 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11885 (format_warning_at_substring_n): New function.
11886 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11887 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11888 format_warning_at_substring with just a shorter name instead of
11889 const function pointer.
11890 (fmtwarn_n): New function.
11891 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11892 appropriate, get rid of all the fmtstr temporaries, move conditionals
11893 with G_() wrapped string literals directly into fmtwarn arguments,
11894 cast dir.len to (int), formatting fixes.
11896 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11898 * doc/invoke.texi: Remove "Cilk Plus" references.
11900 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11901 Richard Biener <rguenther@suse.de>
11904 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11905 for error or warning attributes if CALL_FROM_THUNK_P is set.
11908 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11911 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11912 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11915 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11916 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11918 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11919 instead of -1U in last predictors element's probability member.
11921 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11924 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11925 arguments if they are comparable.
11927 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11929 PR tree-optimization/84634
11930 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11931 masks and masked_loop_p with a single loop_masks, making sure it's
11932 null for bb vectorization.
11934 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11936 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11937 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11938 loop->force_vectorize to check whether there is no alias.
11940 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11943 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11945 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11947 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11948 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11949 instead of a loop around prev_real_insn.
11950 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11953 PR inline-asm/84625
11954 * config/i386/i386.c (ix86_print_operand): Use conditional
11955 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11958 2018-03-02 Richard Biener <rguenther@suse.de>
11960 PR tree-optimization/84427
11961 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11962 (bitmap_set_subtract_values): Rewrite to handle multiple
11965 (prune_clobbered_mems): Likewise.
11966 (phi_translate): Take edge instead of pred/phiblock.
11967 (phi_translate_1): Likewise.
11968 (phi_translate_set): Likewise. Insert all translated
11969 exprs for a value into the set, keeping possibly multiple
11970 expressions per value.
11971 (compute_antic_aux): Adjust for phi_translate changes.
11972 When intersecting union the expressions and prune those
11973 not in the final value set, keeping possibly multiple
11974 expressions per value. Do not use value-insertion
11975 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11976 all expressions. Add verification that the value-sets
11977 only shrink during iteration.
11978 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11979 (do_pre_regular_insertion): Likewise.
11980 (do_pre_partial_partial_insertion): Likewise.
11982 2018-03-02 Richard Biener <rguenther@suse.de>
11985 * config/darwin.c (saved_debug_info_level): New static global.
11986 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11987 (darwin_asm_lto_end): Restore debug info generation settings.
11989 2018-03-01 Martin Liska <mliska@suse.cz>
11992 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11993 volatile arguments.
11995 2018-03-01 Richard Biener <rguenther@suse.de>
11998 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11999 type annotation with LTO when debug was disabled at compile-time.
12001 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12003 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12005 (mips_final_postscan_insn): Likewise.
12007 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12009 PR rtl-optimization/84528
12010 * alias.c (init_alias_target): Add commentary.
12011 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12012 a unique base value if the frame pointer is not eliminated
12013 to the stack pointer.
12015 2018-03-01 Tom de Vries <tom@codesourcery.com>
12017 PR rtl-optimization/83327
12018 * lra-int.h (hard_regs_spilled_into): Declare.
12019 * lra.c (hard_regs_spilled_into): Define.
12020 (init_reg_info): Init hard_regs_spilled_into.
12021 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12022 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12023 (process_bb_lives): Handle hard_regs_spilled_into.
12024 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12025 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12027 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12029 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12030 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12031 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12032 * config/rs6000/aix72.h: New file.
12034 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12036 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12037 instead of warning_at with conditional singular and plural messages
12041 * stor-layout.c (update_alignment_for_field): For
12042 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12043 && !DECL_PACKED (field), do the alignment update, just use
12044 only desired_align instead of MAX (type_align, desired_align)
12046 (place_field): Don't do known_align < desired_align handling
12047 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12048 is non-NULL, instead do it after rli->prev_field handling and
12049 only if not within a bitfield word. For DECL_PACKED (field)
12050 use type_align of BITS_PER_UNIT.
12052 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12054 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12055 superfluous parentheses and trailing spaces.
12057 2018-02-28 Richard Biener <rguenther@suse.de>
12059 PR tree-optimization/84584
12060 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12061 SCoPs with fake exit edge.
12063 2018-02-28 Martin Liska <mliska@suse.cz>
12066 * timevar.c (timer::print): Fix format to properly print 100%
12069 2018-02-28 Richard Biener <rguenther@suse.de>
12071 PR middle-end/84607
12072 * genmatch.c (capture_info::walk_match): Do not mark
12073 captured expressions without operands as expr_p given
12074 they act more like predicates and should be subject to
12075 "lost tail" side-effect preserving.
12077 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12079 PR rtl-optimization/81611
12080 * auto-inc-dec.c (attempt_change): Move dead note from
12081 mem_insn if it's the next use of regno
12082 (find_address): Take address use of reg holding
12083 non-incremented value. Add parm to limit search to the named
12085 (merge_in_block): Attempt to use a mem insn that is the next
12086 use of the original regno.
12088 2018-02-27 Martin Sebor <msebor@redhat.com>
12091 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12092 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12094 2018-02-27 Martin Sebor <msebor@redhat.com>
12096 PR translation/84207
12097 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12098 n argument to unsigned HOST_WIDE_INT.
12099 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12100 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12101 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12102 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12104 2018-02-27 Richard Biener <rguenther@suse.de>
12106 PR tree-optimization/84512
12107 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12108 Do not use the estimate returned from record_stmt_cost for
12109 the scalar iteration cost but sum properly using add_stmt_cost.
12111 2018-02-27 Richard Biener <rguenther@suse.de>
12113 PR tree-optimization/84466
12114 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12115 Adjust last change to less strictly validate use operands.
12117 2018-02-27 Martin Liska <mliska@suse.cz>
12119 PR gcov-profile/84548
12120 * gcov.c (process_file): Allow partial overlap and consider it
12121 also as group functions.
12122 (output_lines): Properly calculate range of lines for a group.
12124 2018-02-27 Martin Liska <mliska@suse.cz>
12126 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12127 'ggc' suffixes. Change first column width.
12128 (timer::print): Fix formatting of the column.
12130 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12132 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12133 preserve inline entry blocks for the sake of debug inline
12134 entry point markers alone.
12135 (remove_unused_locals): Suggest in comments a better place to
12136 force the preservation of inline entry blocks that are
12137 otherwise unused, but do not preserve them.
12139 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12141 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12143 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12146 * config/i386/constraints.md (Bs): Replace
12147 ix86_indirect_branch_register with
12148 TARGET_INDIRECT_BRANCH_REGISTER.
12150 * config/i386/i386.md (indirect_jump): Likewise.
12151 (tablejump): Likewise.
12152 (*sibcall_memory): Likewise.
12153 (*sibcall_value_memory): Likewise.
12154 Peepholes of indirect call and jump via memory: Likewise.
12155 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12156 (*sibcall_value_GOT_32): Likewise.
12157 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12158 (GOT_memory_operand): Likewise.
12159 (call_insn_operand): Likewise.
12160 (sibcall_insn_operand): Likewise.
12161 (GOT32_symbol_operand): Likewise.
12162 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12164 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12166 PR rtl-optimization/83496
12167 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12168 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12169 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12170 redundant insn, if any.
12171 (relax_delay_slots): Likewise.
12172 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12174 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12176 PR tree-optimization/83965
12177 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12178 that grouped statements are part of a reduction chain. Return
12179 true if the statement is not marked as a reduction itself but
12180 is part of a group.
12181 (vect_recog_dot_prod_pattern): Don't check whether the statement
12182 is part of a group here.
12183 (vect_recog_sad_pattern): Likewise.
12184 (vect_recog_widen_sum_pattern): Likewise.
12186 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12189 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12191 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12194 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12196 (ix86_output_indirect_function_return): New prototype.
12197 (ix86_split_simple_return_pop_internal): Likewise.
12198 * config/i386/i386.c (indirect_return_via_cx): New.
12199 (indirect_return_via_cx_bnd): Likewise.
12200 (indirect_thunk_name): Handle return va CX_REG.
12201 (output_indirect_thunk_function): Create alias for
12202 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12203 (ix86_output_indirect_jmp): Remove the bool argument.
12204 (ix86_output_indirect_function_return): New function.
12205 (ix86_split_simple_return_pop_internal): Likewise.
12206 * config/i386/i386.md (*indirect_jump): Don't pass false
12207 to ix86_output_indirect_jmp.
12208 (*tablejump_1): Likewise.
12209 (simple_return_pop_internal): Change it to define_insn_and_split.
12210 Call ix86_split_simple_return_pop_internal to split it for
12211 -mfunction-return=.
12212 (simple_return_indirect_internal): Call
12213 ix86_output_indirect_function_return instead of
12214 ix86_output_indirect_jmp.
12216 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12219 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12220 memset and value initialization afterwards.
12222 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12224 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12226 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12229 * common/config/aarch64/aarch64-common.c
12230 (aarch_option_optimization_table[]): Switch
12231 off fomit-frame-pointer
12233 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12234 Chung-Ju Wu <jasonwucj@gmail.com>
12236 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12238 (store_multiple): Ditto.
12240 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12242 * config.gcc: Add --with-cpu support for nds32 target.
12243 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12244 * config/nds32/nds32.opt: Add -mcpu= option.
12246 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12248 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12249 isel=yes): Warn for these deprecated options.
12251 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12253 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12254 ISA_2_5_MASKS_EMBEDDED.
12256 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12258 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12259 p->max as pointers rather than using iterative_hash_expr.
12261 2018-02-23 Carl Love <cel@us.ibm.com>
12263 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12264 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12266 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12267 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12268 P8V_BUILTIN_VEC_VUNSIGNED2.
12270 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12273 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12274 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12276 * lra-constraints.c (curr_insn_transform): Set up
12277 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12279 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12280 * lra-eliminations.c (spill_pseudos): Ditto.
12281 (process_insn_for_elimination): Ditto.
12282 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12283 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12284 LRA_NON_CLOBBERED_ALT.
12286 2018-02-22 Martin Sebor <msebor@redhat.com>
12288 PR tree-optimization/84480
12289 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12290 to maybe_diag_stxncpy_trunc. Call it.
12291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12292 from gimple_fold_builtin_strcpy. Print inlining stack.
12293 (handle_builtin_stxncpy): Print inlining stack.
12294 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12296 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12299 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12300 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12301 and -fcheck-pointer-bounds are used together.
12302 (indirect_thunk_prefix): New enum.
12303 (indirect_thunk_need_prefix): New function.
12304 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12305 "_nt" instead of "_bnd" for NOTRACK prefix.
12306 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12307 (output_indirect_thunk_function): Likewise.
12309 (ix86_code_end): Update output_indirect_thunk_function calls.
12310 (ix86_output_indirect_branch_via_reg): Replace
12311 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12312 (ix86_output_indirect_branch_via_push): Likewise.
12313 (ix86_output_function_return): Likewise.
12314 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12315 incompatible with -fcf-protection=branch and
12316 -fcheck-pointer-bounds.
12318 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12321 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12322 Change gcc_assert call to output_operand_lossage.
12324 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12326 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12328 2018-02-22 DJ Delorie <dj@redhat.com>
12329 Sebastian Perta <sebastian.perta@renesas.com>
12330 Oleg Endo <olegendo@gcc.gnu.org>
12332 * config/rx/rx.c (rx_rtx_costs): New function.
12333 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12335 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12337 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12339 2018-02-22 Martin Liska <mliska@suse.cz>
12342 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12343 Add "native" as a possible value.
12345 2018-02-22 Martin Liska <mliska@suse.cz>
12348 * config/i386/i386.c (ix86_option_override_internal):
12349 Add "native" as a possible value for -march and -mtune.
12351 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12354 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12355 to all type variants.
12357 PR tree-optimization/84503
12358 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12359 width as info->bitpos + info->bitsize - start.
12360 (merged_store_group::merge_overlapping): Simplify width computation.
12361 (check_no_overlap): New function.
12362 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12363 start + width and last_order of the group, fail if check_no_overlap
12365 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12366 to group if check_no_overlap fails.
12368 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12370 * config/rs6000/altivec.md: Delete contraint arguments to
12371 define_expand, define_split, and define_peephole2, and in
12372 define_insn_and_split if always unused.
12373 * config/rs6000/darwin.md: Ditto.
12374 * config/rs6000/dfp.md: Ditto.
12375 * config/rs6000/rs6000.md: Ditto.
12376 * config/rs6000/sync.md: Ditto.
12377 * config/rs6000/vector.md: Ditto.
12378 * config/rs6000/vsx.md: Ditto.
12380 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12382 * config/rs6000/altivec.md: Write output control strings as braced
12383 blocks instead of double-quoted strings.
12384 * config/rs6000/darwin.md: Ditto.
12385 * config/rs6000/rs6000.md: Ditto.
12386 * config/rs6000/vector.md: Ditto.
12387 * config/rs6000/vsx.md: Ditto.
12389 2018-02-21 Jason Merrill <jason@redhat.com>
12391 PR c++/84314 - ICE with templates and fastcall attribute.
12392 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12394 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12396 * ipa-cp.c (determine_versionability): Fix comment typos.
12398 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12401 * ipa-cp.c (determine_versionability): Do not version functions caling
12404 2018-02-21 Martin Liska <mliska@suse.cz>
12407 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12408 Add "native" as a possible value.
12409 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12410 the macro when native cpu detection is available.
12412 2018-02-21 Martin Liska <mliska@suse.cz>
12415 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12416 Add "native" as a possible value.
12417 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12418 when native cpu detection is available.
12420 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12421 Martin Sebor <msebor@redhat.com>
12423 PR tree-optimization/84478
12424 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12426 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12427 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12428 support which is conservatively correct, for 2 only stay conservative
12429 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12430 argument to the 2 argument get_range_strlen, adjust 6 arg
12431 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12433 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12434 (gimple_fold_builtin_strlen): Pass true as last argument to
12437 2018-02-20 Martin Sebor <msebor@redhat.com>
12439 PR middle-end/84095
12440 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12441 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12442 (builtin_memref::builtin_memref): Factor out parts into
12443 set_base_and_offset and call it.
12445 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12447 PR middle-end/84406
12448 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12449 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12450 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12451 search at the associated MODE_INT.
12453 2018-02-20 Jeff Law <law@redhat.com>
12455 PR middle-end/82123
12456 PR tree-optimization/81592
12457 PR middle-end/79257
12458 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12459 for range data rather than using global data.
12460 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12461 range data rather than using global data.
12462 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12463 pass it to children as needed.
12464 (struct directive::fmtresult): Similarly.
12465 (struct directive::set_width): Similarly.
12466 (struct directive::set_precision): Similarly.
12467 (format_integer, format_directive, parse_directive): Similarly.
12468 (format_none): Accept unnamed vr_values parameter.
12469 (format_percent, format_floating, format_character): Similarly.
12470 (format_string, format_plain): Similarly.
12471 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12472 the EVRP range analyzer for range data rather than using global data.
12473 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12474 gimple-ssa-evrp-analyze.h
12475 (class sprintf_dom_walker): Add after_dom_children member function.
12476 Add evrp_range_analyzer member.
12477 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12478 range analyzer as needed.
12479 (sprintf_dom_walker::after_dom_children): New member function.
12480 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12482 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12483 (evrp_range_analyzer::pop_to_marker): Likewise.
12485 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12487 PR tree-optimization/84419
12488 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12489 with the required type if its current type is compatible but
12492 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12494 PR middle-end/82004
12495 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12496 after vectorization.
12498 2018-02-20 Martin Liska <mliska@suse.cz>
12501 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12502 possible values if we don't have a hint.
12504 2018-02-20 Martin Liska <mliska@suse.cz>
12508 * final.c (shorten_branches): Build align_tab array with one
12510 * opts.c (finish_options): Add alignment option limit check.
12511 (MAX_CODE_ALIGN): Likewise.
12512 (MAX_CODE_ALIGN_VALUE): Likewise.
12513 * doc/invoke.texi: Document maximum allowed option value for
12514 all -falign-* options.
12516 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12519 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12520 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12521 * var-tracking.c (emit_note_insn_var_location): Remove all references
12522 to NOTE_INSN_CALL_ARG_LOCATION.
12523 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12524 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12525 Use copy_rtx_if_shared.
12526 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12527 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12528 (dwarf2out_var_location): Remove handling of
12529 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12531 * final.c (final_scan_insn): Remove all references to
12532 NOTE_INSN_CALL_ARG_LOCATION.
12533 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12534 before dumping final insns.
12535 * except.c (emit_note_eh_region_end): Remove all references to
12536 NOTE_INSN_CALL_ARG_LOCATION.
12537 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12538 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12539 * config/arc/arc.c (hwloop_optimize): Likewise.
12540 * config/arm/arm.c (create_fix_barrier): Likewise.
12541 * config/s390/s390.c (s390_chunkify_start): Likewise.
12542 * config/sh/sh.c (find_barrier): Likewise.
12543 * config/i386/i386.c (rest_of_insert_endbranch,
12544 ix86_seh_fixup_eh_fallthru): Likewise.
12545 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12546 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12547 * config/frv/frv.c (frv_function_prologue): Likewise.
12548 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12550 (note_outside_basic_block_p): Remove all references to
12551 NOTE_INSN_CALL_ARG_LOCATION.
12552 * gengtype.c (adjust_field_rtx_def): Likewise.
12553 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12555 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12556 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12559 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12562 PR tree-optimization/84452
12563 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12564 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12567 2018-02-19 Martin Liska <mliska@suse.cz>
12570 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12572 2018-02-19 Martin Liska <mliska@suse.cz>
12573 Richard Sandiford <richard.sandiford@linaro.org>
12575 PR tree-optimization/82491
12576 * gimple-fold.c (get_base_constructor): Make earlier bail out
12579 2018-02-19 Carl Love <cel@us.ibm.com>
12581 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12582 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12584 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12585 P8V_BUILTIN_VEC_NEG.
12587 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12589 * config/rl78/rl78.md (movdf): New define expand.
12591 2018-02-19 Martin Liska <mliska@suse.cz>
12594 * doc/invoke.texi: Fix typo.
12595 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12597 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12599 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12600 handle rs6000_single_float and rs6000_double_float specially for
12603 2018-02-16 Jeff Law <law@redhat.com>
12605 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12606 (pop_regs): Use it.
12608 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12611 * ipa-inline.c (inline_small_functions): Fix a typo.
12613 2018-02-16 Nathan Sidwell <nathan@acm.org>
12615 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12617 2018-02-16 Carl Love <cel@us.ibm.com>
12619 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12620 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12621 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12622 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12623 expansion to P8V_BUILTIN_VEC_FLOAT2.
12625 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12627 PR rtl-optimization/70023
12628 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12629 src_regno into account.
12631 2018-02-16 Carl Love <cel@us.ibm.com>
12633 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12634 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12635 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12636 * config/rs6000/rs6000.c: Remove case statements for
12637 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12638 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12639 and P9V_BUILTIN_VEC_VINSERT4B.
12640 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12641 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12642 * config/rs6000/vsx.md:
12643 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12646 2018-02-16 Carl Love <cel@us.ibm.com>
12648 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12650 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12652 * config/rs6000/rs6000-c.c: Add the definitions for
12653 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12654 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12655 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12656 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12657 definition for insert4b and define insn *insert3b_internal.
12658 * doc/extend.texi: Add documentation for vec_extract4b.
12660 2018-02-16 Nathan Sidwell <nathan@acm.org>
12662 * doc/extend.texi (Backwards Compatibility): Mention friend
12663 injection. Note for-scope is deprecated.
12664 * doc/invoke.texi (-ffriend-injection): Deprecate.
12666 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12668 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12669 that moved to I2, also allow destinations that are a paradoxical
12670 subreg (instead of a normal reg).
12672 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12675 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12678 2018-02-16 Richard Biener <rguenther@suse.de>
12680 PR tree-optimization/84037
12681 PR tree-optimization/84016
12683 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12684 Adjust vec_construct for the fact we need additional higher latency
12685 128bit inserts for AVX256 and AVX512 vector builds.
12686 (ix86_add_stmt_cost): Scale vector construction cost for
12689 2018-02-16 Richard Biener <rguenther@suse.de>
12691 PR tree-optimization/84417
12692 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12693 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12694 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12696 2018-02-16 Martin Liska <mliska@suse.cz>
12699 * internal-fn.def (ASAN_CHECK): Set proper flags.
12700 (ASAN_MARK): Likewise.
12702 2018-02-16 Julia Koval <julia.koval@intel.com>
12704 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12705 from PTA_CANNONLAKE.
12707 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12710 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12711 Use ++iter rather than iter++ for std::list iterators.
12712 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12713 defer deleting them until all nodes in the forest are processed. Do
12714 free even leaf nodes. Change to_process into auto_vec.
12717 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12718 * vec.h (vec_default_construct): Use memset instead of placement new
12719 if BROKEN_VALUE_INITIALIZATION is defined.
12720 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12721 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12724 PR rtl-optimization/83723
12725 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12726 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12727 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12729 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12731 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12733 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12735 PR rtl-optimization/81443
12736 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12737 from inner REGs to paradoxical SUBREGs.
12739 2018-02-16 Richard Biener <rguenther@suse.de>
12741 PR tree-optimization/84399
12742 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12743 For operands we can analyze at their definition make sure we can
12744 analyze them at each use as well.
12746 2018-02-16 Richard Biener <rguenther@suse.de>
12748 PR tree-optimization/84190
12749 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12750 volatile accesses if the decl isn't volatile.
12752 2018-02-15 Jason Merrill <jason@redhat.com>
12754 PR c++/84314 - ICE with templates and fastcall attribute.
12755 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12756 TYPE_CANONICAL on an existing type.
12758 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12760 PR tree-optimization/84383
12761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12762 dstoff nor call operand_equal_p if dstbase is NULL.
12764 PR tree-optimization/84334
12765 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12766 also a CONSTANT_CLASS_P, punt.
12768 2018-02-14 Jim Wilson <jimw@sifive.com>
12770 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12771 first SMALL_OPERAND check. New local min_second_step. Move assert
12772 to where locals are set. Add TARGET_RVC support.
12773 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12775 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12777 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12779 2018-02-14 Martin Sebor <msebor@redhat.com>
12781 PR tree-optimization/83698
12782 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12783 arrays constrain the offset range to their bounds.
12784 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12785 (builtin_access::overlap): Avoid setting the size of overlap if it's
12787 (maybe_diag_overlap): Also consider arrays when deciding what values
12788 of offsets to include in diagnostics.
12790 2018-02-14 Martin Sebor <msebor@redhat.com>
12793 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12794 that correspond to the kind of a declaration.
12796 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12799 * config/pa/pa.md: Load address of PIC label using the linkage table
12800 if the label is nonlocal.
12802 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12804 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12805 warning message if user requests -maltivec=be.
12806 * doc/invoke.texi: Document deprecation of -maltivec=be.
12808 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12811 * config/rs6000/rs6000-c.c: Update definitions for
12812 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12813 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12815 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12818 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12819 add _get_ssp intrinsics. Remove argument from
12820 __builtin_ia32_rdssp[d|q].
12821 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12822 * config/i386/i386-builtin.def: Remove argument from
12823 __builtin_ia32_rdssp[d|q].
12824 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12825 ix86_expand_special_args_builtin for _rdssp[d|q].
12826 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12827 Clear register before usage.
12828 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12829 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12831 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12833 PR tree-optimization/84357
12834 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12835 operand 1 of an ARRAY_REF too.
12837 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12840 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12841 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12843 (set_of_reg): New struct.
12844 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12845 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12846 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12848 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12849 Split into bitclr, bitset, bitinvert patterns if appropriate.
12850 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12851 use rx_fuse_in_memory_bitop.
12852 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12853 to named insn, correct maximum insn length.
12855 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12858 * machmode.def: Define a complex mode for PARTIAL_INT.
12859 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12861 * doc/rtl.texi: Document CSPImode.
12862 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12864 (msp430_hard_regno_nregs_with_padding): Likewise.
12866 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12869 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12871 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12873 PR rtl-optimization/84169
12874 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12875 we generated a parallel as new i3 and we split that to new i2 and i3
12876 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12877 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12878 those to i2, not i1. Partially rewrite this scan code.
12880 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12883 * stor-layout.c (place_field): For variable length fields, adjust
12884 offset_align afterwards not just based on the field's alignment,
12885 but also on the size.
12887 PR middle-end/84309
12888 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12889 of exps and logs in the use_exp2 case.
12891 2018-02-13 Jeff Law <law@redhat.com>
12893 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12894 entry for "vector".
12896 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12899 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12903 * common.opt (gas-loc-support, gas-locview-support): New.
12904 (ginline-points, ginternal-reset-location-views): New.
12905 * doc/invoke.texi: Document them. Use @itemx where intended.
12906 (gvariable-location-views): Adjust.
12907 * target.def (reset_location_view): New.
12908 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12909 (TARGET_RESET_LOCATION_VIEW): New.
12910 * doc/tm.texi: Rebuilt.
12911 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12912 (dwarf2out_default_as_locview_support): New.
12913 (output_asm_line_debug_info): Use option variables.
12914 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12915 (output_loc_list): Likewise.
12916 (add_high_low_attributes): Check option variables.
12917 Don't output entry view attribute in strict mode.
12918 (gen_inlined_subroutine_die): Check option variables.
12919 (dwarf2out_inline_entry): Likewise.
12920 (init_sections_and_labels): Likewise.
12921 (dwarf2out_early_finish): Likewise.
12922 (maybe_reset_location_view): New, from...
12923 (dwarf2out_var_location): ... here. Call it.
12924 * debug.h (dwarf2out_default_as_loc_support): Declare.
12925 (dwarf2out_default_as_locview_support): Declare.
12926 * hooks.c (hook_int_rtx_insn_0): New.
12927 * hooks.h (hook_int_rtx_insn_0): Declare.
12928 * toplev.c (process_options): Take -gas-loc-support and
12929 -gas-locview-support from dwarf2out. Enable
12930 -gvariable-location-views by default only with locview
12931 assembler support. Enable -ginternal-reset-location-views by
12932 default only if the target defines the corresponding hook.
12933 Enable -ginline-points by default if location views are
12934 enabled; force it disabled if statement frontiers are
12936 * tree-inline.c (expand_call_inline): Check option variables.
12937 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12939 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12941 PR tree-optimization/84321
12942 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12943 handling. Also check whether the anti-range contains any values
12944 that satisfy the mask; switch to a VR_RANGE if not.
12946 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12949 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12951 2018-02-13 Martin Jambor <mjambor@suse.cz>
12954 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12955 of call statements, also set location of a load to a temporary.
12957 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12959 * config/rl78/rl78.c (add_vector_labels): New function.
12960 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12961 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12962 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12963 which checks that no arguments are passed.
12964 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12965 * doc/extend.texi: Documentation for the new attribute.
12967 2018-02-13 Andreas Schwab <schwab@suse.de>
12969 * config/riscv/linux.h (CPP_SPEC): Define.
12971 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12974 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12975 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12976 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12977 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12978 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12979 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12980 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12981 temporarily for AES and PCLMUL builtins.
12983 PR tree-optimization/84339
12984 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12985 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12988 PR middle-end/84309
12989 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12990 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12991 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12993 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12995 * omp-simd-clone.h: New file.
12996 * omp-simd-clone.c: Include omp-simd-clone.h.
12997 (expand_simd_clones): No longer static.
12998 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12999 cgraph.h and omp-simd-clone.h.
13000 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13001 (vect_recog_widen_shift_pattern): Formatting fix.
13002 (vect_pattern_recog_1): Don't check optab for calls.
13005 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13006 operands[2] into a REG before using gen_lowpart on it.
13008 2018-02-12 Jeff Law <law@redhat.com>
13011 * config/sh/sh.c (find_barrier): Consider a sibling call
13014 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13015 successfully back substituting a reg.
13017 2018-02-12 Richard Biener <rguenther@suse.de>
13019 PR tree-optimization/84037
13020 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13021 parameter, move visited init to caller.
13022 (vect_slp_analyze_operations): Separate cost from validity
13023 check, initialize visited once for all instances.
13024 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13026 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13028 (vect_model_promotion_demotion_cost): Likewise.
13029 (vectorizable_bswap): Guard cost modeling with !slp_node
13030 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13032 (vectorizable_call): Likewise.
13033 (vectorizable_conversion): Likewise.
13034 (vectorizable_assignment): Likewise.
13035 (vectorizable_shift): Likewise.
13036 (vectorizable_operation): Likewise.
13037 (vectorizable_store): Likewise.
13038 (vectorizable_load): Likewise.
13039 (vectorizable_condition): Likewise.
13040 (vectorizable_comparison): Likewise.
13042 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13045 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13046 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13047 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13049 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13051 PR middle-end/83665
13052 * params.def (inline-min-speedup): Increase from 8 to 15.
13053 (max-inline-insns-auto): Decrease from 40 to 30.
13054 * ipa-split.c (consider_split): Add some buffer for function to
13055 be considered inlining candidate.
13056 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13059 2018-02-12 Richard Biener <rguenther@suse.de>
13061 PR tree-optimization/84037
13062 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13063 matched stmts if we cannot swap the non-matched ones.
13065 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13067 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13068 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13069 _mm_maskz_scalef_round_ss): New intrinsics.
13070 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13071 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13072 __builtin_ia32_scalefss_round): Remove.
13073 (__builtin_ia32_scalefsd_mask_round,
13074 __builtin_ia32_scalefss_mask_round): New intrinsics.
13075 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13076 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13077 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13078 "<round_constraint>")): Changed to ...
13079 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13080 "<round_scalar_constraint>")): ... this.
13081 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13082 %0, %1, %2<round_op3>}"): Changed to ...
13083 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13084 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13085 %2<round_scalar_mask_op3>}"): ... this.
13086 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13088 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13090 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13091 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13092 (_mm_maskz_sqrt_round_ss): New intrinsics.
13093 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13094 (__builtin_ia32_sqrtsd_mask_round)
13095 (__builtin_ia32_sqrtss_mask_round): New builtins.
13096 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13097 (__builtin_ia32_sqrtss_round): Remove.
13098 (__builtin_ia32_sqrtsd_mask_round)
13099 (__builtin_ia32_sqrtss_mask_round): New builtins.
13100 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13101 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13102 ((match_operand:VF_128 1 "vector_operand"
13103 "xBm,<round_constraint>")): Changed to ...
13104 ((match_operand:VF_128 1 "vector_operand"
13105 "xBm,<round_scalar_constraint>")): ... this.
13106 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13107 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13108 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13109 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13110 %<iptr>1<round_scalar_mask_op3>}): ... this.
13111 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13112 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13114 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13117 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13118 Cast vec_cmpeq result to correct type.
13119 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13120 Cast vec_cmpgt result to correct type.
13122 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13124 * final.c (final_scan_insn_1): Renamed from...
13125 (final_scan_insn): ... this. New wrapper, to recover
13126 seen from the outermost call in recursive ones.
13127 * config/sparc/sparc.c (output_return): Drop seen from call.
13128 (output_sibcall): Likewise.
13129 * config/visium/visium.c (output_branch): Likewise.
13131 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13133 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13136 2018-02-10 Alan Modra <amodra@gmail.com>
13139 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13140 Specify LR as an input.
13142 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13145 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13146 remove_member_access_dummy_vars): New functions.
13147 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13148 lower_omp_1, execute_lower_omp): Use them.
13150 PR rtl-optimization/84308
13151 * shrink-wrap.c (spread_components): Release todo vector.
13153 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13155 PR rtl-optimization/57193
13156 * ira-color.c (struct allocno_color_data): Add member
13157 conflict_allocno_hard_prefs.
13158 (update_conflict_allocno_hard_prefs): New.
13159 (bucket_allocno_compare_func): Add a preference based on
13160 conflict_allocno_hard_prefs.
13161 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13162 (color_allocnos): Remove a dead code. Initiate
13163 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13165 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13168 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13169 constraint from =wa to wa. Avoid a subreg on the output operand,
13170 instead use a pseudo and subreg it in a move.
13171 (p9_xxbrd_<mode>): Changed to ...
13172 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13173 (p9_xxbrd_v2df): New expander.
13174 (p9_xxbrw_<mode>): Changed to ...
13175 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13176 (p9_xxbrw_v4sf): New expander.
13178 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13180 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13182 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13185 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13186 multiply in 32-bit mode.
13187 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13188 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13191 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13193 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13194 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13195 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13196 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13198 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13201 * dwarf2out.c (is_trivial_indirect_ref): New function.
13202 (dwarf2out_late_global_decl): Do not generate a location
13203 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13204 and that are not defined in the current unit.
13206 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13208 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13209 instead of a libcall for UNORDERED.
13211 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13214 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13215 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13217 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13220 * config/s390/s390.c (s390_set_current_function): Invoke
13221 s390_indirect_branch_settings also if fndecl didn't change.
13223 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13225 * config/rs6000/rs6000.md (blockage): Set length to zero.
13227 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13229 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13231 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13234 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13235 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13239 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13240 PARALLEL incoming that failed vt_get_decl_and_offset check.
13242 PR middle-end/84237
13243 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13244 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13246 (get_variable_section): For decls in named .bss* sections pass true as
13247 second argument to bss_initializer_p.
13249 2018-02-09 Marek Polacek <polacek@redhat.com>
13250 Jakub Jelinek <jakub@redhat.com>
13253 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13254 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13255 Sync some changes from cxx_fold_indirect_ref.
13257 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13259 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13261 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13262 (BLOCK_INLINE_ENTRY_LABEL): New.
13263 (dwarf2out_var_location): Disregard inline entry markers.
13264 (inline_entry_data): New struct.
13265 (inline_entry_data_hasher): New hashtable type.
13266 (inline_entry_data_hasher::hash): New.
13267 (inline_entry_data_hasher::equal): New.
13268 (inline_entry_data_table): New variable.
13269 (add_high_low_attributes): Add DW_AT_entry_pc and
13270 DW_AT_GNU_entry_view attributes if a pending entry is found
13271 in inline_entry_data_table. Add old entry_pc attribute only
13272 if debug nonbinding markers are disabled.
13273 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13274 markers are enabled.
13275 (block_within_block_p, dwarf2out_inline_entry): New.
13276 (dwarf2out_finish): Check that no entries remained in
13277 inline_entry_data_table.
13278 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13279 (final_scan_insn, notice_source_line): Likewise.
13280 (rest_of_clean_state): Skip inline entry markers.
13281 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13283 * gimple.c (gimple_build_debug_inline_entry): New.
13284 * gimple.h (enum gimple_debug_subcode): Add
13285 GIMPLE_DEBUG_INLINE_ENTRY.
13286 (gimple_build_debug_inline_entry): Declare.
13287 (gimple_debug_inline_entry_p): New.
13288 (gimple_debug_nonbind_marker_p): Adjust.
13289 * insn-notes.def (INLINE_ENTRY): New.
13290 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13291 inline entry marker notes.
13292 (print_insn): Likewise.
13293 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13294 (INSN_DEBUG_MARKER_KIND): Likewise.
13295 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13296 * tree-inline.c (expand_call_inline): Build and insert
13297 debug_inline_entry stmt.
13298 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13299 inline entry blocks early, if nonbind markers are enabled.
13300 (dump_scope_block): Dump fragment info.
13301 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13302 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13303 (gimple_build_debug_inline_entry): New.
13304 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13305 Enable/disable inline entry points too.
13306 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13307 (DEBUG_INSN): Describe inline entry markers.
13309 * common.opt (gvariable-location-views): New.
13310 (gvariable-location-views=incompat5): New.
13311 * config.in: Rebuilt.
13312 * configure: Rebuilt.
13313 * configure.ac: Test assembler for view support.
13314 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13315 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13316 * dwarf2out.c (var_loc_view): New typedef.
13317 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13318 (dwarf2out_locviews_in_attribute): New.
13319 (dwarf2out_locviews_in_loclist): New.
13320 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13321 (enum dw_line_info_opcode): Add LI_adv_address.
13322 (struct dw_line_info_table): Add view.
13323 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13324 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13325 (zero_view_p): New variable.
13326 (ZERO_VIEW_P): New macro.
13327 (output_asm_line_debug_info): New.
13328 (struct var_loc_node): Add view.
13329 (add_AT_view_list, AT_loc_list): New.
13330 (add_var_loc_to_decl): Add view param. Test it against last.
13331 (new_loc_list): Add view params. Record them.
13332 (AT_loc_list_ptr): Handle loc and view lists.
13333 (view_list_to_loc_list_val_node): New.
13334 (print_dw_val): Handle dw_val_class_view_list.
13335 (size_of_die): Likewise.
13336 (value_format): Likewise.
13337 (loc_list_has_views): New.
13338 (gen_llsym): Set vl_symbol too.
13339 (maybe_gen_llsym, skip_loc_list_entry): New.
13340 (dwarf2out_maybe_output_loclist_view_pair): New.
13341 (output_loc_list): Output view list or entries too.
13342 (output_view_list_offset): New.
13343 (output_die): Handle dw_val_class_view_list.
13344 (output_dwarf_version): New.
13345 (output_compilation_unit_header): Use it.
13346 (output_skeleton_debug_sections): Likewise.
13347 (output_rnglists, output_line_info): Likewise.
13348 (output_pubnames, output_aranges): Update version comments.
13349 (output_one_line_info_table): Output view numbers in asm comments.
13350 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13351 Call maybe_gen_llsym.
13352 (loc_list_from_tree_1): Adjust.
13353 (add_AT_location_description): Create view list attribute if
13354 needed, check it's absent otherwise.
13355 (convert_cfa_to_fb_loc_list): Adjust.
13356 (maybe_emit_file): Call output_asm_line_debug_info for test.
13357 (dwarf2out_var_location): Reset views as needed. Precompute
13358 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13359 attribute. Set view.
13360 (new_line_info_table): Reset next view.
13361 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13362 (dwarf2out_source_line): Likewise. Output view resets and labels to
13363 the assembler, or select appropriate line info opcodes.
13364 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13365 (optimize_string_length): Catch it. Adjust.
13366 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13367 dw_val_class_view_list, and remove it if no longer needed.
13368 (hash_loc_list): Hash view numbers.
13369 (loc_list_hasher::equal): Compare them.
13370 (optimize_location_lists): Check whether a view list symbol is
13371 needed, and whether the locview attribute is present, and
13372 whether they match. Remove the locview attribute if no longer
13374 (index_location_lists): Call skip_loc_list_entry for test.
13375 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13376 Use output_dwarf_version.
13377 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13378 (struct dw_val_node): Add val_view_list.
13379 * final.c (SEEN_NEXT_VIEW): New.
13380 (set_next_view_needed): New.
13381 (clear_next_view_needed): New.
13382 (maybe_output_next_view): New.
13383 (final_start_function): Rename to...
13384 (final_start_function_1): ... this. Take pointer to FIRST,
13385 add SEEN parameter. Emit param bindings in the initial view.
13386 (final_start_function): Reintroduce SEEN-less interface.
13387 (final): Rename to...
13388 (final_1): ... this. Take SEEN parameter. Output final pending
13389 next view at the end.
13390 (final): Reintroduce seen-less interface.
13391 (final_scan_insn): Output pending next view before switching
13392 sections or ending a block. Mark the next view as needed when
13393 outputting variable locations. Notify debug backend of section
13394 changes, and of location view changes.
13395 (rest_of_handle_final): Adjust.
13396 * toplev.c (process_options): Autodetect value for debug variable
13397 location views option. Warn on incompat5 without -gdwarf-5.
13398 * doc/invoke.texi (gvariable-location-views): New.
13399 (gvariable-location-views=incompat5): New.
13400 (gno-variable-location-views): New.
13402 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13404 PR tree-optimization/84136
13405 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13406 that the result of find_edge is non-NULL.
13408 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13411 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13412 storing integer register in SImode. Fix cost of 256 and 512
13413 byte aligned SSE register store.
13415 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13417 * config/i386/i386.c (ix86_multiplication_cost): Fix
13418 multiplication cost for TARGET_AVX512DQ.
13420 2018-02-08 Marek Polacek <polacek@redhat.com>
13422 PR tree-optimization/84238
13423 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13426 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13428 PR tree-optimization/84265
13429 * tree-vect-stmts.c (vectorizable_store): Don't treat
13430 VMAT_CONTIGUOUS accesses as grouped.
13431 (vectorizable_load): Likewise.
13433 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13435 PR tree-optimization/81635
13436 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13437 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13438 (test_round_for_mask): New functions.
13439 (wide_int_cc_tests): Call test_round_for_mask.
13440 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13441 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13442 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13443 range returned by get_range_info.
13445 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13448 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13449 * symtab.c: Include builtins.h
13450 (symtab_node::output_to_lto_symbol_table_p): Move here
13451 from lto-streamer-out.c:output_symbol_p.
13452 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13453 (output_symbol_p): Move all logic to symtab.c
13454 (produce_symtab): Update.
13456 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13458 * config/s390/s390-opts.h (enum indirect_branch): Define.
13459 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13460 (s390_indirect_branch_via_thunk)
13461 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13462 (enum s390_indirect_branch_type): Define.
13463 * config/s390/s390.c (struct s390_frame_layout, struct
13464 machine_function): Remove.
13465 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13466 (indirect_branch_table_label_no, indirect_branch_table_name):
13468 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13469 (enum s390_indirect_branch_option): Define.
13470 (s390_return_addr_from_memory): New function.
13471 (s390_handle_string_attribute): New function.
13472 (s390_attribute_table): Add new attribute handler.
13473 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13474 (s390_indirect_branch_via_thunk): New function.
13475 (s390_indirect_branch_via_inline_thunk): New function.
13476 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13477 sibling call optimization for non z10 compiles.
13478 (s390_emit_call): Force indirect branch target to be a single
13479 register. Add r1 clobber for non-z10 compiles.
13480 (s390_emit_epilogue): Emit return jump via return_use expander.
13481 (s390_reorg): Handle JUMP_INSNs as execute targets.
13482 (s390_option_override_internal): Perform validity checks for the
13483 new command line options.
13484 (s390_indirect_branch_attrvalue): New function.
13485 (s390_indirect_branch_settings): New function.
13486 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13487 (s390_output_indirect_thunk_function): New function.
13488 (s390_code_end): Implement target hook.
13489 (s390_case_values_threshold): Implement target hook.
13490 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13492 * config/s390/s390.h (struct s390_frame_layout)
13493 (struct machine_function): Move here from s390.c.
13494 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13495 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13496 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13497 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13498 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13499 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13500 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13501 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13502 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13503 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13504 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13505 (mnemonic attribute): Add values which aren't recognized
13507 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13508 pattern for branch conversion. Fix mnemonic attribute.
13509 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13510 indirect branch via thunk if requested.
13511 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13512 ("*indirect_jump"): Disable for branch conversion using out of
13514 ("indirect_jump_via_thunk<mode>_z10")
13515 ("indirect_jump_via_thunk<mode>")
13516 ("indirect_jump_via_inlinethunk<mode>_z10")
13517 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13518 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13519 ("casesi_jump_via_inlinethunk<mode>_z10")
13520 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13521 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13522 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13523 ("*indirect2_jump"): Disable for branch conversion.
13524 ("casesi_jump"): Turn into expander and expand patterns for branch
13526 ("return_use"): New expander.
13527 ("*return"): Emit return via thunk and rename it to ...
13528 ("*return<mode>"): ... this one.
13529 * config/s390/s390.opt: Add new options and and enum for the
13532 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13534 * lra-constraints.c (match_reload): Unconditionally use
13535 gen_lowpart_SUBREG, rather than selecting between that
13536 and equivalent gen_rtx_SUBREG code.
13538 2018-02-08 Richard Biener <rguenther@suse.de>
13540 PR tree-optimization/84233
13541 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13542 changed flag instead of boguously re-using phi_inserted.
13544 2018-02-08 Martin Jambor <mjambor@suse.cz>
13546 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13547 static local variables.
13549 2018-02-08 Richard Biener <rguenther@suse.de>
13551 PR tree-optimization/84278
13552 * tree-vect-stmts.c (vectorizable_store): When looking for
13553 smaller vector types to perform grouped strided loads/stores
13554 make sure the mode is supported by the target.
13555 (vectorizable_load): Likewise.
13557 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13559 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13560 Increase LDP/STP opportunities by adding adjacent callee-saves.
13562 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13564 PR rtl-optimization/84068
13565 PR rtl-optimization/83459
13566 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13568 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13570 PR tree-optimization/84224
13571 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13572 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13573 non-zero arguments.
13575 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13578 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13579 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13580 count, remove one USE.
13582 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13584 * doc/install.texi (Configuration): Document the
13585 --with-long-double-format={ibm,ieee} PowerPC configuration
13589 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13590 Convert from define_expand to be define_insn_and_split. Rework
13591 float/double/_Float128 conversions to QI/HI/SImode to work with
13592 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13593 conversions to QI/HImode types did a store and then a load to
13594 truncate the value. For conversions to VSX registers, don't split
13595 the insn, instead emit the code directly. Use the code iterator
13596 any_fix to combine signed and unsigned conversions.
13597 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13598 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13599 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13600 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13601 (fix_<mode>di2_hw): Likewise.
13602 (fixuns_<mode>di2_hw): Likewise.
13603 (fix_<mode>si2_hw): Likewise.
13604 (fixuns_<mode>si2_hw): Likewise.
13605 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13606 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13607 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13608 fix<uns>_trunc<SFDF:mode>si2_p8.
13609 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13610 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13611 (fix<uns>_<mode>_mem): Likewise.
13612 (fctiw<u>z_<mode>_mem): Likewise.
13613 (fix<uns>_<mode>_mem): Likewise.
13614 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13615 the register allocator from doing a direct move to the GPRs to do
13616 a store, and instead use the ISA 3.0 store byte/half-word from
13617 vector register instruction. For IEEE 128-bit floating point,
13618 also optimize stores of 32-bit ints.
13619 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13621 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13623 * genextract.c (push_pathstr_operand): New function to support
13625 (walk_rtx): Call push_pathstr_operand.
13626 (print_path): Support [a-zA-Z].
13628 2018-02-07 Richard Biener <rguenther@suse.de>
13630 PR tree-optimization/84037
13631 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13632 (cse_and_gimplify_to_preheader): Declare.
13633 (vect_get_place_in_interleaving_chain): Likewise.
13634 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13636 (_loop_vec_info::~_loop_vec_info): Delete it.
13637 (cse_and_gimplify_to_preheader): New function.
13638 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13639 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13640 (vectorizable_load): Likewise. For grouped stores always base
13641 the IV on the first element.
13642 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13643 condition before gimplifying.
13645 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13647 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13648 *DIV_EXPR and *MOD_EXPR.
13650 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13653 * config/i386/i386.c (ix86_option_override_internal): Mask out
13654 the CF_SET bit when checking -fcf-protection.
13656 2018-02-07 Tom de Vries <tom@codesourcery.com>
13659 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13662 2018-02-07 Richard Biener <rguenther@suse.de>
13664 PR tree-optimization/84204
13665 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13668 PR tree-optimization/84205
13669 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13670 special-case isl_ast_op_zdiv_r.
13672 PR tree-optimization/84223
13673 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13674 Only add conditions from within the region.
13675 (gather_bbs::after_dom_children): Adjust.
13677 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13680 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13681 * config/avr/avr.md: Only post-reload split REG-REG moves if
13682 either register is GENERAL_REG_P.
13684 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13686 PR tree-optimization/84235
13687 * tree-ssa-scopedtables.c
13688 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13689 if the subtraction is performed in floating point type where NaNs are
13690 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13691 build 1. Formatting fix.
13693 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13696 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13697 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13698 and skip it regardless of bb boundaries. Use CALL_P macro,
13699 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13700 unnecessarily, formatting fix.
13702 2018-02-06 Michael Collison <michael.collison@arm.com>
13704 * config/arm/thumb2.md:
13705 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13706 (*thumb_mov_notscc): Ditto.
13708 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13711 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13712 unsigned_fix, not "s".
13714 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13716 * configure.ac (gcc_fn_eh_frame_ro): New function.
13717 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13718 correct .eh_frame permissions.
13719 * configure: Regenerate.
13721 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13723 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13724 irrelevant options.
13726 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13728 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13729 Display warning message for -mno-speculate-indirect-jumps.
13731 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13733 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13735 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13737 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13739 PR tree-optimization/84225
13740 * tree-eh.c (find_trapping_overflow): Only call
13741 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13743 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13746 * config/i386/i386.c: Reimplement the check of possible options
13747 -mibt/-mshstk conbination. Change error messages.
13748 * doc/invoke.texi: Fix a typo: remove extra '='.
13750 2018-02-06 Marek Polacek <polacek@redhat.com>
13752 PR tree-optimization/84228
13753 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13755 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13758 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13759 emitted arch directives.
13760 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13761 __ARM_FEATURE_COPROC before changing architectures.
13763 2018-02-06 Richard Biener <rguenther@suse.de>
13765 * config/i386/i386.c (print_reg): Fix typo.
13766 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13768 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13770 * configure: Regenerate.
13772 2018-02-05 Martin Sebor <msebor@redhat.com>
13774 PR tree-optimization/83369
13775 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13778 2018-02-05 Martin Liska <mliska@suse.cz>
13780 * doc/invoke.texi: Cherry-pick upstream r323995.
13782 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13784 * ira.c (ira_init_register_move_cost): Adjust comment.
13786 2018-02-05 Martin Liska <mliska@suse.cz>
13788 PR gcov-profile/84137
13789 * doc/gcov.texi: Fix typo in documentation.
13791 2018-02-05 Martin Liska <mliska@suse.cz>
13793 PR gcov-profile/83879
13794 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13795 using dlopen functionality.
13797 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13799 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13800 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13801 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13802 _mm_maskz_range_round_ss): New intrinsics.
13803 (__builtin_ia32_rangesd128_round)
13804 (__builtin_ia32_rangess128_round): Remove.
13805 (__builtin_ia32_rangesd128_mask_round,
13806 __builtin_ia32_rangess128_mask_round): New builtins.
13807 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13808 __builtin_ia32_rangess128_round): Remove.
13809 (__builtin_ia32_rangesd128_mask_round,
13810 __builtin_ia32_rangess128_mask_round): New builtins.
13811 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13812 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13813 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13814 "<round_saeonly_constraint>")): Changed to ...
13815 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13816 "<round_saeonly_scalar_constraint>")): ... this.
13817 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13818 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13819 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13820 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13821 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13823 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13825 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13827 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13828 Remove all values except native, 8540 and 8548.
13830 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13832 * config/i386/i386.c (ix86_output_function_return): Pass
13833 INVALID_REGNUM, instead of -1, as invalid register number to
13834 indirect_thunk_name and output_indirect_thunk.
13836 2018-02-02 Julia Koval <julia.koval@intel.com>
13838 * config.gcc: Add -march=icelake.
13839 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13840 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13841 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13842 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13843 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13844 (processor_target_table): Add icelake.
13845 (ix86_option_override_internal): Handle new PTAs.
13846 (get_builtin_code_for_version): Handle icelake.
13847 (M_INTEL_COREI7_ICELAKE): New.
13848 (fold_builtin_cpu): Handle icelake.
13849 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13850 * doc/invoke.texi: Add -march=icelake.
13852 2018-02-02 Julia Koval <julia.koval@intel.com>
13854 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13855 to wide_int_bitmask.
13856 * wide-int-bitmask.h: New.
13858 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13861 * config/i386/i386.md: Replace Pmode with word_mode in
13862 builtin_setjmp_setup and builtin_longjmp to support x32.
13864 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13868 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13870 (rs6000_supported_cpu_names): New static variable.
13871 (linux_cpu_translation_table): Likewise.
13872 (elf_platform) <cpu>: Define new static variable and use it.
13873 Translate kernel AT_PLATFORM name to canonical name if needed.
13874 Error if platform name is unknown.
13876 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13879 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13881 2018-02-01 Jeff Law <law@redhat.com>
13884 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13885 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13886 the scratch if RELEASE_VIA_POP is false.
13887 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13888 If we have to save a temporary register, decrement SIZE appropriately.
13889 Pass new arguments to release_scratch_register_on_entry.
13890 (ix86_adjust_stack_and_probe): Likewise.
13891 (ix86_emit_probe_stack_range): Pass new arguments to
13892 release_scratch_register_on_entry.
13894 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13896 PR rtl-optimization/84157
13897 * combine.c (change_zero_ext): Use REG_P predicate in
13898 front of HARD_REGISTER_P predicate.
13900 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13902 * config/avr/avr.c (avr_option_override): Move disabling of
13903 -fdelete-null-pointer-checks to...
13904 * common/config/avr/avr-common.c (avr_option_optimization_table):
13907 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13909 PR tree-optimization/81635
13910 * tree-data-ref.c (split_constant_offset_1): For types that
13911 wrap on overflow, try to use range info to prove that wrapping
13914 2018-02-01 Renlin Li <renlin.li@arm.com>
13917 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13918 TAILCALL_ADDR_REGS.
13919 (aarch64_register_move_cost): Likewise.
13920 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13921 TAILCALL_ADDR_REGS.
13922 (REG_CLASS_NAMES): Likewise.
13923 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13924 TAILCALL_ADDR_REGS. Remove IP registers.
13925 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13927 2018-02-01 Richard Biener <rguenther@suse.de>
13929 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13930 for specifying RPO order and allow NULL for that.
13931 * domwalk.c (dom_walker::dom_walker): Likewise.
13932 (dom_walker::walk): Handle NULL RPO order.
13933 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13935 (rewrite_update_dom_walker): Likewise.
13936 (mark_def_dom_walker): Likewise.
13938 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13940 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13941 (aarch64_maybe_expand_sve_subreg_move): Declare.
13942 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13943 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13945 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13946 that are semantically a reverse operation.
13947 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13948 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13949 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13951 (aarch64_can_change_mode_class): For big-endian, forbid changes
13952 between two SVE modes if they have different element sizes.
13954 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13956 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13957 the TImode handling for big-endian targets.
13959 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13961 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13962 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13964 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13965 Remove BSWAP handing for big-endian targets and use the form of
13966 LD1RQ appropariate for the mode.
13968 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13970 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13971 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13972 duplicated element.
13974 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13977 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13978 check for operands that need to go through aarch64_sve_reload_be.
13980 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13982 PR tree-optimization/81661
13983 PR tree-optimization/84117
13984 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13985 * tree-eh.c: Include gimplify.h.
13986 (find_trapping_overflow, replace_trapping_overflow,
13987 rewrite_to_non_trapping_overflow): New functions.
13988 * tree-vect-loop.c: Include tree-eh.h.
13989 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13990 * tree-data-ref.c: Include tree-eh.h.
13991 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13993 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13995 PR rtl-optimization/84123
13996 * combine.c (change_zero_ext): Check if hard register satisfies
13997 can_change_dest_mode before calling gen_lowpart_SUBREG.
13999 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14002 * ira.c (ira_init_register_move_cost): Remove assert.
14004 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14006 PR rtl-optimization/84071
14007 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14008 * doc/tm.texi: Regenerate.
14010 2018-01-31 Richard Biener <rguenther@suse.de>
14012 PR tree-optimization/84132
14013 * tree-data-ref.c (analyze_miv_subscript): Properly
14014 check whether evolution_function_is_affine_multivariate_p
14015 before calling gcd_of_steps_may_divide_p.
14017 2018-01-31 Julia Koval <julia.koval@intel.com>
14020 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14021 * config/i386/i386.md (rdpid_rex64) New.
14022 (rdpid): Make 32bit only.
14024 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14027 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14028 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14030 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14033 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14035 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14037 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14039 PR rtl-optimization/84071
14040 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14041 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14043 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14045 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14046 (arc_attribute_table): Add 'aux' attribute.
14047 (arc_in_small_data_p): Consider aux like variables.
14048 (arc_is_aux_reg_p): New function.
14049 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14050 (arc_get_aux_arg): New function.
14051 (prepare_move_operands): Handle aux-register access.
14052 (arc_handle_aux_attribute): New function.
14053 * doc/extend.texi (ARC Variable attributes): Add subsection.
14055 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14057 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14058 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14059 (arc_attribute_table): Add 'uncached' attribute.
14060 (arc_print_operand): Print '.di' flag for uncached memory
14062 (arc_in_small_data_p): Do not consider for small data the uncached
14064 (arc_is_uncached_mem_p): New function.
14065 * config/arc/predicates.md (compact_store_memory_operand): Check
14066 for uncached memory accesses.
14067 (nonvol_nonimm_operand): Likewise.
14068 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14070 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14073 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14074 falign-loops=): Add Optimization flag.
14076 2018-01-30 Jeff Law <law@redhat.com>
14079 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14080 INT_REGISTERS_SAVED. Check it prior to calling
14081 get_scratch_register_on_entry.
14082 (ix86_adjust_stack_and_probe): Similarly.
14083 (ix86_emit_probe_stack_range): Similarly.
14084 (ix86_expand_prologue): Corresponding changes.
14086 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14089 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14090 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14092 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14095 * lra-constraints.c (curr_insn_transform): Process AND in the
14098 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14100 PR rtl-optimization/83986
14101 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14102 dependence against last_pending_memory_flush in addition to
14103 pending_jump_insns.
14105 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14107 PR tree-optimization/81611
14108 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14111 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14114 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14117 2018-01-30 Richard Biener <rguenther@suse.de>
14118 Jakub Jelinek <jakub@redhat.com>
14120 PR tree-optimization/84111
14121 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14122 inner loops added during recursion, as they don't have up-to-date
14125 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14128 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14129 (can_inline_edge_by_limits_p): ... here.
14130 (can_early_inline_edge_p, check_callers,
14131 update_caller_keys, update_callee_keys, recursive_inlining,
14132 add_new_edges_to_heap, speculation_useful_p,
14133 inline_small_functions,
14134 inline_small_functions, flatten_function,
14135 inline_to_all_callers_1): Update.
14137 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14139 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14142 2018-01-30 Richard Biener <rguenther@suse.de>
14144 PR tree-optimization/83008
14145 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14146 invariant and constant vector uses in stmts when they need
14147 more than one stmt.
14149 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14152 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14153 * configure: Regenerate.
14155 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14157 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14159 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14160 Use gen_rtx_REG rather than gen_lowpart.
14162 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14164 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14165 rather than 0 when creating partial subregs.
14167 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14169 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14172 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14175 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14176 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14177 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14178 flags. This restores the settings used before the 2017-07-24.
14179 Turning off pre increment/decrement/modify allows IVOPTS to
14180 optimize DF/SF loops where the index is an int.
14182 2018-01-29 Richard Biener <rguenther@suse.de>
14183 Kelvin Nilsen <kelvin@gcc.gnu.org>
14186 * tree-vect-stmts.c (vectorizable_call): Don't call
14187 targetm.vectorize_builtin_md_vectorized_function if callee is
14190 2018-01-22 Carl Love <cel@us.ibm.com>
14192 * doc/extend.tex: Fix typo in second arg in
14193 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14195 2018-01-29 Richard Biener <rguenther@suse.de>
14197 PR tree-optimization/84086
14198 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14199 (flush_ssaname_freelist): When SSA names were released reset
14200 the SCEV hash table.
14202 2018-01-29 Richard Biener <rguenther@suse.de>
14204 PR tree-optimization/84057
14205 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14206 removed paths when removing edges.
14208 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14210 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14211 -mfunction-return=@var{choice}.
14213 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14215 PR diagnostic/84034
14216 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14217 Handle CR like TAB.
14218 (layout::print_source_line): Likewise.
14219 (test_get_line_width_without_trailing_whitespace): Add test cases.
14221 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14223 PR middle-end/84040
14224 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14227 2018-01-26 Jim Wilson <jimw@sifive.com>
14229 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14231 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14234 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14236 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14237 and CMP + SUB-immediate -> SUBS.
14239 2018-01-26 Martin Sebor <msebor@redhat.com>
14241 PR tree-optimization/83896
14242 * tree-ssa-strlen.c (get_string_len): Rename...
14243 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14244 Avoid assuming length is constant.
14245 (handle_char_store): Use HOST_WIDE_INT for string length.
14247 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14250 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14251 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14253 2018-01-26 Richard Biener <rguenther@suse.de>
14255 PR rtl-optimization/84003
14256 * dse.c (record_store): Only record redundant stores when
14257 the earlier store aliases at least all accesses the later one does.
14259 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14261 PR rtl-optimization/83985
14262 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14263 REG_CFA_RESTORE insns.
14264 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14265 REG_CFA_RESTORE insns here.
14268 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14269 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14271 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14273 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14274 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14275 (arc_init): Likewise.
14276 (arc_override_options): Likewise.
14277 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14279 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14281 * config/arc/arc.h (TARGET_DBNZ): Define.
14282 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14283 properly set the tune attribute.
14284 (dbnz): Use TARGET_DBNZ guard.
14285 * config/arc/arc.opt (mtune): Add core3 option.
14287 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14289 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14290 recognize new pic like addresses.
14291 (arc_delegitimize_address): Clean up.
14293 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14295 * config/arc/arc-arches.def: Option mrf16 valid for all
14297 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14298 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14299 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14300 * config/arc/arc-tables.opt: Regenerate.
14301 * config/arc/arc.c (arc_conditional_register_usage): Handle
14302 reduced register file case.
14303 (arc_file_start): Set must have build attributes.
14304 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14305 mrf16 option value.
14306 * config/arc/arc.opt (mrf16): Add new option.
14307 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14308 * config/arc/genmultilib.awk: Handle new mrf16 option.
14309 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14310 * config/arc/t-multilib: Regenerate.
14311 * doc/invoke.texi (ARC Options): Document mrf16 option.
14313 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14315 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14316 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14317 (arc_attribute_table): Add 'secure_call' attribute.
14318 (arc_print_operand): Print secure call operand.
14319 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14321 (arc_is_secure_call_p): New function. * config/arc/arc.md
14322 (call_i): Add support for sjli instruction.
14323 (call_value_i): Likewise.
14324 * config/arc/constraints.md (Csc): New constraint.
14326 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14327 John Eric Martin <John.Martin@emmicro-us.com>
14329 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14330 * config/arc/arc.c (_arc_jli_section): New struct.
14331 (arc_jli_section): New type.
14332 (rc_jli_sections): New static variable.
14333 (arc_handle_jli_attribute): New function.
14334 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14335 (arc_file_end): New function.
14336 (TARGET_ASM_FILE_END): Define.
14337 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14338 (arc_add_jli_section): New function.
14339 (jli_call_scan): Likewise.
14340 (arc_reorg): Call jli_call_scan.
14341 (arc_output_addsi): Remove 'S' from printing asm operand.
14342 (arc_is_jli_call_p): New function.
14343 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14345 (movhi_insn): Likewise.
14346 (movsi_insn): Likewise.
14347 (movsi_set_cc_insn): Likewise.
14348 (loadqi_update): Likewise.
14349 (load_zeroextendqisi_update): Likewise.
14350 (load_signextendqisi_update): Likewise.
14351 (loadhi_update): Likewise.
14352 (load_zeroextendhisi_update): Likewise.
14353 (load_signextendhisi_update): Likewise.
14354 (loadsi_update): Likewise.
14355 (loadsf_update): Likewise.
14356 (movsicc_insn): Likewise.
14357 (bset_insn): Likewise.
14358 (bxor_insn): Likewise.
14359 (bclr_insn): Likewise.
14360 (bmsk_insn): Likewise.
14361 (bicsi3_insn): Likewise.
14362 (cmpsi_cc_c_insn): Likewise.
14363 (movsi_ne): Likewise.
14364 (movsi_cond_exec): Likewise.
14365 (clrsbsi2): Likewise.
14366 (norm_f): Likewise.
14376 (call_i): Remove 'S' asm letter, add jli instruction.
14377 (call_value_i): Likewise.
14378 * config/arc/arc.op (mjli-always): New option.
14379 * config/arc/constraints.md (Cji): New constraint.
14380 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14382 (subsf3_fpx): Likewise.
14383 (mulsf3_fpx): Likewise.
14384 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14386 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14387 function attrbutes.
14388 * doc/invoke.texi (ARC): Document mjli-always option.
14390 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14392 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14393 avoid addition with 0 and use incw and decw where possible.
14395 2018-01-26 Richard Biener <rguenther@suse.de>
14397 PR tree-optimization/81082
14398 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14399 association if it requires casting to unsigned.
14400 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14401 from fold_plusminus_mult_expr to catch important cases late when
14402 range info is available.
14404 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14406 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14407 * configure.ac (hidden_linkonce): New test.
14408 * configure: Regenerate.
14409 * config.in: Regenerate.
14411 2018-01-26 Julia Koval <julia.koval@intel.com>
14413 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14414 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14415 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14416 _mm_mask_bitshuffle_epi64_mask): Fix type.
14417 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14418 USI_FTYPE_V4DI_V4DI_USI): Remove.
14419 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14420 __builtin_ia32_vpshufbitqmb256_mask,
14421 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14422 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14423 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14425 2018-01-26 Alan Modra <amodra@gmail.com>
14428 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14429 UNSPEC_VBPERMQ. Sort other unspecs.
14431 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14433 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14435 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14437 PR middle-end/83055
14438 * predict.c (drop_profile): Do not push/pop cfun; update also
14440 (handle_missing_profiles): Fix logic looking for zero profiles.
14442 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14444 PR middle-end/83977
14445 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14446 on functions with #pragma omp declare simd or functions with simd
14448 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14449 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14450 Remove trailing \n from warning_at calls.
14452 2018-01-25 Tom de Vries <tom@codesourcery.com>
14455 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14456 for neutered workers.
14458 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14461 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14462 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14464 2018-01-24 Jeff Law <law@redhat.com>
14467 * i386.c (get_probe_interval): Move to earlier point.
14468 (ix86_compute_frame_layout): If -fstack-clash-protection and
14469 the frame is larger than the probe interval, then use pushes
14470 to save registers rather than reg->mem moves.
14471 (ix86_expand_prologue): Remove conditional for int_registers_saved
14474 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14477 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14478 min/max for never referenced object.
14480 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14482 PR middle-end/83977
14483 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14485 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14486 attributes from DECL_ATTRIBUTES (decl) without affecting
14487 DECL_ATTRIBUTES (current_function_decl).
14488 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14489 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14491 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14493 PR tree-optimization/83979
14494 * fold-const.c (fold_comparison): Use constant_boolean_node
14495 instead of boolean_{true,false}_node.
14497 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14499 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14502 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14504 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14505 Simplify the clause that sets the length attribute.
14506 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14507 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14508 clause that sets the length attribute.
14509 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14511 2018-01-24 Tom de Vries <tom@codesourcery.com>
14514 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14515 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14516 Add strict parameter.
14517 (prevent_branch_around_nothing): Insert dummy insn between branch to
14518 label and label with no ptx insn inbetween.
14519 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14521 2018-01-24 Tom de Vries <tom@codesourcery.com>
14524 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14525 for neutered threads in warp.
14526 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14528 2018-01-24 Richard Biener <rguenther@suse.de>
14530 PR tree-optimization/83176
14531 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14534 2018-01-24 Richard Biener <rguenther@suse.de>
14536 PR tree-optimization/82819
14537 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14538 code generating pluses that are no-ops in the target precision.
14540 2018-01-24 Richard Biener <rguenther@suse.de>
14542 PR middle-end/84000
14543 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14545 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14547 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14548 to merge probabilities.
14549 * predict.c (probably_never_executed): Also mark as cold functions
14550 with global 0 profile and guessed local profile.
14551 * profile-count.c (profile_probability::combine_with_count): New
14553 * profile-count.h (profile_probability::operator*,
14554 profile_probability::operator*=, profile_probability::operator/,
14555 profile_probability::operator/=): Reduce precision to adjusted
14556 and set value to guessed on contradictory divisions.
14557 (profile_probability::combine_with_freq): Remove.
14558 (profile_probability::combine_wiht_count): Declare.
14559 (profile_count::force_nonzero):: Set to adjusted.
14560 (profile_count::probability_in):: Set quality to adjusted.
14561 * tree-ssa-tail-merge.c (replace_block_by): Use
14562 combine_with_count.
14564 2018-01-23 Andrew Waterman <andrew@sifive.com>
14565 Jim Wilson <jimw@sifive.com>
14567 * config/riscv/riscv.c (riscv_stack_boundary): New.
14568 (riscv_option_override): Set riscv_stack_boundary. Handle
14569 riscv_preferred_stack_boundary_arg.
14570 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14571 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14572 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14573 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14574 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14575 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14577 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14580 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14581 of struct ix86_frame.
14582 (ix86_expand_epilogue): Likewise. Add a local variable for
14583 the reg_save_offset field in struct ix86_frame.
14585 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14587 PR tree-optimization/82604
14588 * tree-loop-distribution.c (enum partition_kind): New enum item
14589 PKIND_PARTIAL_MEMSET.
14590 (partition_builtin_p): Support above new enum item.
14591 (generate_code_for_partition): Ditto.
14592 (compute_access_range): Differentiate cases that equality can be
14593 proven at all loops, the innermost loops or no loops.
14594 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14595 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14596 (finalize_partitions, distribute_loop): Don't fuse partition of
14597 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14598 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14599 parloop is enabled.
14601 2018-01-23 Martin Liska <mliska@suse.cz>
14603 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14604 order to ignore the predictor.
14605 (PRED_POLYMORPHIC_CALL): Likewise.
14606 (PRED_RECURSIVE_CALL): Likewise.
14608 2018-01-23 Martin Liska <mliska@suse.cz>
14610 * tree-profile.c (tree_profiling): Print function header to
14611 aware reader which function we are working on.
14612 * value-prof.c (gimple_find_values_to_profile): Do not print
14613 not interesting value histograms.
14615 2018-01-23 Martin Liska <mliska@suse.cz>
14617 * profile-count.h (enum profile_quality): Add
14618 profile_uninitialized as the first value. Do not number values
14619 as they are zero based.
14620 (profile_count::verify): Update sanity check.
14621 (profile_probability::verify): Likewise.
14623 2018-01-23 Nathan Sidwell <nathan@acm.org>
14625 * doc/invoke.texi (ffor-scope): Deprecate.
14627 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14629 PR tree-optimization/83510
14630 * domwalk.c (set_all_edges_as_executable): New function.
14631 (dom_walker::dom_walker): Convert bool param
14632 "skip_unreachable_blocks" to enum reachability. Move setup of
14633 edge flags to set_all_edges_as_executable and only do it when
14634 reachability is REACHABLE_BLOCKS.
14635 * domwalk.h (enum dom_walker::reachability): New enum.
14636 (dom_walker::dom_walker): Convert bool param
14637 "skip_unreachable_blocks" to enum reachability.
14638 (set_all_edges_as_executable): New decl.
14639 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14640 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14642 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14643 but converting true to REACHABLE_BLOCKS.
14644 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14646 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14647 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14648 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14650 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14651 if check_all_array_refs will be called.
14653 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14655 * tree.c (selftest::test_location_wrappers): Add more test
14658 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14660 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14661 (selftest::test_bit_in_range): Likewise.
14663 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14666 * doc/sourcebuild.texi (vect_float): Say that the selector
14667 only describes the situation when -funsafe-math-optimizations is on.
14668 (vect_float_strict): Document.
14670 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14672 PR tree-optimization/83965
14673 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14674 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14675 instead of checking only for a reduction.
14676 (vect_recog_widen_sum_pattern): Likewise.
14678 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14680 * predict.c (probably_never_executed): Only use precise profile info.
14681 (compute_function_frequency): Skip after inlining hack since we now
14682 have quality checking.
14684 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14686 * profile-count.h (profile_probability::very_unlikely,
14687 profile_probability::unlikely, profile_probability::even): Set
14688 precision to guessed.
14690 2018-01-23 Richard Biener <rguenther@suse.de>
14692 PR tree-optimization/83963
14693 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14694 Properly terminate dominator walk when crossing the exit edge not
14695 when visiting its source block.
14697 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14700 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14701 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14703 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14705 PR tree-optimization/83957
14706 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14707 semicolon after for body surrounded by braces.
14709 PR tree-optimization/83081
14710 * profile-count.h (profile_probability::split): New method.
14711 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14712 Use profile_probability::split.
14713 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14714 when splitting a single conditional jump into 2.
14716 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14718 PR tree-optimization/69452
14719 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14722 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14724 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14725 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14726 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14728 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14730 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14732 * config/rl78/rl78.md (movdi): New define_expand.
14733 * config/rl78/rl78.c (rl78_split_movdi): New function.
14735 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14738 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14740 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14741 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14742 128-bit to produce an UNSPEC move to get the double word with the
14743 signbit and then a shift directly to do signbit.
14744 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14745 implementation with a new version that just does either a direct
14746 move or a regular move. Move memory interface to separate insns.
14747 Move insns so they are next to the expander.
14748 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14749 with signbit move. Split big and little endian case.
14750 (signbit<mode>2_dm_mem_le): Likewise.
14751 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14752 (signbit<mode>2_dm2): Likewise.
14754 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14756 * config/rl78/rl78.md (anddi3): New define_expand.
14758 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14760 * config/rl78/rl78.md (umindi3): New define_expand.
14762 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14764 * config/rl78/rl78.md (smindi3): New define_expand.
14766 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14768 * config/rl78/rl78.md (smaxdi3): New define_expand.
14770 2018-01-22 Carl Love <cel@us.ibm.com>
14772 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14773 LVX_V1TI): Add macro expansion.
14774 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14775 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14776 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14777 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14778 Change check to determine if the instruction is a byte reversing
14779 entry. Fix typo in comment.
14780 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14781 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14782 Add def_builtin calls for new builtins.
14783 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14784 Add define_insn expansion.
14786 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14788 * config/rl78/rl78.md (umaxdi3): New define_expand.
14790 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14792 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14793 for non-QImode registers.
14795 2018-01-22 Richard Biener <rguenther@suse.de>
14797 PR tree-optimization/83963
14798 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14799 including the loop exit block.
14800 (scop_detection::merge_sese): Likewise.
14801 (scop_detection::add_scop): Do it here instead.
14803 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14805 * doc/sourcebuild.texi (arm_softfloat): Document.
14807 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14810 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14811 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14812 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14814 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14815 David Edelsohn <dje.gcc@gmail.com>
14818 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14819 Change "crset eq" to "crset 2".
14820 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14821 (*call_indirect_aix<mode>_nospec): Likewise.
14822 (*call_value_indirect_aix<mode>_nospec): Likewise.
14823 (*call_indirect_elfv2<mode>_nospec): Likewise.
14824 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14825 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14826 change assembly output from . to $.
14827 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14828 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14829 (*tablejump<mode>_internal1_nospec): Likewise.
14831 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14834 * config/sh/sh_optimize_sett_clrt.cc:
14835 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14837 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14839 PR tree-optimization/83940
14840 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14841 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14842 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14843 use it to pass back the definition type.
14844 (vect_check_store_rhs): Likewise rhs_dt_out.
14845 (vect_build_gather_load_calls): Add a mask_dt argument and use
14846 it instead of a call to vect_is_simple_use.
14847 (vectorizable_store): Update calls to vect_check_load_store_mask
14848 and vect_check_store_rhs. Use the dt returned by the latter instead
14849 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14850 instead of calls to vect_is_simple_use. Pass the scalar rather
14851 than the vector operand to vect_is_simple_use when handling
14852 second and subsequent copies of an rhs value.
14853 (vectorizable_load): Update calls to vect_check_load_store_mask
14854 and vect_build_gather_load_calls. Use the cached mask_dt and
14855 gs_info.offset_dt instead of calls to vect_is_simple_use.
14857 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14859 PR middle-end/83945
14860 * tree-emutls.c: Include gimplify.h.
14861 (lower_emutls_2): New function.
14862 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14863 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14864 it before further processing.
14867 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14868 UINTVAL (trueop1) instead of INTVAL (op1).
14870 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14874 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14875 INCOMING_FRAME_SP_OFFSET if not defined.
14876 (scan_trace): Add ENTRY argument. If true and
14877 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14878 emit a note to adjust the CFA offset.
14879 (create_cfi_notes): Adjust scan_trace callers.
14880 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14881 INCOMING_FRAME_SP_OFFSET in the CIE.
14882 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14883 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14885 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14886 * doc/tm.texi: Regenerated.
14888 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14890 PR rtl-optimization/83147
14891 * lra-constraints.c (remove_inheritance_pseudos): Use
14892 lra_substitute_pseudo_within_insn.
14894 2018-01-19 Tom de Vries <tom@codesourcery.com>
14895 Cesar Philippidis <cesar@codesourcery.com>
14898 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14900 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14903 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14904 spaces for function labels.
14906 2018-01-19 Martin Liska <mliska@suse.cz>
14908 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14909 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14910 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14911 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14912 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14913 (PRED_CONST_RETURN): Change from 69 to 65.
14914 (PRED_NULL_RETURN): Change from 91 to 71.
14915 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14916 (PRED_LOOP_GUARD): Change from 66 to 73.
14918 2018-01-19 Martin Liska <mliska@suse.cz>
14920 * predict.c (predict_insn_def): Add new assert.
14921 (struct branch_predictor): Change type to signed integer.
14922 (test_prediction_value_range): Amend test to cover
14923 PROB_UNINITIALIZED.
14924 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14925 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14926 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14927 (PRED_LOOP_IV_COMPARE): Likewise.
14928 * predict.h (PROB_UNINITIALIZED): Define new constant.
14930 2018-01-19 Martin Liska <mliska@suse.cz>
14932 * predict.c (dump_prediction): Add new format for
14933 analyze_brprob.py script which is enabled with -details
14935 * profile-count.h (precise_p): New function.
14937 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14939 PR tree-optimization/83922
14940 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14941 there are no statements that need masking.
14942 (vect_active_double_reduction_p): New function.
14943 (vect_analyze_loop_operations): Use it when handling phis that
14944 are not in the loop header.
14946 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14948 PR tree-optimization/83914
14949 * tree-vect-loop.c (vectorizable_induction): Don't convert
14950 init_expr or apply the peeling adjustment for inductions
14951 that are nested within the vectorized loop.
14953 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14955 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14958 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14962 * function.h (gimplify_parameters): Add gimple_seq * argument.
14963 * function.c: Include gimple.h and options.h.
14964 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14965 for the added local temporaries if needed.
14966 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14967 if there are any parameter cleanups, wrap whole body into a
14968 try/finally with the cleanups.
14970 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14973 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14974 Use GET_MODE_CLASS for scalar floating point.
14976 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14980 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14981 Fix call of call_cgraph_insertion_hooks.
14983 2018-01-18 Martin Sebor <msebor@redhat.com>
14985 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14987 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14990 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14993 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14996 * common.opt: (-ffile-prefix-map): New option.
14997 * opts.c (common_handle_option): Defer it.
14998 * opts-global.c (handle_common_deferred_options): Handle it.
14999 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15000 * file-prefix-map.h: New file.
15001 (remap_debug_filename, add_debug_prefix_map): ...here.
15002 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15003 * final.c (debug_prefix_map, add_debug_prefix_map
15004 remap_debug_filename): Move to...
15005 * file-prefix-map.c: New file.
15006 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15007 generalize, get rid of alloca(), use strrchr() instead of strchr().
15008 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15009 Implement in terms of add_prefix_map().
15010 (remap_macro_filename, remap_debug_filename): Implement in term of
15012 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15013 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15014 * dbxout.c: Include file-prefix-map.h.
15015 * varasm.c: Likewise.
15016 * vmsdbgout.c: Likewise.
15017 * xcoffout.c: Likewise.
15018 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15019 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15020 * doc/invoke.texi (-ffile-prefix-map): Document.
15021 (-fdebug-prefix-map): Update description.
15023 2018-01-18 Martin Liska <mliska@suse.cz>
15025 * config/i386/i386.c (indirect_thunk_name): Document that also
15027 (output_indirect_thunk): Document why both instructions
15028 (pause and lfence) are generated.
15030 2018-01-18 Richard Biener <rguenther@suse.de>
15032 PR tree-optimization/83887
15033 * graphite-scop-detection.c
15034 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15035 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15036 (scop_detection::merge_sese): Re-implement with a flood-fill
15037 algorithm that properly finds a SESE region if it exists.
15039 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15042 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15043 pointer_diff optimizations use view_convert instead of convert.
15045 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15047 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15048 Generate different code for -mno-speculate-indirect-jumps.
15049 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15050 (*call_indirect_aix<mode>): Disable for
15051 -mno-speculate-indirect-jumps.
15052 (*call_indirect_aix<mode>_nospec): New define_insn.
15053 (*call_value_indirect_aix<mode>): Disable for
15054 -mno-speculate-indirect-jumps.
15055 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15056 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15057 -mno-speculate-indirect-jumps.
15058 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15060 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15062 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15063 long double type, set the flags for noting the default long double
15064 type, even if we don't pass or return a long double type.
15066 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15069 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15072 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15074 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15075 support for merge[hl].
15076 (fold_mergehl_helper): New helper function.
15077 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15078 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15079 (altivec_vmrglw_direct): Add xxmrglw insn.
15081 2018-01-17 Andrew Waterman <andrew@sifive.com>
15083 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15084 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15086 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15089 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15090 call the lto_location_cache before reading the
15091 DECL_SOURCE_LOCATION of the types.
15093 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15094 Richard Sandiford <richard.sandiford@linaro.org>
15096 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15097 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15098 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15099 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15100 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15102 * config/aarch64/constraints.md (aarch64_movti_operand):
15104 * config/aarch64/predicates.md (Uti): Add new constraint.
15106 2018-01-17 Carl Love <cel@us.ibm.com>
15108 * config/rs6000/vsx.md (define_expand xl_len_r,
15109 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15110 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15112 (define_expand, define_insn): Move the shift left from the
15113 define_insn to the define_expand for lxvl and stxvl instructions.
15114 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15115 and XL_LEN_R definitions to PURE.
15117 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15119 * config/i386/i386.c (indirect_thunk_name): Declare regno
15120 as unsigned int. Compare regno with INVALID_REGNUM.
15121 (output_indirect_thunk): Ditto.
15122 (output_indirect_thunk_function): Ditto.
15123 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15124 in the call to output_indirect_thunk_function.
15126 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15128 PR middle-end/83884
15129 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15130 rather than the size of inner_type to determine the stack slot size
15131 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15133 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15136 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15139 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15141 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15142 endian Linux systems to optionally enable multilibs for selecting
15143 the long double type if the user configured an explicit type.
15144 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15145 have no long double multilibs if not defined.
15146 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15147 warn if the user used -mabi={ieee,ibm}longdouble and we built
15148 multilibs for long double.
15149 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15150 appropriate multilib option.
15151 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15153 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15154 for building long double multilibs.
15155 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15157 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15159 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15162 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15164 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15167 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15170 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15173 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15175 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15176 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15178 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15180 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15181 different rtl trees depending on TARGET_64BIT.
15182 (rs6000_gen_lvx): Likewise.
15184 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15186 * config/visium/visium.md (nop): Tweak comment.
15187 (hazard_nop): Likewise.
15189 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15191 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15192 -mspeculate-indirect-jumps.
15193 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15194 for -mno-speculate-indirect-jumps.
15195 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15196 (*call_value_indirect_elfv2<mode>): Disable for
15197 -mno-speculate-indirect-jumps.
15198 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15199 (indirect_jump): Emit different RTL for
15200 -mno-speculate-indirect-jumps.
15201 (*indirect_jump<mode>): Disable for
15202 -mno-speculate-indirect-jumps.
15203 (*indirect_jump<mode>_nospec): New define_insn.
15204 (tablejump): Emit different RTL for
15205 -mno-speculate-indirect-jumps.
15206 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15207 (tablejumpsi_nospec): New define_expand.
15208 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15209 (tablejumpdi_nospec): New define_expand.
15210 (*tablejump<mode>_internal1): Disable for
15211 -mno-speculate-indirect-jumps.
15212 (*tablejump<mode>_internal1_nospec): New define_insn.
15213 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15216 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15218 * caller-save.c (insert_save): Drop unnecessary parameter. All
15221 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15222 Richard Biener <rguenth@suse.de>
15225 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15226 return early, inline manually is_gimple_sizepos. Make sure if we
15227 call gimplify_expr we don't end up with a gimple constant.
15228 * tree.c (variably_modified_type_p): Don't return true for
15229 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15230 * gimplify.h (is_gimple_sizepos): Remove.
15232 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15234 PR tree-optimization/83857
15235 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15236 vectorizable_live_operation for pure SLP statements.
15237 (vectorizable_live_operation): Handle PHIs.
15239 2018-01-16 Richard Biener <rguenther@suse.de>
15241 PR tree-optimization/83867
15242 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15243 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15245 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15248 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15249 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15250 If off is not INTEGER_CST, issue a may not be aligned warning
15251 rather than isn't aligned. Use isn%'t rather than isn't.
15252 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15254 <case MULT_EXPR>: Improve the case when bottom and one of the
15255 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15256 operand, in that case check if the other operand is multiple of
15257 bottom divided by the INTEGER_CST operand.
15259 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15262 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15263 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15264 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15265 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15266 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15267 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15268 (pa_function_arg_size): New function.
15270 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15272 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15273 in a separate statement.
15275 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15277 PR tree-optimization/83847
15278 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15279 group gathers and scatters.
15281 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15283 PR rtl-optimization/86620
15284 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15286 PR rtl-optimization/83213
15287 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15288 to last if both are JUMP_INSNs.
15290 PR tree-optimization/83843
15291 * gimple-ssa-store-merging.c
15292 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15293 store_immediate_info for bswap/nop orig_stores.
15295 2018-01-15 Andrew Waterman <andrew@sifive.com>
15297 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15299 <UDIV>: Increase cost if !TARGET_DIV.
15301 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15303 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15304 (define_attr "cr_logical_3op"): New.
15305 (cceq_ior_compare): Adjust.
15306 (cceq_ior_compare_complement): Adjust.
15307 (*cceq_rev_compare): Adjust.
15308 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15309 (is_cracked_insn): Adjust.
15310 (insn_must_be_first_in_group): Adjust.
15311 * config/rs6000/40x.md: Adjust.
15312 * config/rs6000/440.md: Adjust.
15313 * config/rs6000/476.md: Adjust.
15314 * config/rs6000/601.md: Adjust.
15315 * config/rs6000/603.md: Adjust.
15316 * config/rs6000/6xx.md: Adjust.
15317 * config/rs6000/7450.md: Adjust.
15318 * config/rs6000/7xx.md: Adjust.
15319 * config/rs6000/8540.md: Adjust.
15320 * config/rs6000/cell.md: Adjust.
15321 * config/rs6000/e300c2c3.md: Adjust.
15322 * config/rs6000/e500mc.md: Adjust.
15323 * config/rs6000/e500mc64.md: Adjust.
15324 * config/rs6000/e5500.md: Adjust.
15325 * config/rs6000/e6500.md: Adjust.
15326 * config/rs6000/mpc.md: Adjust.
15327 * config/rs6000/power4.md: Adjust.
15328 * config/rs6000/power5.md: Adjust.
15329 * config/rs6000/power6.md: Adjust.
15330 * config/rs6000/power7.md: Adjust.
15331 * config/rs6000/power8.md: Adjust.
15332 * config/rs6000/power9.md: Adjust.
15333 * config/rs6000/rs64.md: Adjust.
15334 * config/rs6000/titan.md: Adjust.
15336 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15338 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15339 ix86_indirect_branch_register logic.
15341 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15343 * config/i386/constraints.md (Bs): Update
15344 ix86_indirect_branch_register check. Don't check
15345 ix86_indirect_branch_register with GOT_memory_operand.
15347 * config/i386/predicates.md (GOT_memory_operand): Don't check
15348 ix86_indirect_branch_register here.
15349 (GOT32_symbol_operand): Likewise.
15351 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15353 * config/i386/predicates.md (constant_call_address_operand):
15354 Rewrite ix86_indirect_branch_register logic.
15355 (sibcall_insn_operand): Likewise.
15357 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15359 * config/i386/constraints.md (Bs): Replace
15360 ix86_indirect_branch_thunk_register with
15361 ix86_indirect_branch_register.
15363 * config/i386/i386.md (indirect_jump): Likewise.
15364 (tablejump): Likewise.
15365 (*sibcall_memory): Likewise.
15366 (*sibcall_value_memory): Likewise.
15367 Peepholes of indirect call and jump via memory: Likewise.
15368 * config/i386/i386.opt: Likewise.
15369 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15370 (GOT_memory_operand): Likewise.
15371 (call_insn_operand): Likewise.
15372 (sibcall_insn_operand): Likewise.
15373 (GOT32_symbol_operand): Likewise.
15375 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15377 PR middle-end/83837
15378 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15379 type rather than type addr's type points to.
15380 (expand_omp_atomic_mutex): Likewise.
15381 (expand_omp_atomic): Likewise.
15383 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15386 * config/i386/i386.c (output_indirect_thunk_function): Use
15387 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15388 for __x86_return_thunk.
15390 2018-01-15 Richard Biener <rguenther@suse.de>
15392 PR middle-end/83850
15393 * expmed.c (extract_bit_field_1): Fix typo.
15395 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15398 * config/arm/iterators.md (VF): New mode iterator.
15399 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15400 Remove integer-related logic from pattern.
15401 (neon_vabd<mode>_3): Likewise.
15403 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15405 PR middle-end/82694
15406 * common.opt (fstrict-overflow): No longer an alias.
15407 (fwrapv-pointer): New option.
15408 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15409 also for pointer types based on flag_wrapv_pointer.
15410 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15411 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15412 opts->x_flag_wrapv got set.
15413 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15414 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15415 POINTER_TYPE_OVERFLOW_UNDEFINED.
15416 * match.pd: Likewise in address comparison pattern.
15417 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15419 2018-01-15 Richard Biener <rguenther@suse.de>
15422 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15423 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15424 Reset type names to their identifier if their TYPE_DECL doesn't
15425 have linkage (and thus is used for ODR and devirt).
15426 (save_debug_info_for_decl): Remove.
15427 (save_debug_info_for_type): Likewise.
15428 (add_tree_to_fld_list): Adjust.
15429 * tree-pretty-print.c (dump_generic_node): Make dumping of
15430 type names more robust.
15432 2018-01-15 Richard Biener <rguenther@suse.de>
15434 * BASE-VER: Bump to 8.0.1.
15436 2018-01-14 Martin Sebor <msebor@redhat.com>
15439 * builtins.c (check_access): Avoid warning when the no-warning bit
15442 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15444 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15445 * ira-color (allocno_hard_regs_compare): Likewise.
15447 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15450 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15451 Use .pushsection/.popsection.
15453 2018-01-14 Martin Sebor <msebor@redhat.com>
15456 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15458 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15460 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15461 entry from extra_headers.
15462 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15463 extra_headers, make the list bitwise identical to the i?86-*-* one.
15465 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15467 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15468 -mcmodel=large with -mindirect-branch=thunk,
15469 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15470 -mfunction-return=thunk-extern.
15471 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15472 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15473 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15475 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15477 * config/i386/i386.c (print_reg): Print the name of the full
15478 integer register without '%'.
15479 (ix86_print_operand): Handle 'V'.
15480 * doc/extend.texi: Document 'V' modifier.
15482 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15484 * config/i386/constraints.md (Bs): Disallow memory operand for
15485 -mindirect-branch-register.
15487 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15488 (GOT_memory_operand): Likewise.
15489 (call_insn_operand): Likewise.
15490 (sibcall_insn_operand): Likewise.
15491 (GOT32_symbol_operand): Likewise.
15492 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15493 for -mindirect-branch-register.
15494 (tablejump): Likewise.
15495 (*sibcall_memory): Likewise.
15496 (*sibcall_value_memory): Likewise.
15497 Disallow peepholes of indirect call and jump via memory for
15498 -mindirect-branch-register.
15499 (*call_pop): Replace m with Bw.
15500 (*call_value_pop): Likewise.
15501 (*sibcall_pop_memory): Replace m with Bs.
15502 * config/i386/i386.opt (mindirect-branch-register): New option.
15503 * doc/invoke.texi: Document -mindirect-branch-register option.
15505 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15507 * config/i386/i386-protos.h (ix86_output_function_return): New.
15508 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15509 set function_return_type.
15510 (indirect_thunk_name): Add ret_p to indicate thunk for function
15512 (output_indirect_thunk_function): Pass false to
15513 indirect_thunk_name.
15514 (ix86_output_indirect_branch_via_reg): Likewise.
15515 (ix86_output_indirect_branch_via_push): Likewise.
15516 (output_indirect_thunk_function): Create alias for function
15517 return thunk if regno < 0.
15518 (ix86_output_function_return): New function.
15519 (ix86_handle_fndecl_attribute): Handle function_return.
15520 (ix86_attribute_table): Add function_return.
15521 * config/i386/i386.h (machine_function): Add
15522 function_return_type.
15523 * config/i386/i386.md (simple_return_internal): Use
15524 ix86_output_function_return.
15525 (simple_return_internal_long): Likewise.
15526 * config/i386/i386.opt (mfunction-return=): New option.
15527 (indirect_branch): Mention -mfunction-return=.
15528 * doc/extend.texi: Document function_return function attribute.
15529 * doc/invoke.texi: Document -mfunction-return= option.
15531 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15533 * config/i386/i386-opts.h (indirect_branch): New.
15534 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15535 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15536 with local indirect jump when converting indirect call and jump.
15537 (ix86_set_indirect_branch_type): New.
15538 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15539 (indirectlabelno): New.
15540 (indirect_thunk_needed): Likewise.
15541 (indirect_thunk_bnd_needed): Likewise.
15542 (indirect_thunks_used): Likewise.
15543 (indirect_thunks_bnd_used): Likewise.
15544 (INDIRECT_LABEL): Likewise.
15545 (indirect_thunk_name): Likewise.
15546 (output_indirect_thunk): Likewise.
15547 (output_indirect_thunk_function): Likewise.
15548 (ix86_output_indirect_branch_via_reg): Likewise.
15549 (ix86_output_indirect_branch_via_push): Likewise.
15550 (ix86_output_indirect_branch): Likewise.
15551 (ix86_output_indirect_jmp): Likewise.
15552 (ix86_code_end): Call output_indirect_thunk_function if needed.
15553 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15555 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15556 (ix86_attribute_table): Add indirect_branch.
15557 * config/i386/i386.h (machine_function): Add indirect_branch_type
15558 and has_local_indirect_jump.
15559 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15561 (tablejump): Likewise.
15562 (*indirect_jump): Use ix86_output_indirect_jmp.
15563 (*tablejump_1): Likewise.
15564 (simple_return_indirect_internal): Likewise.
15565 * config/i386/i386.opt (mindirect-branch=): New option.
15566 (indirect_branch): New.
15569 (thunk-inline): Likewise.
15570 (thunk-extern): Likewise.
15571 * doc/extend.texi: Document indirect_branch function attribute.
15572 * doc/invoke.texi: Document -mindirect-branch= option.
15574 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15577 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15579 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15581 * ipa-inline.c (want_inline_small_function_p): Return false if
15582 inlining has already failed with CIF_FINAL_ERROR.
15583 (update_caller_keys): Call want_inline_small_function_p before
15585 (update_callee_keys): Likewise.
15587 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15589 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15591 (rs6000_quadword_masked_address_p): Likewise.
15592 (quad_aligned_load_p): Likewise.
15593 (quad_aligned_store_p): Likewise.
15594 (const_load_sequence_p): Add comment to describe the outer-most loop.
15595 (mimic_memory_attributes_and_flags): New function.
15596 (rs6000_gen_stvx): Likewise.
15597 (replace_swapped_aligned_store): Likewise.
15598 (rs6000_gen_lvx): Likewise.
15599 (replace_swapped_aligned_load): Likewise.
15600 (replace_swapped_load_constant): Capitalize argument name in
15601 comment describing this function.
15602 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15603 and stores that access quad-word aligned addresses and replace
15604 with stvx or lvx instructions when appropriate.
15605 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15606 New function prototype.
15607 (rs6000_quadword_masked_address_p): Likewise.
15608 (rs6000_gen_lvx): Likewise.
15609 (rs6000_gen_stvx): Likewise.
15610 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15611 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15612 when memory address is aligned.
15613 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15614 this split to select lvx instruction when memory address is aligned.
15615 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15616 instruction when memory address is aligned.
15617 (*vsx_le_perm_load_v16qi): Likewise.
15618 (four unnamed splitters): Modify to select the stvx instruction
15619 when memory is aligned.
15621 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15623 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15624 which appears in the queue multiple times.
15626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15627 Alan Hayward <alan.hayward@arm.com>
15628 David Sherwood <david.sherwood@arm.com>
15630 * tree-vectorizer.h (vec_lower_bound): New structure.
15631 (_loop_vec_info): Add check_nonzero and lower_bounds.
15632 (LOOP_VINFO_CHECK_NONZERO): New macro.
15633 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15634 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15635 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15636 fields. Make seg_len the distance travelled, not including the
15638 (dr_direction_indicator): Declare.
15639 (dr_zero_step_indicator): Likewise.
15640 (dr_known_forward_stride_p): Likewise.
15641 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15643 (runtime_alias_check_p): Allow runtime alias checks with
15645 (operator ==): Compare access_size and align.
15646 (prune_runtime_alias_test_list): Rework for new distinction between
15647 the access_size and seg_len.
15648 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15650 (get_segment_min_max): New function.
15651 (create_intersect_range_checks): Use it.
15652 (dr_step_indicator): New function.
15653 (dr_direction_indicator): Likewise.
15654 (dr_zero_step_indicator): Likewise.
15655 (dr_known_forward_stride_p): Likewise.
15656 * tree-loop-distribution.c (data_ref_segment_size): Return
15657 DR_STEP * (niters - 1).
15658 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15660 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15661 (vect_preserves_scalar_order_p): New function, split out from...
15662 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15663 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15664 (vect_vfa_access_size): New function.
15665 (vect_vfa_align): Likewise.
15666 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15667 (dump_lower_bound): New function.
15668 (vect_check_lower_bound): Likewise.
15669 (vect_small_gap_p): Likewise.
15670 (vectorizable_with_step_bound_p): Likewise.
15671 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15672 depencies if the vectorization factor is 1. Convert the checks
15673 for nonzero steps into checks on the bounds of DR_STEP. Try using
15674 a bunds check for variable steps if the minimum required step is
15675 relatively small. Update calls to the dr_with_seg_len
15676 constructor and to vect_compile_time_alias.
15677 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15679 (vect_loop_versioning): Call it.
15680 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15682 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15685 Alan Hayward <alan.hayward@arm.com>
15686 David Sherwood <david.sherwood@arm.com>
15688 * doc/sourcebuild.texi (vect_scatter_store): Document.
15689 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15691 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15693 * genopinit.c (main): Add supports_vec_scatter_store and
15694 supports_vec_scatter_store_cached to target_optabs.
15695 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15696 IFN_MASK_SCATTER_STORE.
15697 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15699 * internal-fn.h (internal_store_fn_p): Declare.
15700 (internal_fn_stored_value_index): Likewise.
15701 * internal-fn.c (scatter_store_direct): New macro.
15702 (expand_scatter_store_optab_fn): New function.
15703 (direct_scatter_store_optab_supported_p): New macro.
15704 (internal_store_fn_p): New function.
15705 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15706 IFN_MASK_SCATTER_STORE.
15707 (internal_fn_mask_index): Likewise.
15708 (internal_fn_stored_value_index): New function.
15709 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15710 for scatter stores.
15711 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15712 * optabs-query.c (supports_vec_scatter_store_p): New function.
15713 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15714 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15715 true for scatter stores.
15716 (vect_gather_scatter_fn_p): Handle scatter stores too.
15717 (vect_check_gather_scatter): Consider using scatter stores if
15718 supports_vec_scatter_store_p.
15719 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15720 scatter stores too.
15721 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15722 internal_fn_stored_value_index.
15723 (check_load_store_masking): Handle scatter stores too.
15724 (vect_get_store_rhs): Make public.
15725 (vectorizable_call): Use internal_store_fn_p.
15726 (vectorizable_store): Handle scatter store internal functions.
15727 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15728 when deciding whether the end of the group has been reached.
15729 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15730 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15731 (mask_scatter_store<mode>): New insns.
15733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15734 Alan Hayward <alan.hayward@arm.com>
15735 David Sherwood <david.sherwood@arm.com>
15737 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15738 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15739 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15741 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15742 Use vect_truncate_gather_scatter_offset if we can't treat the
15743 operation as a normal gather load or scatter store.
15744 (get_group_load_store_type): Take the gather_scatter_info
15745 as argument. Try using a gather load or scatter store for
15746 single-element groups.
15747 (get_load_store_type): Update calls to get_group_load_store_type
15748 and vect_use_strided_gather_scatters_p.
15750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15751 Alan Hayward <alan.hayward@arm.com>
15752 David Sherwood <david.sherwood@arm.com>
15754 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15755 optional tree argument.
15756 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15758 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15759 but continue to use the current value as a fallback.
15760 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15761 to compare the updates.
15762 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15763 (get_load_store_type): Use it when handling a strided access.
15764 (vect_get_strided_load_store_ops): New function.
15765 (vect_get_data_ptr_increment): Likewise.
15766 (vectorizable_load): Handle strided gather loads. Always pass
15767 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15770 Alan Hayward <alan.hayward@arm.com>
15771 David Sherwood <david.sherwood@arm.com>
15773 * doc/md.texi (gather_load@var{m}): Document.
15774 (mask_gather_load@var{m}): Likewise.
15775 * genopinit.c (main): Add supports_vec_gather_load and
15776 supports_vec_gather_load_cached to target_optabs.
15777 * optabs-tree.c (init_tree_optimization_optabs): Use
15778 ggc_cleared_alloc to allocate target_optabs.
15779 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15780 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15782 * internal-fn.h (internal_load_fn_p): Declare.
15783 (internal_gather_scatter_fn_p): Likewise.
15784 (internal_fn_mask_index): Likewise.
15785 (internal_gather_scatter_fn_supported_p): Likewise.
15786 * internal-fn.c (gather_load_direct): New macro.
15787 (expand_gather_load_optab_fn): New function.
15788 (direct_gather_load_optab_supported_p): New macro.
15789 (direct_internal_fn_optab): New function.
15790 (internal_load_fn_p): Likewise.
15791 (internal_gather_scatter_fn_p): Likewise.
15792 (internal_fn_mask_index): Likewise.
15793 (internal_gather_scatter_fn_supported_p): Likewise.
15794 * optabs-query.c (supports_at_least_one_mode_p): New function.
15795 (supports_vec_gather_load_p): Likewise.
15796 * optabs-query.h (supports_vec_gather_load_p): Declare.
15797 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15798 and memory_type field.
15799 (NUM_PATTERNS): Bump to 15.
15800 * tree-vect-data-refs.c: Include internal-fn.h.
15801 (vect_gather_scatter_fn_p): New function.
15802 (vect_describe_gather_scatter_call): Likewise.
15803 (vect_check_gather_scatter): Try using internal functions for
15804 gather loads. Recognize existing calls to a gather load function.
15805 (vect_analyze_data_refs): Consider using gather loads if
15806 supports_vec_gather_load_p.
15807 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15808 (vect_get_gather_scatter_offset_type): Likewise.
15809 (vect_convert_mask_for_vectype): Likewise.
15810 (vect_add_conversion_to_patterm): Likewise.
15811 (vect_try_gather_scatter_pattern): Likewise.
15812 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15813 (vect_vect_recog_func_ptrs): Add it.
15814 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15815 internal_fn_mask_index and internal_gather_scatter_fn_p.
15816 (check_load_store_masking): Take the gather_scatter_info as an
15817 argument and handle gather loads.
15818 (vect_get_gather_scatter_ops): New function.
15819 (vectorizable_call): Check internal_load_fn_p.
15820 (vectorizable_load): Likewise. Handle gather load internal
15822 (vectorizable_store): Update call to check_load_store_masking.
15823 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15824 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15825 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15826 (aarch64_gather_scale_operand_d): New predicates.
15827 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15828 (mask_gather_load<mode>): New insns.
15830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15831 Alan Hayward <alan.hayward@arm.com>
15832 David Sherwood <david.sherwood@arm.com>
15834 * optabs.def (fold_left_plus_optab): New optab.
15835 * doc/md.texi (fold_left_plus_@var{m}): Document.
15836 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15837 * internal-fn.c (fold_left_direct): Define.
15838 (expand_fold_left_optab_fn): Likewise.
15839 (direct_fold_left_optab_supported_p): Likewise.
15840 * fold-const-call.c (fold_const_fold_left): New function.
15841 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15842 * tree-parloops.c (valid_reduction_p): New function.
15843 (gather_scalar_reductions): Use it.
15844 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15845 (vect_finish_replace_stmt): Declare.
15846 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15847 (needs_fold_left_reduction_p): New function, split out from...
15848 (vect_is_simple_reduction): ...here. Accept reductions that
15849 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15850 (vect_force_simple_reduction): Also store the reduction type in
15851 the assignment's STMT_VINFO_REDUC_TYPE.
15852 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15853 (merge_with_identity): New function.
15854 (vect_expand_fold_left): Likewise.
15855 (vectorize_fold_left_reduction): Likewise.
15856 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15857 scalar phi in place for it. Check for target support and reject
15858 cases that would reassociate the operation. Defer the transform
15859 phase to vectorize_fold_left_reduction.
15860 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15861 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15862 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15866 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15867 call to ifc_temp_var.
15869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15870 Alan Hayward <alan.hayward@arm.com>
15871 David Sherwood <david.sherwood@arm.com>
15873 * target.def (legitimize_address_displacement): Take the original
15874 offset as a poly_int.
15875 * targhooks.h (default_legitimize_address_displacement): Update
15877 * targhooks.c (default_legitimize_address_displacement): Likewise.
15878 * doc/tm.texi: Regenerate.
15879 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15880 as an argument, moving assert of ad->disp == ad->disp_term to...
15881 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15882 Try calling targetm.legitimize_address_displacement before expanding
15883 the address rather than afterwards, and adjust for the new interface.
15884 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15885 Match the new hook interface. Handle SVE addresses.
15886 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15887 new hook interface.
15889 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15891 * Makefile.in (OBJS): Add early-remat.o.
15892 * target.def (select_early_remat_modes): New hook.
15893 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15894 * doc/tm.texi: Regenerate.
15895 * targhooks.h (default_select_early_remat_modes): Declare.
15896 * targhooks.c (default_select_early_remat_modes): New function.
15897 * timevar.def (TV_EARLY_REMAT): New timevar.
15898 * passes.def (pass_early_remat): New pass.
15899 * tree-pass.h (make_pass_early_remat): Declare.
15900 * early-remat.c: New file.
15901 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15903 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15906 Alan Hayward <alan.hayward@arm.com>
15907 David Sherwood <david.sherwood@arm.com>
15909 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15910 vfm1 with a bound_epilog parameter.
15911 (vect_do_peeling): Update calls accordingly, and move the prologue
15912 call earlier in the function. Treat the base bound_epilog as 0 for
15913 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15914 this base when peeling for gaps.
15915 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15916 with fully-masked loops.
15917 (vect_estimate_min_profitable_iters): Handle the single peeled
15918 iteration in that case.
15920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15921 Alan Hayward <alan.hayward@arm.com>
15922 David Sherwood <david.sherwood@arm.com>
15924 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15925 single-element interleaving even if the size is not a power of 2.
15926 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15927 accesses for single-element interleaving if the group size is
15930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15931 Alan Hayward <alan.hayward@arm.com>
15932 David Sherwood <david.sherwood@arm.com>
15934 * doc/md.texi (fold_extract_last_@var{m}): Document.
15935 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15936 * optabs.def (fold_extract_last_optab): New optab.
15937 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15938 * internal-fn.c (fold_extract_direct): New macro.
15939 (expand_fold_extract_optab_fn): Likewise.
15940 (direct_fold_extract_optab_supported_p): Likewise.
15941 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15942 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15943 EXTRACT_LAST_REDUCTION.
15944 (get_initial_def_for_reduction): Do not create an initial vector
15945 for EXTRACT_LAST_REDUCTION reductions.
15946 (vectorizable_reduction): Leave the scalar phi in place for
15947 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15948 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15949 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15950 transform phase to vectorizable_condition.
15951 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15953 (vect_finish_stmt_generation): ...here.
15954 (vect_finish_replace_stmt): New function.
15955 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15956 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15958 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15961 Alan Hayward <alan.hayward@arm.com>
15962 David Sherwood <david.sherwood@arm.com>
15964 * doc/md.texi (extract_last_@var{m}): Document.
15965 * optabs.def (extract_last_optab): New optab.
15966 * internal-fn.def (EXTRACT_LAST): New internal function.
15967 * internal-fn.c (cond_unary_direct): New macro.
15968 (expand_cond_unary_optab_fn): Likewise.
15969 (direct_cond_unary_optab_supported_p): Likewise.
15970 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15971 loops using EXTRACT_LAST.
15972 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15973 (extract_last_<mode>): ...this optab.
15974 (vec_extract<mode><Vel>): Update accordingly.
15976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15977 Alan Hayward <alan.hayward@arm.com>
15978 David Sherwood <david.sherwood@arm.com>
15980 * target.def (empty_mask_is_expensive): New hook.
15981 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15982 * doc/tm.texi: Regenerate.
15983 * targhooks.h (default_empty_mask_is_expensive): Declare.
15984 * targhooks.c (default_empty_mask_is_expensive): New function.
15985 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15986 if the target says that empty masks are expensive.
15987 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15989 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15992 Alan Hayward <alan.hayward@arm.com>
15993 David Sherwood <david.sherwood@arm.com>
15995 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15996 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15997 (vect_use_loop_mask_for_alignment_p): New function.
15998 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15999 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16000 niters_skip argument. Make sure that the first niters_skip elements
16001 of the first iteration are inactive.
16002 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16003 Update call to vect_set_loop_masks_directly.
16004 (get_misalign_in_elems): New function, split out from...
16005 (vect_gen_prolog_loop_niters): ...here.
16006 (vect_update_init_of_dr): Take a code argument that specifies whether
16007 the adjustment should be added or subtracted.
16008 (vect_update_init_of_drs): Likewise.
16009 (vect_prepare_for_masked_peels): New function.
16010 (vect_do_peeling): Skip prologue peeling if we're using a mask
16011 instead. Update call to vect_update_inits_of_drs.
16012 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16014 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16015 alignment. Do not include the number of peeled iterations in
16016 the minimum threshold in that case.
16017 (vectorizable_induction): Adjust the start value down by
16018 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16019 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16020 Take the number of skipped iterations into account when calculating
16022 * tree-vect-stmts.c (vect_gen_while_not): New function.
16024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16025 Alan Hayward <alan.hayward@arm.com>
16026 David Sherwood <david.sherwood@arm.com>
16028 * doc/sourcebuild.texi (vect_fully_masked): Document.
16029 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16030 default value to 0.
16031 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16033 (vect_analyze_loop_2): ...here. Don't check the vectorization
16034 factor against the number of loop iterations if the loop is
16037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16038 Alan Hayward <alan.hayward@arm.com>
16039 David Sherwood <david.sherwood@arm.com>
16041 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16042 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16043 (dump_groups): Update accordingly.
16044 (iv_use::mem_type): New member variable.
16045 (address_p): New function.
16046 (record_use): Add a mem_type argument and initialize the new
16048 (record_group_use): Add a mem_type argument. Use address_p.
16049 Remove obsolete null checks of base_object. Update call to record_use.
16050 (find_interesting_uses_op): Update call to record_group_use.
16051 (find_interesting_uses_cond): Likewise.
16052 (find_interesting_uses_address): Likewise.
16053 (get_mem_type_for_internal_fn): New function.
16054 (find_address_like_use): Likewise.
16055 (find_interesting_uses_stmt): Try find_address_like_use before
16056 calling find_interesting_uses_op.
16057 (addr_offset_valid_p): Use the iv mem_type field as the type
16058 of the addressed memory.
16059 (add_autoinc_candidates): Likewise.
16060 (get_address_cost): Likewise.
16061 (split_small_address_groups_p): Use address_p.
16062 (split_address_groups): Likewise.
16063 (add_iv_candidate_for_use): Likewise.
16064 (autoinc_possible_for_pair): Likewise.
16065 (rewrite_groups): Likewise.
16066 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16067 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16068 (get_alias_ptr_type_for_ptr_address): New function.
16069 (rewrite_use_address): Rewrite address uses in calls that were
16070 identified by find_address_like_use.
16072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16073 Alan Hayward <alan.hayward@arm.com>
16074 David Sherwood <david.sherwood@arm.com>
16076 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16078 * gimple-expr.h (is_gimple_addressable: Likewise.
16079 * gimple-expr.c (is_gimple_address): Likewise.
16080 * internal-fn.c (expand_call_mem_ref): New function.
16081 (expand_mask_load_optab_fn): Use it.
16082 (expand_mask_store_optab_fn): Likewise.
16084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16085 Alan Hayward <alan.hayward@arm.com>
16086 David Sherwood <david.sherwood@arm.com>
16088 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16089 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16090 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16091 (cond_umax@var{mode}): Document.
16092 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16093 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16094 (cond_umin_optab, cond_umax_optab): New optabs.
16095 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16096 (COND_IOR, COND_XOR): New internal functions.
16097 * internal-fn.h (get_conditional_internal_fn): Declare.
16098 * internal-fn.c (cond_binary_direct): New macro.
16099 (expand_cond_binary_optab_fn): Likewise.
16100 (direct_cond_binary_optab_supported_p): Likewise.
16101 (get_conditional_internal_fn): New function.
16102 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16103 Cope with reduction statements that are vectorized as calls rather
16105 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16106 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16107 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16108 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16109 (UNSPEC_COND_EOR): New unspecs.
16110 (optab): Add mappings for them.
16111 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16112 (sve_int_op, sve_fp_op): New int attributes.
16114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16115 Alan Hayward <alan.hayward@arm.com>
16116 David Sherwood <david.sherwood@arm.com>
16118 * optabs.def (while_ult_optab): New optab.
16119 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16120 * internal-fn.def (WHILE_ULT): New internal function.
16121 * internal-fn.h (direct_internal_fn_supported_p): New override
16122 that takes two types as argument.
16123 * internal-fn.c (while_direct): New macro.
16124 (expand_while_optab_fn): New function.
16125 (convert_optab_supported_p): Likewise.
16126 (direct_while_optab_supported_p): New macro.
16127 * wide-int.h (wi::udiv_ceil): New function.
16128 * tree-vectorizer.h (rgroup_masks): New structure.
16129 (vec_loop_masks): New typedef.
16130 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16131 and fully_masked_p.
16132 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16133 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16134 (vect_max_vf): New function.
16135 (slpeel_make_loop_iterate_ntimes): Delete.
16136 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16137 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16138 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16139 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16140 internal-fn.h, stor-layout.h and optabs-query.h.
16141 (vect_set_loop_mask): New function.
16142 (add_preheader_seq): Likewise.
16143 (add_header_seq): Likewise.
16144 (interleave_supported_p): Likewise.
16145 (vect_maybe_permute_loop_masks): Likewise.
16146 (vect_set_loop_masks_directly): Likewise.
16147 (vect_set_loop_condition_masked): Likewise.
16148 (vect_set_loop_condition_unmasked): New function, split out from
16149 slpeel_make_loop_iterate_ntimes.
16150 (slpeel_make_loop_iterate_ntimes): Rename to..
16151 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16152 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16153 (vect_do_peeling): Update call accordingly.
16154 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16156 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16157 mask_compare_type, can_fully_mask_p and fully_masked_p.
16158 (release_vec_loop_masks): New function.
16159 (_loop_vec_info): Use it to free the loop masks.
16160 (can_produce_all_loop_masks_p): New function.
16161 (vect_get_max_nscalars_per_iter): Likewise.
16162 (vect_verify_full_masking): Likewise.
16163 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16164 retries, and free the mask rgroups before retrying. Check loop-wide
16165 reasons for disallowing fully-masked loops. Make the final decision
16166 about whether use a fully-masked loop or not.
16167 (vect_estimate_min_profitable_iters): Do not assume that peeling
16168 for the number of iterations will be needed for fully-masked loops.
16169 (vectorizable_reduction): Disable fully-masked loops.
16170 (vectorizable_live_operation): Likewise.
16171 (vect_halve_mask_nunits): New function.
16172 (vect_double_mask_nunits): Likewise.
16173 (vect_record_loop_mask): Likewise.
16174 (vect_get_loop_mask): Likewise.
16175 (vect_transform_loop): Handle the case in which the final loop
16176 iteration might handle a partial vector. Call vect_set_loop_condition
16177 instead of slpeel_make_loop_iterate_ntimes.
16178 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16179 (check_load_store_masking): New function.
16180 (prepare_load_store_mask): Likewise.
16181 (vectorizable_store): Handle fully-masked loops.
16182 (vectorizable_load): Likewise.
16183 (supportable_widening_operation): Use vect_halve_mask_nunits for
16185 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16186 (vect_gen_while): New function.
16187 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16188 (aarch64_uqdec<mode>): New insn.
16190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16191 Alan Hayward <alan.hayward@arm.com>
16192 David Sherwood <david.sherwood@arm.com>
16194 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16195 (reduc_xor_scal_optab): New optabs.
16196 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16197 (reduc_xor_scal_@var{m}): Document.
16198 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16199 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16200 internal functions.
16201 * fold-const-call.c (fold_const_call): Handle them.
16202 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16203 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16204 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16205 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16206 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16207 (UNSPEC_XORV): New unspecs.
16208 (optab): Add entries for them.
16209 (BITWISEV): New int iterator.
16210 (bit_reduc_op): New int attributes.
16212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16213 Alan Hayward <alan.hayward@arm.com>
16214 David Sherwood <david.sherwood@arm.com>
16216 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16217 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16218 * optabs.def (vec_shl_insert_optab): New optab.
16219 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16220 (duplicate_and_interleave): Likewise.
16221 * tree-vect-loop.c: Include internal-fn.h.
16222 (neutral_op_for_slp_reduction): New function, split out from
16223 get_initial_defs_for_reduction.
16224 (get_initial_def_for_reduction): Handle option 2 for variable-length
16225 vectors by loading the neutral value into a vector and then shifting
16226 the initial value into element 0.
16227 (get_initial_defs_for_reduction): Replace the code argument with
16228 the neutral value calculated by neutral_op_for_slp_reduction.
16229 Use gimple_build_vector for constant-length vectors.
16230 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16231 but the first group_size elements have a neutral value.
16232 Use duplicate_and_interleave otherwise.
16233 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16234 Update call to get_initial_defs_for_reduction. Handle SLP
16235 reductions for variable-length vectors by creating one vector
16236 result for each scalar result, with the elements associated
16237 with other scalar results stubbed out with the neutral value.
16238 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16239 Require IFN_VEC_SHL_INSERT for double reductions on
16240 variable-length vectors, or SLP reductions that have
16241 a neutral value. Require can_duplicate_and_interleave_p
16242 support for variable-length unchained SLP reductions if there
16243 is no neutral value, such as for MIN/MAX reductions. Also require
16244 the number of vector elements to be a multiple of the number of
16245 SLP statements when doing variable-length unchained SLP reductions.
16246 Update call to vect_create_epilog_for_reduction.
16247 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16248 and remove initial values.
16249 (duplicate_and_interleave): Make public.
16250 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16251 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16254 Alan Hayward <alan.hayward@arm.com>
16255 David Sherwood <david.sherwood@arm.com>
16257 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16258 (can_duplicate_and_interleave_p): New function.
16259 (vect_get_and_check_slp_defs): Take the vector of statements
16260 rather than just the current one. Remove excess parentheses.
16261 Restriction rejectinon of vect_constant_def and vect_external_def
16262 for variable-length vectors to boolean types, or types for which
16263 can_duplicate_and_interleave_p is false.
16264 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16265 (duplicate_and_interleave): New function.
16266 (vect_get_constant_vectors): Use gimple_build_vector for
16267 constant-length vectors and suitable variable-length constant
16268 vectors. Use duplicate_and_interleave for other variable-length
16269 vectors. Don't defer the update when inserting new statements.
16271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16272 Alan Hayward <alan.hayward@arm.com>
16273 David Sherwood <david.sherwood@arm.com>
16275 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16276 min_profitable_iters doesn't go negative.
16278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16279 Alan Hayward <alan.hayward@arm.com>
16280 David Sherwood <david.sherwood@arm.com>
16282 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16283 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16284 * optabs.def (vec_mask_load_lanes_optab): New optab.
16285 (vec_mask_store_lanes_optab): Likewise.
16286 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16287 (MASK_STORE_LANES): Likewise.
16288 * internal-fn.c (mask_load_lanes_direct): New macro.
16289 (mask_store_lanes_direct): Likewise.
16290 (expand_mask_load_optab_fn): Handle masked operations.
16291 (expand_mask_load_lanes_optab_fn): New macro.
16292 (expand_mask_store_optab_fn): Handle masked operations.
16293 (expand_mask_store_lanes_optab_fn): New macro.
16294 (direct_mask_load_lanes_optab_supported_p): Likewise.
16295 (direct_mask_store_lanes_optab_supported_p): Likewise.
16296 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16298 (vect_load_lanes_supported): Likewise.
16299 * tree-vect-data-refs.c (strip_conversion): New function.
16300 (can_group_stmts_p): Likewise.
16301 (vect_analyze_data_ref_accesses): Use it instead of checking
16302 for a pair of assignments.
16303 (vect_store_lanes_supported): Take a masked_p parameter.
16304 (vect_load_lanes_supported): Likewise.
16305 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16306 vect_store_lanes_supported and vect_load_lanes_supported.
16307 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16308 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16309 parameter. Don't allow gaps for masked accesses.
16310 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16311 and vect_load_lanes_supported.
16312 (get_load_store_type): Take a masked_p parameter and update
16313 call to get_group_load_store_type.
16314 (vectorizable_store): Update call to get_load_store_type.
16315 Handle IFN_MASK_STORE_LANES.
16316 (vectorizable_load): Update call to get_load_store_type.
16317 Handle IFN_MASK_LOAD_LANES.
16319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16320 Alan Hayward <alan.hayward@arm.com>
16321 David Sherwood <david.sherwood@arm.com>
16323 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16325 * config/aarch64/aarch64-protos.h
16326 (aarch64_sve_struct_memory_operand_p): Declare.
16327 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16328 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16329 (VPRED, vpred): Handle SVE structure modes.
16330 * config/aarch64/constraints.md (Utx): New constraint.
16331 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16332 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16333 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16334 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16335 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16336 structure modes. Split into pieces after RA.
16337 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16338 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16340 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16341 SVE structure modes.
16342 (aarch64_classify_address): Likewise.
16343 (sizetochar): Move earlier in file.
16344 (aarch64_print_operand): Handle SVE register lists.
16345 (aarch64_array_mode): New function.
16346 (aarch64_sve_struct_memory_operand_p): Likewise.
16347 (TARGET_ARRAY_MODE): Redefine.
16349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16350 Alan Hayward <alan.hayward@arm.com>
16351 David Sherwood <david.sherwood@arm.com>
16353 * target.def (array_mode): New target hook.
16354 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16355 * doc/tm.texi: Regenerate.
16356 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16357 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16358 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16359 targetm.array_mode.
16360 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16364 Alan Hayward <alan.hayward@arm.com>
16365 David Sherwood <david.sherwood@arm.com>
16367 * fold-const.c (fold_binary_loc): Check the argument types
16368 rather than the result type when testing for a vector operation.
16370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16372 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16373 * doc/tm.texi: Regenerate.
16375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16376 Alan Hayward <alan.hayward@arm.com>
16377 David Sherwood <david.sherwood@arm.com>
16379 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16380 (sve): Document new AArch64 extension.
16381 * doc/md.texi (w): Extend the description of the AArch64
16382 constraint to include SVE vectors.
16383 (Upl, Upa): Document new AArch64 predicate constraints.
16384 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16386 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16387 (msve-vector-bits=): New option.
16388 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16389 SVE when these are disabled.
16390 (sve): New extension.
16391 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16392 modes. Adjust their number of units based on aarch64_sve_vg.
16393 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16394 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16395 aarch64_addr_query_type.
16396 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16397 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16398 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16399 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16400 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16401 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16402 (aarch64_simd_imm_zero_p): Delete.
16403 (aarch64_check_zero_based_sve_index_immediate): Declare.
16404 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16405 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16406 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16407 (aarch64_sve_float_mul_immediate_p): Likewise.
16408 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16409 rather than an rtx.
16410 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16411 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16412 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16413 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16414 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16415 (aarch64_regmode_natural_size): Likewise.
16416 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16417 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16419 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16420 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16421 for VG and the SVE predicate registers.
16422 (V_ALIASES): Add a "z"-prefixed alias.
16423 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16424 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16425 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16426 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16427 (REG_CLASS_NAMES): Add entries for them.
16428 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16429 and the predicate registers.
16430 (aarch64_sve_vg): Declare.
16431 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16432 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16433 (REGMODE_NATURAL_SIZE): Define.
16434 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16436 * config/aarch64/aarch64.c: Include cfgrtl.h.
16437 (simd_immediate_info): Add a constructor for series vectors,
16438 and an associated step field.
16439 (aarch64_sve_vg): New variable.
16440 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16441 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16442 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16443 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16444 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16445 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16446 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16447 (aarch64_get_mask_mode): New functions.
16448 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16449 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16450 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16451 predicate modes and predicate registers. Explicitly restrict
16452 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16453 to store a vector mode if it is recognized by
16454 aarch64_classify_vector_mode.
16455 (aarch64_regmode_natural_size): New function.
16456 (aarch64_hard_regno_caller_save_mode): Return the original mode
16458 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16459 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16460 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16461 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16463 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16464 does not overlap dest if the function is frame-related. Handle
16466 (aarch64_split_add_offset): New function.
16467 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16468 them aarch64_add_offset.
16469 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16470 and update call to aarch64_sub_sp.
16471 (aarch64_add_cfa_expression): New function.
16472 (aarch64_expand_prologue): Pass extra temporary registers to the
16473 functions above. Handle the case in which we need to emit new
16474 DW_CFA_expressions for registers that were originally saved
16475 relative to the stack pointer, but now have to be expressed
16476 relative to the frame pointer.
16477 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16479 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16480 IP0 and IP1 values for SVE frames.
16481 (aarch64_expand_vec_series): New function.
16482 (aarch64_expand_sve_widened_duplicate): Likewise.
16483 (aarch64_expand_sve_const_vector): Likewise.
16484 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16485 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16486 into the register, rather than emitting a SET directly.
16487 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16488 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16489 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16490 (offset_9bit_signed_scaled_p): New functions.
16491 (aarch64_replicate_bitmask_imm): New function.
16492 (aarch64_bitmask_imm): Use it.
16493 (aarch64_cannot_force_const_mem): Reject expressions involving
16494 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16495 (aarch64_classify_index): Handle SVE indices, by requiring
16496 a plain register index with a scale that matches the element size.
16497 (aarch64_classify_address): Handle SVE addresses. Assert that
16498 the mode of the address is VOIDmode or an integer mode.
16499 Update call to aarch64_classify_symbol.
16500 (aarch64_classify_symbolic_expression): Update call to
16501 aarch64_classify_symbol.
16502 (aarch64_const_vec_all_in_range_p): New function.
16503 (aarch64_print_vector_float_operand): Likewise.
16504 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16505 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16506 and the FP immediates 1.0 and 0.5.
16507 (aarch64_print_address_internal): Handle SVE addresses.
16508 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16509 (aarch64_regno_regclass): Handle predicate registers.
16510 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16512 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16513 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16514 (aarch64_convert_sve_vector_bits): New function.
16515 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16516 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16517 rather than an rtx.
16518 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16519 Handle SVE vector and predicate modes. Accept VL-based constants
16520 that need only one temporary register, and VL offsets that require
16521 no temporary registers.
16522 (aarch64_conditional_register_usage): Mark the predicate registers
16523 as fixed if SVE isn't available.
16524 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16525 Return true for SVE vector and predicate modes.
16526 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16527 rather than an unsigned int. Handle SVE modes.
16528 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16530 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16532 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16533 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16534 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16535 (aarch64_sve_float_mul_immediate_p): New functions.
16536 (aarch64_sve_valid_immediate): New function.
16537 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16538 Explicitly reject structure modes. Check for INDEX constants.
16539 Handle PTRUE and PFALSE constants.
16540 (aarch64_check_zero_based_sve_index_immediate): New function.
16541 (aarch64_simd_imm_zero_p): Delete.
16542 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16543 vector modes. Accept constants in the range of CNT[BHWD].
16544 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16545 ask for an Advanced SIMD mode.
16546 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16547 (aarch64_simd_vector_alignment): Handle SVE predicates.
16548 (aarch64_vectorize_preferred_vector_alignment): New function.
16549 (aarch64_simd_vector_alignment_reachable): Use it instead of
16551 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16552 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16554 (MAX_VECT_LEN): Delete.
16555 (expand_vec_perm_d): Add a vec_flags field.
16556 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16557 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16558 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16560 (aarch64_evpc_rev): Rename to...
16561 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16562 (aarch64_evpc_rev_global): New function.
16563 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16564 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16566 (aarch64_evpc_sve_tbl): New function.
16567 (aarch64_expand_vec_perm_const_1): Update after rename of
16568 aarch64_evpc_rev. Handle SVE permutes too, trying
16569 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16570 than aarch64_evpc_tbl.
16571 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16572 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16573 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16574 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16575 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16576 (aarch64_expand_sve_vcond): New functions.
16577 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16578 of aarch64_vector_mode_p.
16579 (aarch64_dwarf_poly_indeterminate_value): New function.
16580 (aarch64_compute_pressure_classes): Likewise.
16581 (aarch64_can_change_mode_class): Likewise.
16582 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16583 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16584 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16585 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16586 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16587 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16588 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16589 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16591 (Dn, Dl, Dr): Accept const as well as const_vector.
16592 (Dz): Likewise. Compare against CONST0_RTX.
16593 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16594 of "vector" where appropriate.
16595 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16596 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16597 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16598 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16599 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16600 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16601 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16602 (v_int_equiv): Extend to SVE modes.
16603 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16605 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16606 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16607 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16608 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16609 (SVE_COND_FP_CMP): New int iterators.
16610 (perm_hilo): Handle the new unpack unspecs.
16611 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16613 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16614 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16615 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16616 (aarch64_equality_operator, aarch64_constant_vector_operand)
16617 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16618 (aarch64_sve_nonimmediate_operand): Likewise.
16619 (aarch64_sve_general_operand): Likewise.
16620 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16621 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16622 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16623 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16624 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16625 (aarch64_sve_float_arith_immediate): Likewise.
16626 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16627 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16628 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16629 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16630 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16631 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16632 (aarch64_sve_float_arith_operand): Likewise.
16633 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16634 (aarch64_sve_float_mul_operand): Likewise.
16635 (aarch64_sve_vec_perm_operand): Likewise.
16636 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16637 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16638 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16639 as well as const_vector.
16640 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16641 in file. Use CONST0_RTX and CONSTM1_RTX.
16642 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16643 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16644 Use aarch64_simd_imm_zero.
16645 * config/aarch64/aarch64-sve.md: New file.
16646 * config/aarch64/aarch64.md: Include it.
16647 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16648 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16649 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16650 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16651 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16652 (sve): New attribute.
16653 (enabled): Disable instructions with the sve attribute unless
16655 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16656 aarch64_expand_mov_immediate.
16657 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16658 CNT[BHSD] immediates.
16659 (movti): Split CONST_POLY_INT moves into two halves.
16660 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16661 Split additions that need a temporary here if the destination
16662 is the stack pointer.
16663 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16664 (*add<mode>3_poly_1): New instruction.
16665 (set_clobber_cc): New expander.
16667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16669 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16670 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16671 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16672 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16673 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16674 Change innermode from fixed_mode_size to machine_mode.
16675 (simplify_subreg): Update call accordingly. Handle a constant-sized
16676 subreg of a variable-length CONST_VECTOR.
16678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16679 Alan Hayward <alan.hayward@arm.com>
16680 David Sherwood <david.sherwood@arm.com>
16682 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16683 (add_offset_to_base): New function, split out from...
16684 (create_mem_ref): ...here. When handling a scale other than 1,
16685 check first whether the address is valid without the offset.
16686 Add it into the base if so, leaving the index and scale as-is.
16688 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16691 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16692 fold_for_warn before checking if arg2 is INTEGER_CST.
16694 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16696 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16697 (store_multiple_operation): Delete.
16698 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16699 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16700 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16701 guarded by TARGET_STRING.
16702 (rs6000_output_load_multiple): Delete.
16703 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16704 OPTION_MASK_STRING / TARGET_STRING handling.
16705 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16706 (const rs6000_opt_masks) <"string">: Change mask to 0.
16707 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16708 (MASK_STRING): Delete.
16709 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16711 (load_multiple): Delete.
16718 (store_multiple): Delete.
16725 (movmemsi_8reg): Delete.
16726 (corresponding unnamed define_insn): Delete.
16727 (movmemsi_6reg): Delete.
16728 (corresponding unnamed define_insn): Delete.
16729 (movmemsi_4reg): Delete.
16730 (corresponding unnamed define_insn): Delete.
16731 (movmemsi_2reg): Delete.
16732 (corresponding unnamed define_insn): Delete.
16733 (movmemsi_1reg): Delete.
16734 (corresponding unnamed define_insn): Delete.
16735 * config/rs6000/rs6000.opt (mno-string): New.
16736 (mstring): Replace by deprecation warning stub.
16737 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16739 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16741 * regrename.c (regrename_do_replace): If replacing the same
16742 reg multiple times, try to reuse last created gen_raw_REG.
16745 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16746 main to workaround a bug in GDB.
16748 2018-01-12 Tom de Vries <tom@codesourcery.com>
16751 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16753 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16755 PR rtl-optimization/80481
16756 * ira-color.c (get_cap_member): New function.
16757 (allocnos_conflict_by_live_ranges_p): Use it.
16758 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16759 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16761 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16764 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16765 (*saddl_se_1): Ditto.
16766 (*ssubsi_1): Ditto.
16767 (*ssubl_se_1): Ditto.
16769 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16771 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16772 rather than wi::to_widest for DR_INITs.
16773 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16774 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16775 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16777 (vect_analyze_group_access_1): Note that here.
16779 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16781 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16782 polynomial type sizes.
16784 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16786 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16787 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16788 (gimple_add_tmp_var): Likewise.
16790 2018-01-12 Martin Liska <mliska@suse.cz>
16792 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16793 (gimple_alloc_sizes): Likewise.
16794 (dump_gimple_statistics): Use PRIu64 in printf format.
16795 * gimple.h: Change uint64_t to int.
16797 2018-01-12 Martin Liska <mliska@suse.cz>
16799 * tree-core.h: Use uint64_t instead of int.
16800 * tree.c (tree_node_counts): Likewise.
16801 (tree_node_sizes): Likewise.
16802 (dump_tree_statistics): Use PRIu64 in printf format.
16804 2018-01-12 Martin Liska <mliska@suse.cz>
16806 * Makefile.in: As qsort_chk is implemented in vec.c, add
16807 vec.o to linkage of gencfn-macros.
16808 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16809 passing the info to record_node_allocation_statistics.
16810 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16812 * ggc-common.c (struct ggc_usage): Add operator== and use
16813 it in operator< and compare function.
16814 * mem-stats.h (struct mem_usage): Likewise.
16815 * vec.c (struct vec_usage): Remove operator< and compare
16816 function. Can be simply inherited.
16818 2018-01-12 Martin Jambor <mjambor@suse.cz>
16821 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16822 * tree-ssa-math-opts.c: Include domwalk.h.
16823 (convert_mult_to_fma_1): New function.
16824 (fma_transformation_info): New type.
16825 (fma_deferring_state): Likewise.
16826 (cancel_fma_deferring): New function.
16827 (result_of_phi): Likewise.
16828 (last_fma_candidate_feeds_initial_phi): Likewise.
16829 (convert_mult_to_fma): Added deferring logic, split actual
16830 transformation to convert_mult_to_fma_1.
16831 (math_opts_dom_walker): New type.
16832 (math_opts_dom_walker::after_dom_children): New method, body moved
16833 here from pass_optimize_widening_mul::execute, added deferring logic
16835 (pass_optimize_widening_mul::execute): Moved most of code to
16836 math_opts_dom_walker::after_dom_children.
16837 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16838 * config/i386/i386.c (ix86_option_override_internal): Added
16839 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16841 2018-01-12 Richard Biener <rguenther@suse.de>
16844 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16845 inline instance vars.
16847 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16850 * config/rx/rx.c (rx_is_restricted_memory_address):
16851 Handle SUBREG case.
16853 2018-01-12 Richard Biener <rguenther@suse.de>
16855 PR tree-optimization/80846
16856 * target.def (split_reduction): New target hook.
16857 * targhooks.c (default_split_reduction): New function.
16858 * targhooks.h (default_split_reduction): Declare.
16859 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16860 target requests first reduce vectors by combining low and high
16862 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16863 (get_vectype_for_scalar_type_and_size): Export.
16864 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16865 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16866 * doc/tm.texi: Regenerate.
16867 * config/i386/i386.c (ix86_split_reduction): Implement
16868 TARGET_VECTORIZE_SPLIT_REDUCTION.
16870 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16873 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16874 in PIC mode except for TARGET_VXWORKS_RTP.
16875 * config/sparc/sparc.c: Include cfgrtl.h.
16876 (TARGET_INIT_PIC_REG): Define.
16877 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16878 (sparc_pic_register_p): New predicate.
16879 (sparc_legitimate_address_p): Use it.
16880 (sparc_legitimize_pic_address): Likewise.
16881 (sparc_delegitimize_address): Likewise.
16882 (sparc_mode_dependent_address_p): Likewise.
16883 (gen_load_pcrel_sym): Remove 4th parameter.
16884 (load_got_register): Adjust call to above. Remove obsolete stuff.
16885 (sparc_expand_prologue): Do not call load_got_register here.
16886 (sparc_flat_expand_prologue): Likewise.
16887 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16888 (sparc_use_pseudo_pic_reg): New function.
16889 (sparc_init_pic_reg): Likewise.
16890 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16891 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16893 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16895 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16896 Add item for branch_cost.
16898 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16900 PR rtl-optimization/83565
16901 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16902 not extend the result to a larger mode for rotate operations.
16903 (num_sign_bit_copies1): Likewise.
16905 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16908 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16910 Use values-Xc.o for -pedantic.
16911 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16913 2018-01-12 Martin Liska <mliska@suse.cz>
16916 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16918 (possible_polymorphic_call_targets): Use it.
16919 (ipa_devirt): Likewise.
16921 2018-01-12 Martin Liska <mliska@suse.cz>
16923 * profile-count.h (enum profile_quality): Use 0 as invalid
16924 enum value of profile_quality.
16926 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16928 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16929 -mext-string options.
16931 2018-01-12 Richard Biener <rguenther@suse.de>
16933 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16934 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16935 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16937 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16939 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16941 * configure.ac (--with-long-double-format): Add support for the
16942 configuration option to change the default long double format on
16944 * config.gcc (powerpc*-linux*-*): Likewise.
16945 * configure: Regenerate.
16946 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16947 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16948 used without modification.
16950 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16952 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16953 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16954 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16955 MISC_BUILTIN_SPEC_BARRIER.
16956 (rs6000_init_builtins): Likewise.
16957 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16959 (speculation_barrier): New define_insn.
16960 * doc/extend.texi: Document __builtin_speculation_barrier.
16962 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16965 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16966 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16967 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16969 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16970 integral modes instead of "ss" and "sd".
16971 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16972 vectors with 32-bit and 64-bit elements.
16973 (vecdupssescalarmodesuffix): New mode attribute.
16974 (vec_dup<mode>): Use it.
16976 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16979 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16980 frame if argument is passed on stack.
16982 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16985 * ree.c (combine_reaching_defs): Optimize also
16986 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16987 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16989 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16991 PR middle-end/83189
16992 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16994 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16996 PR middle-end/83718
16997 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16998 after they are computed.
17000 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17002 PR tree-optimization/83695
17003 * gimple-loop-linterchange.cc
17004 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17005 reset cached scev information after interchange.
17006 (pass_linterchange::execute): Remove call to scev_reset_htab.
17008 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17010 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17011 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17012 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17013 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17014 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17015 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17016 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17017 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17018 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17019 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17020 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17021 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17022 (V_lane_reg): Likewise.
17023 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17025 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17026 (vfmal_lane_low<mode>_intrinsic,
17027 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17028 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17029 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17030 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17031 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17032 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17034 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17036 * config/arm/arm-cpus.in (fp16fml): New feature.
17037 (ALL_SIMD): Add fp16fml.
17038 (armv8.2-a): Add fp16fml as an option.
17039 (armv8.3-a): Likewise.
17040 (armv8.4-a): Add fp16fml as part of fp16.
17041 * config/arm/arm.h (TARGET_FP16FML): Define.
17042 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17044 * config/arm/arm-modes.def (V2HF): Define.
17045 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17046 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17047 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17048 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17049 vfmsl_low, vfmsl_high): New set of builtins.
17050 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17051 (vfml_op): New code attribute.
17052 (VFMLHALVES): New int iterator.
17053 (VFML, VFMLSEL): New mode attributes.
17054 (V_reg): Define mapping for V2HF.
17055 (V_hi, V_lo): New mode attributes.
17056 (VF_constraint): Likewise.
17057 (vfml_half, vfml_half_selector): New int attributes.
17058 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17060 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17061 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17063 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17064 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17065 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17066 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17068 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17069 Document new effective target and option set.
17071 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17073 * config/arm/arm-cpus.in (armv8_4): New feature.
17074 (ARMv8_4a): New fgroup.
17075 (armv8.4-a): New arch.
17076 * config/arm/arm-tables.opt: Regenerate.
17077 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17078 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17079 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17080 Add matching rules for -march=armv8.4-a and extensions.
17081 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17083 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17086 * config/rx/rx.md (BW): New mode attribute.
17087 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17089 2018-01-11 Richard Biener <rguenther@suse.de>
17091 PR tree-optimization/83435
17092 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17093 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17094 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17096 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17097 Alan Hayward <alan.hayward@arm.com>
17098 David Sherwood <david.sherwood@arm.com>
17100 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17102 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17103 (aarch64_print_address_internal): Use it to check for a zero offset.
17105 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17106 Alan Hayward <alan.hayward@arm.com>
17107 David Sherwood <david.sherwood@arm.com>
17109 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17110 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17111 Return a poly_int64 rather than a HOST_WIDE_INT.
17112 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17113 rather than a HOST_WIDE_INT.
17114 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17115 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17116 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17117 final_offset from HOST_WIDE_INT to poly_int64.
17118 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17119 to_constant when getting the number of units in an Advanced SIMD
17121 (aarch64_builtin_vectorized_function): Check for a constant number
17123 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17125 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17126 attribute instead of GET_MODE_NUNITS.
17127 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17128 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17129 GET_MODE_SIZE for fixed-size registers.
17130 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17131 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17132 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17133 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17134 (aarch64_print_operand, aarch64_print_address_internal)
17135 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17136 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17137 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17138 Handle polynomial GET_MODE_SIZE.
17139 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17140 wider than SImode without modification.
17141 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17142 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17143 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17144 passing and returning SVE modes.
17145 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17146 rather than GEN_INT.
17147 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17148 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17149 (aarch64_allocate_and_probe_stack_space): Likewise.
17150 (aarch64_layout_frame): Cope with polynomial offsets.
17151 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17152 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17153 polynomial offsets.
17154 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17155 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17156 poly_int64 rather than a HOST_WIDE_INT.
17157 (aarch64_get_separate_components, aarch64_process_components)
17158 (aarch64_expand_prologue, aarch64_expand_epilogue)
17159 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17160 (aarch64_anchor_offset): New function, split out from...
17161 (aarch64_legitimize_address): ...here.
17162 (aarch64_builtin_vectorization_cost): Handle polynomial
17163 TYPE_VECTOR_SUBPARTS.
17164 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17166 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17167 number of elements from the PARALLEL rather than the mode.
17168 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17169 rather than GET_MODE_BITSIZE.
17170 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17171 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17172 (aarch64_expand_vec_perm_const_1): Handle polynomial
17173 d->perm.length () and d->perm elements.
17174 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17175 Apply to_constant to d->perm elements.
17176 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17177 polynomial CONST_VECTOR_NUNITS.
17178 (aarch64_move_pointer): Take amount as a poly_int64 rather
17180 (aarch64_progress_pointer): Avoid temporary variable.
17181 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17182 the mode attribute instead of GET_MODE.
17184 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17185 Alan Hayward <alan.hayward@arm.com>
17186 David Sherwood <david.sherwood@arm.com>
17188 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17189 x exists before using it.
17190 (aarch64_add_constant_internal): Rename to...
17191 (aarch64_add_offset_1): ...this. Replace regnum with separate
17192 src and dest rtxes. Handle the case in which they're different,
17193 including when the offset is zero. Replace scratchreg with an rtx.
17194 Use 2 additions if there is no spare register into which we can
17195 move a 16-bit constant.
17196 (aarch64_add_constant): Delete.
17197 (aarch64_add_offset): Replace reg with separate src and dest
17198 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17199 Use aarch64_add_offset_1.
17200 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17201 an rtx rather than an int. Take the delta as a poly_int64
17202 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17203 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17204 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17205 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17206 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17207 and aarch64_add_sp.
17208 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17209 aarch64_add_constant.
17211 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17213 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17214 Use scalar_float_mode.
17216 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17218 * config/aarch64/aarch64-simd.md
17219 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17220 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17221 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17222 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17223 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17224 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17225 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17226 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17227 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17228 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17230 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17233 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17234 targ_options->x_arm_arch_string is non NULL.
17236 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17238 * config/aarch64/aarch64.h
17239 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17241 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17244 * expmed.c (emit_store_flag_force): Swap if const op0
17245 and change VOIDmode to mode of op0.
17247 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17249 PR rtl-optimization/83761
17250 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17251 than bytes to mode_for_size.
17253 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17255 PR middle-end/83189
17256 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17257 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17260 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17262 PR middle-end/83575
17263 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17264 when in layout mode.
17265 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17266 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17269 2018-01-10 Michael Collison <michael.collison@arm.com>
17271 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17272 * config/aarch64/aarch64-option-extension.def: Add
17273 AARCH64_OPT_EXTENSION of 'fp16fml'.
17274 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17275 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17276 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17277 * config/aarch64/constraints.md (Ui7): New constraint.
17278 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17279 (VFMLA_SEL_W): Ditto.
17282 (VFMLA16_LOW): New int iterator.
17283 (VFMLA16_HIGH): Ditto.
17284 (UNSPEC_FMLAL): New unspec.
17285 (UNSPEC_FMLSL): Ditto.
17286 (UNSPEC_FMLAL2): Ditto.
17287 (UNSPEC_FMLSL2): Ditto.
17288 (f16mac): New code attribute.
17289 * config/aarch64/aarch64-simd-builtins.def
17290 (aarch64_fmlal_lowv2sf): Ditto.
17291 (aarch64_fmlsl_lowv2sf): Ditto.
17292 (aarch64_fmlalq_lowv4sf): Ditto.
17293 (aarch64_fmlslq_lowv4sf): Ditto.
17294 (aarch64_fmlal_highv2sf): Ditto.
17295 (aarch64_fmlsl_highv2sf): Ditto.
17296 (aarch64_fmlalq_highv4sf): Ditto.
17297 (aarch64_fmlslq_highv4sf): Ditto.
17298 (aarch64_fmlal_lane_lowv2sf): Ditto.
17299 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17300 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17301 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17302 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17303 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17304 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17305 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17306 (aarch64_fmlal_lane_highv2sf): Ditto.
17307 (aarch64_fmlsl_lane_highv2sf): Ditto.
17308 (aarch64_fmlal_laneq_highv2sf): Ditto.
17309 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17310 (aarch64_fmlalq_lane_highv4sf): Ditto.
17311 (aarch64_fmlsl_lane_highv4sf): Ditto.
17312 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17313 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17314 * config/aarch64/aarch64-simd.md:
17315 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17316 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17317 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17318 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17319 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17320 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17321 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17322 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17323 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17324 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17325 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17326 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17327 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17328 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17329 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17330 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17331 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17332 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17333 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17334 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17335 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17336 (vfmlsl_low_u32): Ditto.
17337 (vfmlalq_low_u32): Ditto.
17338 (vfmlslq_low_u32): Ditto.
17339 (vfmlal_high_u32): Ditto.
17340 (vfmlsl_high_u32): Ditto.
17341 (vfmlalq_high_u32): Ditto.
17342 (vfmlslq_high_u32): Ditto.
17343 (vfmlal_lane_low_u32): Ditto.
17344 (vfmlsl_lane_low_u32): Ditto.
17345 (vfmlal_laneq_low_u32): Ditto.
17346 (vfmlsl_laneq_low_u32): Ditto.
17347 (vfmlalq_lane_low_u32): Ditto.
17348 (vfmlslq_lane_low_u32): Ditto.
17349 (vfmlalq_laneq_low_u32): Ditto.
17350 (vfmlslq_laneq_low_u32): Ditto.
17351 (vfmlal_lane_high_u32): Ditto.
17352 (vfmlsl_lane_high_u32): Ditto.
17353 (vfmlal_laneq_high_u32): Ditto.
17354 (vfmlsl_laneq_high_u32): Ditto.
17355 (vfmlalq_lane_high_u32): Ditto.
17356 (vfmlslq_lane_high_u32): Ditto.
17357 (vfmlalq_laneq_high_u32): Ditto.
17358 (vfmlslq_laneq_high_u32): Ditto.
17359 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17360 (AARCH64_FL_FOR_ARCH8_4): New.
17361 (AARCH64_ISA_F16FML): New ISA flag.
17362 (TARGET_F16FML): New feature flag for fp16fml.
17363 (doc/invoke.texi): Document new fp16fml option.
17365 2018-01-10 Michael Collison <michael.collison@arm.com>
17367 * config/aarch64/aarch64-builtins.c:
17368 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17369 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17370 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17371 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17372 (AARCH64_ISA_SHA3): New ISA flag.
17373 (TARGET_SHA3): New feature flag for sha3.
17374 * config/aarch64/iterators.md (sha512_op): New int attribute.
17375 (CRYPTO_SHA512): New int iterator.
17376 (UNSPEC_SHA512H): New unspec.
17377 (UNSPEC_SHA512H2): Ditto.
17378 (UNSPEC_SHA512SU0): Ditto.
17379 (UNSPEC_SHA512SU1): Ditto.
17380 * config/aarch64/aarch64-simd-builtins.def
17381 (aarch64_crypto_sha512hqv2di): New builtin.
17382 (aarch64_crypto_sha512h2qv2di): Ditto.
17383 (aarch64_crypto_sha512su0qv2di): Ditto.
17384 (aarch64_crypto_sha512su1qv2di): Ditto.
17385 (aarch64_eor3qv8hi): Ditto.
17386 (aarch64_rax1qv2di): Ditto.
17387 (aarch64_xarqv2di): Ditto.
17388 (aarch64_bcaxqv8hi): Ditto.
17389 * config/aarch64/aarch64-simd.md:
17390 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17391 (aarch64_crypto_sha512su0qv2di): Ditto.
17392 (aarch64_crypto_sha512su1qv2di): Ditto.
17393 (aarch64_eor3qv8hi): Ditto.
17394 (aarch64_rax1qv2di): Ditto.
17395 (aarch64_xarqv2di): Ditto.
17396 (aarch64_bcaxqv8hi): Ditto.
17397 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17398 (vsha512h2q_u64): Ditto.
17399 (vsha512su0q_u64): Ditto.
17400 (vsha512su1q_u64): Ditto.
17401 (veor3q_u16): Ditto.
17402 (vrax1q_u64): Ditto.
17403 (vxarq_u64): Ditto.
17404 (vbcaxq_u16): Ditto.
17405 * config/arm/types.md (crypto_sha512): New type attribute.
17406 (crypto_sha3): Ditto.
17407 (doc/invoke.texi): Document new sha3 option.
17409 2018-01-10 Michael Collison <michael.collison@arm.com>
17411 * config/aarch64/aarch64-builtins.c:
17412 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17413 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17414 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17415 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17416 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17417 (AARCH64_ISA_SM4): New ISA flag.
17418 (TARGET_SM4): New feature flag for sm4.
17419 * config/aarch64/aarch64-simd-builtins.def
17420 (aarch64_sm3ss1qv4si): Ditto.
17421 (aarch64_sm3tt1aq4si): Ditto.
17422 (aarch64_sm3tt1bq4si): Ditto.
17423 (aarch64_sm3tt2aq4si): Ditto.
17424 (aarch64_sm3tt2bq4si): Ditto.
17425 (aarch64_sm3partw1qv4si): Ditto.
17426 (aarch64_sm3partw2qv4si): Ditto.
17427 (aarch64_sm4eqv4si): Ditto.
17428 (aarch64_sm4ekeyqv4si): Ditto.
17429 * config/aarch64/aarch64-simd.md:
17430 (aarch64_sm3ss1qv4si): Ditto.
17431 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17432 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17433 (aarch64_sm4eqv4si): Ditto.
17434 (aarch64_sm4ekeyqv4si): Ditto.
17435 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17436 (sm3part_op): Ditto.
17437 (CRYPTO_SM3TT): Ditto.
17438 (CRYPTO_SM3PART): Ditto.
17439 (UNSPEC_SM3SS1): New unspec.
17440 (UNSPEC_SM3TT1A): Ditto.
17441 (UNSPEC_SM3TT1B): Ditto.
17442 (UNSPEC_SM3TT2A): Ditto.
17443 (UNSPEC_SM3TT2B): Ditto.
17444 (UNSPEC_SM3PARTW1): Ditto.
17445 (UNSPEC_SM3PARTW2): Ditto.
17446 (UNSPEC_SM4E): Ditto.
17447 (UNSPEC_SM4EKEY): Ditto.
17448 * config/aarch64/constraints.md (Ui2): New constraint.
17449 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17450 * config/arm/types.md (crypto_sm3): New type attribute.
17451 (crypto_sm4): Ditto.
17452 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17453 (vsm3tt1aq_u32): Ditto.
17454 (vsm3tt1bq_u32): Ditto.
17455 (vsm3tt2aq_u32): Ditto.
17456 (vsm3tt2bq_u32): Ditto.
17457 (vsm3partw1q_u32): Ditto.
17458 (vsm3partw2q_u32): Ditto.
17459 (vsm4eq_u32): Ditto.
17460 (vsm4ekeyq_u32): Ditto.
17461 (doc/invoke.texi): Document new sm4 option.
17463 2018-01-10 Michael Collison <michael.collison@arm.com>
17465 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17466 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17467 (AARCH64_FL_FOR_ARCH8_4): New.
17468 (AARCH64_FL_V8_4): New flag.
17469 (doc/invoke.texi): Document new armv8.4-a option.
17471 2018-01-10 Michael Collison <michael.collison@arm.com>
17473 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17474 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17475 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17476 * config/aarch64/aarch64-option-extension.def: Add
17477 AARCH64_OPT_EXTENSION of 'sha2'.
17478 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17479 (crypto): Disable sha2 and aes if crypto disabled.
17480 (crypto): Enable aes and sha2 if enabled.
17481 (simd): Disable sha2 and aes if simd disabled.
17482 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17484 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17485 (TARGET_SHA2): New feature flag for sha2.
17486 (TARGET_AES): New feature flag for aes.
17487 * config/aarch64/aarch64-simd.md:
17488 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17489 conditional on TARGET_AES.
17490 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17491 (aarch64_crypto_sha1hsi): Make pattern conditional
17493 (aarch64_crypto_sha1hv4si): Ditto.
17494 (aarch64_be_crypto_sha1hv4si): Ditto.
17495 (aarch64_crypto_sha1su1v4si): Ditto.
17496 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17497 (aarch64_crypto_sha1su0v4si): Ditto.
17498 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17499 (aarch64_crypto_sha256su0v4si): Ditto.
17500 (aarch64_crypto_sha256su1v4si): Ditto.
17501 (doc/invoke.texi): Document new aes and sha2 options.
17503 2018-01-10 Martin Sebor <msebor@redhat.com>
17505 PR tree-optimization/83781
17506 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17509 2018-01-11 Martin Sebor <msebor@gmail.com>
17510 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17512 PR tree-optimization/83501
17513 PR tree-optimization/81703
17515 * tree-ssa-strlen.c (get_string_cst): Rename...
17516 (get_string_len): ...to this. Handle global constants.
17517 (handle_char_store): Adjust.
17519 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17520 Jim Wilson <jimw@sifive.com>
17522 * config/riscv/riscv-protos.h (riscv_output_return): New.
17523 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17524 (riscv_attribute_table, riscv_output_return),
17525 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17526 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17527 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17529 (riscv_expand_prologue): Add early return for naked function.
17530 (riscv_expand_epilogue): Likewise.
17531 (riscv_function_ok_for_sibcall): Return false for naked function.
17532 (riscv_set_current_function): New.
17533 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17534 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17535 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17536 * doc/extend.texi (RISC-V Function Attributes): New.
17538 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17540 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17541 check for 128-bit long double before checking TCmode.
17542 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17543 128-bit long doubles before checking TFmode or TCmode.
17544 (FLOAT128_IBM_P): Likewise.
17546 2018-01-10 Martin Sebor <msebor@redhat.com>
17548 PR tree-optimization/83671
17549 * builtins.c (c_strlen): Unconditionally return zero for the empty
17551 Use -Warray-bounds for warnings.
17552 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17553 for non-constant array indices with COMPONENT_REF, arrays of
17554 arrays, and pointers to arrays.
17555 (gimple_fold_builtin_strlen): Determine and set length range for
17556 non-constant character arrays.
17558 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17560 PR middle-end/81897
17561 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17564 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17566 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17568 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17571 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17572 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17573 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17574 indexed_or_indirect_operand predicate.
17575 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17576 (*vsx_le_perm_load_v8hi): Likewise.
17577 (*vsx_le_perm_load_v16qi): Likewise.
17578 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17579 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17580 (*vsx_le_perm_store_v8hi): Likewise.
17581 (*vsx_le_perm_store_v16qi): Likewise.
17582 (eight unnamed splitters): Likewise.
17584 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17586 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17587 * config/rs6000/emmintrin.h: Likewise.
17588 * config/rs6000/mmintrin.h: Likewise.
17589 * config/rs6000/xmmintrin.h: Likewise.
17591 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17594 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17596 * tree.c (tree_nop_conversion): Return true for location wrapper
17598 (maybe_wrap_with_location): New function.
17599 (selftest::check_strip_nops): New function.
17600 (selftest::test_location_wrappers): New function.
17601 (selftest::tree_c_tests): Call it.
17602 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17603 (maybe_wrap_with_location): New decl.
17604 (EXPR_LOCATION_WRAPPER_P): New macro.
17605 (location_wrapper_p): New inline function.
17606 (tree_strip_any_location_wrapper): New inline function.
17608 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17611 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17612 stack_realign_offset for the largest alignment of stack slot
17614 (ix86_find_max_used_stack_alignment): New function.
17615 (ix86_finalize_stack_frame_flags): Use it. Set
17616 max_used_stack_alignment if we don't realign stack.
17617 * config/i386/i386.h (machine_function): Add
17618 max_used_stack_alignment.
17620 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17622 * config/arm/arm.opt (-mbranch-cost): New option.
17623 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17626 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17629 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17630 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17632 2018-01-10 Richard Biener <rguenther@suse.de>
17635 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17636 early out so it also covers the case where we have a non-NULL
17639 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17641 PR tree-optimization/83753
17642 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17643 for non-strided grouped accesses if the number of elements is 1.
17645 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17648 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17649 * i386.h (TARGET_USE_GATHER): Define.
17650 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17652 2018-01-10 Martin Liska <mliska@suse.cz>
17655 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17656 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17658 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17659 CLEANUP_NO_PARTITIONING is not set.
17661 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17663 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17664 for vectors, as a partial revert of r254296.
17665 * rtl.h (const_vec_p): Delete.
17666 (const_vec_duplicate_p): Don't test for vector CONSTs.
17667 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17668 * expmed.c (make_tree): Likewise.
17671 * common.md (E, F): Use CONSTANT_P instead of checking for
17673 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17674 checking for CONST_VECTOR.
17676 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17678 PR middle-end/83575
17679 * predict.c (force_edge_cold): Handle in more sane way edges
17680 with no prediction.
17682 2018-01-09 Carl Love <cel@us.ibm.com>
17684 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17686 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17687 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17688 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17689 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17690 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17691 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17692 * config/rs6000/rs6000-protos.h: Add extern defition for
17693 rs6000_generate_float2_double_code.
17694 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17696 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17697 (float2_v2df): Add define_expand.
17699 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17702 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17703 op_mode in the force_to_mode call.
17705 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17707 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17708 instead of checking each element individually.
17709 (aarch64_evpc_uzp): Likewise.
17710 (aarch64_evpc_zip): Likewise.
17711 (aarch64_evpc_ext): Likewise.
17712 (aarch64_evpc_rev): Likewise.
17713 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17714 instead of checking each element individually. Return true without
17716 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17717 whether all selected elements come from the same input, instead of
17718 checking each element individually. Remove calls to gen_rtx_REG,
17719 start_sequence and end_sequence and instead assert that no rtl is
17722 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17724 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17725 order of HIGH and CONST checks.
17727 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17729 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17730 if the destination isn't an SSA_NAME.
17732 2018-01-09 Richard Biener <rguenther@suse.de>
17734 PR tree-optimization/83668
17735 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17737 (canonicalize_loop_form): ... here, renamed from ...
17738 (canonicalize_loop_closed_ssa_form): ... this and amended to
17739 swap successor edges for loop exit blocks to make us use
17740 the RPO order we need for initial schedule generation.
17742 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17744 PR tree-optimization/64811
17745 * match.pd: When optimizing comparisons with Inf, avoid
17746 introducing or losing exceptions from comparisons with NaN.
17748 2018-01-09 Martin Liska <mliska@suse.cz>
17751 * asan.c (shadow_mem_size): Add gcc_assert.
17753 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17755 Don't save registers in main().
17758 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17759 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17760 * config/avr/avr.c (avr_set_current_function): Don't error if
17761 naked, OS_task or OS_main are specified at the same time.
17762 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17764 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17766 * common/config/avr/avr-common.c (avr_option_optimization_table):
17767 Switch on -mmain-is-OS_task for optimizing compilations.
17769 2018-01-09 Richard Biener <rguenther@suse.de>
17771 PR tree-optimization/83572
17772 * graphite.c: Include cfganal.h.
17773 (graphite_transform_loops): Connect infinite loops to exit
17774 and remove fake edges at the end.
17776 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17778 * ipa-inline.c (edge_badness): Revert accidental checkin.
17780 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17783 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17784 symbols; not inline clones.
17786 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17789 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17790 hard registers. Formatting fixes.
17792 PR preprocessor/83722
17793 * gcc.c (try_generate_repro): Pass
17794 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17795 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17798 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17799 Kito Cheng <kito.cheng@gmail.com>
17801 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17802 (riscv_leaf_function_p): Delete.
17803 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17805 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17807 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17809 (do_ifelse): New function.
17810 (do_isel): New function.
17811 (do_sub3): New function.
17812 (do_add3): New function.
17813 (do_load_mask_compare): New function.
17814 (do_overlap_load_compare): New function.
17815 (expand_compare_loop): New function.
17816 (expand_block_compare): Call expand_compare_loop() when appropriate.
17817 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17818 option description.
17819 (-mblock-compare-inline-loop-limit): New option.
17821 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17824 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17825 Reverse order of second and third operands in first alternative.
17826 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17827 of first and second elements in UNSPEC_VPERMR vector.
17828 (altivec_expand_vec_perm_le): Likewise.
17830 2018-01-08 Jeff Law <law@redhat.com>
17832 PR rtl-optimizatin/81308
17833 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17834 (process_switch): If group_case_labels makes a change, then set
17836 (pass_convert_switch::execute): If a switch is converted, then
17837 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17839 PR rtl-optimization/81308
17840 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17843 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17845 PR target/83663 - Revert r255946
17846 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17847 generation for cases where splatting a value is not useful.
17848 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17849 across a vec_duplicate and a paradoxical subreg forming a vector
17850 mode to a vec_concat.
17852 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17854 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17855 -march=armv8.3-a variants.
17856 * config/arm/t-multilib: Likewise.
17857 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17859 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17861 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17863 (cceq_ior_compare_complement): Give it a name so I can use it, and
17864 change boolean_or_operator predicate to boolean_operator so it can
17865 be used to generate a crand.
17866 (eqne): New code iterator.
17867 (bd/bd_neg): New code_attrs.
17868 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17869 a single define_insn.
17870 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17871 decrement (bdnzt/bdnzf/bdzt/bdzf).
17872 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17873 with the new names of the branch decrement patterns, and added the
17874 names of the branch decrement conditional patterns.
17876 2018-01-08 Richard Biener <rguenther@suse.de>
17878 PR tree-optimization/83563
17879 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17882 2018-01-08 Richard Biener <rguenther@suse.de>
17884 PR middle-end/83713
17885 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17887 2018-01-08 Richard Biener <rguenther@suse.de>
17889 PR tree-optimization/83685
17890 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17891 references to abnormals.
17893 2018-01-08 Richard Biener <rguenther@suse.de>
17896 * dwarf2out.c (output_indirect_strings): Handle empty
17897 skeleton_debug_str_hash.
17898 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17900 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17902 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17903 (emit_store_direct): Likewise.
17904 (arc_trampoline_adjust_address): Likewise.
17905 (arc_asm_trampoline_template): New function.
17906 (arc_initialize_trampoline): Use asm_trampoline_template.
17907 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17908 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17909 * config/arc/arc.md (flush_icache): Delete pattern.
17911 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17913 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17914 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17917 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17920 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17921 by not USED_FOR_TARGET.
17922 (make_pass_resolve_sw_modes): Likewise.
17924 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17926 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17929 2018-01-08 Richard Biener <rguenther@suse.de>
17931 PR middle-end/83580
17932 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17934 2018-01-08 Richard Biener <rguenther@suse.de>
17936 PR middle-end/83517
17937 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17939 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17941 PR middle-end/81897
17942 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17943 basic blocks with a small number of successors.
17944 (convert_control_dep_chain_into_preds): Improve handling of
17946 (dump_predicates): Split apart into...
17947 (dump_pred_chain): ...here...
17948 (dump_pred_info): ...and here.
17949 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17950 (can_chain_union_be_invalidated_p): Improve check for invalidation
17952 (uninit_uses_cannot_happen): Avoid unnecessary if
17953 convert_control_dep_chain_into_preds yielded nothing.
17955 2018-01-06 Martin Sebor <msebor@redhat.com>
17957 PR tree-optimization/83640
17958 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17959 subtracting negative offset from size.
17960 (builtin_access::overlap): Adjust offset bounds of the access to fall
17961 within the size of the object if possible.
17963 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17965 PR rtl-optimization/83699
17966 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17967 extract_bit_field_as_subreg to cases in which the extracted
17968 value is also a vector.
17970 * lra-constraints.c (process_alt_operands): Test for the equivalence
17971 substitutions when detecting a possible reload cycle.
17973 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17976 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17977 by default if flag_selective_schedling{,2}. Formatting fixes.
17979 PR rtl-optimization/83682
17980 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17981 if it has non-VECTOR_MODE element mode.
17982 (vec_duplicate_p): Likewise.
17984 PR middle-end/83694
17985 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17986 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17988 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17991 * config/i386/i386-builtin.def
17992 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17993 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17994 Require also OPTION_MASK_ISA_AVX512F in addition to
17995 OPTION_MASK_ISA_GFNI.
17996 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17997 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17998 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17999 to OPTION_MASK_ISA_GFNI.
18000 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18001 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18002 OPTION_MASK_ISA_AVX512BW.
18003 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18004 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18005 addition to OPTION_MASK_ISA_GFNI.
18006 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18007 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18008 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18009 to OPTION_MASK_ISA_GFNI.
18010 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18011 a requirement for all ISAs rather than any of them with a few
18013 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18015 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18016 bitmasks to be enabled with 3 exceptions, instead of requiring any
18017 enabled ISA with lots of exceptions.
18018 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18019 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18020 Change avx512bw in isa attribute to avx512f.
18021 * config/i386/sgxintrin.h: Add license boilerplate.
18022 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18023 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18024 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18025 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18027 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18028 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18029 temporarily sse2 rather than sse if not enabled already.
18032 * config/i386/sse.md (VI248_VLBW): Rename to ...
18033 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18034 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18035 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18036 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18037 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18038 mode iterator instead of VI248_VLBW.
18040 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18042 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18043 (record_modified): Skip clobbers; add debug output.
18044 (param_change_prob): Use sreal frequencies.
18046 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18048 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18049 punt for user-aligned variables.
18051 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18053 * tree-chrec.c (chrec_contains_symbols): Return true for
18056 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18059 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18060 of (x|y) == x for BICS pattern.
18062 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18064 PR tree-optimization/83605
18065 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18066 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18069 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18071 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18072 * config/epiphany/rtems.h: New file.
18074 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18075 Uros Bizjak <ubizjak@gmail.com>
18078 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18079 QIreg_operand instead of register_operand predicate.
18080 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18081 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18082 comments instead of -fmitigate[-_]rop.
18084 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18087 * cgraphunit.c (symbol_table::compile): Switch to text_section
18088 before calling assembly_start debug hook.
18089 * run-rtl-passes.c (run_rtl_passes): Likewise.
18092 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18094 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18095 range_int_cst_p rather than !symbolic_range_p before calling
18096 extract_range_from_multiplicative_op_1.
18098 2018-01-04 Jeff Law <law@redhat.com>
18100 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18101 redundant test in assertion.
18103 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18105 * doc/rtl.texi: Document machine_mode wrapper classes.
18107 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18109 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18110 using tree_to_uhwi.
18112 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18114 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18115 the VEC_PERM_EXPR fold to fail.
18117 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18120 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18121 to switched_sections.
18123 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18126 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18127 test for d.testing.
18129 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18132 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18133 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18135 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18138 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18139 is BLKmode and bitpos not zero or mode change is needed.
18141 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18144 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18147 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18150 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18151 instead of MULT rtx. Update all corresponding splitters.
18152 (*saddl_se): Ditto.
18153 (*ssub<modesuffix>): Ditto.
18154 (*ssubl_se): Ditto.
18155 (*cmp_sadd_di): Update split patterns.
18156 (*cmp_sadd_si): Ditto.
18157 (*cmp_sadd_sidi): Ditto.
18158 (*cmp_ssub_di): Ditto.
18159 (*cmp_ssub_si): Ditto.
18160 (*cmp_ssub_sidi): Ditto.
18161 * config/alpha/predicates.md (const23_operand): New predicate.
18162 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18163 Look for ASHIFT, not MULT inner operand.
18164 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18166 2018-01-04 Martin Liska <mliska@suse.cz>
18168 PR gcov-profile/83669
18169 * gcov.c (output_intermediate_file): Add version to intermediate
18171 * doc/gcov.texi: Document new field 'version' in intermediate
18172 file format. Fix location of '-k' option of gcov command.
18174 2018-01-04 Martin Liska <mliska@suse.cz>
18177 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18179 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18181 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18183 2018-01-03 Martin Sebor <msebor@redhat.com>
18185 PR tree-optimization/83655
18186 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18187 checking calls with invalid arguments.
18189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18191 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18192 (vectorizable_mask_load_store): Delete.
18193 (vectorizable_call): Return false for masked loads and stores.
18194 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18195 instead of gimple_assign_rhs1.
18196 (vectorizable_load): Handle IFN_MASK_LOAD.
18197 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18201 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18203 (vectorizable_mask_load_store): ...here.
18204 (vectorizable_load): ...and here.
18206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18208 * tree-vect-stmts.c (vect_build_all_ones_mask)
18209 (vect_build_zero_merge_argument): New functions, split out from...
18210 (vectorizable_load): ...here.
18212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18214 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18216 (vectorizable_mask_load_store): ...here.
18217 (vectorizable_store): ...and here.
18219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18221 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18223 (vectorizable_mask_load_store): ...here.
18225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18227 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18228 (vect_model_store_cost): Take a vec_load_store_type instead of a
18230 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18231 (vect_model_store_cost): Take a vec_load_store_type instead of a
18233 (vectorizable_mask_load_store): Update accordingly.
18234 (vectorizable_store): Likewise.
18235 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18239 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18240 IFN_MASK_LOAD calls here rather than...
18241 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18244 Alan Hayward <alan.hayward@arm.com>
18245 David Sherwood <david.sherwood@arm.com>
18247 * expmed.c (extract_bit_field_1): For vector extracts,
18248 fall back to extract_bit_field_as_subreg if vec_extract
18251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18252 Alan Hayward <alan.hayward@arm.com>
18253 David Sherwood <david.sherwood@arm.com>
18255 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18256 they are variable or constant sized.
18257 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18258 slots for constant-sized data.
18260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18261 Alan Hayward <alan.hayward@arm.com>
18262 David Sherwood <david.sherwood@arm.com>
18264 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18265 handling COND_EXPRs with boolean comparisons, try to find a better
18266 basis for the mask type than the boolean itself.
18268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18270 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18271 is calculated and how it can be overridden.
18272 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18273 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18275 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18279 Alan Hayward <alan.hayward@arm.com>
18280 David Sherwood <david.sherwood@arm.com>
18282 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18283 Remove the mode argument.
18284 (aarch64_simd_valid_immediate): Remove the mode and inverse
18286 * config/aarch64/iterators.md (bitsize): New iterator.
18287 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18288 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18289 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18290 aarch64_simd_valid_immediate.
18291 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18292 (aarch64_reg_or_bic_imm): Likewise.
18293 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18294 with an insn_type enum and msl with a modifier_type enum.
18295 Replace element_width with a scalar_mode. Change the shift
18296 to unsigned int. Add constructors for scalar_float_mode and
18297 scalar_int_mode elements.
18298 (aarch64_vect_float_const_representable_p): Delete.
18299 (aarch64_can_const_movi_rtx_p)
18300 (aarch64_simd_scalar_immediate_valid_for_move)
18301 (aarch64_simd_make_constant): Update call to
18302 aarch64_simd_valid_immediate.
18303 (aarch64_advsimd_valid_immediate_hs): New function.
18304 (aarch64_advsimd_valid_immediate): Likewise.
18305 (aarch64_simd_valid_immediate): Remove mode and inverse
18306 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18307 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18308 and aarch64_float_const_representable_p on the result.
18309 (aarch64_output_simd_mov_immediate): Remove mode argument.
18310 Update call to aarch64_simd_valid_immediate and use of
18311 simd_immediate_info.
18312 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18320 (mode_nunits): Likewise CONST_MODE_NUNITS.
18321 * machmode.def (ADJUST_NUNITS): Document.
18322 * genmodes.c (mode_data::need_nunits_adj): New field.
18323 (blank_mode): Update accordingly.
18324 (adj_nunits): New variable.
18325 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18327 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18328 listed in adj_nunits.
18329 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18330 listed in adj_nunits. Don't emit case statements for such modes.
18331 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18332 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18333 nothing if adj_nunits is nonnull.
18334 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18335 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18336 (emit_mode_fbit): Update use of print_maybe_const_decl.
18337 (emit_move_size): Likewise. Treat the array as non-const
18339 (emit_mode_adjustments): Handle adj_nunits.
18341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18343 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18344 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18345 (VECTOR_MODES): Use it.
18346 (make_vector_modes): Take the prefix as an argument.
18348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18349 Alan Hayward <alan.hayward@arm.com>
18350 David Sherwood <david.sherwood@arm.com>
18352 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18353 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18354 for MODE_VECTOR_BOOL.
18355 * machmode.def (VECTOR_BOOL_MODE): Document.
18356 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18357 (make_vector_bool_mode): New function.
18358 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18360 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18361 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18363 * stor-layout.c (int_mode_for_mode): Likewise.
18364 * tree.c (build_vector_type_for_mode): Likewise.
18365 * varasm.c (output_constant_pool_2): Likewise.
18366 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18367 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18368 for MODE_VECTOR_BOOL.
18369 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18370 of mode class checks.
18371 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18372 instead of a list of mode class checks.
18373 (expand_vector_scalar_condition): Likewise.
18374 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18377 Alan Hayward <alan.hayward@arm.com>
18378 David Sherwood <david.sherwood@arm.com>
18380 * machmode.h (mode_size): Change from unsigned short to
18382 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18383 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18384 or if measurement_type is not polynomial.
18385 (fixed_size_mode::includes_p): Check for constant-sized modes.
18386 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18387 return a poly_uint16 rather than an unsigned short.
18388 (emit_mode_size): Change the type of mode_size from unsigned short
18389 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18390 (emit_mode_adjustments): Cope with polynomial vector sizes.
18391 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18393 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18395 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18396 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18397 * caller-save.c (setup_save_areas): Likewise.
18398 (replace_reg_with_saved_mem): Likewise.
18399 * calls.c (emit_library_call_value_1): Likewise.
18400 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18401 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18402 (gen_lowpart_for_combine): Likewise.
18403 * convert.c (convert_to_integer_1): Likewise.
18404 * cse.c (equiv_constant, cse_insn): Likewise.
18405 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18406 (cselib_subst_to_values): Likewise.
18407 * dce.c (word_dce_process_block): Likewise.
18408 * df-problems.c (df_word_lr_mark_ref): Likewise.
18409 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18410 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18411 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18412 (rtl_for_decl_location): Likewise.
18413 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18414 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18415 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18416 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18417 (expand_expr_real_1): Likewise.
18418 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18419 (pad_below): Likewise.
18420 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18421 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18422 * ira.c (get_subreg_tracking_sizes): Likewise.
18423 * ira-build.c (ira_create_allocno_objects): Likewise.
18424 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18425 (ira_sort_regnos_for_alter_reg): Likewise.
18426 * ira-costs.c (record_operand_costs): Likewise.
18427 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18428 (resolve_simple_move): Likewise.
18429 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18430 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18431 (lra_constraints): Likewise.
18432 (CONST_POOL_OK_P): Reject variable-sized modes.
18433 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18434 (add_pseudo_to_slot, lra_spill): Likewise.
18435 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18436 * optabs-query.c (get_best_extraction_insn): Likewise.
18437 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18438 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18439 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18440 * recog.c (offsettable_address_addr_space_p): Likewise.
18441 * regcprop.c (maybe_mode_change): Likewise.
18442 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18443 * regrename.c (build_def_use): Likewise.
18444 * regstat.c (dump_reg_info): Likewise.
18445 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18446 (find_reloads, find_reloads_subreg_address): Likewise.
18447 * reload1.c (eliminate_regs_1): Likewise.
18448 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18449 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18450 (simplify_binary_operation_1, simplify_subreg): Likewise.
18451 * targhooks.c (default_function_arg_padding): Likewise.
18452 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18453 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18454 (verify_gimple_assign_ternary): Likewise.
18455 * tree-inline.c (estimate_move_cost): Likewise.
18456 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18457 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18458 (get_address_cost_ainc): Likewise.
18459 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18460 (vect_supportable_dr_alignment): Likewise.
18461 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18462 (vectorizable_reduction): Likewise.
18463 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18464 (vectorizable_operation, vectorizable_load): Likewise.
18465 * tree.c (build_same_sized_truth_vector_type): Likewise.
18466 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18467 * var-tracking.c (emit_note_insn_var_location): Likewise.
18468 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18469 (ADDR_VEC_ALIGN): Likewise.
18471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18472 Alan Hayward <alan.hayward@arm.com>
18473 David Sherwood <david.sherwood@arm.com>
18475 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18477 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18478 or if measurement_type is polynomial.
18479 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18480 * combine.c (make_extraction): Likewise.
18481 * dse.c (find_shift_sequence): Likewise.
18482 * dwarf2out.c (mem_loc_descriptor): Likewise.
18483 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18484 (extract_bit_field, extract_low_bits): Likewise.
18485 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18486 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18487 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18488 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18489 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18490 * reload.c (find_reloads): Likewise.
18491 * reload1.c (alter_reg): Likewise.
18492 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18493 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18494 * tree-if-conv.c (predicate_mem_writes): Likewise.
18495 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18496 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18497 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18498 * valtrack.c (dead_debug_insert_temp): Likewise.
18499 * varasm.c (mergeable_constant_section): Likewise.
18500 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18503 Alan Hayward <alan.hayward@arm.com>
18504 David Sherwood <david.sherwood@arm.com>
18506 * expr.c (expand_assignment): Cope with polynomial mode sizes
18507 when assigning to a CONCAT.
18509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18510 Alan Hayward <alan.hayward@arm.com>
18511 David Sherwood <david.sherwood@arm.com>
18513 * machmode.h (mode_precision): Change from unsigned short to
18515 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18517 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18518 or if measurement_type is not polynomial.
18519 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18520 in which the mode is already known to be a scalar_int_mode.
18521 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18522 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18524 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18525 for GET_MODE_PRECISION.
18526 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18527 for GET_MODE_PRECISION.
18528 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18530 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18531 (expand_field_assignment, make_extraction): Likewise.
18532 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18533 (get_last_value): Likewise.
18534 * convert.c (convert_to_integer_1): Likewise.
18535 * cse.c (cse_insn): Likewise.
18536 * expr.c (expand_expr_real_1): Likewise.
18537 * lra-constraints.c (simplify_operand_subreg): Likewise.
18538 * optabs-query.c (can_atomic_load_p): Likewise.
18539 * optabs.c (expand_atomic_load): Likewise.
18540 (expand_atomic_store): Likewise.
18541 * ree.c (combine_reaching_defs): Likewise.
18542 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18543 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18544 * tree.h (type_has_mode_precision_p): Likewise.
18545 * ubsan.c (instrument_si_overflow): Likewise.
18547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18548 Alan Hayward <alan.hayward@arm.com>
18549 David Sherwood <david.sherwood@arm.com>
18551 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18552 polynomial numbers of units.
18553 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18554 (valid_vector_subparts_p): New function.
18555 (build_vector_type): Remove temporary shim and take the number
18556 of units as a poly_uint64 rather than an int.
18557 (build_opaque_vector_type): Take the number of units as a
18558 poly_uint64 rather than an int.
18559 * tree.c (build_vector_from_ctor): Handle polynomial
18560 TYPE_VECTOR_SUBPARTS.
18561 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18562 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18563 (build_vector_from_val): If the number of units is variable,
18564 use build_vec_duplicate_cst for constant operands and
18565 VEC_DUPLICATE_EXPR otherwise.
18566 (make_vector_type): Remove temporary is_constant ().
18567 (build_vector_type, build_opaque_vector_type): Take the number of
18568 units as a poly_uint64 rather than an int.
18569 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18571 * cfgexpand.c (expand_debug_expr): Likewise.
18572 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18573 (store_constructor, expand_expr_real_1): Likewise.
18574 (const_scalar_mask_from_tree): Likewise.
18575 * fold-const-call.c (fold_const_reduction): Likewise.
18576 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18577 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18578 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18579 (fold_relational_const): Likewise.
18580 (native_interpret_vector): Likewise. Change the size from an
18581 int to an unsigned int.
18582 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18583 TYPE_VECTOR_SUBPARTS.
18584 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18585 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18586 duplicating a non-constant operand into a variable-length vector.
18587 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18588 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18589 * ipa-icf.c (sem_variable::equals): Likewise.
18590 * match.pd: Likewise.
18591 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18592 * print-tree.c (print_node): Likewise.
18593 * stor-layout.c (layout_type): Likewise.
18594 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18595 * tree-cfg.c (verify_gimple_comparison): Likewise.
18596 (verify_gimple_assign_binary): Likewise.
18597 (verify_gimple_assign_ternary): Likewise.
18598 (verify_gimple_assign_single): Likewise.
18599 * tree-pretty-print.c (dump_generic_node): Likewise.
18600 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18601 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18602 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18603 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18604 (vect_shift_permute_load_chain): Likewise.
18605 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18606 (expand_vector_condition, optimize_vector_constructor): Likewise.
18607 (lower_vec_perm, get_compute_type): Likewise.
18608 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18609 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18610 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18611 (vect_recog_mask_conversion_pattern): Likewise.
18612 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18613 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18614 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18615 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18616 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18617 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18618 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18619 (supportable_widening_operation): Likewise.
18620 (supportable_narrowing_operation): Likewise.
18621 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18623 * varasm.c (output_constant): Likewise.
18625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18626 Alan Hayward <alan.hayward@arm.com>
18627 David Sherwood <david.sherwood@arm.com>
18629 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18630 so that both the length == 3 and length != 3 cases set up their
18631 own permute vectors. Add comments explaining why we know the
18632 number of elements is constant.
18633 (vect_permute_load_chain): Likewise.
18635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18636 Alan Hayward <alan.hayward@arm.com>
18637 David Sherwood <david.sherwood@arm.com>
18639 * machmode.h (mode_nunits): Change from unsigned char to
18641 (ONLY_FIXED_SIZE_MODES): New macro.
18642 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18643 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18644 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18646 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18647 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18648 or if measurement_type is not polynomial.
18649 * genmodes.c (ZERO_COEFFS): New macro.
18650 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18652 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18653 Use ZERO_COEFFS when emitting initializers.
18654 * data-streamer.h (bp_pack_poly_value): New function.
18655 (bp_unpack_poly_value): Likewise.
18656 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18657 for GET_MODE_NUNITS.
18658 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18659 for GET_MODE_NUNITS.
18660 * tree.c (make_vector_type): Remove temporary shim and make
18661 the real function take the number of units as a poly_uint64
18662 rather than an int.
18663 (build_vector_type_for_mode): Handle polynomial nunits.
18664 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18665 * emit-rtl.c (const_vec_series_p_1): Likewise.
18666 (gen_rtx_CONST_VECTOR): Likewise.
18667 * fold-const.c (test_vec_duplicate_folding): Likewise.
18668 * genrecog.c (validate_pattern): Likewise.
18669 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18670 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18671 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18672 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18673 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18674 * rtlanal.c (subreg_get_info): Likewise.
18675 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18676 (vect_grouped_load_supported): Likewise.
18677 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18678 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18679 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18680 (simplify_const_unary_operation, simplify_binary_operation_1)
18681 (simplify_const_binary_operation, simplify_ternary_operation)
18682 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18683 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18684 instead of CONST_VECTOR_NUNITS.
18685 * varasm.c (output_constant_pool_2): Likewise.
18686 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18687 explicit-encoded elements in the XVEC for variable-length vectors.
18689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18691 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18694 Alan Hayward <alan.hayward@arm.com>
18695 David Sherwood <david.sherwood@arm.com>
18697 * coretypes.h (fixed_size_mode): Declare.
18698 (fixed_size_mode_pod): New typedef.
18699 * builtins.h (target_builtins::x_apply_args_mode)
18700 (target_builtins::x_apply_result_mode): Change type to
18701 fixed_size_mode_pod.
18702 * builtins.c (apply_args_size, apply_result_size, result_vector)
18703 (expand_builtin_apply_args_1, expand_builtin_apply)
18704 (expand_builtin_return): Update accordingly.
18706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18708 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18709 * cselib.c (cselib_hash_rtx): Likewise.
18710 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18711 CONST_VECTOR encoding.
18713 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18714 Jeff Law <law@redhat.com>
18717 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18718 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18719 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18720 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18723 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18724 explicitly probe *sp in a noreturn function if there were any callee
18725 register saves or frame pointer is needed.
18727 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18730 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18731 BLKmode for ternary, binary or unary expressions.
18734 * var-tracking.c (delete_vta_debug_insn): New inline function.
18735 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18736 insns from get_insns () to NULL instead of each bb separately.
18737 Use delete_vta_debug_insn. No longer static.
18738 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18739 delete_vta_debug_insns callers.
18740 * rtl.h (delete_vta_debug_insns): Declare.
18741 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18742 instead of variable_tracking_main.
18744 2018-01-03 Martin Sebor <msebor@redhat.com>
18746 PR tree-optimization/83603
18747 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18748 arguments past the endof the argument list in functions declared
18749 without a prototype.
18750 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18751 Avoid checking when arguments are null.
18753 2018-01-03 Martin Sebor <msebor@redhat.com>
18756 * doc/extend.texi (attribute const): Fix a typo.
18757 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18758 issuing -Wsuggest-attribute for void functions.
18760 2018-01-03 Martin Sebor <msebor@redhat.com>
18762 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18763 offset_int::from instead of wide_int::to_shwi.
18764 (maybe_diag_overlap): Remove assertion.
18765 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18766 * gimple-ssa-sprintf.c (format_directive): Same.
18767 (parse_directive): Same.
18768 (sprintf_dom_walker::compute_format_length): Same.
18769 (try_substitute_return_value): Same.
18771 2018-01-03 Jeff Law <law@redhat.com>
18773 PR middle-end/83654
18774 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18775 non-constant residual for zero at runtime and avoid probing in
18776 that case. Reorganize code for trailing problem to mirror handling
18779 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18781 PR tree-optimization/83501
18782 * tree-ssa-strlen.c (get_string_cst): New.
18783 (handle_char_store): Call get_string_cst.
18785 2018-01-03 Martin Liska <mliska@suse.cz>
18787 PR tree-optimization/83593
18788 * tree-ssa-strlen.c: Include tree-cfg.h.
18789 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18790 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18791 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18793 (strlen_dom_walker::before_dom_children): Call
18794 gimple_purge_dead_eh_edges. Dump tranformation with details
18796 (strlen_dom_walker::before_dom_children): Update call by adding
18797 new argument cleanup_eh.
18798 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18800 2018-01-03 Martin Liska <mliska@suse.cz>
18803 * cif-code.def (VARIADIC_THUNK): New enum value.
18804 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18807 2018-01-03 Jan Beulich <jbeulich@suse.com>
18809 * sse.md (mov<mode>_internal): Tighten condition for when to use
18810 vmovdqu<ssescalarsize> for TI and OI modes.
18812 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18814 Update copyright years.
18816 2018-01-03 Martin Liska <mliska@suse.cz>
18819 * ipa-visibility.c (function_and_variable_visibility): Skip
18820 functions with noipa attribure.
18822 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18824 * gcc.c (process_command): Update copyright notice dates.
18825 * gcov-dump.c (print_version): Ditto.
18826 * gcov.c (print_version): Ditto.
18827 * gcov-tool.c (print_version): Ditto.
18828 * gengtype.c (create_file): Ditto.
18829 * doc/cpp.texi: Bump @copying's copyright year.
18830 * doc/cppinternals.texi: Ditto.
18831 * doc/gcc.texi: Ditto.
18832 * doc/gccint.texi: Ditto.
18833 * doc/gcov.texi: Ditto.
18834 * doc/install.texi: Ditto.
18835 * doc/invoke.texi: Ditto.
18837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18839 * vector-builder.h (vector_builder::m_full_nelts): Change from
18840 unsigned int to poly_uint64.
18841 (vector_builder::full_nelts): Update prototype accordingly.
18842 (vector_builder::new_vector): Likewise.
18843 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18844 (vector_builder::operator ==): Likewise.
18845 (vector_builder::finalize): Likewise.
18846 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18847 Take the number of elements as a poly_uint64 rather than an
18849 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18850 from unsigned int to poly_uint64.
18851 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18852 (vec_perm_indices::new_vector): Likewise.
18853 (vec_perm_indices::length): Likewise.
18854 (vec_perm_indices::nelts_per_input): Likewise.
18855 (vec_perm_indices::input_nelts): Likewise.
18856 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18857 number of elements per input as a poly_uint64 rather than an
18858 unsigned int. Use the original encoding for variable-length
18859 vectors, rather than clamping each individual element.
18860 For the second and subsequent elements in each pattern,
18861 clamp the step and base before clamping their sum.
18862 (vec_perm_indices::series_p): Handle polynomial element counts.
18863 (vec_perm_indices::all_in_range_p): Likewise.
18864 (vec_perm_indices_to_tree): Likewise.
18865 (vec_perm_indices_to_rtx): Likewise.
18866 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18867 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18868 (tree_vector_builder::new_binary_operation): Handle polynomial
18869 element counts. Return false if we need to know the number
18870 of elements at compile time.
18871 * fold-const.c (fold_vec_perm): Punt if the number of elements
18872 isn't known at compile time.
18874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18876 * vec-perm-indices.h (vec_perm_builder): Change element type
18877 from HOST_WIDE_INT to poly_int64.
18878 (vec_perm_indices::element_type): Update accordingly.
18879 (vec_perm_indices::clamp): Handle polynomial element_types.
18880 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18881 (vec_perm_indices::all_in_range_p): Likewise.
18882 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18884 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18885 polynomial vec_perm_indices element types.
18886 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18887 * fold-const.c (fold_vec_perm): Likewise.
18888 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18889 * tree-vect-generic.c (lower_vec_perm): Likewise.
18890 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18891 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18892 element type to HOST_WIDE_INT.
18894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18895 Alan Hayward <alan.hayward@arm.com>
18896 David Sherwood <david.sherwood@arm.com>
18898 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18899 rather than an int. Use plus_constant.
18900 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18901 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18904 Alan Hayward <alan.hayward@arm.com>
18905 David Sherwood <david.sherwood@arm.com>
18907 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18908 a HOST_WIDE_INT to a poly_int64.
18910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18911 Alan Hayward <alan.hayward@arm.com>
18912 David Sherwood <david.sherwood@arm.com>
18914 * calls.c (load_register_parameters): Cope with polynomial
18915 mode sizes. Require a constant size for BLKmode parameters
18916 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18917 forces a parameter to be padded at the lsb end in order to
18918 fill a complete number of words, require the parameter size
18919 to be ordered wrt UNITS_PER_WORD.
18921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18922 Alan Hayward <alan.hayward@arm.com>
18923 David Sherwood <david.sherwood@arm.com>
18925 * reload1.c (spill_stack_slot_width): Change element type
18926 from unsigned int to poly_uint64_pod.
18927 (alter_reg): Treat mode sizes as polynomial.
18929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18930 Alan Hayward <alan.hayward@arm.com>
18931 David Sherwood <david.sherwood@arm.com>
18933 * reload.c (complex_word_subreg_p): New function.
18934 (reload_inner_reg_of_subreg, push_reload): Use it.
18936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18937 Alan Hayward <alan.hayward@arm.com>
18938 David Sherwood <david.sherwood@arm.com>
18940 * lra-constraints.c (process_alt_operands): Reject matched
18941 operands whose sizes aren't ordered.
18942 (match_reload): Refer to this check here.
18944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18945 Alan Hayward <alan.hayward@arm.com>
18946 David Sherwood <david.sherwood@arm.com>
18948 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18949 that the mode size is in the set {1, 2, 4, 8, 16}.
18951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18952 Alan Hayward <alan.hayward@arm.com>
18953 David Sherwood <david.sherwood@arm.com>
18955 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18956 Use plus_constant instead of gen_rtx_PLUS.
18958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18959 Alan Hayward <alan.hayward@arm.com>
18960 David Sherwood <david.sherwood@arm.com>
18962 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18963 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18964 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18965 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18966 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18967 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18968 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18969 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18970 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18971 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18973 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18974 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18975 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18976 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18977 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18978 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18979 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18980 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18981 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18982 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18984 * expr.c (emit_move_resolve_push): Treat the input and result
18985 of PUSH_ROUNDING as a poly_int64.
18986 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18987 (emit_push_insn): Likewise.
18988 * lra-eliminations.c (mark_not_eliminable): Likewise.
18989 * recog.c (push_operand): Likewise.
18990 * reload1.c (elimination_effects): Likewise.
18991 * rtlanal.c (nonzero_bits1): Likewise.
18992 * calls.c (store_one_arg): Likewise. Require the padding to be
18993 known at compile time.
18995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18996 Alan Hayward <alan.hayward@arm.com>
18997 David Sherwood <david.sherwood@arm.com>
18999 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19000 Use plus_constant instead of gen_rtx_PLUS.
19002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19003 Alan Hayward <alan.hayward@arm.com>
19004 David Sherwood <david.sherwood@arm.com>
19006 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19007 rather than an int.
19009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19010 Alan Hayward <alan.hayward@arm.com>
19011 David Sherwood <david.sherwood@arm.com>
19013 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19014 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19015 via stack temporaries. Treat the mode size as polynomial too.
19017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19018 Alan Hayward <alan.hayward@arm.com>
19019 David Sherwood <david.sherwood@arm.com>
19021 * expr.c (expand_expr_real_2): When handling conversions involving
19022 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19023 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19024 as a poly_uint64 too.
19026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19027 Alan Hayward <alan.hayward@arm.com>
19028 David Sherwood <david.sherwood@arm.com>
19030 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19033 Alan Hayward <alan.hayward@arm.com>
19034 David Sherwood <david.sherwood@arm.com>
19036 * combine.c (can_change_dest_mode): Handle polynomial
19037 REGMODE_NATURAL_SIZE.
19038 * expmed.c (store_bit_field_1): Likewise.
19039 * expr.c (store_constructor): Likewise.
19040 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19041 and polynomial REGMODE_NATURAL_SIZE.
19042 (gen_lowpart_common): Likewise.
19043 * reginfo.c (record_subregs_of_mode): Likewise.
19044 * rtlanal.c (read_modify_subreg_p): Likewise.
19046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19047 Alan Hayward <alan.hayward@arm.com>
19048 David Sherwood <david.sherwood@arm.com>
19050 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19051 numbers of elements.
19053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19054 Alan Hayward <alan.hayward@arm.com>
19055 David Sherwood <david.sherwood@arm.com>
19057 * match.pd: Cope with polynomial numbers of vector elements.
19059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19060 Alan Hayward <alan.hayward@arm.com>
19061 David Sherwood <david.sherwood@arm.com>
19063 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19064 in a POINTER_PLUS_EXPR.
19066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19067 Alan Hayward <alan.hayward@arm.com>
19068 David Sherwood <david.sherwood@arm.com>
19070 * omp-simd-clone.c (simd_clone_subparts): New function.
19071 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19072 (ipa_simd_modify_function_body): Likewise.
19074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19075 Alan Hayward <alan.hayward@arm.com>
19076 David Sherwood <david.sherwood@arm.com>
19078 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19079 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19080 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19081 (expand_vector_condition, vector_element): Likewise.
19082 (subparts_gt): New function.
19083 (get_compute_type): Use subparts_gt.
19084 (count_type_subparts): Delete.
19085 (expand_vector_operations_1): Use subparts_gt instead of
19086 count_type_subparts.
19088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19089 Alan Hayward <alan.hayward@arm.com>
19090 David Sherwood <david.sherwood@arm.com>
19092 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19093 (vect_compile_time_alias): ...this new function. Do the calculation
19094 on poly_ints rather than trees.
19095 (vect_prune_runtime_alias_test_list): Update call accordingly.
19097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19098 Alan Hayward <alan.hayward@arm.com>
19099 David Sherwood <david.sherwood@arm.com>
19101 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19103 (vect_schedule_slp_instance): Likewise.
19105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19106 Alan Hayward <alan.hayward@arm.com>
19107 David Sherwood <david.sherwood@arm.com>
19109 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19110 constant and extern definitions for variable-length vectors.
19111 (vect_get_constant_vectors): Note that the number of units
19112 is known to be constant.
19114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19115 Alan Hayward <alan.hayward@arm.com>
19116 David Sherwood <david.sherwood@arm.com>
19118 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19119 of units as polynomial. Choose between WIDE and NARROW based
19122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19123 Alan Hayward <alan.hayward@arm.com>
19124 David Sherwood <david.sherwood@arm.com>
19126 * tree-vect-stmts.c (simd_clone_subparts): New function.
19127 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19130 Alan Hayward <alan.hayward@arm.com>
19131 David Sherwood <david.sherwood@arm.com>
19133 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19134 vectors as polynomial. Use build_index_vector for
19135 IFN_GOMP_SIMD_LANE.
19137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19138 Alan Hayward <alan.hayward@arm.com>
19139 David Sherwood <david.sherwood@arm.com>
19141 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19142 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19143 for variable-length vectors.
19144 (vectorizable_mask_load_store): Treat the number of units as
19145 polynomial, asserting that it is constant if the condition has
19146 already been enforced.
19147 (vectorizable_store, vectorizable_load): Likewise.
19149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19150 Alan Hayward <alan.hayward@arm.com>
19151 David Sherwood <david.sherwood@arm.com>
19153 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19154 of units as polynomial. Punt if we can't tell at compile time
19155 which vector contains the final result.
19157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19158 Alan Hayward <alan.hayward@arm.com>
19159 David Sherwood <david.sherwood@arm.com>
19161 * tree-vect-loop.c (vectorizable_induction): Treat the number
19162 of units as polynomial. Punt on SLP inductions. Use an integer
19163 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19164 cast of such a series for variable-length floating-point
19167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19168 Alan Hayward <alan.hayward@arm.com>
19169 David Sherwood <david.sherwood@arm.com>
19171 * tree.h (build_index_vector): Declare.
19172 * tree.c (build_index_vector): New function.
19173 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19174 of units as polynomial, forcibly converting it to a constant if
19175 vectorizable_reduction has already enforced the condition.
19176 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19177 to create a {1,2,3,...} vector.
19178 (vectorizable_reduction): Treat the number of units as polynomial.
19179 Choose vectype_in based on the largest scalar element size rather
19180 than the smallest number of units. Enforce the restrictions
19183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19184 Alan Hayward <alan.hayward@arm.com>
19185 David Sherwood <david.sherwood@arm.com>
19187 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19188 number of units as polynomial.
19190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19191 Alan Hayward <alan.hayward@arm.com>
19192 David Sherwood <david.sherwood@arm.com>
19194 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19195 * target.def (autovectorize_vector_sizes): Return the vector sizes
19196 by pointer, using vector_sizes rather than a bitmask.
19197 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19198 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19199 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19201 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19202 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19203 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19204 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19205 * omp-general.c (omp_max_vf): Likewise.
19206 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19207 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19208 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19209 * tree-vect-slp.c (vect_slp_bb): Likewise.
19210 * doc/tm.texi: Regenerate.
19211 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19213 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19214 the vector size as a poly_uint64 rather than an unsigned int.
19215 (current_vector_size): Change from an unsigned int to a poly_uint64.
19216 (get_vectype_for_scalar_type): Update accordingly.
19217 * tree.h (build_truth_vector_type): Take the size and number of
19218 units as a poly_uint64 rather than an unsigned int.
19219 (build_vector_type): Add a temporary overload that takes
19220 the number of units as a poly_uint64 rather than an unsigned int.
19221 * tree.c (make_vector_type): Likewise.
19222 (build_truth_vector_type): Take the number of units as a poly_uint64
19223 rather than an unsigned int.
19225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19226 Alan Hayward <alan.hayward@arm.com>
19227 David Sherwood <david.sherwood@arm.com>
19229 * target.def (get_mask_mode): Take the number of units and length
19230 as poly_uint64s rather than unsigned ints.
19231 * targhooks.h (default_get_mask_mode): Update accordingly.
19232 * targhooks.c (default_get_mask_mode): Likewise.
19233 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19234 * doc/tm.texi: Regenerate.
19236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19237 Alan Hayward <alan.hayward@arm.com>
19238 David Sherwood <david.sherwood@arm.com>
19240 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19241 * omp-general.c (omp_max_vf): Likewise.
19242 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19243 (expand_omp_simd): Handle polynomial safelen.
19244 * omp-low.c (omplow_simd_context): Add a default constructor.
19245 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19246 (lower_rec_simd_input_clauses): Update accordingly.
19247 (lower_rec_input_clauses): Likewise.
19249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19250 Alan Hayward <alan.hayward@arm.com>
19251 David Sherwood <david.sherwood@arm.com>
19253 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19254 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19255 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19256 (vect_analyze_slp_cost): Likewise.
19257 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19258 (vect_model_load_cost): Likewise.
19260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19261 Alan Hayward <alan.hayward@arm.com>
19262 David Sherwood <david.sherwood@arm.com>
19264 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19265 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19266 from an unsigned int * to a poly_uint64_pod *.
19267 (calculate_unrolling_factor): New function.
19268 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19271 Alan Hayward <alan.hayward@arm.com>
19272 David Sherwood <david.sherwood@arm.com>
19274 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19275 from an unsigned int to a poly_uint64.
19276 (_loop_vec_info::slp_unrolling_factor): Likewise.
19277 (_loop_vec_info::vectorization_factor): Change from an int
19279 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19280 (vect_get_num_vectors): New function.
19281 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19282 (vect_get_num_copies): Use vect_get_num_vectors.
19283 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19284 to an unsigned int *.
19285 (vect_analyze_data_refs): Change min_vf from an int * to a
19287 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19288 than an unsigned HOST_WIDE_INT.
19289 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19290 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19291 to an unsigned int *.
19292 (vect_analyze_data_ref_dependences): Likewise.
19293 (vect_compute_data_ref_alignment): Handle polynomial vf.
19294 (vect_enhance_data_refs_alignment): Likewise.
19295 (vect_prune_runtime_alias_test_list): Likewise.
19296 (vect_shift_permute_load_chain): Likewise.
19297 (vect_supportable_dr_alignment): Likewise.
19298 (dependence_distance_ge_vf): Take the vectorization factor as a
19299 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19300 (vect_analyze_data_refs): Change min_vf from an int * to a
19302 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19303 vfm1 as a poly_uint64 rather than an int. Make the same change
19304 for the returned bound_scalar.
19305 (vect_gen_vector_loop_niters): Handle polynomial vf.
19306 (vect_do_peeling): Likewise. Update call to
19307 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19308 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19310 * tree-vect-loop.c (vect_determine_vectorization_factor)
19311 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19312 (vect_get_known_peeling_cost): Likewise.
19313 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19314 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19315 (vect_transform_loop): Likewise. Use the lowest possible VF when
19316 updating the upper bounds of the loop.
19317 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19318 rather than an int.
19319 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19320 polynomial unroll factors.
19321 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19322 (vect_make_slp_decision): Likewise.
19323 (vect_supported_load_permutation_p): Likewise, and polynomial
19325 (vect_analyze_slp_cost): Handle polynomial vf.
19326 (vect_slp_analyze_node_operations): Likewise.
19327 (vect_slp_analyze_bb_1): Likewise.
19328 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19329 than an unsigned HOST_WIDE_INT.
19330 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19331 (vectorizable_load): Handle polynomial vf.
19332 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19334 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19337 Alan Hayward <alan.hayward@arm.com>
19338 David Sherwood <david.sherwood@arm.com>
19340 * match.pd: Handle bit operations involving three constants
19341 and try to fold one pair.
19343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19345 * tree-vect-loop-manip.c: Include gimple-fold.h.
19346 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19347 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19348 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19349 Add a path that uses a step of VF instead of 1, but disable it
19351 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19352 and niters_no_overflow parameters. Update calls to
19353 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19354 Create a new SSA name if the latter choses to use a ste other
19355 than zero, and return it via niters_vector_mult_vf_var.
19356 * tree-vect-loop.c (vect_transform_loop): Update calls to
19357 vect_do_peeling, vect_gen_vector_loop_niters and
19358 slpeel_make_loop_iterate_ntimes.
19359 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19360 (vect_gen_vector_loop_niters): Update declarations after above changes.
19362 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19364 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19365 128-bit round to integer instructions.
19366 (ceil<mode>2): Likewise.
19367 (btrunc<mode>2): Likewise.
19368 (round<mode>2): Likewise.
19370 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19372 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19373 unaligned VSX load/store on P8/P9.
19374 (expand_block_clear): Allow the use of unaligned VSX
19375 load/store on P8/P9.
19377 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19379 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19381 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19382 swap associated with both a load and a store.
19384 2018-01-02 Andrew Waterman <andrew@sifive.com>
19386 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19387 * config/riscv/riscv.md (clear_cache): Use it.
19389 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19391 * web.c: Remove out-of-date comment.
19393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19395 * expr.c (fixup_args_size_notes): Check that any existing
19396 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19397 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19398 (emit_single_push_insn): ...here.
19400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19402 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19403 (const_vector_encoded_nelts): New function.
19404 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19405 (const_vector_int_elt, const_vector_elt): Declare.
19406 * emit-rtl.c (const_vector_int_elt_1): New function.
19407 (const_vector_elt): Likewise.
19408 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19409 of CONST_VECTOR_ELT.
19411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19413 * expr.c: Include rtx-vector-builder.h.
19414 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19415 directly on the tree encoding.
19416 (const_vector_from_tree): Likewise.
19417 * optabs.c: Include rtx-vector-builder.h.
19418 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19419 sequence of "u" values.
19420 * vec-perm-indices.c: Include rtx-vector-builder.h.
19421 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19422 directly on the vec_perm_indices encoding.
19424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19426 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19427 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19428 * rtx-vector-builder.h: New file.
19429 * rtx-vector-builder.c: Likewise.
19430 * rtl.h (rtx_def::u2): Add a const_vector field.
19431 (CONST_VECTOR_NPATTERNS): New macro.
19432 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19433 (CONST_VECTOR_DUPLICATE_P): Likewise.
19434 (CONST_VECTOR_STEPPED_P): Likewise.
19435 (CONST_VECTOR_ENCODED_ELT): Likewise.
19436 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19437 (unwrap_const_vec_duplicate): Likewise.
19438 (const_vec_series_p): Check for a non-duplicated vector encoding.
19439 Say that the function only returns true for integer vectors.
19440 * emit-rtl.c: Include rtx-vector-builder.h.
19441 (gen_const_vec_duplicate_1): Delete.
19442 (gen_const_vector): Call gen_const_vec_duplicate instead of
19443 gen_const_vec_duplicate_1.
19444 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19445 (gen_const_vec_duplicate): Use rtx_vector_builder.
19446 (gen_const_vec_series): Likewise.
19447 (gen_rtx_CONST_VECTOR): Likewise.
19448 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19449 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19450 Build a new vector rather than modifying a CONST_VECTOR in-place.
19451 (handle_special_swappables): Update call accordingly.
19452 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19453 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19454 Build a new vector rather than modifying a CONST_VECTOR in-place.
19455 (handle_special_swappables): Update call accordingly.
19457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19459 * simplify-rtx.c (simplify_const_binary_operation): Use
19460 CONST_VECTOR_ELT instead of XVECEXP.
19462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19464 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19465 the selector elements to be different from the data elements
19466 if the selector is a VECTOR_CST.
19467 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19468 ssizetype for the selector.
19470 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19472 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19473 before testing each element individually.
19474 * tree-vect-generic.c (lower_vec_perm): Likewise.
19476 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19478 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19479 * selftest-run-tests.c (selftest::run_tests): Call it.
19480 * vector-builder.h (vector_builder::operator ==): New function.
19481 (vector_builder::operator !=): Likewise.
19482 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19483 (vec_perm_indices::all_from_input_p): New function.
19484 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19485 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19486 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19487 instead of reading the VECTOR_CST directly. Detect whether both
19488 vector inputs are the same before constructing the vec_perm_indices,
19489 and update the number of inputs argument accordingly. Use the
19490 utility functions added above. Only construct sel2 if we need to.
19492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19494 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19495 the broadcast of the low byte.
19496 (expand_mult_highpart): Use an explicit encoding for the permutes.
19497 * optabs-query.c (can_mult_highpart_p): Likewise.
19498 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19499 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19500 (vectorizable_bswap): Likewise.
19501 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19502 explicit encoding for the power-of-2 permutes.
19503 (vect_permute_store_chain): Likewise.
19504 (vect_grouped_load_supported): Likewise.
19505 (vect_permute_load_chain): Likewise.
19507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19509 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19510 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19511 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19512 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19513 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19514 (vect_gen_perm_mask_any): Likewise.
19516 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19518 * int-vector-builder.h: New file.
19519 * vec-perm-indices.h: Include int-vector-builder.h.
19520 (vec_perm_indices): Redefine as an int_vector_builder.
19521 (auto_vec_perm_indices): Delete.
19522 (vec_perm_builder): Redefine as a stand-alone class.
19523 (vec_perm_indices::vec_perm_indices): New function.
19524 (vec_perm_indices::clamp): Likewise.
19525 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19526 (vec_perm_indices::new_vector): New function.
19527 (vec_perm_indices::new_expanded_vector): Update for new
19528 vec_perm_indices class.
19529 (vec_perm_indices::rotate_inputs): New function.
19530 (vec_perm_indices::all_in_range_p): Operate directly on the
19531 encoded form, without computing elided elements.
19532 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19533 encoding. Update for new vec_perm_indices class.
19534 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19535 the given vec_perm_builder.
19536 (expand_vec_perm_var): Update vec_perm_builder constructor.
19537 (expand_mult_highpart): Use vec_perm_builder instead of
19538 auto_vec_perm_indices.
19539 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19540 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19541 or double series encoding as appropriate.
19542 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19543 vec_perm_indices instead of auto_vec_perm_indices.
19544 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19545 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19546 (vect_permute_store_chain): Likewise.
19547 (vect_grouped_load_supported): Likewise.
19548 (vect_permute_load_chain): Likewise.
19549 (vect_shift_permute_load_chain): Likewise.
19550 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19551 (vect_transform_slp_perm_load): Likewise.
19552 (vect_schedule_slp_instance): Likewise.
19553 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19554 (vectorizable_mask_load_store): Likewise.
19555 (vectorizable_bswap): Likewise.
19556 (vectorizable_store): Likewise.
19557 (vectorizable_load): Likewise.
19558 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19559 vec_perm_indices instead of auto_vec_perm_indices. Use
19560 tree_to_vec_perm_builder to read the vector from a tree.
19561 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19562 vec_perm_builder instead of a vec_perm_indices.
19563 (have_whole_vector_shift): Use vec_perm_builder and
19564 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19565 truncation to calc_vec_perm_mask_for_shift.
19566 (vect_create_epilog_for_reduction): Likewise.
19567 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19568 from auto_vec_perm_indices to vec_perm_indices.
19569 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19570 instead of changing individual elements.
19571 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19572 the vector in d.perm.
19573 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19574 from auto_vec_perm_indices to vec_perm_indices.
19575 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19576 instead of changing individual elements.
19577 (arm_vectorize_vec_perm_const): Use new_vector to install
19578 the vector in d.perm.
19579 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19580 Update vec_perm_builder constructor.
19581 (rs6000_expand_interleave): Likewise.
19582 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19583 (rs6000_expand_interleave): Likewise.
19585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19587 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19588 to qimode could truncate the indices.
19589 * optabs.c (expand_vec_perm_var): Likewise.
19591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19593 * Makefile.in (OBJS): Add vec-perm-indices.o.
19594 * vec-perm-indices.h: New file.
19595 * vec-perm-indices.c: Likewise.
19596 * target.h (vec_perm_indices): Replace with a forward class
19598 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19599 * optabs.h: Include vec-perm-indices.h.
19600 (expand_vec_perm): Delete.
19601 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19602 (expand_vec_perm_const): Declare.
19603 * target.def (vec_perm_const_ok): Replace with...
19604 (vec_perm_const): ...this new hook.
19605 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19606 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19607 * doc/tm.texi: Regenerate.
19608 * optabs.def (vec_perm_const): Delete.
19609 * doc/md.texi (vec_perm_const): Likewise.
19610 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19611 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19612 expand_vec_perm for constant permutation vectors. Assert that
19613 the mode of variable permutation vectors is the integer equivalent
19614 of the mode that is being permuted.
19615 * optabs-query.h (selector_fits_mode_p): Declare.
19616 * optabs-query.c: Include vec-perm-indices.h.
19617 (selector_fits_mode_p): New function.
19618 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19619 is defined, instead of checking whether the vec_perm_const_optab
19620 exists. Use targetm.vectorize.vec_perm_const instead of
19621 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19622 fit in the vector mode before using a variable permute.
19623 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19624 vec_perm_indices instead of an rtx.
19625 (expand_vec_perm): Replace with...
19626 (expand_vec_perm_const): ...this new function. Take the selector
19627 as a vec_perm_indices rather than an rtx. Also take the mode of
19628 the selector. Update call to shift_amt_for_vec_perm_mask.
19629 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19630 Use vec_perm_indices::new_expanded_vector to expand the original
19631 selector into bytes. Check whether the indices fit in the vector
19632 mode before using a variable permute.
19633 (expand_vec_perm_var): Make global.
19634 (expand_mult_highpart): Use expand_vec_perm_const.
19635 * fold-const.c: Includes vec-perm-indices.h.
19636 * tree-ssa-forwprop.c: Likewise.
19637 * tree-vect-data-refs.c: Likewise.
19638 * tree-vect-generic.c: Likewise.
19639 * tree-vect-loop.c: Likewise.
19640 * tree-vect-slp.c: Likewise.
19641 * tree-vect-stmts.c: Likewise.
19642 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19644 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19645 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19646 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19647 (aarch64_vectorize_vec_perm_const): ...this new function.
19648 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19649 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19650 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19651 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19652 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19653 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19654 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19656 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19657 check for NEON modes.
19658 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19659 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19660 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19661 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19663 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19664 the old VEC_PERM_CONST conditions.
19665 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19666 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19667 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19668 (ia64_vectorize_vec_perm_const_ok): Merge into...
19669 (ia64_vectorize_vec_perm_const): ...this new function.
19670 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19671 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19672 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19673 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19674 * config/mips/mips.c (mips_expand_vec_perm_const)
19675 (mips_vectorize_vec_perm_const_ok): Merge into...
19676 (mips_vectorize_vec_perm_const): ...this new function.
19677 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19678 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19679 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19680 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19681 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19682 (rs6000_expand_vec_perm_const): Delete.
19683 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19685 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19686 (altivec_expand_vec_perm_const_le): Take each operand individually.
19687 Operate on constant selectors rather than rtxes.
19688 (altivec_expand_vec_perm_const): Likewise. Update call to
19689 altivec_expand_vec_perm_const_le.
19690 (rs6000_expand_vec_perm_const): Delete.
19691 (rs6000_vectorize_vec_perm_const_ok): Delete.
19692 (rs6000_vectorize_vec_perm_const): New function.
19693 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19694 an element count and rtx array.
19695 (rs6000_expand_extract_even): Update call accordingly.
19696 (rs6000_expand_interleave): Likewise.
19697 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19698 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19699 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19700 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19701 (rs6000_expand_vec_perm_const): Delete.
19702 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19703 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19704 (altivec_expand_vec_perm_const_le): Take each operand individually.
19705 Operate on constant selectors rather than rtxes.
19706 (altivec_expand_vec_perm_const): Likewise. Update call to
19707 altivec_expand_vec_perm_const_le.
19708 (rs6000_expand_vec_perm_const): Delete.
19709 (rs6000_vectorize_vec_perm_const_ok): Delete.
19710 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19711 reference to the SPE evmerge intructions.
19712 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19713 an element count and rtx array.
19714 (rs6000_expand_extract_even): Update call accordingly.
19715 (rs6000_expand_interleave): Likewise.
19716 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19717 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19723 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19724 vector mode and that that mode matches the mode of the data
19726 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19727 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19728 directly using expand_vec_perm_1 when forcing selectors into
19730 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19732 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19734 * optabs-query.h (can_vec_perm_p): Delete.
19735 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19736 * optabs-query.c (can_vec_perm_p): Split into...
19737 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19738 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19739 particular selector is valid.
19740 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19741 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19742 (vect_grouped_load_supported): Likewise.
19743 (vect_shift_permute_load_chain): Likewise.
19744 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19745 (vect_transform_slp_perm_load): Likewise.
19746 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19747 (vectorizable_bswap): Likewise.
19748 (vect_gen_perm_mask_checked): Likewise.
19749 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19750 implementations of variable permutation vectors into account
19751 when deciding which selector to use.
19752 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19753 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19754 with a false third argument.
19755 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19756 to test whether the constant selector is valid and can_vec_perm_var_p
19757 to test whether a variable selector is valid.
19759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19761 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19762 * optabs-query.c (can_vec_perm_p): Likewise.
19763 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19764 instead of vec_perm_indices.
19765 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19766 (vect_gen_perm_mask_checked): Likewise,
19767 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19768 (vect_gen_perm_mask_checked): Likewise,
19770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19772 * optabs-query.h (qimode_for_vec_perm): Declare.
19773 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19774 (qimode_for_vec_perm): ...this new function.
19775 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19777 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19779 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19780 does not have a conditional at the top.
19782 2018-01-02 Richard Biener <rguenther@suse.de>
19784 * ipa-inline.c (big_speedup_p): Fix expression.
19786 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19789 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19792 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19796 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19797 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19798 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19799 cond_taken_branch_cost 3->4.
19801 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19803 PR tree-optimization/83581
19804 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19805 TODO_cleanup_cfg if any changes have been made.
19807 PR middle-end/83608
19808 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19809 convert_modes if target mode has the right side, but different mode
19812 PR middle-end/83609
19813 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19814 last argument when extracting from CONCAT. If either from_real or
19815 from_imag is NULL, use expansion through memory. If result is not
19816 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19817 the parts directly to inner mode, if even that fails, use expansion
19820 PR middle-end/83623
19821 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19822 check for bswap in mode rather than HImode and use that in expand_unop
19825 Copyright (C) 2018 Free Software Foundation, Inc.
19827 Copying and distribution of this file, with or without modification,
19828 are permitted in any medium without royalty provided the copyright
19829 notice and this notice are preserved.